article stringlengths 1.98k 169k | summary stringlengths 1.01k 4.15k | section_headings listlengths 2 38 | keywords listlengths 0 12 | year stringclasses 11 values | title stringlengths 30 189 |
|---|---|---|---|---|---|
In embryonic stem cells ( ESCs ) , the Tip60 histone acetyltransferase activates genes required for proliferation and silences genes that promote differentiation . Here we show that the class II histone deacetylase Hdac6 co-purifies with Tip60-p400 complex from ESCs . Hdac6 is necessary for regulation of most Tip60-p400 target genes , particularly those repressed by the complex . Unlike differentiated cells , where Hdac6 is mainly cytoplasmic , Hdac6 is largely nuclear in ESCs , neural stem cells ( NSCs ) , and some cancer cell lines , and interacts with Tip60-p400 in each . Hdac6 localizes to promoters bound by Tip60-p400 in ESCs , binding downstream of transcription start sites . Surprisingly , Hdac6 does not appear to deacetylate histones , but rather is required for Tip60-p400 binding to many of its target genes . Finally , we find that , like canonical subunits of Tip60-p400 , Hdac6 is necessary for robust ESC differentiation . These data suggest that Hdac6 plays a major role in the modulation of Tip60-p400 function in stem cells .
ESC self-renewal and differentiation are controlled by multiple pathways: exogenous factors that act through well-defined signaling pathways that are also employed in adult cells , and a network of nuclear factors that regulate the ESC transcriptome ( Hanna et al . , 2010 ) . Regulators of gene expression can be further sub-divided into ( i ) sequence-specific transcription factors , including ESC-specific master regulators , ( ii ) non-coding RNAs that act both in cis and in trans to regulate specific subsets of genes , and ( iii ) chromatin regulatory complexes , most of which are expressed in multiple cell and tissue types , and often act very broadly in the genome to covalently modify histones , remodel nucleosomes , or modify higher-order chromatin folding ( Hanna et al . , 2010; Young , 2011 ) . A number of chromatin regulators have been identified from RNA-interference screens or traditional knockout studies that are important for various features of ESC identity . However , for most chromatin regulatory complexes , several key questions remain , including how they find their genomic targets , how their catalytic activities lead to alteration of gene expression , and how the activities of these factors are altered to facilitate differentiation . In mammals , several chromatin remodeling complexes are modular , with distinct forms expressed in different cell or tissue types , or sometimes within the same cells ( Wang et al . , 1996; Ho et al . , 2009; Ramírez and Hagman , 2009; Fazzio and Panning , 2010; Hanna et al . , 2010 ) . For example , the mammalian SWI/SNF-family complex BAF ( Brg1/Brahma Associated Factor ) consists of several related complexes with many shared subunits , plus a few subunits that are specific to each particular cell type . In particular , when neural progenitors differentiate into neurons in mouse , two BAF subunits are replaced with two paralogous subunits that shift BAF from a factor promoting self-renewal to one that promotes differentiation ( Lessard et al . , 2007; Yoo et al . , 2009; Hanna et al . , 2010; Young , 2011 ) . Another unique combination of subunits , different from those observed in differentiated cells , comprises BAF complex from ESCs ( esBAF ) ( Ho et al . , 2009 ) . Similarly , multiple forms of PRC1 ( Polycomb Repressive Complex 1 ) have been purified from human and mouse cells that each contain the Ring1a/b ubiquitin ligase , but have different arrays of accessory proteins that confer distinct target specificity and activities ( Gao et al . , 2012; Tavares et al . , 2012 ) . Tip60-p400 has been purified from cancer cell lines as a 17 subunit chromatin remodeling complex with two chromatin remodeling activities: the Tip60 ( also known as Kat5 ) subunit acetylates the N-terminal tails of histones H2A , H4 , and a number of transcription factors , while the p400 subunit mediates exchange of H2A–H2B dimers for H2AZ–H2B dimers within nucleosomes ( Doyon et al . , 2004; Cai et al . , 2005; Squatrito et al . , 2006 ) . In somatic cells , Tip60-p400 serves mainly as a transcriptional co-activator that functions with numerous sequence-specific transcription factors to activate gene expression ( Brady et al . , 1999; Baek et al . , 2002; Frank et al . , 2003; Legube et al . , 2004 ) . In contrast , while Tip60-p400 promotes expression of some genes required for cellular proliferation and cell cycle regulation in ESCs , its most prominent function is to silence genes that are active during differentiation ( Fazzio et al . , 2008a , 2008b ) . RNAi-mediated knockdown ( KD ) of several Tip60-p400 subunits in ESCs individually induces a phenotype in which differentiation and ESC markers are expressed simultaneously , proliferation is reduced , the cell cycle is altered , and cells exhibit diminished self-renewal and pluripotency ( Fazzio et al . , 2008a ) . Consistent with these phenotypes , mice homozygous for a Tip60 deletion allele die at the pre-implantation stage ( Hu et al . , 2009 ) . It remains unknown why Tip60-p400 functions mainly as a repressor of differentiation gene expression in ESCs rather than an activator of expressed genes , as it does in most cell types examined . Similarly , treatment of ESCs with Trichostatin A ( TSA ) , a drug that broadly inhibits class I and II HDACs and results in elevated acetylation of most lysines targeted by HATs , promotes morphological changes similar to those observed upon KD of Tip60-p400 subunits ( McCool et al . , 2007; Karantzal et al . , 2008 ) . Therefore , maintenance of proper levels of histone acetylation appears to be essential to perpetuate the pluripotent state , as neither significant increases nor decreases in histone acetylation appear to be compatible with ESC self-renewal . However , TSA also inhibits several HDAC family members known to target acetylated lysines on non-histone proteins , leaving open the possibility that these targets play an equal or greater role in maintenance of ESC self-renewal . Furthermore , deletion or KD of several individual HDACs in ESCs produces phenotypes that differ substantially from those of Tip60-p400 subunits ( McBurney et al . , 2003; Dovey et al . , 2010 ) , suggesting that different HDACs perform different functions in ESC self-renewal . In this study , we interrogate the composition of Tip60-p400 complex in mouse ESCs in order to identify unique interacting proteins that might account for its altered functional repertoire in this cell type . We find that the class II histone deacetylase ( HDAC ) , Hdac6 , is a stable interaction partner with Tip60-p400 in ESCs , but not mouse embryo fibroblasts ( MEFs ) . Subsequent analyses revealed that Hdac6 also interacts with Tip60-p400 in adult neural stem cells from the brain and some cancer cell lines , but is sequestered away from Tip60-p400 in the cytoplasm of most differentiated cell types , as previously reported ( Verdel et al . , 2000; Hubbert et al . , 2002; Kawaguchi et al . , 2003; Valenzuela-Fernández et al . , 2008 ) . We show that Hdac6 is necessary for regulation of most Tip60-target genes in ESCs , particularly differentiation genes repressed by Tip60-p400 in ESCs . Surprisingly , while its deacetylase domains are required for silencing of differentiation genes , Hdac6 does not regulate gene expression by deacetylating histones near the promoters of Tip60-p400 targets . Instead , the catalytic domains of Hdac6 are required for its interaction with Tip60-p400 . Furthermore , we find that Hdac6 is necessary for normal Tip60-p400 enrichment at its gene targets , just downstream of their transcription start sites ( TSSs ) , suggesting that Hdac6 helps recruit Tip60-p400 complex to many target gene promoters . Finally , we show that Hdac6 is necessary for several major functions of Tip60-p400 in ESCs , as both Tip60- and Hdac6-deficient ESCs have defects in formation of single colonies , reduced proliferation rates , and defects in differentiation . However , unlike Tip60 KD , Hdac6 KD does not prevent ESC self-renewal . Thus , Hdac6 is a component of a novel , stem cell-specific , form of Tip60-p400 complex that is necessary for gene regulation and normal differentiation in ESCs .
Tip60-p400 complex has roles in both activation and repression of transcription in most cell types where it has been examined . However , in ESCs , Tip60-p400 is required for repression of many more genes than it activates , raising the possibility that a unique form of Tip60-p400 complex that might be expressed in ESCs that shifts the balance of its activity toward a more repressive role . To test this possibility , we targeted a 36 amino acid 6-histidine-3-FLAG ( H3F ) tag to the C-terminus of one copy of the endogenous Tip60 gene in murine ESCs ( Figure 1—figure supplement 1 ) , performed double affinity purifications from tagged or untagged cells ( Figure 1A ) , and identified proteins that co-purified with Tip60 using LC-MS/MS ( Table 1 ) . By this approach , we identified 16 of 17 known subunits ( Ikura et al . , 2000; Cai et al . , 2003 , 2005; Doyon et al . , 2004; Altaf et al . , 2009 ) of Tip60-p400 , suggesting that expression of the tagged form of Tip60 from its endogenous locus allowed for normal complex formation . Furthermore , we observed a number of novel Tip60-p400-interacting proteins , including chromatin regulatory proteins and transcription factors . To test for cell type specificity of Tip60-p400 complexes we generated a knock-in mouse harboring Tip60-H3F , isolated embryonic fibroblasts and repeated the purification . We observed several bands within Tip60 purifications from ESCs that were not observed in purifications from Tip60-H3F MEFs or untagged cells ( Figure 1A ) , consistent with the possibility that ESCs express a distinct form of Tip60-p400 complex . 10 . 7554/eLife . 01557 . 003Figure 1 . Identification of Tip60-p400-interacting proteins in ESCs . ( A ) Silver stained gel of purified Tip60 complex from Tip60-H3F ESCs and MEFs , along with untagged control cells . ( B ) Validation of Hdac6 interaction . Western blots for Hdac6 , p400 , and Dmap1 following immunoprecipitation with anti-FLAG antibody from nuclear extracts derived from the indicated ESC lines . ( C ) Tip60 complexes purified from Tip60-H3F ESC and MEF nuclear extracts were subjected to Western blotting for Hdac6 , Dmap1 , and FLAG . ( D ) Tip60-H3F was immunoprecipitated from MEF whole cell extracts as above and blotted for the indicated proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 00310 . 7554/eLife . 01557 . 004Figure 1—figure supplement 1 . Targeting of H3F tag to C-terminus of endogenous Tip60 gene . Shown are the 3' end of the Tip60 gene plus downstream sequence ( below ) and the targeting construct for introducing the C-terminal tandem 6-His-3-FLAG ( H3F ) tag ( above ) . The counter-selection cassette within the targeting construct is omitted for brevity . See ‘Materials and methods’ for details of ESC targeting . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 00410 . 7554/eLife . 01557 . 005Figure 1—figure supplement 2 . Reciprocal co-immunoprecipitation of Tip60 and Hdac6 . ( Above ) Immunoprecipitations from untagged or Hdac6-H3F ESCs with or without stable expression of Tip60-GFP ( as indicated ) were subjected to Western blotting for Tip60-GFP or Hdac6 . ( Below ) Mild overexpression of Tip60 in Tip60-GFP cells . RT-qPCR of Tip60 or Hdac6 in untagged and Hdac6-H3F cells stably expressing Tip60-GFP . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 00510 . 7554/eLife . 01557 . 006Figure 1—figure supplement 3 . Hdac6 does not interact with RNA Polymerase II in ESCs . Western blot of purified Tip60 complex from ESCs for RNA polymerase II ( RNA Pol II ) and Dmap1 . ( Tip60-p400 subunit Dmap1 is shown to confirm the presence of subunits of Tip60-p400 complex after purification . ) DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 00610 . 7554/eLife . 01557 . 007Figure 1—figure supplement 4 . Hdac6 interaction with Tip60-p400 is independent of DNA . DNase I ( left ) or ethidium bromide ( right ) was added into nuclear extracts during immunoprecipitation , and the Tip60-interacting proteins were analyzed by Western blotting with the indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 00710 . 7554/eLife . 01557 . 008Figure 1—figure supplement 5 . Hdac6 interaction with Tip60-p400 is resistant to high salt concentrations . Tip60-H3F was purified from ESCs as above and subjected to washes with buffer containing the salt concentrations indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 00810 . 7554/eLife . 01557 . 009Table 1 . Proteins co-purifying with Tip60-H3F in ESCsDOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 009NameDescription# of peptidesMW ( kD ) Peptides/MWGel slicesStk38Serine/threonine-protein kinase 3878541 . 447 , 14 , 15 , 16 , 17Ruvbl1RuvB-like 142500 . 847 , 8 , 13 , 15 , 16 , 17Ruvbl2RuvB-like 238510 . 7515 , 16 , 17Acta1 or other isoActin21420 . 5018Sun2Protein unc-84 homolog B36820 . 4411 , 12 , 13Hdac6Histone deacetylase 6381260 . 301 , 2 , 3 , 4 , 5 , 6 , 7 , 8Kat5Histone acetyltransferase KAT517590 . 2913 , 14 , 15ActbActin , cytoplasmic 112420 . 295 , 8 , 9TrrapTransformation/transcription domain-associated protein722920 . 251 , 2Epc1Enhancer of polycomb homolog 122900 . 249 , 10Brd8Bromodomain-containing protein 8241030 . 236 , 7Yeats4YEATS domain-containing protein 46270 . 2220Epc2Enhancer of polycomb homolog 220910 . 2210 , 11H2BHistone H2B3140 . 2121Ing3Inhibitor of growth protein 310470 . 2116 , 17Ep400E1A-binding protein p400703370 . 211 , 2Dmap1DNA methyltransferase 1-associated protein 110530 . 1914 , 15Hspa8Heat shock cognate 71 kDa protein13690 . 1913Lima1LIM domain and actin-binding protein 115840 . 189 , 10Vps72Vacuolar protein sorting-associated protein 72 homolog7410 . 1716 , 17Actl6aActin-like protein 6A8470 . 1717 , 18Actg1Actin , cytoplasmic 16420 . 1419H2afvHistone H2A . V2140 . 1422Meaf6Chromatin modification-related protein MEAF63220 . 1420Mbtd1MBT domain-containing protein 18710 . 1113 , 14Rps1840S ribosomal protein S182180 . 1121Tubb5Tubulin beta-5 chain5500 . 1016Tuba1a or other isoTubulin alpha chain5500 . 1016Trim28Transcription intermediary factor 1-beta8890 . 099 , 10Morf4l2Mortality factor 4-like protein 23340 . 0919MrgbpMRG-binding protein2240 . 0820Rangap1Ran GTPase-activating protein 15640 . 0812Spna2Spectrin alpha chain , brain182850 . 063SetxProbable helicase senataxin182980 . 062 , 3Sf3b1Splicing factor 3B subunit 13540 . 066SfpqSplicing factor , proline- and glutamine-rich4750 . 0510Rab5cRas-related protein Rab-5C1230 . 0420Lrrfip2Leucine-rich repeat flightless-interacting protein 22470 . 0416Stat2Signal transducer and activator of transcription 22500 . 048NonoNon-POU domain-containing octamer-binding protein2550 . 0415TprNucleoprotein TPR62740 . 023HnrnpfHeterogeneous nuclear ribonucleoprotein F1460 . 0217Spnb2Spectrin beta chain , brain 152740 . 023FlnaFilamin-A42810 . 013FliiProtein flightless-1 homolog21450 . 016HdxHighly divergent homeobox1770 . 0111Morf4l1Mortality factor 4-like protein 10410N/AProteins in bold represent known Tip60-p400 subunits found in Tip60-H3F purification from ESCs . The protein in bold italic represents the known Tip60-p400 subunit not found in purification from ESCs . We were intrigued by the finding that Hdac6 co-purified with Tip60-p400 in ESCs ( Table 1 ) . Hdac6 is a class II histone deacetylase ( HDAC ) that is expressed in many different cell types but is usually localized to the cytoplasm ( Verdel et al . , 2000; Hubbert et al . , 2002; Kawaguchi et al . , 2003; Valenzuela-Fernández et al . , 2008 ) , as are its well-established substrates: α-tubulin ( Hubbert et al . , 2002 ) , Hsp90 ( Kovacs et al . , 2005 ) , and cortactin ( Zhang et al . , 2007 ) . Moreover , despite its homology to proteins that deacetylate histone tails , Hdac6 has not been found to target histones in vivo ( Haggarty et al . , 2003 ) . To confirm that Hdac6 is a bona fide Tip60-p400-interacting protein , we performed reciprocal co-immunoprecipitation experiments in ESCs , observing the Tip60-Hdac6 interaction no matter which protein was immunoprecipitated ( Figure 1B , Figure 1—figure supplement 2 ) . Previously , both Hdac6 and Tip60 were separately found to interact with RNA Polymerase II ( Pol II ) in human CD4+ T-cells ( Wang et al . , 2009 ) , raising the possibility that the Hdac6-Tip60 interaction we observed in ESCs might be mediated by Pol II . However , there were no peptides corresponding to Pol II in our LC-MS/MS data , and we could not detect Pol II in Tip60-H3F immunoprecipitates ( Figure 1—figure supplement 3 ) , arguing against this explanation . Furthermore , the Tip60-Hdac6 interaction was independent of DNA and resistant to high salt ( Figure 1—figure supplement 4 , Figure 1—figure supplement 5 ) , verifying that Hdac6 is a stable interaction partner within Tip60-p400 complex . Finally , we tested whether Hdac6 interacts with Tip60-p400 complex in a differentiated cell type , MEFs . Unlike ESCs , Tip60-H3F immunoprecipitated from MEF nuclear extracts or whole cell lysates did not pull down Hdac6 ( Figure 1C–D ) , consistent with the possibility that Hdac6 interacts with Tip60-p400 complex in only a subset of cell types . The reported cytoplasmic localization of Hdac6 in multiple types of cells ( Verdel et al . , 2000; Hubbert et al . , 2002; Kawaguchi et al . , 2003; Valenzuela-Fernández et al . , 2008 ) raised the question of whether its interaction with Tip60-p400 complex was physiologically relevant . To address this issue , we first confirmed that Hdac6 exhibited significant nuclear localization in ESCs , in contrast to MEFs , in which Hdac6 was mainly cytoplasmic ( Figure 2A ) . Next , we prepared cytoplasmic and nuclear protein fractions to examine the cellular localization of Hdac6 in ESCs and several adult cell types by Western blotting . Interestingly , while differentiated cells ( MEFs , whole brain ) exhibited the reported cytoplasmic sequestration , high levels of Hdac6 in undifferentiated cells , including ESCs , NSCs , and hematopoietic stem and progenitor cells ( HPCs ) , were found in the nucleus ( Figure 2B ) . Consistent with these data , differentiation of either ESCs or NSCs caused a dramatic decrease in nuclear Hdac6 , accompanied by increased Hdac6 within the cytoplasm ( Figure 2C–D ) . To test whether nuclear localization of Hdac6 promoted its interaction with Tip60-p400 complex , we immunoprecipitated Tip60 from NSCs isolated from Tip60-H3F knock-in mice . Indeed , Hdac6 was present in Tip60-H3F immunoprecipitates from NSC nuclear extracts ( Figure 2E ) . These data suggest that the interaction of Hdac6 with Tip60-p400 in undifferentiated cells is lost during the course of differentiation of some types of stem cells , due to nuclear exclusion of Hdac6 . However , cytoplasmic sequestration in differentiated cells cannot be the sole factor preventing Hdac6 from associating with Tip60-p400 , since we did not observe Hdac6 within Tip60-H3F immunoprecipitates from MEF whole cell lysates ( in which nuclear and cytoplasmic proteins are mixed; Figure 1D ) . Together , these data show that Hdac6 exhibits significant nuclear localization in some types of embryonic and adult stem and progenitor cells , where it associates with Tip60-p400 complex . Furthermore , stem cell differentiation promotes re-localization of Hdac6 to the cytoplasm , where it is sequestered away from Tip60-p400 . 10 . 7554/eLife . 01557 . 010Figure 2 . Hdac6 is partially nuclear in multiple types of undifferentiated cells and interacts with Tip60-p400 in NSCs and cancer cell lines . ( A ) ESCs ( top ) or MEFs ( bottom ) were subjected to immunofluorescence using an antibody recognizing Hdac6 . DAPI staining is shown to identify nuclei . ( B ) High levels of nuclear Hdac6 in stem cells but not differentiated cells . Western blots of Hdac6 and known Tip60-p400 subunit Dmap1 in cytoplasmic ( C ) and nuclear ( N ) fractions of indicated cells are shown , with β-actin serving as a loading control . ( C ) Hdac6 relocalizes to the cytoplasm during ESC differentiation . Cytoplasmic ( C ) and nuclear ( N ) fractions from ESCs differentiated for the indicated number of days were Western blotted for the indicated proteins . ( D ) Hdac6 relocalizes to the cytoplasm during NSC differentiation . Cytoplasmic ( C ) and nuclear ( N ) fractions from undifferentiated NSCs ( day 0 ) or NSCs differentiated for 7 days were Western blotted for the indicated proteins . ( E ) Hdac6 interacts with Tip60-p400 in NSCs . Shown are Western blots for the proteins indicated of input or Tip60-p400 complex immunoprecipitated from NSCs . ( F ) Hdac6 is nuclear localized in cancer cell lines . Cells were fractionated and Western blotted as in ( B ) . ( G ) Hdac6 interacts with Tip60-p400 in a cancer cell line . 293 . T cells were transfected with the indicated constructs . Nuclear extracts were prepared , subjected to immunoprecipitation with an anti-HA antibody , and Western blotted as indicated . ( H ) Re-localization of Hdac6 to the nucleus is an early event during transformation . MEFs were infected with pBABE-puro retrovirus expressing SV40 large T antigen or empty vector , harvested after 5 days ( including 3 days of selection ) , and fractionated as in ( B ) . Western blots are shown for the proteins indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 010 While Hdac6 localizes to the cytoplasm of normal somatic cells , nuclear Hdac6 has been observed in some cancers , in particular within tumors that are poorly differentiated ( Subramanian et al . , 2011; Riolo et al . , 2012 ) . We therefore tested several human cancer cell lines and found that Hdac6 exhibited significant nuclear localization in each of the cell lines ( Figure 2F ) . We further tested whether Tip60-p400 interacts with Hdac6 in cancer cells , by transfecting epitope-tagged Tip60 and Hdac6 constructs into 293 . T cells ( to facilitate immunoprecipitation ) , and found that Tip60 co-precipitated with Hdac6 ( Figure 2G ) , suggesting that , as with ESCs and NSCs , Hdac6 functions within Tip60-p400 complex in some cancer cells . Interestingly , loss-of-function of either Hdac6 or genes encoding Tip60-p400 subunits within cancer cell lines has previously been shown to elicit defects in anchorage-independent growth and hypersensitivity to DNA damaging agents ( Feng et al . , 2003; Lee et al . , 2008; Wang et al . , 2012 ) , supporting this hypothesis . Finally , we asked whether re-localization of Hdac6 to the nucleus in cancer cells is an early event during transformation , at a time when it might be more likely to contribute to cancer progression . We found that shortly after introduction of SV40 large T antigen into MEFs , a large fraction of Hdac6 re-localized to the nucleus ( Figure 2H ) , suggesting that nuclear localization of Hdac6 is an early event during transformation . Together , these data suggest that Hdac6 functions within Tip60-p400 complex in some types of cancer , and support the idea that stem cells and some cancer cells share several common phenotypes and regulatory pathways . To test whether Hdac6 is necessary for gene regulation by Tip60-p400 complex in ESCs , we used DNA microarrays to examine the changes in mRNA levels upon Tip60 or Hdac6 KD . We observed highly correlated gene expression profiles in ESCs knocked down individually for Tip60 and Hdac6 ( R = 0 . 63 ) , suggesting a significant overlap in their sets of target genes , although Hdac6 KD generally had weaker effects on expression of common targets ( Figure 3A–B ) . Next , we performed unsupervised hierarchical clustering of mRNA expression data comparing ESCs depleted of Tip60 , Hdac6 , or both to control KD ESCs . We observed four main clusters of genes differentially expressed in Tip60 KD cells: genes upregulated upon Tip60 KD but unaffected by Hdac6 KD ( Figure 3C , cluster 1 ) , genes upregulated in both single KDs ( cluster 2 ) , genes downregulated in Tip60 KD cells but unaffected by Hdac6 KD ( cluster 3 ) , and genes downregulated in both single KDs ( cluster 4 ) . Genes downregulated upon Tip60 KD were significantly overrepresented ( relative to the number expected by chance ) among Hdac6-independent Tip60 targets , while genes upregulated upon Tip60 KD were significantly overrepresented among Hdac6-dependent target genes ( Figure 3D ) , suggesting that Tip60-dependent repression in ESCs usually requires Hdac6 . Double KD of Tip60 and Hdac6 was nearly identical to the Tip60 single KD , consistent with the model that Hdac6 functions within Tip60-p400 complex ( Figure 3C ) . We next tested individual genes from each cluster by RT-qPCR , which generally confirmed the microarray results ( Figure 3E ) . We further validated these results using two Hdac6 mutant ESC lines: one that harbors a hypomorphic allele that expresses reduced Hdac6 levels ( Hdac6reduced ) , and one that contains an Hdac6 deletion ( Hdac6null; Figure 3—figure supplement 1 , Figure 3—figure supplement 2 ) . Both lines exhibited misregulation of Tip60/Hdac6 target genes similar to that observed upon Hdac6 KD , with greater effects usually observed in the more severe Hdac6null line ( Figure 3—figure supplement 3 ) . Finally , we examined which classes of genes were enriched in each regulatory cluster . While Hdac6-independent targets of Tip60-p400 were enriched for genes involved in cellular growth , homeostasis , and the cell cycle ( Figure 3—figure supplement 4 , clusters 1 and 3 ) , the majority of Hdac6-dependent Tip60-p400 target genes were differentiation-induced genes ( Figure 3—figure supplement 4 , cluster 2 ) , consistent with the idea that Hdac6 is broadly important for Tip60-p400-dependent repression of developmental genes , although it also plays a smaller role in the activation of some Tip60-dependent proliferation genes ( Figure 3—figure supplement 4 , cluster 4 ) . 10 . 7554/eLife . 01557 . 011Figure 3 . Overlapping effects of Hdac6 KD and Tip60 KD on gene expression in ESCs . ( A ) Scatter plot of gene expression ( Log2 ( fold change ) relative to control KD ESCs ) upon Tip60 KD relative to Hdac6 KD . Genes misregulated upon Tip60 KD ( adjusted p<0 . 1 ) are shown in red . ( B ) Western blot showing the levels of p400 , Dmap1 , Hdac6 and Tip60 ( FLAG ) upon Tip60 or Hdac6 KD . β-actin is shown as a loading control . ( C ) Unsupervised hierarchical clustering of genes misregulated ( adjusted p<0 . 1 ) upon Tip60 KD . Up-regulated genes are indicated in yellow and downregulated genes are indicated in blue . The first cluster ( #1 ) includes 200 genes that were upregulated upon Tip60 KD but not Hdac6 KD , the second cluster ( #2 ) includes 867 genes upregulated in both Tip60 KD and Hdac6 KD ESCs , the third cluster ( #3 ) includes 277 genes that were downregulated only in Tip60 KD cells and the forth cluster ( #4 ) includes 424 genes that were downregulated in both Tip60 KD and Hdac6 KD ESCs . ( D ) Hdac6-dependent target genes are biased toward genes repressed by Tip60 . Genes repressed or activated by Tip60 were split based on their Hdac6-dependence , and each group was plotted as the Log2 ( ratio ) of genes observed in each group relative to the expected number of genes if Hdac6-dependence was randomly distributed . Asterisk indicates statistically significant enrichment ( p<10–20 ) . ( E ) Validation of microarray datasets . The expression levels of genes from each cluster were measured by RT-qPCR in the indicated KDs and expressed as Log2 ( fold change ) values relative to control ( GFP ) KD ESCs after normalization . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01110 . 7554/eLife . 01557 . 012Figure 3—figure supplement 1 . Hdac6 mutant ESCs . Diagram of Hdac6 mutant ESC line Hdac6tm1a ( EUCOMM ) Wtsi before ( Hdac6reduced ) and after ( Hdac6null ) CRE expression . Since , prior to CRE expression , this line produces a low level of Hdac6 protein , we infer that some transcripts fail to include the gene trap allele and therefore produce full-length , wild-type transcript . SA: splice acceptor; pA: cleavage and polyadenylation sequence . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01210 . 7554/eLife . 01557 . 013Figure 3—figure supplement 2 . Hypomorphic and null Hdac6 mutant ESCs . Whole cell extracts from normal ( WT ) ESCs or Hdac6 mutant ESCs as diagrammed in Figure 3—figure supplement 1 were prepared , followed by Western blotting for the proteins indicated . Note higher levels of acetylated tubulin upon CRE expression . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01310 . 7554/eLife . 01557 . 014Figure 3—figure supplement 3 . Hdac6 mutant ESCs exhibit alterations in Tip60-target gene expression consistent with KD phenotypes . Cells described in Figure 3—figure supplement 1 were harvested for RNA and subjected to RT-qPCR for the genes indicated from Tip60/Hdac6 target gene cluster 2 ( cluster 4 gene shown as a control ) . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01410 . 7554/eLife . 01557 . 015Figure 3—figure supplement 4 . GO-term enrichment of gene clusters . The significance of enrichment [−Log10 ( p-value ) ] of Gene Ontology ( GO ) terms over-represented in each cluster of genes in Figure 3C . GO terms that were partially redundant with those listed were eliminated for brevity . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 015 Despite overlapping roles in gene regulation , the function of Hdac6 within Tip60-p400 complex remained unclear . Hdac6 is not thought to bind chromatin or regulate gene expression in most cell types ( Verdel et al . , 2000; Hubbert et al . , 2002; Haggarty et al . , 2003 ) , therefore we considered the possibility that Hdac6 modulates Tip60-p400 function prior to chromatin binding by the complex . To determine whether Hdac6 associates with chromatin-bound Tip60-p400 , we tested whether Hdac6 co-localizes with Tip60-p400 on chromatin . To this end , we examined the genome-wide distributions of Tip60 and Hdac6 in ESCs using chromatin immunoprecipitation followed by deep sequencing of the precipitated DNA ( ChIP-seq ) . To facilitate these analyses , we generated an ESC line in which the H3F tag utilized above was fused to the C-terminus of Hdac6 at the endogenous Hdac6 locus ( Figure 4—figure supplement 1 ) . This line allowed us to directly compare the genomic binding profiles of Tip60 and Hdac6 with untagged control cells using the same antibody , thereby eliminating differences in background . We found that Tip60 was enriched at the 5’ ends of many genes in ESCs , with two peaks of binding flanking the promoter regions of most targets , one at approximately 400 base pairs upstream and another at approximately 100 base pairs downstream of the TSS ( Figure 4A–B ) . This two peak pattern of Tip60 binding and the gene set bound by Tip60 were very similar to previous mapping data examining the distribution of the p400 subunit of Tip60-p400 complex in ESCs ( Fazzio et al . , 2008a ) . Interestingly , we found that while Hdac6 was also enriched at p400- ( Figure 4A–B ) and Tip60-target genes ( Figure 4C ) near their TSSs , its pattern of binding at each gene was somewhat different , forming one peak of enrichment at approximately 100 base pairs downstream of the TSS that overlapped with the downstream Tip60 peak ( Figure 4A–B ) . Like p400 ( Fazzio et al . , 2008a ) , we found that Hdac6 was enriched at genes marked by H3K4me3 , including bivalent genes also marked by H3K27me3 ( Figure 4—figure supplement 2 ) . Furthermore , we found that , on average , Hdac6 and Tip60 were both enriched to significantly higher levels at the promoter regions of genes that are misregulated upon Hdac6 or Tip60 KD compared to the genes unaffected by KD of these factors ( Figure 4D ) , suggesting that many of these genes are direct targets . Interestingly , while the levels of Hdac6 binding were elevated at genes from clusters 2 and 4 ( Figure 3C ) , whose expression levels are regulated by Hdac6 , they were also elevated at cluster 3 genes , whose expression levels are Tip60-dependent but Hdac6-independent ( Figure 4E ) , suggesting that Tip60-p400 acts independently of Hdac6 to activate these genes . Together , these data show that Hdac6 binds in an asymmetrical pattern with respect to the transcription start site at its genomic targets . In addition , the overlap between Hdac6 and one of the two peaks of Tip60-p400 binding is consistent with a model in which Hdac6 functions within chromatin-bound Tip60-p400 complex to regulate common target genes in ESCs . Consistent with our finding that Hdac6 does not interact with Tip60-p400 in differentiated cells , we found that Hdac6-dependent Tip60-target genes in ESCs were not bound by Tip60 in MEFs ( Figure 4—figure supplement 3 ) . 10 . 7554/eLife . 01557 . 016Figure 4 . Tip60 and Hdac6 co-localize on chromatin . ( A ) Heat map representation of ChIP-seq data for H3F-tagged Tip60 and Hdac6 comprising the 2 kb surrounding the transcriptional start sites ( TSS ) of 10 , 507 genes for which published p400 ChIP-chip data ( Fazzio et al . , 2008a ) ( p400 enrichment ) were available . ChIP-seq data from E14 ( Untagged ) cells is shown as a control . All panels are sorted by decreasing p400 binding for the 1 kb surrounding the TSS , ranging from high levels of p400 binding ( red ) to genes unbound by p400 ( white ) . ( B ) Tip60 and Hdac6 binding correlate with p400 binding . Genes in the p400 ChIP-chip dataset were grouped by the intensity of p400 enrichment: The groups of genes exhibiting the top 10% of p400 enrichment ( top 10% ) , the 11th–30th percentile ( next 20% ) , the 31st–60th percentile ( middle 30% ) and the rest of genes in dataset ( last 40% ) . Upper panel: averaged Tip60 enrichment for groups of genes at each level of p400 binding are shown relative to the TSS . Lower panel: averaged Hdac6 binding data for genes in the same groups . ( C ) Correlation of Tip60 and Hdac6 binding . Shown is a Venn diagram delineating the overlap between the gene sets bound by Tip60 and Hdac6 . The p-value was calculated by summing the hypergeometric probabilities of Tip60/Hdac6 overlap below the number observed and subtracting from one . ( D ) Hdac6 and Tip60 are enriched at genes regulated by these factors . Upper panel: Tip60 binding segregated by genes that are upregulated , downregulated , or unchanged by Tip60 KD . Bottom panel: Hdac6 binding at genes segregated as in upper panel . ( E ) Tip60 and Hdac6 are enriched at genes within clusters 2 , 3 , and 4 . Tip60 ( upper panel ) and Hdac6 ( bottom panel ) binding data are shown for genes segregated by cluster , as in Figure 2C . Asterisks mark clusters exhibiting statistically significantly higher promoter-proximal ( −500 to +500 ) binding of indicated factor than does the set of genes not regulated by Tip60 and Hdac6: * ( p<0 . 05 ) ; ** ( p<0 . 01 ) ; *** ( p<10-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01610 . 7554/eLife . 01557 . 017Figure 4—figure supplement 1 . Targeting of the H3F tag to the C-terminus of the endogenous Hdac6 gene . Shown are the 3' end of the Hdac6 gene plus downstream sequence ( below ) and the targeting construct for introducing the C-terminal tandem 6-His-3-FLAG tag ( above ) . The counter-selection cassette within the targeting construct is omitted for brevity . See ‘Materials and methods’ for details of ESC targeting . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01710 . 7554/eLife . 01557 . 018Figure 4—figure supplement 2 . Hdac6 is enriched at genes marked by H3K4me3 . Heatmaps showing the enrichment of various histone modifications or variants within promoter-proximal regions ( TSS ± 2 kb ) of all genes , sorted by Hdac6 enrichment . H3K4me3 is strongly enriched at genes with high levels of Hdac6 binding , including bivalent genes marked by both H3K4me3 and H3K27me3 , similar to previous findings for the p400 subunit of Tip60-p400 . Histone modification data are from the following sources: H3K4me1/me2/me3 , H3K27me3 , H3K36me3 ( Mikkelsen et al . , 2007 ) ; H2AZ ( Hu et al . , 2013 ) ; and H3K9me3 ( Teif et al . , 2012 ) DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 01810 . 7554/eLife . 01557 . 019Figure 4—figure supplement 3 . Hdac6-dependent Tip60 targets in ESCs are not bound by Tip60 in MEFs . Tip60 enrichment near the promoters of the genes indicated was determined by ChIP-qPCR in ESCs and MEFs . We examined binding to Cdkn1a , a known Tip60 target in MEFs , as a positive control for ChIP in MEFs . Data shown are the mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 019 We next tested several possible models by which Hdac6 might function within Tip60-p400 complex to regulate gene expression . First , we considered the possibility that Hdac6 was necessary for complex formation or stability . To test this model , we purified Tip60-H3F in control and Hdac6 KD ESCs , but found that the composition of Tip60-p400 complex was similar in the presence or absence of Hdac6 ( Figure 5—figure supplement 1 ) , arguing against this explanation . Alternatively , Hdac6 might be necessary for localization of Tip60-p400 to its target genes in ESCs . Finally , Tip60-p400 may recruit Hdac6 to its target genes , where it regulates gene expression by deacetylating histones . To distinguish between these latter two possibilities , we performed a series of experiments . First , we tested whether the treatment of ESCs with tubastatin A ( TubA ) , a chemical inhibitor that prevents Hdac6-dependent deacetlyation by binding within the catalytic channel of Hdac6’s deacetylase domains ( Butler et al . , 2010 ) , had similar effects on gene expression as Hdac6 KD . Like Hdac6 KD or Tip60 KD ESCs , most target genes were de-repressed upon TubA treatment ( Figure 5A ) , consistent with the possibility that histone deacetylation by Hdac6 was necessary for repression of its gene targets . TubA treatment did not affect expression or nuclear localization of Hdac6 ( Figure 5—figure supplement 2 ) , suggesting that it acts directly on the Hdac6 deacetylase domains . 10 . 7554/eLife . 01557 . 020Figure 5 . The Hdac6 deacetylase domains are necessary for Tip60-p400 binding but do not reverse histone acetylation catalyzed by Tip60 . ( A ) Treatment of ESCs with Hdac6 catalytic domain inhibitor Tubastatin A causes de-repression of Tip60-p400 target genes . RT-qPCRs for indicated genes are shown for GFP KD or Tip60 KD ESCs treated with either Tubastatin A ( TubA ) or DMSO vehicle . Data are expressed as Log2 ( fold change ) values relative to DMSO treated GFP KD ESCs after normalization . Shown are the mean ± SD values of three technical replicates from one representative experiment of two biological replicates performed . ( B ) H4 acetylation levels for several common Tip60/Hdac6 target genes in Tip60 KD and Hdac6 KD ESCs were measured by ChIP-qPCR , using an antibody specific for tetra-acetylated histone H4 . H4 acetylation levels in cells knocked down as indicated are expressed as a fraction of the input . Shown are the mean ± SD values of three technical replicates from one representative experiment of two biological replicates performed . ( C ) ESCs were treated overnight with the indicated amounts of TubA in their growth medium , Tip60-H3F was immunoprecipitated as above , and co-immunoprecipitating proteins were examined by Western blotting . ( D ) Tip60-H3F was immunoprecipitated from ESCs grown under normal conditions , and the beads were washed in buffer with or without TubA . The Hdac6 eluted in the TubA wash or remaining bound to beads is shown by Western blotting , along with canonical Tip60 subunit Dmap1 . ( E ) HA-tagged wild type , deacetylase domain 1 ( HD1 ) mutant , or double deacetylase domain mutant ( HD1 + HD2 ) Hdac6 was stably expressed in Tip60-H3F ESCs , Tip60-H3F was immunoprecipitated from nuclear extracts , and co-precipitating proteins were examined by Western blotting . Co-IP of canonical Tip60 complex subunit Dmap1 is shown as a control . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02010 . 7554/eLife . 01557 . 021Figure 5—figure supplement 1 . Hdac6 KD does not cause general disruption of Tip60-p400 complex . ( Left ) Silver stain of Tip60-p400 complex purified from control or Hdac6 KD Tip60-H3F ESCs or untagged ESCs , as indicated . ( Right ) Tip60-p400 complex purified from control or Hdac6 KD ESCs Western blotted for indicated proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02110 . 7554/eLife . 01557 . 022Figure 5—figure supplement 2 . Tubastatin A has no effect on Hdac6 expression or localization . ESCs were treated as shown , and nuclear and cytoplasmic fractions were Western blotted for the proteins indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02210 . 7554/eLife . 01557 . 023Figure 5—figure supplement 3 . Hdac6 KD reduces H2AK5 acetylation at Tip60-target genes . H2AK5 acetylation levels for several common Tip60/Hdac6 target genes in Tip60 and Hdac6 KD ESCs was measured by ChIP-qPCR , expressed as Log2 ratios relative to control ( GFP ) KD . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02310 . 7554/eLife . 01557 . 024Figure 5—figure supplement 4 . Hdac6 KD has no effect on bulk H4 acetylation levels , but strongly increases tubulin acetylation . ESCs knocked down as indicated were Western blotted with antibodies to tetra-acetylated histone H4 or H3 ( left ) , or acetylated tubulin ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02410 . 7554/eLife . 01557 . 025Figure 5—figure supplement 5 . Hdac6 KD does not induce Tip60-p400 acetylation or ubiquitination . Tip60-p400 complex purified from control or Hdac6 KD Tip60-H3F ESCs or untagged ESCs Western blotted with antibodies recognizing acetylated lysine ( left ) or ubiquitin ( right ) . Note that the background band in the ubiquitin western blot is also found in the untagged control , and is therefore non-specific . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 025 Next , we tested whether Hdac6 KD resulted in increased acetylation of histone tails at Hdac6 target promoters by examining acetylation of the N-terminal tails of histones H4 and H2A , two major targets of the Tip60 acetyltransferase activity ( Yamamoto and Horikoshi , 1997; Kimura and Horikoshi , 1998; Yan et al . , 2000; Doyon et al . , 2004; Altaf et al . , 2010 ) . ChIP-qPCRs using antibodies recognizing histone H4 acetylated at all four N-terminal lysines or acetylated lysine-5 of histone H2A were performed in control , Tip60 KD , and Hdac6 KD ESCs . Surprisingly , like Tip60 KD , Hdac6 KD resulted in a decrease in both H4 and H2A acetylation at most shared targets of Tip60 and Hdac6 that were examined ( Figure 5B , Figure 5—figure supplement 3 ) . These findings show that Hdac6 does not silence differentiation genes by counteracting Tip60-dependent histone acetylation , despite the fact that treatment with TubA resulted in gene expression changes similar to Hdac6 KD . Hdac6 KD had no significant effect on bulk histone H4 acetylation , although tubulin acetylation was strongly enhanced by Hdac6 KD ( Figure 5—figure supplement 4 ) , consistent with a model in which Hdac6 is required for Tip60 function only at common target promoters . A few subunits of Tip60-p400 complex are known to be acetylated ( Choudhary et al . , 2009 ) , raising the possibility that Hdac6 regulates Tip60-p400 function by deacetylating the complex , which then leads to altered biochemical activity or chromatin binding by the complex . We tested this possibility by Western blotting Tip60-p400 complex purified from control or Hdac6 KD ESCs with an antibody specific for acetyl-lysine , but we did not observe detectable acetylation of any subunit in either KD ( Figure 5—figure supplement 5 ) . Additionally , we performed mass spectrometry on p400 , the most highly acetylated subunit in Tip60-p400 complex ( Choudhary et al . , 2009 ) , purified from control or Hdac6 KD ESCs , but did not observe any notable differences in acetylation levels ( data not shown ) . Thus , while the loss of Hdac6 function does not lead to an observable increase in acetylation of Tip60-p400 complex or histone tails , it does lead to de-repression of Tip60-p400 target genes . Next , we considered the possibility that Hdac6 binds Tip60-p400 through one or both of its deacetylase domains . In this scenario , despite the fact that Hdac6 does not appear to deacetylate Tip60-p400 , TubA could cause de-repression of Tip60- and p400-target genes simply by preventing Hdac6 from binding Tip60-p400 . We tested this possibility by immunoprecipitating Tip60-p400 complex from Tip60-H3F ESCs treated with either vehicle alone or TubA and determining whether Hdac6 co-precipitates with Tip60 . Interestingly , TubA treatment completely abolished Hdac6 association with Tip60-p400 ( Figure 5C ) . We reasoned that if TubA directly inhibited binding of Hdac6 to Tip60-p400 , then addition of the drug to purified Hdac6-containing Tip60-p400 complex should disrupt this interaction . In contrast , if Hdac6 deacetylates some unknown protein , which then activates it to bridge the interaction of Hdac6 with Tip60-p400 , the treatment of cells with TubA should disrupt the Hdac6 interaction with Tip60-p400 , while in vitro TubA treatment of Tip60-p400 complex should not disrupt the Hdac6 interaction . To distinguish between these two possibilities , we immunoprecipitated Tip60-p400 from untreated Tip60-H3F ESCs and , after washing unbound proteins away from beads , subjected them to five additional washes with or without TubA . After harvesting the protein eluted in the TubA washes or remaining bound to beads , we examined the distribution of Hdac6 by Western blotting . Interestingly , we found that the TubA washes removed a large fraction of Hdac6 from bead-bound Tip60-p400 complex ( Figure 5D ) , suggesting that TubA directly disrupts the interaction of Hdac6 with Tip60-p400 , rather than preventing deacetylation of histones or other proteins . Consistent with these findings , we found that mutation of the Hdac6 deacetylase domains had the same effect as TubA treatment: mutation of deacetylase domain 1 partially disrupted Hdac6’s interaction with Tip60-p400 in ESCs , while mutation of both deacetylase domains abolished this interaction ( Figure 5E ) . Together , these data indicate that Hdac6 interacts with Tip60-p400 via its deacetylase domains and that TubA directly disrupts this interaction . The finding that Hdac6 does not appear to deacetylate histones or Tip60-p400 subunits , but that its deacetylase domains are necessary for interaction with Tip60-p400 complex , was consistent with a model in which Hdac6 regulates gene expression by helping recruit Tip60-p400 complex to its targets in vivo . We tested this possibility by comparing Tip60 enrichment in control KD and Hdac6 KD ESCs , predicting that if Hdac6 is necessary for Tip60-p400 recruitment , Hdac6 KD should result in diminished Tip60 enrichment at many of its target genes . We observed a striking difference in Tip60 binding upon Hdac6 KD compared to control cells: Tip60 enrichment downstream of the TSS was strongly reduced at many target genes in Hdac6 KD ESCs , along with a modest decrease in enrichment upstream of the TSS ( Figure 6A–B ) . Importantly , the reduction of Tip60 binding upon Hdac6 KD was most severe at genes whose expression is regulated by Tip60-p400 , compared to genes not regulated by the complex ( Figure 6C , Figure 6—figure supplement 1 ) . To confirm these findings , we tested the effects of Hdac6 KD or mutation on the p400 subunit of Tip60-p400 complex , finding that Hdac6 loss also strongly reduced p400 binding ( Figure 6D–E , Figure 6—figure supplement 2 ) . 10 . 7554/eLife . 01557 . 026Figure 6 . Hdac6 is necessary for normal Tip60 binding to its targets on chromatin . ( A ) Heat map representations of Tip60 binding as measured by ChIP-seq in Tip60-H3F ESCs upon GFP KD or Hdac6 KD . Data are sorted by p400 binding as in Figure 3A . ( B ) Averaged Tip60 binding upon GFP KD or Hdac6 KD are shown relative to the TSS . ( C ) Hdac6 KD mainly reduces Tip60 enrichment at genes that are misregulated upon Tip60 KD . Average Tip60 binding profiles upon GFP KD or Hdac6 KD are shown for genes misregulated upon Tip60 KD ( adjusted p<0 . 1 ) and genes that are unaffected . ( D ) Heat map representations of p400 binding as measured by ChIP-seq using an antibody against endogenous p400 in control and Hdac6 KD ESCs . Genes are sorted exactly as in ( A ) . ( E ) Average p400 binding profiles upon GFP KD or Hdac6 KD are shown relative to the TSS . ( F ) Tip60 KD does not affect Hdac6 binding . Average Hdac6 binding profiles upon GFP KD or Hdac6 KD are plotted as in ( B and E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02610 . 7554/eLife . 01557 . 027Figure 6—figure supplement 1 . Reduced levels of Tip60 binding to differentiation genes upon Hdac6 KD . ( Left ) Example genomic loci showing Tip60 binding as measured by ChIP-seq upon GFP KD or Hdac6 KD . The first two panels show that Tip60 binding to the promoter regions of highly expressed genes ( Rps9 , Rpl27a ) is relatively unaffected by Hdac6 KD . The second two panels show that Tip60 binding to the promoter regions of differentiation genes ( Nodal , Nefl ) is strongly reduced downstream of the TSS upon Hdac6 KD . ( Right ) Anti-FLAG ChIP-qPCR data are shown for the indicated ESCs at the promoter-proximal regions of the target genes shown . Note that Speer2 is a negative control locus to which Tip60-p400 does not bind . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02710 . 7554/eLife . 01557 . 028Figure 6—figure supplement 2 . Reduced levels of p400 binding to differentiation genes upon Hdac6 KD . Effects of Tip60 or Hdac6 KD ( left ) or mutation ( right ) on p400 binding to target genes . Anti-p400 ChIP-qPCR data are shown for the indicated KDs/mutations at the promoter-proximal regions of target genes . Data are normalized to IgG background ChIPs . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 02810 . 7554/eLife . 01557 . 029Figure 6—figure supplement 3 . Hdac6 binding is unaffected by Tip60 KD . Anti-FLAG ChIP-qPCR data are shown for the indicated ESCs at the promoter-proximal regions of target genes . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 029 Although we found that Hdac6 was required for Tip60 binding downstream of the TSS at most of its target genes , it remained possible that Tip60 was also required for Hdac6 to bind the same genes . To test this possibility , we mapped Hdac6 binding in control and Tip60 KD ESCs . In contrast to Tip60’s requirement for Hdac6 , Tip60 KD had no effect on Hdac6 localization ( Figure 6F , Figure 6—figure supplement 3 ) . Together , these data support a model in which Hdac6 binds to Tip60-p400 via its deacetylase domains , helping to recruit the complex to target promoters . In addition , these data rule out the possibility that Tip60-p400 represses differentiation genes by recruiting Hdac6 to these sites , since common targets of Tip60-p400 and Hdac6 were misregulated upon Tip60 KD even though Hdac6 localization was unaffected . Previously , we showed that ESCs depleted of Tip60-p400 subunits exhibited two prominent phenotypes: a failure to self-renew under conditions favoring ESC growth and a defect in the formation of embryoid bodies ( EBs ) under conditions favoring differentiation ( Fazzio et al . , 2008a ) . In contrast , while deletion of Hdac6 in ESCs was reported to cause a defect in colony formation , as well as a slight proliferation defect , Hdac6 KO ESCs could nonetheless continue to self-renew ( Zhang et al . , 2003 ) , and mice derived from Hdac6 KO ESCs are viable . However , it was unclear from previous reports whether Hdac6 KO or KD might result in impaired ESC differentiation in vitro , similar to that of Tip60 KD . To test this possibility , we first confirmed that Hdac6 KD ESCs exhibited colony formation and proliferation defects similar to those described for Hdac6 KO ESCs . Indeed , we observed a significant decrease in colony formation and size in both Hdac6 KD and Tip60 KD ESCs ( Figure 7A ) . Furthermore , Hdac6 KD ESCs had a small but reproducible decrease in proliferation rate , while Tip60 KD caused a somewhat more severe proliferation defect ( Figure 7B ) , consistent with previous studies ( Zhang et al . , 2003; Fazzio et al . , 2008a ) . Next , we tested whether EB formation or cellular differentiation might be impaired upon Hdac6 KD , as we previously observed upon KD of genes encoding known Tip60-p400 subunits Tip60 , p400 , or Dmap1 ( Fazzio et al . , 2008a ) . EBs are thought to roughly mimic the embryonic state , as cells proliferate within spherical aggregates that subsequently develop cystic structures and differentiate into cells from all three germ layers ( Martin and Evans , 1975 ) . Indeed we found that , similar to Tip60 KD ESCs , Hdac6 KD ESCs formed EBs that were significantly smaller and more heterogeneous than control KD ESCs ( Figure 7C ) . Furthermore , both Hdac6 and Tip60 KD EBs exhibited delayed induction of several differentiation markers during a time course of ESC differentiation ( Figure 8 ) . Thus , while Hdac6 loss results in only modest phenotypes in self-renewing ESCs , it is necessary for normal EB formation and cellular differentiation in vitro , consistent with its role in Tip60-p400 recruitment to differentiation genes . 10 . 7554/eLife . 01557 . 030Figure 7 . Hdac6 is necessary for ESC colony and EB formation . ( A ) Colony formation assays . Indicated KD ESCs were plated at clonal density and grown for seven days , at which time they were stained with crystal violet for visualization . ( B ) Growth curve . ESCs were infected with shRNA expressing viruses as shown and cultured in normal ESC medium . Cells were counted at the times indicated . ( C ) EB formation assay . Left and upper right: Brightfield images of EBs formed by hanging-drop cultures of ESCs knocked down as indicated , then cultured in differentiation medium , as described in ‘Materials and methods’ . Scale bars equal 400 μm . Lower right: box plot quantification of the range of EB sizes by diameter . The upper and lower limits of the box correspond to the 75th and 25th percentiles of each KD , respectively , and the dark line corresponds to the median of each box . At least 88 EBs were measured for each KD . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 03010 . 7554/eLife . 01557 . 031Figure 8 . Impaired differentiation of Hdac6 KD ESCs . Indicated KDs were differentiated for 0 , 2 , 4 , or 6 days . At the indicated time points , RNA was isolated and RT-qPCR quantification of several differentiation markers of each primary germ layer was performed . Data shown are mean ± SD of three technical replicates from one representative experiment of two biological replicates performed . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 031
In this study , we showed that Hdac6 interacts with Tip60-p400 complex in ESCs , NSCs , and some cancer cell types , and is necessary for the proper regulation of most genes regulated by Tip60-p400 in ESCs . For a majority of Tip60-p400 target genes , we found that Hdac6 facilitated Tip60-p400 binding to its gene targets . Interestingly , while Hdac6 does not appear to deacetylate histones in vivo , we found that its catalytic domains were necessary for interaction with Tip60-p400 complex . Hdac6 has well established deacetylase activity directed against several cytoplasmic proteins , most notably tubulin ( Hubbert et al . , 2002 ) , leaving open the possibility that , in stem cells , it may deacetylate some nuclear protein ( s ) that remain to be discovered . Tip60-p400 binds chromatin near the 5′ ends of genes in two peaks surrounding the TSS: an upstream peak that does not overlap with Hdac6 and is moderately sensitive to its loss , and a downstream peak that overlaps with and more strongly requires Hdac6 ( Figure 9 ) . Recruitment of Tip60-p400 is the only apparent function of Hdac6 in gene regulation , since Hdac6 binding is maintained upon Tip60 KD , while gene silencing is not . These data contrast with previously observed genetic interactions of mammalian Tip60 with other HDACs , in which Tip60 recruitment counteracts the repressive effects of HDACs on common targets ( Baek et al . , 2002 ) , although similarities in gene expression profiles between Tip60 loss of function and chemical inhibition of HDACs have been reported in Drosophila ( Schirling et al . , 2010 ) . 10 . 7554/eLife . 01557 . 032Figure 9 . Model for Hdac6- and Tip60-p400-dependent repression of differentiation genes in ESCs . In the presence of Hdac6 , Tip60-p400 binds both upstream and downstream of TSSs and differentiation genes are silenced . In the absence of Hdac6 , Tip60-p400 binding is reduced causing de-repression of differentiation genes . Tip60-p400 binding downstream of target TSSs appears to be more strongly affected by Hdac6 KD . Note that Hdac6 also appears to affect the activation of some genes ( not depicted ) , also by recruitment of Tip60-p400 complex . DOI: http://dx . doi . org/10 . 7554/eLife . 01557 . 032 We found that Hdac6 KD or mutation recapitulates most phenotypes of ESCs lacking Tip60 or p400: de-repression of many differentiation genes , impaired colony formation , a slower proliferation rate , impaired EB formation , and delayed kinetics of differentiation . Consistent with the phenotypes shared by Hdac6 KD and Tip60 KD ESCs , both Hdac6 KD and loss-of-function mutations of Tip60-p400 subunits were previously shown to impair anchorage-independent growth and increase DNA damage sensitivity in cancer cells ( Feng et al . , 2003; Lee et al . , 2008; Wang et al . , 2012 ) . However , unlike Tip60 ( Hu et al . , 2009 ) , Hdac6 KO ESCs can self-renew and are competent for mouse development ( Zhang et al . , 2008 ) , strongly suggesting that the defect observed for Hdac6 KD ESCs in vitro is overcome in the embryo by compensatory mechanisms . Since Hdac6 is only partially required for Tip60-p400 function in ESCs , the levels to which Tip60 and Hdac6 target genes are misregulated within the ICM of Hdac6−/− embryos may not be severe enough to induce a developmental arrest . In addition , since Hdac6 appears to function within the nucleus in only a subset of cell types ( stem and progenitor cells ) , the effect of Hdac6 loss on gene regulation is likely much more limited than that of Tip60 loss throughout the embryo . The dynamic regulation of Hdac6 localization during stem cell differentiation suggests that nuclear exclusion of Hdac6 may play a major role in gene regulation by reducing Tip60-p400 binding to specific sets of genes in differentiated cells . Consistent with this model , Hdac6-dependent Tip60-target genes in ESCs were not bound by Tip60 in MEFs ( Figure 4—figure supplement 3 ) , suggesting that Hdac6 does not play any role in Tip60-p400 regulation in these cells . It remains to be determined whether alternative proteins substitute for Hdac6 in differentiated cells to recruit Tip60-p400 to different promoters , or if alternative mechanisms of recruitment ( e . g . , binding to histone modifications or interaction with sequence-specific DNA-binding proteins ) are more important upon differentiation . Together , these studies identify a new , stem cell-specific mechanism by which Tip60-p400 is regulated , and uncover a role for Hdac6 in gene regulation . These findings differ substantially from established models of Hdac6 function as mainly a regulator of cellular motility and clearance of misfolded proteins via deacetylation of a small set of cytoplasmic targets , and suggest that cell type must be carefully considered when examining the phenotypes observed upon Hdac6 loss of function . In addition , these studies lend support to the idea that different types of stem cells , as well as some cancer cells , share common features of chromatin structure that allow them to maintain their developmental potency , and provide one possible mechanistic link underlying this commonality . Re-localization of Hdac6 to the cytoplasm during stem cell differentiation could lead to expression of an alternate set of Tip60-p400 target genes . Conversely , after Hdac6 is restored to the nucleus upon cellular transformation , it may re-direct Tip60-p400 to a subset of its target genes that are normally stem cell-specific , potentially eliciting changes in gene expression that promote cancer development .
ESCs were grown under feeder-free conditions and used for all ESC experiments . The Tip60-H3F and Hdac6-H3F lines were made by targeting into E14 ( Hooper et al . , 1987 ) and Hdac6tm1a ( EUCOMM ) Wtsi ESCs were obtained from EUCOMM ( clone EPD0519_4_C03 ) . This clone expresses low levels of Hdac6 protein before introduction of CRE , while CRE addition converts the mutation to a deletion . Therefore , we refer to the Hdac6 alleles as Hdac6reduced and Hdac6null before and after CRE introduction , respectively . CRE was introduced by Lenti-LucS ( Addgene plasmid 22778 ) lentiviral infection , and the cells were harvested 3 days later for RT-qPCR or ChIP . Tip60-EGFP and Hdac6-HD mutant-HA expression: ES cells were infected with pLJM1puro lentiviral vectors containing mouse Tip60 cDNA fused to EGFP . The infected cells were selected by puromycin 3 days post-infection and Tip60 expression was checked by Western-blot . Hdac6 was cloned into pBabe HAII-hygro retroviral vector with or without previously described HD1 and HD2 mutations ( Zhang et al . , 2006 ) . MEFs were infected with pBabe retroviral vectors containing wild-type or mutant mouse Hdac6 cDNAs . The infected cells were selected by hygromycin 3 days post-infection and Hdac6 expression was checked by Western-blot . For imaging , embryoid bodies ( EBs ) were generated by placing 300 cells , after infection with shRNA-expressing lentiviruses as indicated , in 30 μl of medium lacking LIF and incubating in hanging-drop cultures . For examination of differentiation markers , 106 cells were suspended in non-cell culture treated petri dishes for 2 days , and transferred to gelatin-coated cell culture dishes for another 4 days . RNA was isolated at the indicated time points . Mouse embryonic fibroblasts ( MEFs ) were isolated from Tip60-H3F/+ or wild-type littermates using standard protocols ( Coles et al . , 2007 ) and mouse NSCs were isolated as previously described ( Li et al . , 2008 ) . Colony formation assay: control , Tip60 , or Hdac6 KD ESCs were plated 3 days after lentiviral shRNA infection . 2000 cells were seeded into each 10 cm dish and cultured for 7 days . The cells were fixed and stained with fixation-staining solution ( 6% glutaraldehyde , 0 . 5% crystal violet ) for 30 min at room temperature followed by three washes with water . For examination of Hdac6 redistribution during stem cell differentiation , ESCs were plated in N2B27 medium ( Ying et al . , 2003 ) and NSCs were plated in DMEM with 10% fetal bovine serum ( FBS ) for the number of days indicated . The cancer cell lines were grown in DMEM with 10% fetal bovine serum . Lentiviral shRNA expression vectors from the TRC library ( Thermo Fisher , Waltham , MA , USA ) were obtained from the UMMS RNAi Core Facility . After screening through multiple shRNAs for each gene to be knocked down , we identified the most effective hairpin for each gene for subsequent experiments , listed as follows:pLKO . 1/shGFP: GCAAGCTGACCCTGAAGTTCATpLKO . 1/shTip60: CGGAGTATGACTGCAAAGGTTpLKO . 1/shHdac6: CGCTGACTACATTGCTGCTTT Lentiviral vectors and Lenti-X packaging plasmids were transfected into 293 . T cells using Fugene6 ( Roche , Branford , CT , USA ) . At 48 , 60 and 72 hr after transfection , lentivirus-containing media were collected and concentrated over a 20% sucrose cushion by centrifugation at 14 , 000 rpm for 4 hr in an SW-28 rotor . Concentrated virus was re-suspended in 200 μl PBS , aliquoted , and stored at −80°C . Cells were lysed using an NE-PER Extraction kit ( Thermo Fisher ) to isolate cytoplasmic and nuclear fractions . Western blotting was performed with antibodies against Hdac6 ( Cat . 07-732; Millipore , Billerica , MA , USA ) , acetyl-Histone H4 ( Cat . 06-866; Millipore ) , β-actin ( Cat . A5316; Sigma , St . Louis , MO , USA ) , Flag-M2 ( Cat . F1804; Sigma ) , p400 ( Cat . A300-541A; Bethyl Labs , Montgomery , TX , USA ) , Dmap1 ( Cat . 10411-1-AP; Proteintech Group , Chicago , IL , USA ) , GFP ( Cat . ab290; Abcam , Cambridge , MA , USA ) , Pol II ( Cat . sc-899; Santa Cruz Biotechnology , Santa Cruz , CA , USA ) , Acetylated Lysine ( Cat . 9441S; Cell Signaling Technologies , Danvers , MA , USA ) , ubiquitin ( P4D1 ) , and HA ( 12CA5 ) . For immunoprecipitation , the aliquots of nuclear extract were incubated with specific antibodies conjugated with protein G magnetic beads ( New England Biolabs , Ipswich , MA , USA ) or FLAG-M2 Agarose beads ( Sigma ) in IP buffer ( 50 mM Tris-HCl pH7 . 4 , 250 mM NaCl , 0 . 1% Triton X-100 , plus 1X HALT protease inhibitors ( Thermo Fisher ) overnight at 4°C . To examine the effect of inhibitors of Hdac6 on its interaction with Tip60 complex , beads were subjected to five additional washes with or without 10 μM tubastatin A ( ChemieTek , Indianapolis , IN , USA ) . Tip60 complex was purified from nuclear extracts of Tip60-H3F ESCs as described previously for Mbd3 ( Yildirim et al . , 2011 ) . Briefly , nuclear extracts were subjected to sequential affinity purification steps using FLAG-M2 Agarose ( Sigma ) and TALON Agarose beads ( Clontech Laboratories , Mountain View , CA , USA ) . The proteins purified from untagged control and Tip60-H3F cells were separated by SDS-PAGE and were either stained with SimplyBlue SafeStain ( Invitrogen , Grand Island , NY , USA ) after TCA precipitation and re-suspension in sample buffer or SilverXpress ( Invitrogen ) for visualization in Figure 1 . Affinity-purified samples were separated by SDS-PAGE gel , in-gel digested and analyzed by LC-MS and LC-MS/MS as described previously ( Chu et al . , 2006 ) . Briefly , 1 ml aliquot of the digestion mixture was injected into a Dionex Ultimate 3000 RSLCnano UHPLC system with an autosampler ( Dionex Corporation , Sunnyvale , CA , USA ) , and the eluant was connected directly to a nanoelectrospray ionization source of an LTQ Orbitrap XL mass spectrometer ( Thermo Fisher ) . LC-MS data were acquired in an information-dependent acquisition mode , cycling between a MS scan ( m/z 310-2000 ) acquired in the Orbitrap , followed by low-energy CID analysis in the linear ion trap . The centroided peak lists of the CID spectra were generated by PAVA ( Guan and Burlingame , 2010 ) and searched against a database that consisted of the Swiss-Prot protein database , to which a randomized version had been concatenated , using Batch-Tag , a program in the in-house version of the University of California San Francisco Protein Prospector version 5 . 9 . 2 . A precursor mass tolerance of 15 ppm and a fragment mass tolerance of 0 . 5 Da were used for protein database search . Protein hits are reported with a Protein Prospector protein score ≥22 , protein discriminant score ≥0 . 0 and a peptide expectation value ≤0 . 01 ( Chalkley et al . , 2005 ) . This set of protein identification parameters threshold did not return any substantial false positive protein hit from the randomized half of the concatenated database . The cells from ∼80% confluent 10 cm dishes were crosslinked by adding fixation solution ( 1% formaldehyde , 0 . 1M NaCl , 1 mM EDTA , 50 mM HEPES⋅KOH pH 7 . 6 ) for 10 min at room temperature . Crosslinking was quenched with 125 mM Glycine for 5 min . The cells were washed twice with cold PBS containing protease inhibitors ( Roche ) , and pelleted at 1000×g for 5 min at 4°C . The cell pellets were either flash frozen in liquid nitrogen and stored at −80°C or immediately sonicated . The pellets were resuspended in Lysis buffer 1 ( 50 mM HEPES⋅KOH pH 7 . 6 , 140 mM NaCl , 1 mM EDTA , 10% ( vol/vol ) Glycerol , 0 . 5% NP-40 , 0 . 25% Triton X-100 ) including protease inhibitors and incubated for 10 min at 4°C . After centrifugation at 1350×g for 5 min , the pellets were resuspended in Lysis buffer 2 ( 10 mM Tris-HCl pH 8 . 0 , 200 mM NaCl , 1 mM EDTA , 0 . 5 mM EGTA ) containing protease inhibitors and incubated for another 10 min at 4°C . The pellets were collected after centrifugation at 1350×g for 5 min and resuspended in Lysis buffer 2 for sonication . The samples were sonicated in a Bioruptor ( UCD-200; Diagenode , Delville , NJ , USA ) set to high for three cycles ( 10 min per cycle with 30 s on/ 30 s off ) to generate 300–1000 base-pair fragments . The supernatants were collected after a 13 , 000 rpm spin for 10 min at 4°C . 50-μl Protein G Magnetic beads ( New England Biolabs ) were washed twice with PBS with 5 mg/ml BSA and 10 μg of anti-Flag M2 antibody ( Sigma ) coupled in 500 μl PBS with 5 mg/ml BSA overnight at 4°C . Immunoprecipitation was performed with antibody-coupled beads and sonicated supernatants in ChIP buffer ( 20 mM Tris-HCl pH8 . 0 , 150 mM NaCl , 2 mM EDTA , 1% Triton X-100 ) overnight at 4°C . The magnetic beads were washed twice with ChIP buffer , once with ChIP buffer including 500 mM NaCl , four times with RIPA buffer ( 10 mM Tris-HCl pH8 . 0 , 0 . 25M LiCl , 1 mM EDTA , 0 . 5% NP-40 , 0 . 5% Na⋅Deoxycholate ) , and once with TE buffer ( pH 8 . 0 ) . Chromatin was eluted twice from washed beads by adding elution buffer ( 20 mM Tris-HCl pH8 . 0 , 100 mM NaCl , 20 mM EDTA , 1% SDS ) and incubating for 15 min at 65°C . The crosslinking was reversed at 65°C for 6 hr and RNase A ( Sigma ) was added for 1 hr at 37°C followed by proteinase K ( Ambion , Carlsbad , CA , USA ) treatment overnight at 50°C . ChIP-enriched DNA was purified using Phenol/Chloroform/Isoamyl alcohol extractions in phase-lock tubes . Then , chromatin was analyzed by qPCR using a SYBR FAST universal kit ( KAPA Biosystems , Woburn , MA , USA ) with specific primers ( Supplementary file 1A ) . 5 μg of total RNA from control ( GFP ) , Tip60 , Hdac6 KD or double KD ESCs was subjected to RNA amplification and labeling using the Low Input Quick Amp Labeling Kit protocol ( Agilent , Santa Clara , CA , USA ) with minor modifications . Briefly , cRNA was amplified by in vitro transcription with amino-allyl UTP ( 3:2 ratio for amino-ally UTP: UTP ) overnight at 37°C . Then , cRNA was purified using Zymo RNA purification columns and labeled with Cy3 ( GE Healthcare , Uppsala , Sweden ) at room temperature for 60 min in the dark . The fluorescence intensity of Cy3 was determined by NanoDrop and 50 picomoles of cRNA was used for fragmentation and hybridization on Agilent 4X44K mouse whole-genome microarrays . Slides were scanned on Agilent DNA microarray scanner G2565CA and fluorescence data were obtained using Agilent Feature Extraction software at the UMass Medical School genomics core facility . The expression profiles from two biological replicates were analyzed as previously described ( Yildirim et al . , 2011 ) . Enrichment of Gene Ontology terms and categories was performed with DAVID 6 . 7 ( Huang et al . , 2009a , 2009b ) . Microarray data can be obtained from GEO ( http://www . ncbi . nlm . nih . gov/geo/ ) , accession: GSE42329 . | Embryonic stem cells are cells that are able to transform into many other types of cells , such as blood cells and skin cells , as well as being able to divide in order to produce more stem cells . Mature cells lack this ability , which is called pluripotency , which is why there is so much interest in using embryonic stem cells to replace or regenerate human cells that have been lost or damaged through injury or illness . The various processes that result in self-renewal ( the production of new stem cells ) or differentiation ( the production of other types of cells ) are controlled by a wide variety of pathways , including some that only apply to the regulation of gene expression in stem cells . A number of these processes are known to involve chromatin – the densely packed structure formed by DNA and proteins called histones . Now Chen et al . study the means by which chromatin controls the stem cell fates by examining how a large enzyme called Tip60-p400 that interacts with histones – one of the main components of chromatin – in both mature cells and embryonic stem cells . Tip60-p400 is known to switch on genes that cause stem cells to undergo self-renewal , and to switch off the genes that allow stem cells to transform into other cell types , but the molecular mechanisms responsible for these effects have not yet been identified . Chen et al . studied the activity of Tip60-p400 in mouse embryonic stem cells , and found that another enzyme , Hdac6 , had to be present for Tip60-p400 to regulate the genes in the stem cells . Hdac6 is mostly found in the cytoplasm of cells that have differentiated into other cell types , and in the nucleus of stem cells , which is where the DNA resides . In cells from mice that lack Hdac6 , Chen et al . also found that stem cells fail to replicate or differentiate properly in culture , underscoring the importance of this particular enzyme , and filling in another piece of the puzzle of stem cell biology . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"chromosomes",
"and",
"gene",
"expression"
] | 2013 | Hdac6 regulates Tip60-p400 function in stem cells |
During mammalian development , the challenge for the embryo is to override intrinsic cellular plasticity to drive cells to distinct fates . Here , we unveil novel roles for the HIPPO signaling pathway in controlling cell positioning and expression of Sox2 , the first marker of pluripotency in the mouse early embryo . We show that maternal and zygotic YAP1 and WWTR1 repress Sox2 while promoting expression of the trophectoderm gene Cdx2 in parallel . Yet , Sox2 is more sensitive than Cdx2 to Yap1/Wwtr1 dosage , leading cells to a state of conflicted cell fate when YAP1/WWTR1 activity is moderate . Remarkably , HIPPO signaling activity resolves conflicted cell fate by repositioning cells to the interior of the embryo , independent of its role in regulating Sox2 expression . Rather , HIPPO antagonizes apical localization of Par complex components PARD6B and aPKC . Thus , negative feedback between HIPPO and Par complex components ensure robust lineage segregation .
During embryogenesis cells gradually differentiate , adopting distinct gene expression profiles and fates . In mammals , the first cellular differentiation is the segregation of trophectoderm and inner cell mass . The trophectoderm , which comprises the polarized outer surface of the blastocyst , will mainly produce cells of the placenta , while the inner cell mass will produce pluripotent cells , which are progenitors of both fetus and embryonic stem cells . Understanding how pluripotent inner cell mass cells are segregated from non-pluripotent cells therefore reveals how pluripotency is induced in a naturally occurring setting . Progenitors of inner cell mass are first morphologically apparent at the 16 cell stage as unpolarized cells residing inside the morula ( reviewed in Frum and Ralston , 2017 ) . However , at this stage , pluripotency genes such as Pou5f1 ( Oct4 ) and Nanog , do not specifically label inside cells ( Dietrich and Hiiragi , 2007; Niwa et al . , 2005; Palmieri et al . , 1994; Strumpf et al . , 2005 ) . Thus , the first cell fate decision has been studied mainly from the perspective of trophectoderm specification because the transcription factor CDX2 , which is essential for trophectoderm development ( Strumpf et al . , 2005 ) , is expressed specifically in outer cells of the 16 cell embryo ( Ralston and Rossant , 2008 ) , and has provided a way to distinguish future trophectoderm cells from non-trophectoderm cells . Knowledge of CDX2 as a marker of trophectoderm cell fate enabled the discovery of mechanisms that sense cellular differences in polarity and position in the embryo , and then respond by regulating expression of Cdx2 ( Nishioka et al . , 2009 ) . However , the exclusive study of Cdx2 regulation does not provide direct knowledge of how pluripotency is established because the absence of Cdx2 expression does not necessarily indicate acquisition of pluripotency . As such , our understanding of the first cell fate decision in the early mouse embryo is incomplete . In contrast to other markers of pluripotency , Sox2 is expressed specifically in inside cells at the 16 cell stage , and is therefore the first marker of pluripotency in the embryo ( Guo et al . , 2010; Wicklow et al . , 2014 ) . The discovery of how Sox2 expression is regulated in the embryo therefore provides unique insight into how pluripotency is first established in vivo . Genes promoting expression of Sox2 in the embryo have been described ( Cui et al . , 2016; Wallingford et al . , 2017 ) . However , it is currently unclear how expression of Sox2 becomes restricted to inside cells . We previously showed that Sox2 is restricted to inside cells by a Cdx2-independent mechanism ( Wicklow et al . , 2014 ) , which differs from Oct4 and Nanog , which are restricted to the inner cell mass by CDX2 ( Niwa et al . , 2005; Strumpf et al . , 2005 ) . Thus , Sox2 and Cdx2 are regulated in parallel , leading to complementary inside/outside expression patterns . However , it is not known whether Sox2 is regulated by the same pathway that regulates Cdx2 or whether a distinct pathway could be in use . The expression of Cdx2 is regulated by members of the HIPPO signaling pathway . In particular , the HIPPO pathway kinases LATS1/2 become active in unpolarized cells located deep inside the embryo , where they antagonize activity of the YAP1/WWTR1/TEAD4 transcriptional complex that is thought to promote expression of Cdx2 ( Anani et al . , 2014; Cockburn et al . , 2013; Hirate et al . , 2013; Kono et al . , 2014; Korotkevich et al . , 2017; Leung and Zernicka-Goetz , 2013; Lorthongpanich et al . , 2013; Mihajlović and Bruce , 2016; Nishioka et al . , 2009; Nishioka et al . , 2008; Posfai et al . , 2017; Rayon et al . , 2014; Watanabe et al . , 2017; Yagi et al . , 2007; Zhu et al . , 2017 ) . In this way , the initially ubiquitous expression of Cdx2 becomes restricted to outer trophectoderm cells . However , the specific requirements for Yap1 and Wwtr1 in the regulation of Cdx2 has been inferred from overexpression of wild type and dominant-negative variants , neither of which provide the standard of gene expression analysis that null alleles can provide . Nonetheless , the roles of Yap1 and Wwtr1 in regulating expression of Sox2 have not been investigated . Here , we evaluate the roles of maternal and zygotic YAP1/WWTR1 in regulating expression of Sox2 and cell fate during blastocyst formation .
To identify the mechanisms regulating Sox2 expression during blastocyst formation , we focused on how Sox2 expression is normally repressed in the trophectoderm to achieve inside cell-specific expression . We previously showed that SOX2 is specific to inside cells in the absence of the trophectoderm factor CDX2 ( Wicklow et al . , 2014 ) , suggesting that mechanisms that repress Sox2 in the trophectoderm act upstream of Cdx2 . Rho-associated , coiled-coil containing protein kinases ( ROCK1 and 2 ) are thought to act upstream of Cdx2 because embryos developing in the presence of a ROCK-inhibitor ( Y-27632 , ROCKi ) exhibit reduced Cdx2 expression ( Kono et al . , 2014 ) . Additionally , quantitative RT-PCR showed that Sox2 mRNA levels are elevated in ROCKi-treated embryos ( Kono et al . , 2014 ) , suggesting that ROCK1/2 activity leads to transcriptional repression of Sox2 . However , the role of ROCK1/2 in regulating the spatial expression of Sox2 has not been investigated . To evaluate the roles of ROCK1/2 in patterning Sox2 expression , we collected 8-cell stage embryos prior to embryo compaction ( E2 . 5 ) , and then cultured these either in control medium or in the presence of ROCKi for 24 hr ( Figure 1A ) . Embryos cultured in control medium exhibited normal cell polarity , evidenced by the apical localization of PARD6B and basolateral localization of E-cadherin ( CDH1 ) in outside cells ( Figure 1B , C ) as expected ( Vestweber et al . , 1987; Vinot et al . , 2005 ) . Additionally , SOX2 was detected only in inside cells in control embryos ( Figure 1C , D ) . By contrast , embryos cultured in ROCKi exhibited defects in cell polarity ( Figure 1B’ , C’ ) , consistent with prior studies ( Kono et al . , 2014 ) . Interestingly , in ROCKi-treated embryos , we observed ectopic SOX2 expression in cells located on the outer surface of the embryo ( Figure 1C’ , D ) , indicating that ROCK1/2 participates in repressing expression of Sox2 in the trophectoderm . To scrutinize the identity of outside-positioned SOX2-positive cells in ROCKi-treated embryos , we co-stained an additional cohort of control and ROCKi-treated embryos with CDX2 and SOX2 and compared the overlap of lineage marker expression . In control embryos , CDX2 was detected only in outside cells ( Figure 1—figure supplement 1A ) as expected at this stage ( Ralston and Rossant , 2008; Strumpf et al . , 2005 ) . In ROCKi-treated embryos , CDX2 expression levels were reduced ( Figure 1—figure supplement 1A’ ) as was the proportion of outside cells in which CDX2 was detected ( Figure 1E ) , as previously reported ( Kono et al . , 2014 ) . However , among outside cells , a substantial proportion coexpressed CDX2 and SOX2 in ROCKi-treated embryos compared with controls ( Figure 1E and Figure 1—figure supplement 1A ) , suggesting that ROCK inhibition leads to an increase in outside cells of mixed lineage . Since SOX2 expression does not regulate expression of CDX2 ( Wicklow et al . , 2014 ) , these observations suggest that ROCK1/2 activity regulates these genes through parallel mechanisms . We next sought to identify mediators that act downstream of ROCK1/2 to repress expression of Sox2 in the trophectoderm . Several direct and indirect targets of ROCK1/2 kinases in the early embryo have been described ( Alarcon and Marikawa , 2018; Shi et al . , 2017 ) . Among these is YAP1 , a transcriptional partner of TEAD4 ( Nishioka et al . , 2009 ) , since ROCK activity is required for the nuclear localization of YAP1 ( Kono et al . , 2014 ) . Notably , Tead4 is required to repress expression of Sox2 in the trophectoderm ( Wicklow et al . , 2014 ) , consistent with the possibility that YAP1 partners with TEAD4 to repress Sox2 expression in the trophectoderm . To test this hypothesis , we overexpressed a constitutively active variant of YAP1 ( YAP1CA ) . Substitution of alanine at serine 112 leads YAP1 to be constitutively nuclear and constitutively active ( YAP1CA hereafter ) ( Dong et al . , 2007; Nishioka et al . , 2009; Zhao et al . , 2007 ) . We injected mRNAs encoding YAP1CA and GFP into one of two blastomeres at the 2-cell stage , and then cultured these to the blastocyst stage ( Figure 1F ) . This mosaic approach to overexpression permitted comparison of Yap1CA-overexpressing with non-injected cells , which served as internal negative controls . We first examined localization of YAP1 in these embryos at the morula stage , with the expectation that YAP1 would be detected in nuclei of both inside and outside cells in YAP1CA-overexpressing cells ( Nishioka et al . , 2009 ) . As expected , YAP1 was observed in nuclei of all Yap1CA-overexpressing cells ( Figure 1—figure supplement 1B , C ) . We next evaluated the consequences of ectopic nuclear YAP1 on expression of SOX2 in inside cells . We observed a conspicuous decrease in the proportion of Yap1CA-overexpressing inside cells expressing detectable SOX2 ( Figure 1G , H ) . Therefore , nuclear YAP1 is sufficient to repress Sox2 expression in the inner cell mass , indicative of a likely role for YAP1 in repressing expression of Sox2 in the trophectoderm downstream of ROCK1/2 . To functionally test of the role of YAP1 in repressing expression of Sox2 , we injected one of two blastomeres with mRNA encoding LATS2 kinase , which inactivates YAP1 and , presumably , the related protein WWTR1 by phosphorylation , causing their cytoplasmic retention ( Nishioka et al . , 2008 ) . We then examined expression of SOX2 after culturing embryos to the blastocyst stage ( Figure 2A ) , predicting that LATS2 kinase would induce the ectopic expression of Sox2 in outside cells . Surprisingly , we observed that almost all Lats2-overexpressing cells ended up within the inner cell mass by the blastocyst stage ( Figure 2B , C ) , in contrast to cells injected with GFP mRNA only , which contributed to both inner cell mass and trophectoderm . Notably , SOX2 was detected in all Lats2-overexpressing cells observed within the inner cell mass ( Figure 2D ) , suggesting that Lats2-overexpressing cells were not only localized to the inner cell mass but also exhibited position-appropriate regulation of Sox2 . The strikingly increased prevalence of Lats2-overexpressing cells in the inner cell mass was also associated with a stark decrease in the number of Lats2-overexpressing cells detected within the trophectoderm and a decrease in the number of outside cells compared to embryos injected with GFP mRNA alone ( Figure 2C , E ) , suggesting that Lats2-overexpressing outside cells either internalize or undergo cell death . Furthermore , we observed cellular fragments within the trophectoderm of Lats2-overexpressing embryos ( Figure 2B , yellow arrowheads ) , as well as increased TUNEL staining in Lats2-overexpressing embryos compared to embryos injected with GFP mRNA only ( Figure 2—figure supplement 1A–B , D ) , consistent with increased death of Lats2-overexpressing cells . In addition to detecting SOX2 in all Lats2-overexpressing cells located inside the embryo , SOX2 was also detected in rare Lats2-overexpressing cells that remained on the embryo surface ( Figure 2D ) . Therefore , LATS2 is sufficient to induce expression of SOX2 in cells regardless of their position within the embryo . We predicted that , if Lats2 overexpression drove cells to adopt inner cell mass fate by influencing YAP1 and WWTR1 activity , then co-overexpression of Yap1CA would enable Lats2-overexpressing cells to contribute to trophectoderm . Consistent with this prediction , cooverexpression of Lats2 and Yap1CA led to a significant decrease in the proportion of Lats2-overexpressing cells contributing to the inside cell position , and a concomitant increase in the proportion of Lats2-overexpressing cells remaining in the outside position ( Figure Figure 2—figure supplement 2A–D ) . Moreover , cooverexpression of Lats2 and Yap1CA reduced the number of TUNEL positive nuclei , consistent with Yap1CA rescuing survival of outside-positioned Lats2-overexpressing cells ( Figure 2—figure supplement 1C–D ) . Collectively , these observations strongly suggest that LATS2 promotes inside cell positioning by regulating the activities of YAP1 and , likely , the related protein WWTR1 . To pinpoint when Lats2-overexpressing cells come to occupy the inside of the embryo , we performed a time course , examining the position of injected and non-injected cells from the 16-cell to the blastocyst stage ( ~80 cells ) . Surprisingly , between the 16 and 32-cell stages , the proportion of injected and non-injected cells in the total , outside , and inside cell populations were comparable whether embryos had been injected with Lats2 and GFP or GFP mRNA alone ( Figure 2F–H ) . In embryos injected with GFP mRNA alone , the proportion of injected and non-injected cells making up the total , outside , and inside cell populations remained constant throughout the time course . In contrast , starting around the 32-cell stage , the average proportion of Lats2-overexpressing cells making up the inside population began to increase dramatically . This increase was associated with a decrease in the proportion Lats2-overexpressing cells making up the outside population , consistent with internalization of Lats2-overexpressing cells after the 32-cell stage ( Figure 2G ) . After the 32-cell stage , Lats2-injected cells became underrepresented as a proportion of the total cell population ( Figure 2H ) , lending further support to the idea that Lats2-overexpressing cells that fail to internalize undergo cell death . Interestingly , the inside-skewed contribution of Lats2-overexpressing cells did not influence the ability of non-injected cells to contribute to the ICM ( Figure 2I ) , arguing that Lats2-overexpression drives inside positioning cell-autonomously . We therefore conclude that Lats2 overexpression acts on cell position and survival around the time of blastocyst formation . Our observation that Lats2-overexpression induces both the expression of SOX2 and cell repositioning to inner cell mass prompted us to investigate whether SOX2 itself drives cell repositioning downstream of Lats2 . In support of this hypothesis , SOX2 activity has been proposed to bias inner cell mass fate ( Goolam et al . , 2016; White et al . , 2016 ) . We therefore investigated whether Sox2 is required for the inner cell mass-inducing activity of LATS2 by overexpressing Lats2 in embryos lacking maternal and zygotic Sox2 ( Figure 3A ) , as previously described ( Wicklow et al . , 2014 ) . However , we observed that Lats2-overexpressing cells were equally likely to occupy inside position in the presence and absence of Sox2 ( Figure 3B , C ) . Furthermore , Lats2-overexpressing cells were equally unlikely to occupy outside position in the presence and absence of Sox2 ( Figure 3D ) . Therefore , although Lats2 overexpression is sufficient to induce expression of Sox2 , LATS2 acts on cell positioning/survival independently of Sox2 . Trophectoderm cell fate has been proposed to be determined by apically localized membrane components that maintain the position of future trophectoderm cells on the embryo surface ( Anani et al . , 2014; Korotkevich et al . , 2017; Maître et al . , 2016; Maître et al . , 2015; Samarage et al . , 2015; Zenker et al . , 2018 ) . For example , the apical membrane components aPKC and PARD6B are required for maintaining outside cell position and trophectoderm fate ( Alarcon , 2010; Dard et al . , 2009; Hirate et al . , 2015; Plusa et al . , 2005 ) . Because Lats2 overexpression led cells to adopt an inside position , this raised the testable possibility that LATS2 antagonizes localization of aPKC and PARD6B . Since Lats2 overexpression leads to cell positioning starting around the 32-cell stage , we examined the localization of aPKCz and PARD6B in embryos just prior to the 32-cell stage . At this stage , apical membrane components PARD6B and aPKCz were detected at the apical membrane of non-injected outside cells and outside cells injected with GFP only ( Figure 4A–D ) . By contrast , most Lats2-overexpressing outside cells lacked detectable aPKCz and PARD6B ( Figure 4A–D ) . Therefore , LATS2 is sufficient to antagonize localization of key apical domain proteins in outside cells , providing a compelling mechanism for the observed repositioning of Lats2-overexpressing outside cells . We also examined other markers of apicobasal polarization in Lats2-overexpressing outside cells prior to the 32-cell stage . Curiously , other markers of apicobasal polarization were properly localized in all cells examined . For example , CDH1 was restricted to the basolateral membrane ( Figure 4E ) , while filamentous Actin and phospho-ERM were restricted to the apical domain in outside cells of both Lats2-overexpressing and non-injected outside cells ( Figure 4F , G ) . Thus , we propose that Lats2-overexpressing outside cells initially possess hallmarks of apicobasal polarization , but aPKC and PARD6B fail to properly localize , leading to the eventual depolarization and internalization of outside cells . Our overexpression data suggested that the activities of YAP1 and WWTR1 are important for regulating cell fate and gene expression . Next , we aimed to test the requirements for Yap1 and Wwtr1 in embryogenesis . Yap1 null embryos survive until E9 . 0 ( Morin-Kensicki et al . , 2006 ) , suggesting that oocyte-expressed ( maternal ) Yap1 ( Yu et al . , 2016 ) , or the Yap1 paralogue Wwtr1 ( Varelas et al . , 2010 ) are important for preimplantation development . However , embryos lacking maternal and zygotic Wwtr1 and Yap1 have not been scrutinized . To generate embryos lacking maternal and zygotic Wwtr1 and Yap1 , we deleted Wwtr1 and Yap1 from the female germ line using mice carrying conditional alleles of Wwtr1 and Yap1 ( Xin et al . , 2013; Xin et al . , 2011 ) and the female germ line-specific Zp3Cre ( de Vries et al . , 2000 ) . We then crossed these females to males heterozygous for deleted alleles of Wwtr1 and Yap1 ( see Materials and methods ) . From these crosses , we obtained embryos lacking maternally provided Wwtr1 and Yap1 and either heterozygous or null for Wwtr1 and/or Yap1 ( Supplementary file 1 ) . At E3 . 25 ( ≤32 cells ) , SOX2 and CDX2 are normally mutually exclusive ( Figure 5A ) . However , with decreasing number of wild type zygotic alleles of Wwtr1 and Yap1 , we observed worsening phenotypes ( Figure 5B–F ) . In the complete absence of Wwtr1 and Yap1 , we observed a severe loss of CDX2 and expansion of SOX2 in outside cells ( Figure 5D–F ) , phenocopying Lats2 overexpression . However , in embryos of intermediate genotypes , we observed expanded SOX2 and persistent , yet lower , expression levels of CDX2 ( Figure 5C , E–F ) . Thus , regulation of Sox2 expression is more sensitive to Wwtr1 and Yap1 dosage than is Cdx2 . Moreover , these observations indicate that intermediate doses of Wwtr1 and Yap1 produce outside cells expressing markers of mixed cell lineage at E3 . 25 . Based on our observations of Lats2-overexpressing embryos , we anticipated that defects in cell positioning in embryos lacking maternal and zygotic Wwtr1 and Yap1 could arise after the 32-cell stage . We therefore examined embryos lacking Wwtr1 and Yap1 at E3 . 75 , when embryos possess more than 32 cells . Indeed , we observed skewed lineage contributions , correlating with the dosage of Wwtr1 and Yap1 ( Figure 6A–D ) . Embryos with one or fewer wild type alleles of Wwtr1 or Yap1 exhibited an increase in the number of inside cells and a reduction in the number of outside cells ( Figure 6A–B ) , consistent with altered cell positioning . Although the average total number of cells was also reduced in these embryos ( Figure 6C ) , the reduction in total cell number did not alone account for the loss of cells on the outside of the embryo ( Supplementary file 2 ) . This observation suggested that , similar to Lats2-overexpressing cells , cells with reduced Wwtr1 and Yap1 exhibit an increased frequency of outside cell death , in addition to increased outside cell internalization . Consistent with this , embryos with one or fewer wild type alleles of Wwtr1 or Yap1 exhibited an increase in the ratio of inside to outside cells ( Figure 6D ) and an increase in cells undergoing apoptosis by TUNEL assay ( Figure 6G and Figure 6—figure supplement 1A , B ) . Critically , the fewer outside cells in embryos lacking Wwtr1 and Yap1 , which appeared stretched over the mass of inside cells , exhibited ectopic expression of SOX2 ( Figure 6E–F ) . Therefore , WWTR1/YAP1 repress inner cell mass fate , downstream of LATS kinases . Intriguingly , our data also indicate that WWTR1 is a more potent repressor of Sox2 at E3 . 75 than YAP1 since embryos with a single wild type allele of Wwtr1 had significantly fewer cells expressing ectopic SOX2 then embryos with a single wild type allele of Yap1 ( Figure 6—figure supplement 1F , G ) . Since loss of Wwtr1 and Yap1 phenocopied Lats2 overexpression in terms of Sox2 expression , cell death , and cell repositioning , we next evaluated the apical domain and cell polarization in outside cells of embryos lacking Wwtr1 and Yap1 at E3 . 75 . As expected , observed greatly reduced aPKC at the apical membrane of outside cells in embryos with one or fewer doses of Wwtr1 or Yap1 ( Figure 6H and Figure 6—figure supplement 1C ) . In addition , we evaluated the localization of the tight junction protein ZO-1 , which suggested failure in tight junction formation in embryos with one or fewer doses of Wwtr1 and Yap1 ( Figure 6I and Figure 6—figure supplement 1D ) . Notably , however , other markers of apicobasal polarity , such as CDH1 and pERM were correctly localized in outside cells of mutant embryos at this stage ( Figure 6J and Figure 6—figure supplement 1E ) , consistent with some normal cell polarization . Our observations indicate that WWTR1 and YAP1 play a crucial role in the formation of the apical domain and maintaining the positioning and survival of outside cells while repressing expression of Sox2 .
During preimplantation development , lineage-specific transcription factors are commonly expressed in ‘noisy’ domains before refining to a lineage-appropriate pattern ( Simon et al . , 2018 ) . For example , Oct4 and Nanog are expressed in both inner cell mass and trophectoderm until after blastocyst formation ( Dietrich and Hiiragi , 2007; Strumpf et al . , 2005 ) . Similarly , CDX2 is detected in inner cell mass , as well as trophectoderm , until blastocyst stages ( McDole and Zheng , 2012; Ralston and Rossant , 2008; Strumpf et al . , 2005 ) . In striking contrast to these genes , SOX2 is never detected in outside cells ( Wicklow et al . , 2014 ) , indicating that robust mechanisms must exist to minimize noise and prevent its aberrant expression in trophectoderm . Here , we identify YAP1/WWTR1 as key components that repress Sox2 expression in outside cells of the embryo . Notably , manipulations known to antagonize YAP1/WWTR1 activity , including chemical inhibition of ROCK and overexpression of LATS2 , lead to ectopic expression of SOX2 in outside cells , reinforcing the notion that YAP1/WWTR1 activity are crucial for repression of Sox2 in outside cells . Additionally , we find that Sox2 expression is more sensitive than is Cdx2 to YAP1/WWTR1 activity , since intermediate doses of active YAP1/WWTR1 yield cells that coexpress both SOX2 and CDX2 ( Figure 7A ) . This observation is consistent with the fact that CDX2 is initially detected in inside cells of the embryo during blastocyst formation ( Dietrich and Hiiragi , 2007; McDole and Zheng , 2012; Ralston and Rossant , 2008 ) , where SOX2 is also expressed ( Wicklow et al . , 2014 ) . Thus , inside cells could initially possess intermediate doses of active YAP1/WWTR1 at this early stage . By contrast , outside cells have greatly reduced YAP1/WWTR1 activity , owing to elevated LATS activity . In this way , the HIPPO pathway ensures robust developmental transitions , by rapidly nudging SOX2-expressing cells into their correct and final positions inside the embryo ( Figure 7B ) . Consistent with our proposed model , the timing of HIPPO-induced cell internalization coincides with loss of cell fate plasticity around the 32-cell stage ( Posfai et al . , 2017 ) . This timing also coincides with the formation of mature tight junctions among outside cells ( Sheth et al . , 1997 ) , which reinforce and intensify differences in HIPPO signaling activity between inside and outside compartments of the embryo ( Hirate and Sasaki , 2014; Leung and Zernicka-Goetz , 2013 ) . Our observations indicate that HIPPO signaling can , in turn , interfere with trophectoderm epithelialization . Therefore , we propose that HIPPO engages in a negative feedback loop with cell polarity components ( Figure 7B ) . We propose two mechanisms by which HIPPO signaling eliminates cells from the trophectoderm , both of which are downstream of YAP1/WWTR1 ( Figure 7C ) . First , a small proportion of conflicted cells undergo cell death . This is in line with the observed increase in the level of apoptosis detected after the 32-cell stage ( Copp , 1978 ) . We showed that cell lethality due to elevated HIPPO can be rescued by increasing levels of nuclear YAP1 , suggesting that YAP1 activity normally provides a pro-survival signal to trophectoderm cells , consistent with the proposed role of YAP1 in promoting proliferation in non-eutherian mammals ( Frankenberg , 2018 ) . Moreover , deletion of Sox2 did not rescue survival of outside cells in which HIPPO signaling was artificially elevated , arguing that HIPPO resolves cell fate conflicts independently of lineage-specific genes . The second way that conflicted cells are eliminated from the trophectoderm is that cells with elevated HIPPO signaling drive their own internalization . This is consistent with the observation that cells in which Tead4 has been knocked down become internalized ( Mihajlović et al . , 2015 ) . However , in contrast to Tead4 loss of function , which preserves the apical domain in outside cells ( Mihajlović et al . , 2015; Nishioka et al . , 2008 ) , we observed that Yap1/Wwtr1 loss of function leads loss of apical PARD6D/aPKC . These observations suggest that YAP1/WWTR1 could partner with proteins other than TEAD4 to regulate apical domain formation . Consistent with this proposal , TEAD1 has been proposed to play an essential role in the early embryo ( Sasaki , 2017 ) . Nevertheless , since PARD6B/aPKC are essential for outside cell positioning ( Dard et al . , 2009; Hirate et al . , 2015; Plusa et al . , 2005 ) , the loss of the apical domain could affect cell positioning in several ways . For instance , loss of PARD6B/aPKC would eventually lead to cell depolarization ( Alarcon , 2010 ) , which could influence any of the processes normally governing the allocation of inside cells , such as oriented cleavage , cell contractility , or apical constriction ( Korotkevich et al . , 2017; Maître et al . , 2016; Samarage et al . , 2015 ) . Identifying the downstream mechanisms by which HIPPO drives cells to inner cell mass will be a stimulating topic of future study . Our studies also revealed that SOX2 does not play a role in cell positioning . This observation sheds light on a recent study , which showed that SOX2 dwells longer in select nuclei of four-cell stage embryos that are destined to contribute to the inner cell mass ( White et al . , 2016 ) . We propose that SOX2 is associated with future pluripotent state but does not alone contribute to all aspects of pluripotency , such as inside positioning . It is therefore still unclear why it is important to establish the inside cell-specific SOX2 expression during embryogenesis . Identification pathways that function downstream of YAP1/WWTR1 and in parallel to SOX2 to promote formation of pluripotent cells will provide meaningful insights into the natural origins of mammalian pluripotent stem cell progenitors .
All animal research was conducted in accordance with the guidelines of the Michigan State University Institutional Animal Care and Use Committee . Wild type embryos were derived from CD-1 mice ( Charles River ) . The following alleles or transgenes were used in this study , and maintained in a CD-1 background: Sox2tm1 . 1Lan ( Smith et al . , 2009 ) , Yaptm1 . 1Eno ( Xin et al . , 2011 ) , Wwtr1tm1 . 1Eno ( Xin et al . , 2013 ) , Tg ( Zp3-cre ) 93Knw ( de Vries et al . , 2000 ) . Null alleles were generated by breeding mice carrying floxed alleles and mice carrying ubiquitously expressed Cre , 129-Alpltm ( cre ) Nagy ( Lomelí et al . , 2000 ) . Mice were maintained on a 12 hr light/dark cycle . Embryos were collected by flushing the oviduct or uterus with M2 medium ( Millipore ) . For embryo culture , KSOM medium ( Millipore ) was equilibrated overnight prior to embryo collection . Y-27632 ( Millipore ) was included in embryo culture medium at a concentration of 80 µM with 0 . 4% DMSO , or 0 . 4% DMSO as control , where indicated . Embryos were cultured at 37°C in a 5% CO2 incubator under light mineral oil . Lats2 and Yap1S112A ( Yap1CA ) mRNA was synthesized from cDNAs cloned into the pcDNA3 . 1-poly ( A ) 83 plasmid ( Yamagata et al . , 2005 ) using the mMESSAGE mMACHINE T7 transcription kit ( Invitrogen ) . EGFP or nls-GFP mRNA were synthesized from EGFP cloned into the pCS2 plasmid or the nls-GFP plasmid ( Ariotti et al . , 2015 ) using the mMESSAGE mMACHINE SP6 transcription kit ( Invitrogen ) . mRNAs were cleaned and concentrated prior to injection using the MEGAclear Transcription Clean-Up Kit ( Invitrogen ) . Lats2 and YAP1CA mRNAs were injected into one blastomere of two-cell stage embryos at a concentration of 500 ng/µl , mixed with 350 ng/µl EGFP or nls-GFP mRNA diluted in 10 mM Tris-HCl ( pH 7 . 4 ) , 0 . 1 mM EDTA . Embryos were fixed with 4% formaldehyde ( Polysciences ) for 10 min , permeabilized with 0 . 5% Triton X-100 ( Sigma Aldrich ) for 30 min , and then blocked with blocking solution ( 10% Fetal Bovine Serum ( Hyclone ) , 0 . 1% Triton X-100 ) for 1 hr at room temperature , or overnight at 4°C . Primary Antibodies used were: mouse anti-CDX2 ( Biogenex , CDX2-88 ) , goat anti-SOX2 ( Neuromics , GT15098 ) , rabbit anti-PARD6B ( Santa Cruz , sc-67393 ) , rabbit anti-PARD6B ( Novus Biologicals , NBP1-87337 ) , mouse anti-PKCζ ( Santa Cruz Biotechnology , sc-17781 ) , rat anti-CDH1 ( Sigma Aldrich , U3254 ) , mouse anti-ZO1 ( Thermo Fisher Scientific , 33–9100 ) , mouse anti-YAP ( Santa Cruz Biotechnology , sc101199 ) , rabbit anti phospho-YAP ( Cell Signaling Technologies , 4911 ) , chicken anti-GFP ( Aves , GFP-1020 ) . Stains used were: Phallodin-633 ( Invitrogen ) , DRAQ5 ( Cell Signaling Technologies ) and DAPI ( Sigma Aldrich ) . Secondary antibodies conjugated to DyLight 488 , Cy3 or Alexa Flour 647 fluorophores were obtained from Jackson ImmunoResearch . Embryos were imaged using an Olympus FluoView FV1000 Confocal Laser Scanning Microscope system with 20x UPlanFLN objective ( 0 . 5 NA ) and 5x digital zoom . For each embryo , z-stacks were collected , with 5 µm intervals between optical sections . All embryos were imaged prior to knowledge of their genotypes . For each embryo , z-stacks were analyzed using Photoshop or Fiji , which enabled the virtual labeling , based on DNA stain , of all individual cell nuclei . Using this label to identify individual cells , each cell in each embryo was then assigned to relevant phenotypic categories , without knowledge of embryo genotype . Phenotypic categories included marker expression ( e . g . , SOX2 or CDX2 positive or negative ) , protein localization ( e . g . , aPKC or CDH1 apical , basal , absent , or unlocalized ) , and cell position , where cells making contact with the external environment were considered ‘outside’ and cells surrounded by other cells were considered ‘inside’ cells . Embryos were fixed , permeabilized , and blocked as described for immunofluorescence . Zonae pellucida were removed using Tyrode’s Acid treatment prior to performing the TUNEL assay ( In Situ Cell Death Detection Kit , Fluorescein , Millipore-Sigma ) . Embryos were incubated in 200 µl of a 1:10 dilution of enzyme in label solution for 2 hr at 37°C . Embryos were then washed twice with blocking solution for 10 min each , and then mounted in a 1 to 400 dilution of DRAQ5 in blocking solution to stain DNA . To determine embryo genotypes , embryos were collected after imaging and genomic DNA extracted using the Extract-N-Amp kit ( Sigma ) in a final volume of 10 µl . Genomic extracts ( 1–2 µl ) were then subjected to PCR using allele-specific primers ( Supplementary file 3 ) . | As an embryo develops , its cells divide , grow and migrate in specific patterns to build an organized collection of cells that go on to form our tissues and organs . One of the first steps – well before the embryo has implanted into the womb – is to allocate cells to make part of the placenta . Once this process is complete , the remaining cells continue building the organism . These cells are pluripotent , meaning they can develop into any part of the body . Scientists think that the embryo manages to sort ‘placenta cells’ from pluripotent ones with the help of certain proteins , which the mother has packaged into her eggs . To investigate this further , Frum et al . used genetic tools to track a specific gene called Sox2 that identifies pluripotent cells as soon as they are formed in mouse embryos . The experiments revealed that the mother places two closely related proteins known as YAP1 and WWTR1 within each egg , which help to make placenta cells different from pluripotent cells . Moreover , both proteins enable the embryo to segregate these two cell types to two different locations: placenta cells are moved to the outer layer of the embryo , while pluripotent cells are moved to the inside . Current technologies allow researchers to create pluripotent cells in the laboratory . But these approaches often result in error , failing to replicate the embryo’s natural ability . By studying how embryos form and arrange pluripotent cells , scientists hope to advance stem cell technology ( which emerge from pluripotent cells ) . This may help to find new ways to heal damaged tissues and organs , or to treat or even prevent many diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2018 | HIPPO signaling resolves embryonic cell fate conflicts during establishment of pluripotency in vivo |
In pig production , inefficient feed digestion causes excessive nutrients such as phosphorus and nitrogen to be released to the environment . To address the issue of environmental emissions , we established transgenic pigs harboring a single-copy quad-cistronic transgene and simultaneously expressing three microbial enzymes , β-glucanase , xylanase , and phytase in the salivary glands . All the transgenic enzymes were successfully expressed , and the digestion of non-starch polysaccharides ( NSPs ) and phytate in the feedstuff was enhanced . Fecal nitrogen and phosphorus outputs in the transgenic pigs were reduced by 23 . 2–45 . 8% , and growth rate improved by 23 . 0% ( gilts ) and 24 . 4% ( boars ) compared with that of age-matched wild-type littermates under the same dietary treatment . The transgenic pigs showed an 11 . 5–14 . 5% improvement in feed conversion rate compared with the wild-type pigs . These findings indicate that the transgenic pigs are promising resources for improving feed efficiency and reducing environmental impact .
Annual global pig production is approximately 1 . 2 billion heads , with more than half produced in China ( USDA , 2016 ) . Grains are the main feedstuff of the pig industry; however , its production capacity in China and many other countries is insufficient . The livestock industry often allows animals to achieve maximal growth in order to fully utilize their economic outputs , yet inefficient feed digestion can cause serious nutrient emissions to the environment . Two nutrients that have received the most attention from environmental groups are nitrogen ( N ) and phosphorus ( P ) which are often supplied in excessive amounts in the diet in order to ensure maximal growth . In pig production , only 1/3 of feed N and P is metabolically utilized from cereal and soybean-based diets . The deposition rate of N is only 25–32% in grower-finisher pigs ( Shirali et al . , 2012 ) . It has been reported that N excretion is up to 20 kg/sow/year and 25 kg/boar/year ( DEFRA , 2007 ) . Approximately 51% of N intake is excreted in urine , which is mainly from protein metabolism and underutilized amino acids and non-protein nitrogen ( NPN ) ( Shirali et al . , 2012 ) . Fecal N excretion comes from undigested protein fractions and endogenous tissue losses such as digestive enzyme secretions and desquamation of intestinal cells , which accounts for 17% of the N intake ( Dourmad et al . , 1999 ) . Only approximately 30% of P is retained in a grower-finisher pig on a cereal-soybean meal-based diet . In total , 70% of ingested P is excreted either through the feces or urine ( Dourmad et al . , 1999 ) . The N and P from animal excreta pollute the water , soil , or air of intensive pig production sites ( Osada et al . , 2011; Philippe et al . , 2011; Carter and Kim , 2013 ) . Surface water becomes eutrophic following excessive P and N inputs , thereby causing overgrowth of blue-green algae and death of aquatic animals ( Jongbloed and Lenis , 1998; Poulsen , 2000 ) . Considering all these aspects , improving nutrient utilization in feed is of great significance to maximize feed grain utilization as well as for environmental conservation . Non-starch polysaccharides ( NSPs ) are primarily present in plant cell walls ( McDougall et al . , 1996; Sarkar et al . , 2009 ) . In cereal grains , arabinoxylans and β-glucans are found in the cell walls of the protein-rich aleurone layer and starchy endosperm and can act as a barrier to nutrient hydrolysis and absorption ( Bacic and Stone , 1981 ) . Similarly , the cell wall polysaccharides of soybean , canola seed , and peas may also be responsible for this nutrient-encapsulating effect ( Omogbenigun et al . , 2004 ) . Therefore , NSPs are the main anti-nutrient factors of cereal and bran ( Fangel et al . , 2012; Sarkar et al . , 2009 ) . Due to a lack of endogenous NSP-degrading enzymes ( NSPases ) , pigs are inherently incapable of digesting NSPs ( Hooda et al . , 2010 ) , but can partially degrade this material through the action of the natural microbial community in their intestinal tract . High-P emission from monogastric animals such as pigs and poultry arises from their poor physiological ability to hydrolyze plant phytates , which account for up to 80% of P in common cereal grains , oil seed meals , and by-products ( Ravindran et al . , 1994 ) . Phytates are negatively charged saturated cyclic acids that can bind to positively charged molecules in the diet such as minerals and protein , thereby reducing nutrient digestibility and increasing discharge of the unabsorbed nutrients to the environment ( Dersjant-Li et al . , 2015 ) . Various methods have been employed to address the issues of inefficient utilization of feed nutrients in the pig industry . For example , dietary supplementation of phytate- or NSP-degrading enzymes has been proposed to reduce P or N emissions from pig farms ( Kiarie et al . , 2010; Zijlstra et al . , 2010 ) , as well as increasing pig body weight gain and feed conversion efficiency ( Diebold et al . , 2004; Willamil et al . , 2012; Woyengo and Nyachoti , 2011 ) . Recent advancements in genetic engineering and animal cloning technologies have facilitated in the establishment of genetically modified pigs with economically significant traits . Transgenic ( TG ) pig lines secreting salivary bacterial phytases have been generated previously . The P content of fecal matter from TG weaner and grower-finisher pigs fed on soybean meals was decreased by as much as 75% and 56% , respectively , compared to their non-TG counterparts . Endogenous salivary phytase significantly promoted the digestion of P from dietary phytates ( Golovan et al . , 2001 ) . To our knowledge , no pig lines that express multi-NSP-degrading enzymes have been established to date . Here , we established stable transgenic pig lines that co-expressed NSP-degrading enzymes ( β-glucanase and xylanase ) and phytase in saliva . Multiple enzymes coordinately degrade NSPs and phytates in feed grains . We also report the grain digestibility , nutrient emission , growth performances , and feed conversion rate of the TG pigs compared with their wild-type littermates .
Through characterization of multiple codon-optimized β-glucanase genes fused with the N-terminal porcine parotid secretory protein ( PSP ) signal peptide , we have determined that Bispora sp . MEY-1 endo-β-glucanase from ( BG17A ) and Bacillus licheniformisβ−1 , 3–1 , 4-glucanase ( EG1314 ) exhibited optimal activity and stability in porcine cells , and the pH condition is compatible to that of the pig digestive tract ( Zhang et al . , 2015 ) . We previously reported that the fused BG17A and EG1314 , which was linked by a self-cleaving 2A peptide , had a broader optimal pH range and higher stability in an acidic environment than either of them alone ( Zhang et al . , 2015 ) . After codon optimization and fusion with the pig PSP signal peptide , three xylanases ( XYNB , XYL11 , and XYF63 ( also known as XYN11F63 ) ) were transfected into PK15 cells and subjected to enzymatic activity assay . Among the three xylanases , XYNB presented the highest enzymatic activity ( Figure 1—figure supplement 1A ) and stability ( Figure 1—figure supplement 1B ) . In addition , XYNB showed greater resistance to peptic and tryptic hydrolysis than the other two xylanases ( Figure 1—figure supplement 1C–E ) . As for the two phytases , Citrobacter freundii APPA ( CAPPA ) only had two narrow peaks at the optimal pH levels of 2 . 5 and 5 . 0 , respectively ( Figure 1—figure supplement 2A ) , whereas Escherichia coli APPA ( EAPPA ) exhibited a broad optimal pH ranging from 1 . 5 to 5 . 0 ( Figure 1—figure supplement 2B ) . EAPPA was more tolerant of pepsin and trypsin than CAPPA . There was almost no reduction in activity of EAPPA after a 2 hr pepsin treatment , whereas 52 . 2% of the biological activity was left for CAPPA after treatment ( Figure 1—figure supplement 2C ) . When treated with trypsin alone , EAPPA and CAPPA retained 98 . 2% and 39 . 7% of their activity , respectively; when treated with trypsin +EDTA , EAPPA and CAPPA retained 31 . 8% and 13 . 7% of their activity , respectively ( Figure 1—figure supplement 2D ) . Based on these results , two β-glucanases genes ( bg17A and eg1314 ) , a xylanase gene ( xynB ) , and a phytase gene ( eappA ) showed better performance that the other candidate transgenes . A polycistronic cassette of fusion enzymes was constructed by head-to-tail ligation of four selected genes and flanked on the 3' end by an Hemagglutinin ( HA ) tag . The DNA sequences of the self-cleaving peptides E2A , T2A , and P2A were used as linkers between the coding DNA sequences of two neighboring enzymes ( Figure 1—figure supplement 3A ) . Before construction of the final TG vector , the fusion enzyme sequences were ligated downstream of the CMV promoter , and the expression level and enzymatic activity of each fusion enzyme were measured in porcine cells . We were able to detect the expression of all the four active enzymes in the PK15 cells , although the expression level and enzymatic activity of each recombinant fusion enzyme was lower than its original monomeric counterpart ( Figure 1—figure supplement 3B and C ) . The lower transfection efficiency of the large transgene construct likely accounted for the observed lower expression and enzymatic activity in cells . We then employed a mouse PSP promoter to replace the CMV promoter to control the salivary gland-specific expression of the fusion gene , and this expression cassette was inserted , together with a CMV promoter-driven neo-EGFP fusion gene , into the piggyBac transposon vector to form a TG vector , namely , pPB-mPSP-BgEgXyAp-neoGFP ( Figure 1A ) . The resulting TG piggyBac transposon vector ( pPB-mPSP-BgEgXyAp-neoGFP ) and a piggyBac transposase expression vector ( hyPBase ) were co-transfected into the porcine fetal fibroblasts ( PFFs ) of a male Duroc pig . Transfected PFFs were selected with G418 for approximately two weeks , and the resulting EGFP-expressing cell colonies were pooled and identified by PCR for the presence of the transgene . Four colonies with great EGFP expression were used as donor cells for somatic cell nuclear transfer . A total of 4008 reconstructed embryos were generated and transferred to 16 recipient sows ( Supplementary file 11 ) . Thirty-three live and two stillborn cloned piglets were born , of which 25 founders were positive for transgene by PCR detection ( Figure 1—figure supplement 4A and B ) . Bright green fluorescence signals were observed in hoof , tongue , heart , muscle , and submandibular gland ( Figure 1B and C ) . Among the 25 TG founders , five piglets ( 601 , 603 , 701 , 705 , and 709 ) harbored the fragments of the ampicillin-resistance gene of the transgene vector ( Figure 1—figure supplement 4A ) , implying the occurrence of a random , but not transposon-mediated transgene integration into host cell genome . The other 20 piglets harbored the intact transgene expression cassette ( total length: 19 , 886 bp ) . Southern blotting , quantitative PCR , inverse PCR , and sequencing results further demonstrated that 19 piglets carried a single copy of the transgene ( Figure 1D; Figure 1—figure supplement 4C; Figure 1—figure supplement 5; Supplementary file 1 ) , of which two carried a single copy of the transgene that was inserted into intron 1 of Legumain ( line 1 ) ( Figure 1—figure supplement 4C; Figure 1—figure supplement 5; Supplementary file 1 ) , 17 carried a single copy of the transgene that was integrated into intron 5 of CEP112 ( line 2 , of which eight piglets survived ) ( Figure 1D; Figure 1—figure supplement 5; Supplementary file 1 ) . One ( 708 ) carried three copies of the transgene , and the integration site was the intergenic region between LOC100525528 and CXCL2 ( Figure 1D; Supplementary file 1 ) . There were a total of 16 piglets survived to weaning . Thirteen of them are positive for transgene , including nine piggyBac-mediated ( one of line 1 and 8 of line 2 ) and four randomly integrated transgenic pigs . Eight of the transgenic pigs survived to sexual maturity . RT-PCR analysis indicated that the BgEgXyAp , bg17 , eg1314 , xynB , and eappA transgenes were unambiguously expressed in the parotid , submandibular , and sublingual glands , whereas these were undetectable in the other tissues of the TG founders such as the lungs , heart , liver , stomach , spleen , kidney , duodenum , colon , and muscle ( Figure 2A; Figure 2—figure supplement 1A ) . Quantitative PCR analysis indicated that the highest BgEgXyAp transgene expression levels were observed in the parotid gland , followed by the submandibular and sublingual glands , and trace or undetectable level were observed in the other tissues of the TG founders ( Figure 2—figure supplement 1B ) . Ectopic expression was not observed in this study . Western blot analysis demonstrated the expression of β-glucanase , xylanase , and phytase in the saliva of the TG founders ( Figure 2B ) . During the feeding period , the TG pigs ( line 1 and line 2 ) produced 0 . 3–2 . 3 U/mg of β-glucanase , 0 . 6–2 . 4 U/mg of xylanase , and 0 . 5–5 . 7 U/mg of phytase in the saliva ( Figure 2C–E ) . The total salivary protein concentrations of the TG and WT pigs are shown in Figure 2F . Pigs of TG line two that harbored a single-copy transgene within CEP112 intron were used in growth trials and feed evaluations . The TG line two pigs were crossed with WT Duroc pigs , which generated 116 F1 progeny , of which 57 tested positive for transgene . Furthermore , 404 of the F2 progeny were sired , of which 231 were positive for transgene . To understand the effect of the three enzymes of the TG pigs on nutrient digestion , we investigated the pattern of salivary secretion and enzyme production in the TG pigs . Saliva was collected from the unilateral parotid glands of the TG pigs and analyzed in terms of enzyme yield . The average β-glucanase , xylanase , and phytase yields were 2 , 331 . 8 , 2 , 413 . 4 , and 2 , 935 . 2 U per kilogram meal , respectively , in grower pigs , and 920 . 8 , 939 . 0 , and 1 , 042 . 2 U per kilogram meal , respectively , in finisher pigs ( Table 1 ) . The volume of saliva collected from the parotid gland of the finisher pigs was significantly lower compared to that of the grower pigs , which may be attributable to the shorter feeding time ( Ft ) in finisher pigs . Furthermore , saliva and enzyme production at different time points were evaluated . The results show that the pigs only secreted saliva from parotid gland during Ft . At the other time points , including 10 min before or after feeding ( Bf or Af ) , insignificant amount of saliva was secreted ( Figure 2—figure supplement 2A ) . The enzymes expressed by the transgene showed high enzymatic activity at Bf , Ft , and Af . At rest time ( Rt ) , the saliva collected from either parotid or mouth showed reduced enzymatic activities . Of note , a significantly lower enzyme activity was observed in the saliva samples collected at Rt ( Figure 2—figure supplement 2B ) . The grower-finisher founder TG pigs ( weight range: 30 kg to 50 kg ) were fed corn-soybean ( CS ) or wheat-corn-soybean-bran ( WCSB ) diets ( Supplementary file 2 ) to investigate the effects of a salivary cocktail of β-glucanase , xylanase , and phytase on feed utilization . The traditional CS diet contains a low level of NSPs and total P with a high proportion of phytates ( 70 . 3% ) , and the WCSB diet contains a relatively high concentration of NSPs and total P with 63 . 2% phytates . For each diet , 6 TG pigs and six age-matched and weight-matched non-TG Duroc boars ( WTs ) were fed; and 6 WTs were fed the same diet with supplementary multi-enzyme preparations of β-glucanase , xylanase , and phytase ( namely , WT ( + ) group ) . After dietary treatment , nutrient digestion among the experimental groups was measured and compared . For both diets , the apparent total tract digestibility ( ATTD ) of dry matter ( DM ) , P , N , and calcium ( Ca ) significantly increased in TG pigs compared with that of WT pigs ( Figure 3A ) . The fecal outputs of N and P , relative to input by feed , were significantly decreased in the TG pigs compared with that of the WT pigs ( Figure 3B ) . Fecal N and P excretion were decreased by 24 . 9% and 45 . 8% , respectively , with the CS diet , and 23 . 2% and 34 . 8% , respectively , with the WCSB diet . A significant reduction in total P and Ca ( feces plus urine ) was also observed in the TG pigs with both diets . Almost all tested parameters of the TG pigs showed some improvement compared with that of the WT ( + ) pigs that were fed the same diets supplemented with multi-enzyme preparations , although the differences were not statistically significant among groups ( Supplementary file 3 and Supplementary file 4 ) . The serum components of the TG pigs fed on a diet in Supplementary file 5 containing a low N level and high proportion of phytates ( 78 . 4% ) ( LNHP ) were analyzed . Serum alkaline phosphatase activity in the TG pigs was lower than the WT littermates . Serum P and glucose levels of the TG pigs were greater than that of the WT littermates . The serum D-xylose levels of the TG pigs were slightly higher than the WT littermates . No differences in serum Ca , Zn , urea N , uric acid , and total protein concentrations between TG and WT pigs were observed ( Table 2 ) . To assess the growth performance , eight F1 TG pigs ( females ) and 17 WT littermates ( females ) were fed a LNHP diet ( Supplementary file 5 ) during the growing period from 30 kg to 50 kg weights was measured . The TG pigs exhibited a higher average daily gain ( ADG ) rate and lower feed conversion rate ( FCR ) than the WT pigs during this stage ( Supplementary file 6 ) . Growth performance of the F2 TG pigs was also measured . A total of 74 F2 TG pigs ( 23 boars , 51 gilts ) and 52 WT littermates ( 21 boars , 31 gilts ) were raised together and fed the same diets shown in Supplementary file 7 . In this study , TG boars showed increased average daily feed intake ( ADFI ) ( p=0 . 077 ) compared to the WT boars that were fed the same diets ( Figure 4A ) . Similar results were observed in the gilts . Significantly improved ADG rates and lower FCR were observed in the TG boars and gilts compared to the WT gilts during the entire feeding period ( Figure 4B and C ) . It took an average of 110 days for TG boars to grow from 30 kg to 115 kg , whereas the WT boars needed 145 days of feeding on the same diets . Similar results were observed in the gilts . It took 121 days for the TG gilts to grow from 30 kg to 115 kg , whereas the WT females required 150 days ( Figure 4D ) . Taken together , the TG pigs showed a 7 . 0–7 . 6% higher ADFI than the WT pigs over the same grower and finisher phases ( 30 kg to 115 kg ) , but gained 23 . 0–24 . 4% more body weight ( BW ) daily than the WT pigs over the same grower and finisher phases ( 30 kg to 115 kg ) . The time to reach 115 kg BW was shortened by 19 . 2–21 . 9% ( 29 . 6 days to 35 . 1 days ) . The FCR decreased by 11 . 5–14 . 5% during the grower and finisher periods .
Previous studies have shown that TG pigs that secrete the microbial enzyme phytase from their salivary glands have significantly reduced P levels in their manure ( Golovan et al . , 2001; Meidinger et al . , 2013 ) . The major objective of the present study was to enhance the digestive utilization of feed grain and decrease the N and P emissions from pig manure . We obtained TG pigs by introducing the microbial genes bg17A , eg1314 , xynB , and eappA , which encode for endo-glucanase ( the glycoside hydrolase family 7 ) , endo-β−1 , 3–1 , 4-glucanase ( the glycoside hydrolase family 16 ) , endo-xylanase , and 6-phytase , respectively , into the genome of pigs so that these could produce phytate- and NSP-degrading enzymes . These enzymes , which are inherently secreted by microbial communities , were optimized to adapt to the digestive tract environment of pigs and expressed specifically in the porcine salivary gland to initiate the digestion process of NSPs and phytates in the mouth . The feeding trials demonstrated that the TG pigs possessed a high digestive capacity for N , DM , phytate P , and other nutrients , enhanced growth performance , and decreased manure nutrient emissions . We did not find any negative side effects in these TG animals such as changes in spirit , behavior , reproductive capacity , viability , growth performance , blood physiology , and biochemistry . We totally obtained 33 live cloned piglets . Twenty-five of them were positive for the transgene . Although we used the selected single-cell-colonies for SCNT , these colonies might include few cells from other colonies with different genotypes as all colonies were cultured and screened on the same dish . Therefore , the transgene-negative pigs will be produced if impure cell colonies were used for animal cloning . Many of the piglets born alive died from visible malformation ( 30 . 3% ) and weakness ( 18 . 2% ) likely related to various internal malformation in the digestive , circulatory or cardiopulmonary systems . A previous study reported that WT and TG cloned pigs showed 10 . 9–17 . 8% and 25 . 2–39 . 8% malformation rates , respectively , which resulted in a considerable loss ( 58% ) of cloned piglets before weaning ( Liu et al . , 2015a; Schmidt et al . , 2015 ) . In our result , the malformation and mortality rate was in agreement with the previous report . Notably , the non-TG cloned pigs in our study also showed a high death rate , suggesting that the problem of high mortality rate of transgenic cloned pigs may relate to cloning technique . The insertion sites of transgene are the intron region of Legumain and CEP112 for line 1 and line 2 , respectively . Legumain is a cysteine protease that involves in antigen processing for class II MHC presentation in lysosomes ( Chen et al . , 1997 ) . Legumain is overexpressed in some solid tumors ( Dall and Brandstetter , 2016 ) . Previous studies reported that inhibition/knock-down of legumain could suppress tumor progression ( Briggs et al . , 2010; Liu et al . , 2015b ) . Legumain knockout mice showed a phenotype of decreased weight gain ( Shirahama-Noda et al . , 2003 ) . CEP112 gene encodes a centrosomal protein of 112 kDa which is localized around spindle poles , and its structure and function are still unclear ( Jakobsen et al . , 2011; Kumar et al . , 2013 ) . The expression and function of the two genes in TG pigs may not be compromised significantly , as the foreign transgene is integrated into the intron region . No significant phenotypic difference was observed between the TG pigs and their WT littermates . In this study , transgenes encoding microbial enzymes were driven by the salivary gland-specific PSP promoter . TG pig line two harbors a single copy of a multi-transgene , which facilitates the establishment of a breeding population with identical genotypes . Compared to the previously reported phytase TG pig lines that harbor a high copy number of a foreign gene in an insertion locus ( 2 to 35 copies of the transgene , producing 5 to 6 , 000 U/mL of phytase ) ( Golovan et al . , 2001 ) , our multi-TG pigs secreted a lower amount of enzymes into the saliva . In fact , our multi-TG pigs demonstrated similar digestibility with the high-expression phytase TG pigs , as well as reduced emission of P in manure . For grower-finisher pigs , our study displayed an approximately 45 . 7% reduction in fecal P when fed CS meals ( 70 . 3% phytates ) , versus 56% reportedly in high-expression phytase TG pigs fed on soybean meals ( 53% phytates ) ( Golovan et al . , 2001 ) . The three major salivary glands showed high expression levels and enzymes activities , which could be arranged in decreasing order as follows: parotid gland >sublingual gland>submaxillary glands , which agree with the findings of a previous report ( Golovan et al . , 2001 ) . The enzyme activities per milliliter of parotid saliva between grower pigs and finisher pigs were similar . The expression levels of glucanase , xylanase , and phytase in the bilateral parotid glands of the TG pigs were 4 , 663 . 68 , 4 , 826 . 76 , and 5 , 870 . 38 U/kg diet intake for grower pigs , and 1 , 841 . 64 , 1 , 878 . 06 , and 2 , 084 . 38 U/kg diet intake for finisher pigs . These levels of expression are higher than the amounts used as dietary supplements . The TG pigs demonstrated increased P retention and reduced percentages of manure P excretions compared with age-matched WT pigs fed on the same diets , implying the significant digestive effect of TG phytase . The P retention rate was 54 . 7–67 . 3% , and reduction in total P output ranged from 21 . 3% to 44 . 2% . These results are slightly lower than that of previously reported TG phytase pigs , which had P retention rates of 57 . 2–77 . 8% , and total P output decreased by 27 . 5–62 . 0% ( Golovan et al . , 2001; Meidinger et al . , 2013 ) . This discrepancy may be caused by the different expression levels of transgenes and the ingredients of the diets . However , the new pig model demonstrated that N digestibility was significantly enhanced in all diets tested . WCSB diets contain high levels of NSPs , which account for the highest fraction of polymeric carbohydrate in protein-rich materials ( aleurone layer ) . Glucanase and xylanase secreted by TG pigs can effectively degrade the glucans and xylans in the cell walls of the aleurone layer , in which the matrix proteins and phytate globoids would be exposed and subjected to degradation by phytase and endogenous proteases . These results agree with that of high-NSP diets supplemented with NSPase ( Diebold et al . , 2004; Willamil et al . , 2012 ) . For the corn-soybean diet , the effects of enzyme supplementation on nutrient digestibility of pigs , as reported in the literature , were highly variable . Some studies show positive responses to enzyme supplementation ( Ji et al . , 2008; Kim et al . , 2003 ) , whereas others do not ( Li et al . , 1999; Willamil et al . , 2012 ) . TG pigs fed on CS diets also showed significantly enhanced N digestibility . As the main ingredient of CS diets , corn contains high levels of phytates . The synergistic effect of NSPase ( glucanase and xylanase ) and phytase involves the digest of NSPs and phytates in the aleurone layer and endosperm cells of corn , thereby alleviating the anti-nutrition effect of phytates by reducing the binding of phytates to proteins and digestive enzymes in the digestive tract . In terms of N emission , the grower TG pigs emitted 24 . 0% less fecal N than the WT pigs . The NSPs in the cell wall are degraded by NSPases ( glucanase and xylanase ) that are secreted in the upper digestive tract . Moreover , NSPases can reduce the compensated secretion of endogenous fluids to decrease endogenous nutrient losses ( Kerr and Shurson , 2013 ) . TG phytase facilitates in the release of phytate-chelating proteases and other digestive enzymes to further degrade nutrients . The TG pigs displayed better growth performance than conventional pigs fed on diets without supplemental P during the grower and finisher phases . Both TG boars and gilts showed a faster rate in BW gain , and exhibited greater feed efficiency than the respective WT littermates . Blood serum measurements also demonstrated enhanced serum P , glucose , and xylose levels in TG pigs , which signified enhanced digestive utilization of these nutrients . A significant decrease in serum alkaline phosphatase levels is an indicator of well-developed bone ( Meidinger et al . , 2013; Sefer et al . , 2012; Selle and Ravindran , 2008 ) . These results are clear manifestations of the improved nutrient digestibility and growth performance of TG pigs . In summary , the multi-TG pigs reported in the present study exhibited significantly enhanced digestibility of feed N , P , Ca , and other minerals . The pigs produced manure with significantly reduced P and N emissions into the environment as well as exhibited improved growth performances . This genetic strategy offers a very valuable biological solution for inefficient feed digestion and environmental emissions due to the global expansion of the livestock industry .
The codons of two β-glucanase genes , bgl7A from Bispora sp . MEY-1 ( Luo et al . , 2010 ) and eg1314 from Bacillus licheniformis EGW039 ( CGMCC 0635 ) ( Teng et al . , 2006 ) , three xylanase genes , xyl11 , xyn63 , and xynB , from Aspergillus niger CBS513 . 88 ( Liu et al . , 2010 ) , Penicillium sp . F63 CGMCC1669 ( Deng et al . , 2006 ) , and Aspergillus niger CGMCC1067 ( Guo et al . , 2013 ) , ( respectively ) , as well as two phytase genes , eappA ( GenBank accession No . AF537219 . 1 ) from Escherichia coli and cappA ( GenBank accession No . AF537219 . 1 ) from Citrobacter freundii were optimized per codon usage bias in pigs . The original signal peptide of these genes were replaced by the signal peptide of the porcine parotid secretory protein ( PSP ) . Optimized genes were synthesized by Genscript ( Nanjing , China ) and ligated into the eukaryotic expression vectors pCDNA3 . 1+or pcDNA6 . 0 ( Invitrogen , Carlsbad , CA , USA ) . The bgl7A , eg1314 , xynB , and eappA genes were fused in a head-to-tail tandem array , with E2A , T2A , and P2A used as linkers between them . Flag-tag and HA-tag were added to the C terminal of Eg1314 and EAPPA , respectively , to facilitate the detection of protein expression . The fusion of the four genes , named BgEgXyAp , was cloned into pCDNA3 . 1 ( + ) to examine its expression and enzyme activity levels by transient transfection using Lipofectamine 2000 ( Invitrogen Carlsbad , CA , USA ) in porcine cells . A 12 . 1 kb upstream genomic sequence of murine PSP , as a promoter to drive the expression of the fusion gene specifically in salivary gland , were cloned and ligated to BgEgXyAp , and then introduced into the transposon piggyBac vector pPB-lox-neoEGFP-loxp ( a gift from The Wellcome Trust Sanger Institute , Cambridgeshire , UK ) to form the final transgene construct . The final construct was confirmed by sequence analysis . Primary PFFs were isolated from 35-day-old male fetuses of Duroc pigs . PFFs were cultured in Dulbecco's Modified Eagle Medium ( DMEM , Gibco 12491–023 ) ( Thermo Fisher Scientific , Suwanee , GA , USA ) supplemented with 12% fetal bovine serum ( FBS , Gibco 10100–147 ) ( Thermo Fisher Scientific , Suwanee , GA , USA ) and 1% ( v:v ) penicillin/streptomycin ( 10 , 000 U/mL penicillin , 10 , 000 μg/mL streptomycin; GIBCO-BRL , Grand Island , NY , USA ) at 39°C in an incubator with 5% CO2 . The transgene was mixed with a transposase , pCMV-hyPBase ( a gift from the University of Hawaii , Honolulu , HI ) , and transfected into PFFs by electroporation ( BTX , San Diego , CA ) . The transfected cells were split 1:6 into fresh culture medium . After 24 hr , 300 μg/mL G418 ( Gibco ) was added to the medium to select transfected cell colonies , and the plates were incubated in media containing G418 for about 15 days . The surviving cell colonies with EGFP expression were isolated within colony cylinders ( Bellco Glass , Vineland , NJ , USA ) , and propagated in a fresh 24-well plate . Four colonies that proliferated well , with bright fluorescence , were then expanded and screened for the presence of the BgEgXyAp transgene . SCNT was performed as previously described ( Lai et al . , 2006 ) . The reconstructed embryos were surgically transferred to the oviduct of the recipient gilts the day after estrus was observed . The pregnancy status of the surrogates was detected using an ultrasound scanner at 26 d after the embryo was transferred . This status was then monitored weekly before the expected due date . The cloned piglets were born by natural birth . Genomic DNA was isolated from pig ear skin biopsies and used in PCR identification of the transgene . The PCR primes are listed in Supplementary file 8 . For RT-PCR , reverse transcription of mRNA was conducted to generate cDNA , which was then used as template for PCR . The RT-PCR primers are listed in Supplementary file 9 . Relative quantitative real-time PCR and absolute quantitative real-time PCR used in detecting expression levels and copy numbers of transgene , respectively , were performed as described elsewhere ( Wu et al . , 2013 ) . The Q-PCR primers are listed in Supplementary file 9 . Genomic DNA was isolated from the ears of transgenic founders and wild-type ( WT ) controls by phenol-chloroform extraction . Fifteen micrograms of DNA were digested with HindIII , fractionated in a 0 . 8% agarose gel , and transferred onto a nylon membrane ( GE Healthcare , Pittsburgh , PA , USA ) . The membrane was then hybridized with a probe . The probe primers are listed in Supplementary file 8 . Hybridization and washing were performed with DIG-High Prime DNA Labeling and Detection Starter Kit II ( Roche , Basel , Switzerland ) . Prehybridization was conducted at 42°C for 30 min , hybridization at 50°C for 8 hr , then incubated for 30 min in blocking solution and further incubated for 30 min in an anti-digoxigenin-AP antibody solution . After incubation , the membrane was exposed for 5–20 min to 1 mL of ready-to-use CSPD , and images were captured with an EC3 imaging system ( UVP , LLC , Upland , CA , USA ) . Pig saliva was concentrated using an Amicon Ultra 15 mL centrifugal filter ( Millipore ) . After protein quantification , total protein samples were separated on a 10% sodium dodecyl sulfate polyacrylamide gel ( SDS-PAGE ) , and transferred to a polyvinylidene difluoride ( PVDF ) ( Millipore , Temecula , CA , USA ) membrane . The membranes were blocked with 5% non-fat dry milk , subsequently incubated with the corresponding antibodies , and then developed with an enhanced chemiluminescence solution ( Thermo Fisher Scientific , Suwanee , GA , USA ) . Chemiluminescent signals were captured by a cooled charged-coupled device ( CCD ) camera . For the primary antibodies , rabbit polyclonal anti-BG17A and anti-XYNB antibodies , which were prepared by Genscript ( Nanjing , China ) , were respectively used to detect β-glucanases and xylanase . A mouse monoclonal anti-HA antibody ( Abcam , Cambridge , UK ) was used to detect phytase . Information on the antibodies used in this study is presented in Supplementary file 10 . Horseradish peroxidase-conjugated anti-mouse IgG or horseradish peroxidase-conjugated anti-rabbit IgG ( Abcam , Cambridge , UK ) was used as secondary antibody . Oral saliva was collected by clamping an absorbent cotton into the buccal cavity , and allowing the piglets to soak with saliva and chew . The resulting liquid was squeezed out using an injector , centrifuged , and then used for assays . To collect saliva from the parotid gland , a fistula was surgically installed under anesthesia with propofol at the unilateral ( right cheek ) parotid duct of each pig . Saliva from the parotid gland was collected using a medical drainage bag during feeding time , before feeding ( within 30 min ) , after feeding ( within 30 min ) and at selected time points dur The supernatants from transfected cells and saliva from transgenic and non-transgenic pigs were used as total protein samples for the enzymatic activity assays . β-glucanase and xylanase activity assays were based on estimating the amount of reducing sugars released from the relevant substrates in the reactions using 3 , 5-dinitrosalicylic acid ( DNS ) reagent , as previously described ( Liu et al . , 2010; Luo et al . , 2010 ) . One unit of activity was defined as the quantity of enzyme that releases reducing sugar at the rate of 1 μmol/min . Phytase activity in saliva was determined by means of vanadium molybdenum yellow spectrophotometry . The reaction was performed in a final volume of 600 μL solution containing 0 . 25 M of acetate buffer ( pH 5 . 5 ) , 5 mM sodium phytate , and 50 μL enzyme preparation at 39°C for 30 min , followed by termination of reaction by adding 400 mL of an ammonium molybdate-ammonium vanadate-nitric acid mixture . After mixing and centrifugation , the absorbance was measured at a wavelength of 415 nm . One unit of phytase activity was defined as the amount of activity that liberates one micromole of phosphate per minute at 39°C . Six TG pigs and 12 age- and body weight-matched non-TG pigs were used in dietary treatment experiments . The TG group and six non-TG pigs were fed a diet ( CS or WCSB ) . The other six non-TG pigs were fed the same diet with exogenous feed enzymes . The pigs were housed in small groups of three to four animals per pen . Each pig was kept in individual metabolic cage ( Length × width × height: 1 . 40 × 0 . 67 × 1 . 15 m ) with stainless steel mesh floors for collection of urine and feces sample at the indicated time points . The cage was laid side by side for pig to communicate with each other . The facilities were provided with forced ventilation and heat lamp for thermal regulation , and each cage had one feeder and one water nipple for ad libitum access to feed and water . The ingredients of the selected experimental diets are listed in Supplementary file 2 . The diets were provided in pellet form . All pigs were fed on a commercial grower diet for an adaptation period of one week in the cage , and then fed experimental diets for a pre-test period of six days prior to the start of the experiment . During the dietary treatment period , the animals were fed the experimental diets for four days . Pig feed intake was based on BW ( BW ×0 . 04 ) . Equal quantities of diets were added to the feeders twice daily in the morning ( 09:00 ) and afternoon ( 16:00 ) , and the initial and final BW of each experimental diet were recorded . Fecal samples were collected and processed as previously described ( Kim et al . , 2005 ) . Feeds and dried feces were ground using a grinder , passed through a 0 . 425 mm size screen , and analyzed for DM ( AOAC , 2005; method 930 . 15 ) . Gross energy ( GE ) was analyzed according to ISO: 9831–1998 using a bomb calorimetry ( Parr 6300; Parr Instrument Co . , Louisville , KY , USA ) . Other nutrients in the diets and feces were analyzed using the Chinese National Standard analytical method ( GB/T ) . The following methods were used: GB/T 6432–1994 for CP , GB/T 20806–2006 for NDF , GB/T 20805–2006 for ADF , GB/T 6434–2006 for CF , GB/T 18335–2003 for calcium , GB/T 6437–2002 for P , GB/T 6438–2007 for ash , and GB/T 21912–2008 for titanium dioxide . Apparent digestibility coefficients were calculated as described elsewhere ( Grela et al . , 20112018 ) . The growth of the F2 TG pigs of both genders was compared to that of age-matched WT littermates fed on the same diets listed in Supplementary file 7 . A total of 23 TG boars ( 32 . 9 ± 3 . 80 kg ) and 21 WT littermates boars ( 31 . 5 ± 2 . 97 kg ) were grouped according to weight , and randomly allocated to seven pens fitted with MK3 FIRE feeders ( FIRE , Osborne Industries Inc . , Osborne , KS , USA ) . Similarly , 51 TG gilts ( 31 . 8 ± 3 . 55 ) and 31 WT gilts ( 30 . 2 ± 1 . 64 ) were randomly allocated to 11 pens that were fitted with MK3 FIRE feeders . Individual feed intake and BW were recorded when a pig with an ear transponder visited the FIRE feeders . All pigs were allowed free access to water throughout the measurement phase . The data were analyzed using the GLM procedure ( SAS Inst . Inc . , Cary , NC , USA ) . For the apparent total tract nutrient digestibility values , fecal nutrient output , and the growth performance , analysis of covariance ( ANCOVA ) was used . The BW of the tested pigs at the start of the corresponding experimental period was used as the covariable . Least square means were calculated and the differences between means were tested using Turkey-Kramer adjustment for multiple comparisons when appropriate . For salivary protein and saliva secretion by the parotid gland , one-way ANOVA followed by Duncan's multiple-comparison tests were used . For serum biochemical endpoints and saliva enzymes secretion , an unpaired two-sample t-test ( two-tailed ) was used . The level of significance was set at p<0 . 05 , and trends were discussed at p<0 . 1 . | The bodily waste that pigs produce contains high levels of chemicals that can damage the environment , such as nitrogen and phosphorus . For example , when excessive amounts of these two compounds make their way into the water , they can cause blue-green algae to grow too much , which asphyxiates other life in the water . Pigs produce a lot of nitrogen and phosphorus because they cannot efficiently digest their food . In particular , the animals lack the enzymes required to break down two types of molecules present in their feedstuff: phytates and non-starch polysaccharides ( NSPs ) . Zhang , Li et al . take four microbial genes which code for the enzymes needed to digest NSPs and phytates , and they add these DNA sequences into the genomes of pigs . The animals then produce enzymes in their saliva that transform NSPs and phytates into molecules which can be used by their digestive system . The pigs thus get more energy from their food , and they grow faster and bigger . They also produce less nitrogen and phosphorus in their waste . Over 1 . 2 billion pigs are farmed each year , and they are the most economically important meat source in the world . Raising animals that can digest their food better would reduce the need for pig feed , increase productivity and reduce environmental pollution . However , discussions with policy makers and with the public will be necessary before these results can be adopted by the farming industry . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"tools",
"and",
"resources",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Novel transgenic pigs with enhanced growth and reduced environmental impact |
Experience alters brain structure , but the underlying mechanism remained unknown . Structural plasticity reveals that brain function is encoded in generative changes to cells that compete with destructive processes driving neurodegeneration . At an adult critical period , experience increases fiber number and brain size in Drosophila . Here , we asked if Toll receptors are involved . Tolls demarcate a map of brain anatomical domains . Focusing on Toll-2 , loss of function caused apoptosis , neurite atrophy and impaired behaviour . Toll-2 gain of function and neuronal activity at the critical period increased cell number . Toll-2 induced cycling of adult progenitor cells via a novel pathway , that antagonized MyD88-dependent quiescence , and engaged Weckle and Yorkie downstream . Constant knock-down of multiple Tolls synergistically reduced brain size . Conditional over-expression of Toll-2 and wek at the adult critical period increased brain size . Through their topographic distribution , Toll receptors regulate neuronal number and brain size , modulating structural plasticity in the adult brain .
Structural brain plasticity and neurodegeneration reveal generative and destructive processes operating in the brain . Plasticity reflects adaptations of the brain to environmental change , involving adult neurogenesis , growth of neurites and synapses , which correlate with learning , experience , physical exercise and anti-depressant treatment ( Holtmaat and Svoboda , 2009; Deng et al . , 2010 ) ; conversely , neuroinflammation , neurodegeneration , loss of neurons , neurites and synapses , correlate with ageing , stress , depression and disease ( Wohleb et al . , 2016 ) . Structural brain plasticity affects the brain topographically , influencing the specific regions involved in experience-dependent processing . These manifestations suggest that brain function is encoded in physical changes to cells . Structural plasticity occurs in the Drosophila brain ( Sugie et al . , 2018 ) . Breeding adult flies in constant darkness decreases , and in constant light increases brain volume ( Barth and Heisenberg , 1997; Barth et al . , 1997 ) . Breeding adult flies in isolation vs . crowded conditions , or in single sex vs . mixed groups , also causes brain volume changes ( Technau , 1984; Heisenberg et al . , 1995 ) . The affected modules include the optic lobe , the mushroom body calyx and central complex ( Technau , 1984; Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . Changes in brain volume are prominent in a critical period spanning from adult eclosion to day 5 , and correlate with changes in fiber number ( Technau , 1984; Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . The molecular mechanisms underlying structural brain plasticity are unknown , and discovering them is crucial to understand the normal functionality of the brain as well as its pathological responses to disease . Primary candidates to regulate brain plasticity are the neurotrophins . In the mammalian brain , neurotrophins ( BDNF , NGF , NT3 , NT4 ) regulate cell proliferation , cell survival , circuit connectivity , synaptic transmission and potentiation ( Lu et al . , 2005 ) . Alterations in neurotrophins underlie brain disease , and anti-depressants increase the levels of the neurotrophin BDNF ( Krishnan and Nestler , 2008; Wohleb et al . , 2016 ) . NTs have dual functions , as they promote plasticity via p75NTR activating NF-κΒ , and via Trk receptors activating AKT , ERK and CREB downstream , and they promote neurodegeneration via p75NTR and JNK signalling ( Lu et al . , 2005 ) . Drosophila neurotrophins ( DNTs ) also regulate neuronal survival and death , connectivity and synaptic structural plasticity ( Zhu et al . , 2008; Sutcliffe et al . , 2013; McIlroy et al . , 2013; Foldi et al . , 2017; Ulian-Benitez et al . , 2017 ) . However , there are no canonical tyrosine-kinase-Trk and p75NTR receptors in Drosophila , and instead , DNTs are ligands for the Kekkons , kinase-less members of the Trk family , and Tolls ( McIlroy et al . , 2013; Foldi et al . , 2017; Ulian-Benitez et al . , 2017 ) . Drosophila Toll and mammalian Toll-Like-Receptors ( TLRs ) are best known for their universal function in innate immunity ( Leulier and Lemaitre , 2008 ) , but also have non-immune functions in development and in the central nervous system ( CNS ) ( Anthoney et al . , 2018 ) . In neurons , Tolls and TLRs can promote neuronal survival via MyD88 and neuronal death via Sarm , both in flies and mammals ( Kim et al . , 2007; McIlroy et al . , 2013; Mukherjee et al . , 2015; Foldi et al . , 2017 ) . In humans , alterations in TLR function underlie brain diseases from stroke and neurodegeneration to multiple sclerosis and neuroinflammation ( Okun et al . , 2011; Hanamsagar et al . , 2012 ) . Most attention has focused on TLR functions in microglia , their response to damage or infection , and in neuroinflammation ( Fiebich et al . , 2018 ) . However , TLRs are also in neurons , but functions in neurons and neural progenitor cells are largely unknown . Importantly , TLRs can influence neurogenesis , neuronal survival and death , neurite growth , synaptic transmission and behaviour , including learning and memory ( Ma et al . , 2006; Rolls et al . , 2007; Okun et al . , 2010b; Okun et al . , 2011; Qi et al . , 2011; Okun et al . , 2012; Madar et al . , 2015; Liu et al . , 2016b; Patel et al . , 2016; Hung et al . , 2018; Min et al . , 2018 ) . These findings suggest that TLRs could regulate structural brain plasticity , but this remains little explored . Tolls regulate cell number plasticity in the Drosophila ventral nerve cord ( VNC ) through a three-tier mechanism ( Foldi et al . , 2017 ) . In embryos and larvae , Toll-6 and Toll-7 maintain neuronal survival via MyD88 and NF-κB ( McIlroy et al . , 2013; Foldi et al . , 2017 ) . However , in pupae , they can also promote apoptosis via Weckle ( Wek ) , Sarm and JNK ( Foldi et al . , 2017 ) . Furthermore , different Tolls lead to different outcomes , for instance , Toll-1 is more pro-apoptotic than Toll-6 ( Foldi et al . , 2017 ) . Whether a neuron lives or dies in the CNS depends on the ligand and its cleavage state it receives , the Toll or combination of Tolls it expresses , and the downstream adaptors available for signalling ( Foldi et al . , 2017 ) . Thus , cell number control is context dependent . The ability of DNTs and Tolls to regulate cell number by promoting both cell survival and cell death is crucial for the modulation of structural brain plasticity , homeostasis and neurodegeneration . Here , we asked whether Toll receptors influence developmental and structural plasticity in the Drosophila brain .
To find out whether Toll receptors are expressed in the brain , we looked for Toll transcripts in embryos and dissected CNSs from larvae to adult brains , using reverse-transcription PCR ( RT-PCR ) ( Figure 1—figure supplement 1 ) . Toll-3 transcripts were absent from larval L2 CNSs; Toll-4 and −9 mRNAs were barely detected in all sample types; whereas Toll-1 , –2 , −5 , –6 , −7 , –8 were expressed in embryos , larval ( L2 , L3 ) CNSs , and pupal and adult fly heads ( Figure 1—figure supplement 1 ) . Thus , all Tolls are expressed in pupal and adult brains , with Toll-1 , –2 , −5 , –6 , −7 , –8 most prominently . To visualise the spatial distribution of Tolls in the brain , we generated GAL4 reporter lines for the Tolls . Using CRISPR/Cas9-accelerated homologous recombination to insert a pTV cassette ( Baena-Lopez et al . , 2013 ) , we generated a knock-in/knock-out Toll-2pTV allele , and we used the pTV-attP landing site to generate a Toll-2pTV-GAL4 driver line . Toll-4GAL4 and Toll-5GAL4 were generated by CRISPR/Cas9 , inserting T2AGAL4 immediately upstream of the start codon . Unfortunately , we could not get transformants for Toll-9 . Toll-3GAL4 , Toll-6GAL4 and Toll-7GAL4 were made using Recombinase-Mediated Cassette Exchange ( RMCE ) of MIMIC insertions into the intronless coding regions of these genes . Toll-8GAL4 is TolloMD806 , which has a P-element insertion just 180 bp upstream of the start codon , within the 5’UTR of Toll-8 . The GAL4 driver lines were used to visualise membrane tethered FlyBow ( for Toll-2 , –4 , −5 , –7 ) and tdTomato ( for Toll-3 , –6 , −8 ) reporters , and all necessarily reproduced the endogenous expression patterns of the Toll genes . Toll-1 was visualised using commercially available and previously validated anti-Toll-1 antibodies ( Lund et al . , 2010; Khadilkar et al . , 2017 ) . In the adult brain , Toll-1 was found in all photoreceptor cells ( Figure 1A , G ) . Toll-2 , –5 , −6 , –7 and −8 were all expressed in Kenyon cells , with Toll-2 and −6 comprising most cells ( Figure 1A , B , C ) . Toll-5 and −7 were expressed in the protocerebral bridge ( Figure 1A , B ) . Toll-2 , –5 , −6 , –7 and −8 were differentially expressed in the antennal lobes ( Figure 1D ) . Toll-1 , –2 , −3 , –5 , −6 , –7 , −8 were expressed in distinct and overlapping fan shaped body neuropile layers ( Figure 1E ) ; Toll-1 , –2 and −7 in distinct ellipsoid body neuropile rings ( Figure 1F ) , and Tol-1 , –2 , −6 and −8 in the sub-esophageal ganglion ( SOG ) ( Figure 1A ) . Toll-2 , –3 , −5 , –6 , −7 , –8 were expressed in optic lobes , with Toll-3 and −6 having a prominent expression in the lamina ( Figure 1A , G ) and Toll-2 a broad expression throughout the optic lobes ( Figure 1A , G ) . In summary , these patterns revealed: ( 1 ) a map of Toll expression profiles coincident with anatomical brain domains; ( 2 ) profiles specific to each Toll; ( 3 ) complementary patterns in neuropiles of the visual system and central complex ( fan shaped body , ellipsoid body and protocerebral bridge ) ; ( 4 ) overlapping distributions in optic lobes , antennal lobes , Kenyon cells and mushroom bodies . Tolls could influence brain structure and connectivity by virtue of their topographic profiles ( Figure 1I ) . To ask whether Tolls may influence brain development and/or adult structural brain plasticity , we focused on Toll-2 , as it is most broadly expressed . In neurons , Drosophila Tolls and mammalian TLRs promote neuronal survival via MyD88 and neuronal death via Sarm ( Kim et al . , 2007; McIlroy et al . , 2013; Mukherjee et al . , 2015; Foldi et al . , 2017; Figure 2C ) . Thus , to investigate whether Toll-2 is required for cell survival in brain development , we first verified whether MyD88 was expressed in the brain . MyD88NP6394 flies bear a GAL4 insertion within the transcribed 5’UTR exon , and thus it necessarily represents the endogenous expression pattern of the gene ( from now on called MyD88GAL4 ) . MyD88GAL4 >tdTomato revealed MyD88+ cells throughout the optic lobes , central brain , and mushroom bodies ( Figure 2A ) . Toll-2 appears to be expressed in all Kenyon cells , whereas MyD88 is only in the subset that projects along the core α , β lobes ( Figure 2A ) . Using the nuclear reporter histone-YFP ( his-YFP ) revealed that more cells expressed Toll-2 than MyD88 ( Figure 2B ) . In the optic lobes , MyD88 >hisYFP includes large , sparsely distributed cells , that may or may not also be Toll-2+ ( Figure 2B ) . To identify the Toll-2+ and MyD88+ cells , Toll-2 >hisYFP and MyD88 >hisYFP adult brains were labelled with pan-neuronal anti-Elav and pan-glial anti-Repo . There were many MyD88+ Elav+ neurons as well as MyD88+ Repo+ glia ( Figure 2D ) . By contrast , none of the Toll-2+ cells were Repo+ , whilst most Toll-2+ cells were Elav+ ( Figure 2E ) . Thus , MyD88+ cells comprise both neurons and glia , that are most likely regulated by multiple Tolls , and Toll-2+ cells are mostly neurons . To ask whether Toll-2 might regulate cell survival in brain development , we visualised apoptotic cells using anti-Dcp1 antibodies upon Toll-2 knock-down . In brain development , the peak of cell death occurs 24 hr after puparium formation ( Hara et al . , 2018 ) . So , we quantified apoptosis in day one pupal brains , using purposely developed DeadEasy Central Brain software . Toll-2 RNAi knock-down in MyD88+ cells increased apoptosis in the pupal central brain ( Figure 2G ) , meaning that Toll-2 is required to maintain cell survival . To verify whether apoptosis resulted in cell loss , we counted automatically MyD88 >hisYFP cells in the central brain . Using two independent UAS-Toll-2 RNAi lines of flies , Toll-2 knock-down with MyD88GAL4 decreased cell number in the central brain of both pupae and adult flies ( Figure 2H ) . Thus , Toll-2 loss of function in MyD88+ cells increased apoptosis and caused cell loss . On the other hand , sustained over-expression of Toll-2 with MyD88GAL4 throughout development did not affect cell number in the pupal or adult brains ( Figure 2H ) . In larvae and pupae , Tolls can also induce apoptosis via Sarm , and different Tolls have distinct pro-apoptotic drive ( Foldi et al . , 2017 ) . As Toll-2 gain of function did not reduce cell number , this meant that Toll-2 does not induce apoptosis in the pupal or adult brain . Together , these data showed that Toll-2 maintains the survival of MyD88+ neurons during brain development . To test if Toll-2 maintains neuronal survival via the MyD88 pathway , we knocked-down MyD88 with MyD88GAL4 . Similarly to Toll-2 loss of function , MyD88 knock-down also resulted in cell loss in the pupal central brain ( Figure 2H ) . However , by the adult stage , cell number was restored vs . controls ( Figure 2H ) . This was in contrast to the persistent cell loss caused by Toll-2-RNAi into the adult , suggesting that MyD88 carries out further functions too . These data showed that MyD88 is required to maintain cell survival during brain development , downstream of at least Toll-2 . To analyse the effect in Kenyon Cells ( KCs ) , we developed another plug-in - DeadEasy KCs- to count KCs labelled with Toll-2 >hisYFP . Toll-2pTV/Toll-2Δ7-35 mutations increased KC number in pupal brains , but neither sustained gain nor loss of Toll-2 function with Toll-2 >Toll-2RNAi knock-down or Toll-2pTV/Toll-2Δ7-35 mutations affected KC number in adult brains ( Figure 2I ) . Over-expression of Toll-2 with another mushroom body driver , MBGAL4 , did not affect KC number in pupal or adult brains either ( Figure 2—figure supplement 1B , C ) . Thus , KCs are resilient to alterations in Toll-2 function alone . To further test whether Toll-2 is neuroprotective , we induced Toll-2pTV homozygous mutant MARCM clones . Genetic complementation tests over the previously described null allele 18wΔ7-35 ( 18 w is a synonym of Toll-2 , thus hereby will be referred to as Toll-2Δ7-35 ) and a deficiency for the locus , Df ( 2R ) BSC594 , showed that Toll-2pTV is a strong hypomorphic loss of function allele ( Figure 3—figure supplement 1A ) . Toll-2pTV mutant clones were induced from dividing cells in the pupal brain , where Toll-2 is widely expressed ( Figure 3A ) . They were induced using hsFlp , and resulting mutant neurons were visualised in adult brains with elav >mCD8 GFP ( Figure 3B ) . Loss of Toll-2 function caused extensive neuronal loss , neuronal atrophy , loss of neurites - axons and dendrites- and axonal misrouting ( Figure 3C–H ) . Loss of dendrites could be clearly observed in the lamina ( Figure 3C ) ; axonal degeneration and misrouting in medulla and lobula ( Figure 3D ) ; and loss of entire axonal neuropiles in the medulla , SOG and fan shaped body ( Figure 3D , E , H ) . Whether mushroom bodies were affected was less clear ( Figure 3F , G ) , perhaps because our heat shock regime missed mushroom body neuroblast divisions , as we could not observe many mushroom body projections in control brains either ( Figure 3G ) . Dramatic neuronal deficits could be found throughout many brain domains ( Figure 3C–H ) . The loss of neurons in mutant clones was consistent with Toll-2 maintaining neuronal survival , but could also reflect a function promoting progenitor cell proliferation; neurite atrophy meant that Toll-2 loss of function prevented neuronal differentiation or caused neurodegeneration . Toll-2 mutants are semi-lethal , have reduced lifespan and impaired climbing ( Figure 3—figure supplement 1B , C ) – phenotypes commonly associated with neurodegeneration . Toll-2 is expressed in the visual system , ventral nerve cord and central complex , which is the higher control center for locomotion and spatial navigation in the brain ( Strauss and Heisenberg , 1993 ) . Thus , we tested the performance of Toll-2 mutants in the Buridan arena , which could reveal whether loss of Toll-2 affected vertical vs . horizontal locomotion , visual processing , or motivation to walk . Wild-type flies free to walk in a circular lit-up arena walked back and forth between two diametrically opposed dark stripes ( 38 . 5% ) , but most often wondered randomly ( 61 . 5% , Figure 3I ) . Toll-2pTV/Df ( 2R ) BSC22 and Toll-2Δ7-35/Df ( 2R ) BSC22 mutants also most often walked randomly or along the perimeter ( 63 . 7–92% ) , but overall walked less than controls , and some walked very little ( 7 . 7–18 . 1% ) . Adult specific Toll-2 knock-down in neurons , with tubulinGAL80ts to switch on GAL4 and drive elav >Toll-2RNAi after adult fly eclosion , reproduced the behavioural phenotypes of the mutants ( Figure 3I ) . Importantly , this shows that Toll-2 is required in adult neurons . Both wild-type and mutant flies could walk between the black stripes , meaning that loss of Toll-2 function does not impair vision . Quantitative analysis of the flies’ walking behaviour did not reveal significant differences between the genotypes in their preference to walk between the dark stripes , or away from the centre of the arena . Hence , we cannot conclude that Toll-2 mutant flies have impaired visual processing . Interestingly , all wild-type flies walked more than Toll-2 mutants . In fact , loss of Toll-2 function significantly affected locomotion: all genotypes could walk at the same speed ( Figure 3J ) , but Toll-2 mutants spent less time walking than controls , thus overall travelled shorter distances ( Figure 3K ) , and although they had as many or more walking bouts , these were brief ( Figure 3L , M ) . Importantly , these phenotypes were consistent across different Toll-2 mutant alleles , and also when Toll-2 was conditionally knocked-out in adult post-mitotic neurons only ( Figure 3J–M ) . As Toll-2 mutants could achieve the same speeds as wild-type flies , but walked less , either motor circuit function and/or the motivation to walk were impaired . To conclude , Toll-2 loss of function resulted in neurodegeneration and impaired behaviour . Thus , Toll-2 is required for the formation and integrity of brain neural circuits . To ask whether Toll-2 might affect structural plasticity in the adult brain , we altered its function at the adult critical period ( i . e . day 0–5 post-eclosion ) , when the brain is most plastic ( Technau , 1984; Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . We used tubulinGAL80ts to silence GAL4 , then switched on GAL4 at post-eclosion adult day 0 and analyzed the brains two days later . Over-expression of Toll-2 with MyD88GAL4 increased the number of histone-YFP+ cells in the central brain ( Figure 4A ) , meaning that Toll-2 can regulate cell number at the adult critical period . Conditional Toll-2-RNAi-knock-down also caused a mild increase in cell number ( Figure 4A ) , which could be due to compensatory adjustments by other Tolls . In mushroom body KCs , neither Toll-2 knock-down nor over-expression with Toll-2ptvGAL4 restricted to the critical period had any effect on KC number ( Figure 4B ) . The optic lobes are particularly susceptible to structural plasticity ( Heisenberg et al . , 1995; Barth et al . , 1997 ) , thus we drove conditional over-expression with tubulinGAL80ts and Toll-2 >histone YFP , and automatically counted YFP+ cells with purposely adapted DeadEasy Optic Lobes software ( Figure 4C ) . Conditional Toll-2-RNAi knock-down had no effect , whereas over-expression of Toll-2 increased the number of YFP+ medulla neurons ( Figure 4C ) . Thus , Toll-2 can increase cell number in the adult optic lobes . The increase in cell number by Toll-2 gain of function in the central brain and optic lobes is consistent with a neuroprotective function , but could also involve cell proliferation . Either way , these data showed that Toll-2 is not pro-apoptotic in the adult brain , and instead can positively regulate cell number during the adult critical period . Experience increased the volume of multiple domains of the adult brain ( Technau , 1984; Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . To test whether stimulating neuronal activity during the critical period affects cell number in the adult brain , we activated neurons using the heat sensitive TrpA1 cation channel , whilst preventing leakage and excito-toxicity , and automatically counted His-YFP+ cells ( Figure 4D ) . Neuronal activation with TrpA1 increased Toll-2 >hisYFP medulla neuron number ( Figure 4D ) . Importantly , this increase could be rescued with Toll-2-RNAi knock-down ( Figure 4D ) . This meant that neuronal activity alters cell number in the optic lobes , via a Toll-2 dependent mechanism . To conclude , both Toll-2 and neuronal activity can increase cell number in the adult brain . KC number is robustly unaltered , but cell number in the central brain and optic lobes is plastic , and plasticity depends on Toll-2 . The alterations in cell number caused by loss and gain of Toll-2 function strongly implied that downstream nuclear signalling pathways were most likely involved . To test this , we visualised the distribution of the different adaptors and downstream targets of Toll signalling . Toll-receptor signalling positively regulates cell survival via Wek , MyD88 , NF-κB , and cell death via Wek , Sarm , JNK ( Kim et al . , 2007; McIlroy et al . , 2013; Mukherjee et al . , 2015; Foldi et al . , 2017 ) ; Toll-6 also functions upstream of pro-survival ERK , and FoxO , with nuclear FoxO promoting apoptosis whereas cytoplasmic export of FoxO by PI3Kinase signalling promotes cell survival ( Siegrist et al . , 2010; McLaughlin et al . , 2016; Foldi et al . , 2017 ) . Sarm inhibits MyD88 , and drives the pro-apoptotic functions of Tolls ( Carty et al . , 2006; Mukherjee et al . , 2015; Foldi et al . , 2017 ) . Thus , we used: 1 ) available GAL lines sarmNP7460 and sarmNP0257 , to drive expression of the FlyBow reporter; ( 2 ) the JNK signalling reporter TRE-Red; ( 3 ) anti-FoxO antibodies: ( 4 ) GFP-tagged forms of the pro-survival NF-κB transcription factors Dorsal ( Dl ) and Dif ( Anthoney et al . , 2018 ) , which function downstream of MyD88 . Both dorsal and dif produce cytoplasmic isoforms that lack the nuclear localization signal , and nuclear isoforms that have it ( Zhou et al . , 2015 ) . We used transgenic flies bearing Bacmids in which the nuclear isoforms were tagged with GFP , Dif-GFP-FPTB and Dl-GFP-FPTB . All of these signalling reporters were found both in pupal and adult brains ( Figure 4—figure supplement 1 ) . Thus , Toll signalling adaptors MyD88 and Sarm , and their downstream targets that normally regulate gene expression , cell survival and cell death , could function in pupal and adult brains . So far , data were consistent with Toll-2 maintaining neuronal survival via the canonical MyD88 pathway in the brain . However , multiple Tolls can regulate this pathway , and Tolls can also promote apoptosis via non-canonical signalling pathways ( Foldi et al . , 2017 ) , thus altering the levels of Toll-2 could cause compensation by other Tolls , compounding the phenotypes . Thus , we tested how signalling downstream of Tolls would affect cell number in the adult brain . To activate the pro-survival pathways downstream of Tolls in either MyD88+ central brain cells or Toll-2+ KCs , we knocked-down pro-apoptotic JNK signalling , activated pro-survival NF-κB signalling by knocking-down the inhibitor cactus , and induced the nuclear export of FoxO by over-expressing the activated form of PI3Kinase ( McLaughlin et al . , 2016; Foldi et al . , 2017 ) , and used tubulinGAL80ts to induce GAL4 expression conditionally , in adults only ( Figure 4E ) . Pro-survival signalling at the adult critical period increased cell number in the central brain ( Figure 4E ) , consistent with the neuroprotective function of Toll-2 . To drive Toll-dependent pro-apoptotic signalling at the adult critical period , we knocked-down MyD88 and over-expressed wek , which links Tolls to pro-apoptotic Sarm and JNK signalling ( Foldi et al . , 2017 ) . This genotype results in dramatic cell loss in pupae ( Foldi et al . , 2017 ) . Unexpectedly , over-expression of wek , or MyD88 RNAi knock-down , or both together , did not cause cell loss in adult day two central brains . Instead , they all increased cell number , with the combination of wek gain of function and MyD88 loss of function having the greatest effect ( Figure 4E ) . This surprising result indicated that: ( 1 ) MyD88 loss of function enabled an alternative pathway to increase cell number; ( 2 ) Wek did not simply induce apoptosis in the adult brain; ( 3 ) removal of MyD88 facilitated a hitherto unknown Wek function . Perhaps Wek-induced apoptosis provoked non-autonomous compensatory proliferation of some cells , or Wek itself induced cell proliferation , and this was antagonized by MyD88 . In KCs , pro-survival signalling had no effect on Toll-2+ cells ( Figure 4F ) , and over-expression of wek with MyD88 knock-down caused a rather mild , not significant increase in cell number ( Figure 4F ) . This mild increase indicated that different subsets of KCs could be regulated via different mechanisms and/or redundant functions between multiple Tolls might enable compensatory adjustments . The above surprising results raised an important question: did Toll-2 gain of function increase cell proliferation in the adult brain ? To test whether Toll-2 might induce proliferation in the adult brain , we asked whether over-expression of the cell cycle inhibitor , Retinoblastoma-protein factor ( Rbf280 ) , could influence Toll-2+ cell number . We over-expressed Rbf280 with Toll-2GAL4 and this caused lethality at day one pupa , precluding analysis of adult brains . In pupae , over-expression of Rbf280 decreased Toll-2+ KC number ( Toll-2 >hisYFP , Rbf280 , Figure 5A ) , showing that Toll-2 functions in mushroom body neuroblasts , that divide to produce KCs . To test whether Toll-2 might influence proliferation in other brain domains , we asked whether blocking proliferation with Rbf280 could rescue the increase in his-YFP+ cell number caused by Toll-2 gain of function . And it could , Rbf280 rescued the excess in cell number caused by Toll-2 over-expression , both in central brain ( with MyD88 >hisYFP ) and optic lobes ( with Toll-2 >hisYFP , Figure 5A ) . These data meant that Toll-2 can induce cell proliferation in pupal and adult brains . To further test whether Toll-2 could induce cell proliferation in the adult brain , we asked whether over-expression of Toll-2 restricted to the adult critical period could drive generation of mitotic recombination MARCM clones . Mitotic recombination was induced with heat-shock-Flipase at adult day one and neurons were visualised at adult day two with elavGAL4 >UASmCD8 GFP; all neurons also over-expressed Toll-2 , but control samples did not . We found no clones in control brains ( n = 17 ) , but amongst Toll-2 over-expressing brains , 4/17 had GFP+ clones ( Figure 5B , C ) . Importantly , the clones resulted in differentiated neurons that sent projections to different medulla layers , SOG , central brain and mushroom bodies ( Figure 5BC ) . These data show that gain of Toll-2 function induced cell proliferation in adult brains . To visualise proliferating cells , we used a common readout of cell division - the G1 to S transition - with the S-phase marker PCNA-GFP . At 33 hr after puparium formation , there were some PCNA-GFP+ cells in normal control brains ( Figure 5D ) , as previously reported ( Siegrist et al . , 2010 ) . Heat-shock induced conditional over-expression of Toll-2 in the pupa ( PCNAGFP , hsGAL4 >Toll-2 ) increased the number of PCNA-GFP+ cells , in the central brain and mushroom bodies ( Figure 5D–F , H ) . Thus , Toll-2 signalling can promote G1 to S transition in the pupal brain . To test whether Toll-2 could promote cell cycling also in the adult brain , we heat-shocked flies at the adult critical period . PCNA-GFP+ cells were seen in control adult brains , and over-expression of Toll-2 increased their incidence ( Figure 5G , H ) . However , increases were not statistically significant . Perhaps not all adult progenitors were in G1 , as with PCNA-GFP those in G2 would have been missed , and perhaps Toll-2 also provoked cycling from S-phase to G2 or to G2 to mitosis ( M ) . To test whether there might be progenitors in G2 in the adult brain , we visualised Cdc25/String ( Stg ) , which is expressed in G2 and triggers G2/M transition and entry into mitosis ( Edgar and O'Farrell , 1989 ) . Using flies bearing a Stg-GFP fusion protein ( Buszczak et al . , 2007 ) , we found Stg-GFP+ cells throughout the adult brain , both in controls and in brains over-expressing Toll-2 ( Figure 5I ) . These data showed that there are cycling cells in G2 or G2/M in the adult brain . To test whether Toll-2 could induce cell cycling from progenitors arrested in either G1 or G2 , we used Fly-FUCCI , at the adult critical period . FUCCI reveals cycling cells in G1 , G1/S , G2 and G2/M phases , but not non-cycling cells in G0 ( Zielke et al . , 2014 ) . It drives expression of degron fusion proteins to cell cycle factors E2F and cyclin-B , which get destroyed as cells enter S-phase or G1 , respectively . Cells that are only E2F-GFP+ are in G1 , cells that are only CycB-RFP+ are in S phase , and cells that are both E2F-GFP+ and CycB-RFP+ are in G2 or M . Over-expressing FUCCI at the adult critical period only , we found cells that were E2F-GFP+ ( G1 ) , some that were CycB-RFP+ ( S ) and some that were both E2F-GFP+ and CycB-RFP+ ( G2/M ) ( Figure 6A–C ) . This meant that the normal adult brain bears cycling cells , presumably progenitor cells resting in G1 or G2 . Over-expression of Toll-2 together with FUCCI at the adult critical period , increased the number of cells in G1 , S and G2/M phases of the cell cycle compared to controls ( Figure 6A–D ) . This means that there are progenitor cells in the adult brain , and Toll-2 can induce their cycling . To conclude , the above data showed that Toll-2 can induce cell cycling and proliferation in the developing pupal brain and in the adult brain . Furthermore , data indicated that this involved Toll-2 repressing the MyD88 pathway and activating a Wek pathway downstream . The finding that over-expression of Toll-2 or wek increased cell number , even further if MyD88 was also knocked-down , suggested that Wek and MyD88 antagonise each other to regulate cell cycling downstream of Toll-2 in the adult brain . Progenitor cells had been previously reported in the Drosophila adult brain ( Kato et al . , 2009; Fernández-Hernández et al . , 2013; Foo et al . , 2017 ) . Thus , we revisited the identity of Toll-2+ and MyD88+ cells , and asked whether they included progenitor cells . In the adult brain , most Toll-2 >hisYFP+ were Elav+ positive neurons , and none were Repo+ glia , but there were some Toll-2 >hisYFP+ Elav-negative and Repo-negative cells , in optic lobes and central brain ( Figure 6E , F ) . Amongst them , were large Toll-2+ and MyD88+ cells ( Figure 6G , H ) . The large MyD88 >HisYFP+ cells were never Elav+ , and some but not all were Repo+ ( Figure 6G ) . Using neuroblast markers anti-Miranda ( Mira ) and anti-Dpn , we found labelled cells in the normal adult brain , as previously reported ( Fernández-Hernández et al . , 2013; Foo et al . , 2017 ) . Many MyD88 >hisYFP+ cells in the central brain and optic lobes were also Dpn+ ( Figure 6I ) , and some MyD88 >hisYFP+ cells also had Mira ( Figure 6J ) . Most prominently , many of the large MyD88 >hisYFP+ cells were Dpn+ ( Figure 6I ) . Altogether , these data showed that Toll-2 and MyD88 are expressed in progenitor cells in the adult brain at the critical period . To test whether these MyD88+ progenitor cells coincided with those revealed by PCNA-GFP , Stg-GFP and FUCCI , and identify their cell cycle state , we stained MyD88 >FUCCI cells with anti-Dpn . Upon Toll-2 over-expression in the adult , some Dpn+ cells were GFP—RFP+ cells ( i . e . in S-phase ) , some GFP+ RFP+ ( i . e . in G2/M ) , and many were GFP+ RFP— ( i . e . in G1 ) ( Figure 6K ) . Thus , as MyD88 knock-down increased cell number , this meant that MyD88 is expressed in adult progenitor cells , where it normally prevents cell cycling . Toll-2 can overcome this repression . Then , how does Toll-2 regulate cell cycling in the adult brain ? Yorkie ( Yki ) is a positive regulator of cell proliferation , and it functions downstream of Toll-1 both in immunity and cell competition ( Koontz et al . , 2013; Liu et al . , 2016a; Katsukawa et al . , 2018 ) . Yki can regulate the G1-S cyclin cycE , but is best known for regulating G2-M stg and entry into mitosis , for which Yki shuttles highly dynamically in and out of the nucleus ( Huang et al . , 2005; Manning et al . , 2018 ) . To test whether cell proliferation at the adult critical period might involve Yki , and whether yki could be a target of Toll-2 signalling , we first visualised Yki in the brain . Using a Yki-GFP protein fusion ( Fletcher et al . , 2018 ) , we found many Yki-GFP+ cell nuclei in the normal , adult brain at the critical period ( Figure 7A–C ) . Conditional over-expression of Toll-2 at the critical period increased Yki-GFP+ nuclei throughout the brain ( Figure 7C , D ) . Yki-GFP+ nuclei were numerous , but because of cytoplasmic signal in other cells , automatic cell counting with DeadEasy was not possible . Thus , to reliably count Yki-GFP+ nuclei manually , we focused on three regions of interest ( ROI ) : the optic lobe medulla; the sub-esophageal ganglion ( SOG ) and a top left anterior corner in the central brain ( CB ) . There were Yki-GFP+ nuclei in control brains , that resembled the MyD88+ cells , most noticeably in the optic lobes ( Figure 7A , B ) . Over-expression of Toll-2 altered the distribution of Yki-GFP+ nuclei: in the medulla , most brains had fewer Yki-GFP+ nuclei than controls , but 20% brains had more; in the SOG and CB , over-expression of Toll-2 resulted in more Yki-GFP+ nuclei ( Figure 7C , D ) . Both the bimodal distribution and increase in YFP+ nuclei , mean that over-expression of Toll-2 induced cell cycling , which provoked shuttling of Yki both in and out of the nucleus . Thus , these data showed that nuclear Yki is present in the adult brain at the critical period , and its shuttling is regulated by Toll-2 . Yki shuttles into the nucleus to promote entry into mitosis , and stg is one of its targets ( Huang et al . , 2005; Manning et al . , 2018 ) . We showed that Toll-2 overexpression increased cell number , Stg and the incidence of nuclear Yki-GFP in the adult brain . Thus , to further test whether Toll-2 might induce proliferation via Yki in the brain , we used genetic epistasis . We asked whether the increase in MyD88+ cell number caused by over-expression of Toll-2 in the central brain ( tubGAL80ts , MyD88 >Toll-2 ) could be rescued by knocking-down yki . And it did: conditional yki-RNAi knock-down rescued the excess of cell number caused by Toll-2 gain of function , restoring cell number to control levels in the central brain ( Figure 7E , E’ ) . Thus , Yki functions downstream of Toll-2 to increase cell number in the adult brain . Since both over-expression of wek and knock-down of MyD88 increased cell number , we asked whether Wek might regulate cell number downstream of Toll-2 , to antagonise MyD88 . We tested this using genetic epistasis . Indeed , wek RNAi knock-down in MyD88+ cells rescued the increase in cell number caused by Toll-2 over-expression ( Figure 7F ) , demonstrating that Toll-2 increases cell number via Wek . To conclude , there are quiescent MyD88+ progenitors in the adult brain . MyD88 prevents their proliferation and promotes quiescence , whereas Toll-2 signaling via Wek overcomes their repression to induce proliferation , which requires Yki ( Figure 7G ) . The above data showed that Toll-2 can regulate cell survival and cell proliferation in the developing and adult brain . There are nine Tolls , and most of them are expressed in the brain . Thus , redundancy between the Tolls might obscure the effects of altering the levels of only one . Different Tolls could also bring about different cellular outcomes - for instance , Toll-1 is more pro-apoptotic than Toll-6 and −7 in pupa ( Foldi et al . , 2017 ) – compounding the phenotypes . Importantly , the spatial , distinct expression patterns of Tolls could fine-tune the size of distinct anatomical domains . Thus , we asked what effect might down-regulating multiple Tolls at once have in the brain . We tested two different RNAi lines for each Toll . Upon sustained knock-down using Toll-2pTVGAL4 , all combinations except one resulted in pupal lethality , precluding analysis in the adult brain . Thus , we analysed pupal brains . Simultaneous knock-down of Toll-2 , –7 , −1 or Toll-2 , –7 , −8 , or Toll-2 , –7 , −6 with Toll-2pTVGAL4 resulted in smaller brains than controls ( Figure 8A ) . There was noticeable loss of cells in the optic lobes of Toll-2pTVGAL4 > Toll-2 , –7 , −8-RNAi , and Toll-2 , –7 , −6-RNAi ( Figure 8A ) . Cell density varied across modules , rendering automatic cell counting with DeadEasy inaccurate , thus we measured brain size instead . Overall brains were smaller in all these genotypes compared to controls , with Toll-2pTVGAL4 > Toll-2 , –7 , −6-RNAi having the smallest brains ( Figure 8A , B ) . Optic lobe , central brain and KC cluster area were all reduced compared to controls ( Figure 8A , B and Figure 8—figure supplement 1A , B ) . Knock-down of multiple Tolls resulted in deeper KC clusters along the A/P axis ( Figure 8D ) , revealing that KC clusters were disorganized . In Toll-2pTVGAL4 > Toll-2 , –7 −6-RNAi brains , the reduction in KC cluster area also correlated with a decrease in KC number ( Figure 8C , E and Figure 8—figure supplement 1B ) . Altogether , these data showed that during development: ( 1 ) down-regulation of multiple Tolls disrupts brain structure and reduces brain size . ( 2 ) Multiple Tolls regulate cell number in the optic lobes and central brain . ( 3 ) Tolls function in concert to regulate KC cluster size , organisation and KC number . Next , we asked whether conditional knock-down of multiple Tolls might affect the adult brain . For this , due to genetics limitations , we could only test two Tolls at a time . Conditional knock-down of both Toll-2 and −6 with MyD88GAL4 , restricted to the adult critical period , increased cell number in the central brain ( Figure 8—figure supplement 1C , D ) . This was reminiscent of the increase in cell number caused by MyD88 knockdown ( Figure 4E ) , and could also involve the pro-apoptotic function of Toll-6 ( Foldi et al . , 2017 ) . Toll-2pTVGAL4 flies are heterozygous mutant for Toll-2 , and together with RNAi knock-down of Toll-6 , –7 with Toll-2GAL4 did not affect brain size , but with Toll-2 , –6 knock-down most brains were smaller than controls , and around 25% were bigger ( Figure 8—figure supplement 1E , F ) . Most remarkably , whereas conditional manipulation of Toll-2 alone did not affect KCs ( Figure 4B , F ) , conditional knock-down of both Toll-2 , –6 or Toll-6 , –7 in the adult altered KCs ( Figure 8F–H ) . KC number and cluster depth increased with Toll-6 , –7-RNAi , and decreased slightly ( albeit not significantly ) with Toll-2 , –6-RNAi knock-down ( Figure 8F–H ) . These data meant that: ( 1 ) different Tolls have distinct functions at the adult critical period; ( 2 ) Tolls have distinct as well as redundant or overlapping functions in KCs; and ( 3 ) KC organization and number can be altered in the adult by Tolls . Experience increases brain size by around 5% compared to un-stimulated controls ( Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . Thus , in view of the above results , we asked whether conditional manipulation of Toll-2 or its signalling pathways could influence brain size at the adult critical period . Most remarkably , over-expression of Toll-2 in MyD88+ cells at the adult critical period ( tubGAL80ts , MyD88 >Toll-2 ) increased brain size compared to controls ( Figure 8I , J ) . Even if only female brains were analysed , there was considerable variability in brain size in controls , precluding statistical significance ( Figure 8J , grey box-plots ) . Still , both the median ( box-plots in Figure 8J ) and mean differed: Toll-2 over-expressing brains were on average 5 . 4% ( using UAS-Toll-2EP709 ) and 7 . 4% ( using UAS-Toll-2attP2 ) larger than control brains , respectively ( i . e . mean control brain size was 138 , 585μ2 vs . 146 , 177μ2 of tubGAL80ts , MyD88 >Toll-2EP709 and 148 , 861μ2 of tubGAL80ts , MyD88 >Toll-2attP2 brains ) . Furthermore , whereas 40% of control brains were larger than the mean , around 80% of Toll-2 over-expressing brains were larger than the control mean ( >140 , 000 ) ( Figure 8K , left ) . Conditional MyD88-RNAi ( e . g . tubGAL80ts , MyD88 >MyD88 RNAi ) knock-down or over-expression of wek resulted in even larger brains , with virtually all brains being larger than both the control median and mean ( 14 . 7% larger for wek and 20% for MyD88-RNAi , Figure 8I , J , K ) . These results were consistent with both MyD88 promoting quiescence and Wek promoting cell proliferation . Surprisingly , however , although the concerted over-expression of wek and MyD88-RNAi increased cell number ( Figure 4E ) , it did not increase brain size ( Figure 8J , K ) , perhaps because together they can also induce apoptosis ( Foldi et al . , 2017 ) . Whereas conditional yki-RNAi knock-down in MyD88+ cells ( tubGAL80ts , MyD88 >yki RNAi ) resulted in highly variable brain size , yki-RNAi could rescue the increase in brain size caused by Toll-2 gain of function ( Figure 8J , K ) . This was consistent with Yki functioning downstream of Toll-2 to induce cell proliferation in the adult brain . To conclude , these data showed that , like experience , Toll-2 can positively regulate brain size at the adult critical period . This involved counteracting MyD88 and activating Wek and Yki signalling downstream .
At least seven of the nine Toll-receptors are expressed topographically , mapping the distinct modules that form the brain . Toll receptors regulate cell number and brain size in development and structural brain plasticity in the adult , through their ability to promote either cell survival or death , progenitor cell quiescence or proliferation . Evidence indicates that Tolls can underlie the changes that experience brings about in the adult brain ( Technau , 1984; Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) , and that structural plasticity and neurodegeneration are two faces of Toll-driven cellular responses in the brain ( Figure 9 ) . Toll-2 promotes neuronal survival and proliferation , both in development and in the adult brain . Toll-2 is neuroprotective as loss of function caused neurodegeneration: it increased apoptosis and caused neuronal loss , and Toll-2 mutant neurons that survived had dendrite loss , axon atrophy and misrouting . Toll-2 loss of function also impaired climbing and walking , and decreased lifespan , phenotypes characteristic of neurodegeneration . Toll-2 promotes cell survival through the canonical MyD88-NFκB pathway , as previously found for the pro-survival functions of Toll-6 and −7 in development ( McIlroy et al . , 2013; Foldi et al . , 2017 ) . Both in flies and mammals , Tolls and TLRs promote cell survival via MyD88 and cell death via Sarm , which activates the pro-apoptotic function of JNK and inhibits MyD88 ( Kim et al . , 2007; McIlroy et al . , 2013; Mukherjee et al . , 2015; Foldi et al . , 2017 ) . Distinct Tolls and TLRs can preferentially promote cell survival or cell death , as for instance , Toll-1 is more pro-apoptotic than Toll-6 ( Foldi et al . , 2017 ) , and in mammals TLR4 promotes neuronal survival and TLR8 neuronal death ( Ma et al . , 2006; Zhu et al . , 2016 ) . We showed that Toll-2 over-expression did not cause cell loss , meaning that Toll-2 is not pro-apoptotic in the brain . Altogether , data showed that Toll-2 is neuroprotective in the brain . A novel molecular pathway underlies the ability of Toll-2 to regulate cell proliferation in development and structural brain plasticity in the adult . A remarkable finding was that Toll-2 gain of function in the pupal or adult brain not only maintained neurons alive , but also induced cell cycling - which was visualized with standard cell proliferation markers PCNA-GFP for S-phase , FUCCI for G1 , G2 , G2/M , and Stg and nuclear Yki for G2/M and M phases - , it increased cell number and brain size . We showed that there are progenitor cells in the adult brain that are kept quiescent by MyD88 , and loss of MyD88 at the adult critical period increased neuronal number and brain size . As MyD88 is the general adaptor for canonical Toll signalling , this implies that Toll signalling maintains progenitor cells quiescent . Cell proliferation was induced when the repression by MyD88 was overcome by Toll-2 over-expression and signalling via the alternative adaptor , Wek . Conditional over-expression of Toll-2 or wek , or knock-down of MyD88 , at the adult critical period increased neuronal number and brain size . Furthermore , the effect of Toll-2 was dependent on both Wek and Yki , a well-known inducer of cell proliferation ( Koontz et al . , 2013 ) . Over-expression of Toll-2 induced nuclear translocation and shuttling of Yki , correlating with nuclear localization of Stg , target of Yki , in the brain . Furthermore , genetic epistasis analysis showed that the increase in cell number caused by Toll-2 over-expression could be rescued with either yki-RNAi or wek-RNAi knockdown . Thus , Toll-receptor signalling can switch between promoting quiescence via MyD88 to promoting cell proliferation via Wek . Whether Wek might activate cell proliferation directly , or activate Yki or JNK first , was not solved . Wek also induces cell death , by linking Tolls to Sarm , which activates pro-apoptotic JNK signalling ( Foldi et al . , 2017 ) . JNK can also induce cell proliferation , and Yki is activated downstream of JNK and Toll in other contexts ( Enomoto et al . , 2015; Gerlach et al . , 2018; Katsukawa et al . , 2018 ) . Thus , Wek could activate cell proliferation directly and independently of Yki , or it could activate Yki either directly , or via Sarm and/or JNK . Any of these options is possible , as in immunity and in cell competition ( which also involves regulation of cell number ) , Toll-1 can regulate Yki downstream via JNK-dependent and independent pathways ( Liu et al . , 2016a; Katsukawa et al . , 2018 ) . Either way , our data showed that Toll-2 can prevent or induce progenitor cell proliferation , through alternative MyD88 or Wek signalling pathways downstream ( Figure 9C ) . Knock-down of multiple Tolls through development severely altered brain structure and reduced brain size . Most likely , Tolls promote either cell survival or cell proliferation or both during brain development , and together they modulate brain formation . How each of them may influence the adult brain , is more difficult to dissect . Through their ability to elicit multiple cellular outcomes , Tolls can have distinct , redundant , synergistic , antagonistic or compensatory functions . For instance , whereas Toll-1 and −6 can have pro-apoptotic functions in the pupal VNC ( Foldi et al . , 2017 ) , Toll-2 was not pro-apoptotic in the brain . Altering Toll-2 function alone did not affect Kenyon Cells , but simultaneous persistent knock-down of three Tolls reduced KC number and disorganized KC clusters; and whereas conditional knock-down of Toll-2 , –6 at the adult critical period reduced KC number , conditional knock-down of Toll-6 , –7 increased KC number . Thus , although all Tolls could access the same downstream signalling pathways , each Toll modulates these pathways in their own way . As a consequence , knock-down of one or more Tolls most likely induced complex responses by other Tolls in the same or neighbouring cells , compounding the phenotypes . What enables Tolls to elicit different cellular outcomes is an intriguing question . Adult neurogenesis is much debated . Neurogenesis occurs in the adult brain of humans and other animals ( Cayre et al . , 1996; Deng et al . , 2010 ) , but the extent of it is unknown , and solving this is important to understand how the brain works and brain disease . In Drosophila , developmental neural stem cells are eliminated by the end of pupal life ( Ito and Hotta , 1992; Maurange et al . , 2008; Siegrist et al . , 2010; Hara et al . , 2018 ) . However , neurogenesis has been reported in the adult brain: naturally occurring cell death induces cell proliferation in the adult brain ( Kato et al . , 2009 ) ; cell proliferation was reported with mitotic recombination clones ( Fernández-Hernández et al . , 2013 ) ; mir-31 mutations induce glial and neuroblast proliferation in the adult brain ( Foo et al . , 2017 ) ; injury induces gliogenesis and neurogenesis ( Kato et al . , 2009; Fernández-Hernández et al . , 2013; Moreno et al . , 2015 ) ; the partner of Yki , scalloped , is expressed in the adult brain ( Rohith and Shyamala , 2017 ) ; immuno-histochemistry and single cell transcriptomics revealed that neuroblast or intermediate neural progenitor ( INP ) markers eyeless , dichate , grainy-head , dpn , miranda and inscutable are expressed in the adult brain ( Callaerts et al . , 2001; Fernández-Hernández et al . , 2013; Foo et al . , 2017; Zhu et al . , 2017; Croset et al . , 2018; Davie et al . , 2018; Konstantinides et al . , 2018 ) ; and interference with the normal regulation of cell survival and cell death – processes that Tolls can influence - results in ectopic and/or persistent neuroblasts in the adult brain ( Siegrist et al . , 2010; Hara et al . , 2018 ) . Consistently with these findings , in the adult brain there are MyD88+ Dpn+ Stg-GFP+ Yki-GFP+ FUCCI+ cells in S-phase or G2/M in the optic lobes , and in S , G1 and G2/M in the central brain . MyD88+ progenitor cells are normally quiescent and Toll-2 gain of function can induce their cell cycling and proliferation at the adult critical period . Adult progenitor cells may be distinct from developmental neuroblasts . In fact , the fate of quiescent INPs has not been determined , suggesting some could also exist in the adult brain ( Walsh and Doe , 2017 ) . In other insects , adult progenitor cells differ from developmental neuroblasts , and instead originate from hemocytes ( Simões and Rhiner , 2017 ) . In the mammalian brain , adult progenitors originate from glia ( Falk and Götz , 2017 ) . Some of the large Toll-2+ and MyD88+ Dpn+ cells also had the glial marker Repo . Thus , adult progenitor cells may not originate from canonical developmental larval neuroblasts . Experience alters brain structure topographically , altering the regions involved in experience-dependent processing ( Figure 9A ) . For instance , rearing flies in constant light or constant darkness alters the size of brain modules involved in vision ( e . g . optic lobes ) ( Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . Neuronal activity increased cell number in the optic lobes in a Toll-2 dependent manner , and conditional over-expression of Toll-2 or wek in the adult brain , at the critical period , increased both cell number and brain size . The anatomical segregation of the seven Toll receptors enables them to modulate cell number within distinct brain modules . This implies that: ( 1 ) in development , Tolls could adjust brain neuronal populations topographically to the motor and sensory circuits , enabling appropriate behaviour . ( 2 ) In evolution , Tolls could have driven changes in brain shape , enabling adaptation to distinct environments and behavioural diversification . ( 3 ) In adults , Tolls can enable structural brain plasticity , by adjusting brain neuronal populations topographically to experience-dependent inputs , and drive behavioural adaptation throughout the life-course . TLRs could operate in analogous ways in the human brain ( Okun et al . , 2009; Okun et al . , 2011 ) . TLRs are expressed throughout the mammalian brain ( Ma et al . , 2006; Cameron et al . , 2007; Rolls et al . , 2007; Lathia et al . , 2008; Okun et al . , 2009; Okun et al . , 2010a; Okun et al . , 2011; Okun et al . , 2012 ) . Some TLRs have neuro-protective and other TLRs pro-apoptotic functions , for instance TLR-2 and −4 promote cell survival , and TLR3 and −8 apoptosis , neurite retraction and neurodegeneration ( Ma et al . , 2006; Hung et al . , 2018 ) . TLRs can regulate neural stem cell proliferation , formation or elimination of neurites and neurons , including during learning and long-term memory ( Ma et al . , 2006; Rolls et al . , 2007; Okun et al . , 2010a; Okun et al . , 2012; Madar et al . , 2015; Hung et al . , 2018 ) . To conclude , through their topographic distribution , Tolls modulate cell number and brain size , in development and in structural plasticity in the adult ( Figure 9 ) . They do so by engaging different molecular pathways that regulate neuronal survival or death , and progenitor quiescence via MyD88 or proliferation via Wek . It will be compelling to test if the combination of TLR topography and diversity of signalling options downstream , also underlies neurodegeneration and structural plasticity in the human brain .
Toll-2pTV flies were generated by CRISPR/Cas9 mediated homologous recombination replacement of the coding region of Toll-2 for the pTV cassette ( Baena-Lopez et al . , 2013 ) . For PCR amplification of genomic regions , primers with ideally no-off targets were designed using the tools in http://www . ncbi . nlm . nih . gov/tools/primer-blast/ . Genomic DNA was extracted from Oregon R wild-type flies as Polymerase Chain Reaction ( PCR ) template . For the 5’ homology arm ( HA ) , 4 . 993 kb upstream of the Toll-2 start codon were PCR-amplified using primers 3 and 4 ( Supplementary file 3 ) , and for the 3’homology arm ( HA ) , 3 . 419 kb downstream if theToll-2 CDS starting from the stop codon were amplified by PCR using primers 1 and 2 ( Supplementary file 3 ) . Conventional ligation cloning was used to insert the 5’HA into NotI and KpnI enzyme restriction sites in the pTV vector ( Baena-Lopez et al . , 2013 ) , and the 3’HA into SpeI and AscI sites . Insertion of both 5’HA and 3’HA into the correct locus , as well as the integrity of the pTV cassette were verified by PCR diagnostics . The gRNA targeting Toll-2 was created by annealing primers 5 and 6 in Supplementary file 3 followed by conventional ligation cloning into the BbsI site of vector pU6 . 3 ( Port et al . , 2014 ) . Cloning was verified by sequencing . The resulting Toll-2-gRNA construct was inserted into the attP2 landing site by FC31 transgenesis . Transgenic flies were crossed to nos-Cas9 transgenic flies , the Toll-2pTV construct was injected into progeny flies ( BestGene ) , and transformants were identified using w+ as a genetic selection marker . PCR analysis indicates that the Toll-2pTV insertion deleted half of the coding region of Toll-2 , including the start codon , generating a mutant allele . Toll-2pTV carries mCherry expressed under the control of the endogenous Toll-2 promoter . The GAL4 sequence in the original pTV cassette was damaged ( Baena-Lopez et al . , 2013 ) , so GAL4-attB ( AddGene ) was integrated into the attP-landing site within the pTV cassette , using FC31 transgenesis ( injections by BestGene Inc ) . This placed GAL4 coding sequences just before the Toll-2 start codon . UAS-FlyBow was used for genetic selection . Toll-4GAL4 and Toll-5GAL4 were generated by CRISPR/Cas9 facilitated homologous recombination , inserting T2AGAL4 ( AddGene ) just upstream of their start codon , whilst retaining intact coding regions , and following the protocol described in Gratz et al . ( 2015 ) . For the 5’ homology arms ( HA ) and 3’HA , 1 kb fragments were amplified from template genomic DNA from wild-type Oregon R flies using primers 9–12 ( Supplementary file 3 ) for Toll-4 and primers 15–18 ( Supplementary file 3 ) for Toll-5 , and cloned using conventional ligation cloning into AgeI and NotI restriction sites for the 5’HAs and into AscI and SpeI sites for the 3’HAs . Toll-4 and Toll-5 gRNAs were generated as above , using primers 13 , 14 and 19 , 20 ( Supplementary file 3 ) , respectively , and cloned into the BbsI site of the pU6 . 3 vector . Toll-4-T2AGAL4 and Toll-5-T2AGAL4 and gRNA-pU63 constructs were injected into nos-Cas9 flies ( BestGene Inc ) and transformants were identified with 3xP3-DsRed genetic selection marker , which was subsequently removed by CRE-Lox mediated recombination from flanking LoxP sites in the T2A-GAL4 vector . Toll-3GAL4MI02994 , Toll-6GAL4MIO2127 and Toll-7GAL4MI13963 were generated by Recombination Mediated Cassette Exchange ( RMCE ) ( Venken et al . , 2011 ) , by injecting the GAL4-attB plasmid into Toll-3MI02994 , Toll-6MIO2127 and Toll-7MI13963 MIMIC lines and using FC31 transgenesis ( injections by BestGene Inc ) . Toll-3 , –6 and −7 are all intron-less genes , thus GAL4 is expressed under the control of their endogenous promoters . UASToll-2 transgenic flies were generated by PCR-amplifying the coding region of the intron-less Toll-2 gene using primers 7 , 8 ( Supplementary file 3 ) , followed by insertion into the pUAS-gw-attB vector using Gateway cloning and FC31 transgenesis into the attP2 landing site ( BestGene Inc ) on the third chromosome . The genetic selection marker w+ was used to identify transformant flies . Non-quantitative Reverse Transcription PCR ( RT-PCR ) was carried out on wild-type Oregon-R embryos , dissected central nervous systems ( CNSs ) from second and third instar-wandering larvae and pupae , and whole heads . Total RNA was extracted with Trizol ( Ambion ) and RNA integrity and concentration was confirmed using a Nano-drop . RNA samples were DNase treated to remove residual genomic DNA contamination . 300 ng of RNA was used for cDNA synthesis following GoScript Reverse Transcriptase treatment . Samples were diluted 1:3 with Nuclease free H20 , and a no RT sample of 300 ng of RNA made up to 60 ul with Nuclease free H20 . Standard PCR reaction was performed to amplify each of the Toll receptor cDNA using GoTaq PCR protocol . PCR primers were designed using primer-BLAST ( http://www . ncbi . nlm . nih . gov/tools/primer-blast/ ) and 2 uM of forward and reverse primers specific to each sample were used ( primers 21–38 in Supplementary file 3 ) . GAPDH is a general housekeeping gene that was used as a positive control during every round of PCR . To over-express or knock-down Toll-2 in the pupa or adult brain only , we used the temperature sensitive form of the GAL4 repressor , GAL80ts , ubiquitously expressed under the control of tubulin promoter . Flies bearing Toll-2pTVGAL4 UAS-histone-YFP/+; tubGAL80ts were used to count cell number in Kenyon cells and optic lobes , and flies bearing MyD88GAL4 UAShistoneYFP/+; tubGAL80ts were used to count cells in the central brain . At 18°C GAL80ts represses GAL4 expression , whereas at 30°C , GAL80ts is inactivated enabling GAL4 expression . To conditionally manipulate Toll-2 expression in pupa , embryo collections were left to develop at 18 °C , pupae were harvested 0–3 hr after puparium formation ( APF ) and moved to a 30 °C incubator , where they were kept for 72 hr . Fluorescent pupae that expressed UAS-histone-YFP under the control of MyD88GAL4 or Toll-2GAL4 , were selected to be dissected , fixed and scanned . The adult critical period for structural plasticity spans from adult day 1 to 5 ( Technau , 1984; Heisenberg et al . , 1995; Barth and Heisenberg , 1997; Barth et al . , 1997 ) . To conditionally manipulate Toll-2 expression at the adult critical period and analyse the consequences in the central brain and kenyon cells , embryo collections were left to develop at 18 °C , adult progeny flies were selected and harvested 0–3 hr after eclosion , moved to a 30 °C incubator where they were kept for 48 hr , and then brains were dissected , fixed and scanned . To conditionally manipulate Toll-2 expression at the adult critical period and analyse the consequences in the optic lobes , embryo collections were left to develop at 18 °C , adult progeny flies were selected and harvested 0–3 hr after eclosion , and kept at 18°C for another 96 hr . Fluorescent pupae were selected at 96 hr APF , moved to 30 °C for further 6 hr , followed by 24 hr at 25 °C , and then brains were dissected , fixed and scanned . For all cell counting experiments in the adult brain , only female flies were used . To test whether Toll-2 can induce proliferation , we used the S-phase marker PCNA-GFP , whereby GFP is expressed under the control of the PCNA promoter containing E2F binding sites . Control ( PCNA-GFP/+;+/+;hs-Gal4/+ ) or test ( PCNA-GFP/+; UAS Toll-2EP709/+; hs-Gal4/+ ) embryos were allowed to develop at 25°C , pupae were harvested at 0–1 hr APF and after further 24 hr , they were heat shocked at 37 °C for 30 min in a water bath to induce Toll-2 expression . To allow for cell cycle onset , pupae were transferred again to 25°C for a further 9 hr , after which brains were selected and dissected , fixed and stained with anti-GFP antibodies . To test induction of proliferation in the adult brain , female flies of the same genotypes as above were selected with genetic markers , harvested 0–1 hr after eclosion , heat shocked at 37 °C for 30 min in a water bath , returned to 25 °C for another 9 hr , and then their brains were dissected , fixed and stained with anti-GFP antibodies . To activate neurons with temperature , we used the heat-activated cation channel TrpA1 expressed under GAL4 control . Toll-2pTVGAL4 , UAS-histone-YFP/+; tubGAL80ts/UAS-TrpA1 and Toll-2pTVGAL4 , UAS-histone-YFP/UAS-Toll-2RNAi; tubGAL80ts/UAS-TrpA1 Flies were kept at 18°C from embryo until 4 days after eclosion ( 96 hr AE ) . Flies were then transferred to 30°C incubator for 5 hr to switch off Tubulin Gal80ts and switch on Gal4 , allowing for translation . To prevent phototoxicity , flies were then shifted between a 30°C water bath and an 18°C incubator every 10 min for an hour . Finally , flies were moved to a 23°C incubator for another 24 hr and then dissected , fixed and scanned . Immunostainings in the pupal and adult brain were carried out following standard procedures . Primary antibodies used were: rabbit-anti-GFP at 1:250 ( Molecular Probes ) ; rabbit anti-DsRed at 1:100 ( Clontek ) ; goat anti-Toll-1 at 1:10 ( Santa Cruz ) ; anti-FoxO at 1:500 ( gift of P . Leopold ) ; Guinea pig -anti-Dpn at 1:1000 ( gift of Y . Jan ) ; rat anti-Mira at 1:10 ( abcam ) ; mouse-anti-Repo 1:10 ( Developmental Studies Hybridoma Bank , DSHB ) ; rat-anti-Elav at 1:250 ( DSHB ) . Secondary antibodies used: Alexa Donkey-anti-Rabbit 488 at 1:250; Alexa Goat-anti-Rabbit 546 at 1:250; Alexa Goat-anti Guinea-pig 633 at 1:250; Alexa Goat-anti-Rat 647 1:250; Alexa Goat-anti-mouse 647 at 1:250 . All microscopy images were captured using laser scanning confocal microscopy with either a Zeiss LSM710 or a Leica SP8 laser scanning confocal microscopes . For image data , brains were scanned with at a resolution of either 512 × 512 or 1024 × 1024 . With the Leica SP8 , we used the 20x oil objective for whole adult brains and pupal brains , and zoom 0 . 9 for pupal brain only; 40x oil objective for optic lobes; and 63x oil objective for Kenyon cells , with step size: pupa: 0 . 25 µm , adult: 0 . 2 µm; acquisition speed was 400 Hz , airy 1 , and no line averaging was applied; sections were 0 . 96 µm or 1 µm apart , unless otherwise indicated . With the Zeiss LSM710 , for whole adult brains , we used a 25x objective , zoom 0 . 6 , 1 or 2 rounds of averaging , speed 7–9 , and step size 1 µm . To automatically count cell number in the adult brain , we purposely developed software plug-ins to be run with ImageJ , which we called DeadEasy Central Brain , DeadEasy Optic Lobes and DeadEasy Kenyon Cells . These are image processing programs that identify and count cells labeled with the nuclear marker Histone-YFP in a 3D stacks of confocal images throughout the brain . Brains expressing UAS-Histone-YFP under the control of GAL4 , were dissected and fixed , and without staining , scanned throughout at the confocal microscope , using settings described above . The plug-ins identify cells in the following way: ( 1 ) DeadEasy Optic Lobes: First , noise is reduced using a median filter . Then , particles whose diameter is greater than the minimum cell radius and whose brightness is above a given threshold found empirically , are retained as cells . A black majority filter is then applied to better separate cells and eliminate noise further . Finally , the very small and very large particles that are not considered cells are eliminated and the identified cells are labeled and counted . ( 2 ) DeadEasy Central Brain: In this algorithm , unlike the previous one , instead of the medium filter , the small particles in the original image are eliminated by using a minimum filter followed by a maximum filter , then the same steps as for the previous algorithm are followed . ( 3 ) DeadEasy_Mushroom_body: In this algorithm a medium 3D filter is first applied in order to eliminate noise . Then the separation between cells is accentuated , by using a minimum 3D filter followed by a maximum 3D one . Then , the regions whose radius is smaller than the radius of a cell are eliminated . To identify the cells in each region that appear too crowded , we proceed to identify recursively the spheres of variable radius that can be located in each region . It begins with spheres of greater radius , then decrease in size . Once the cells are identified they are labeled and counted . Statistical analysis was then carried out as described below . Brain size was measured using ImageJ . Statistical analyses were carried out using Graph-Pad Prism . Confidence interval was 95% , setting significance at p<0 . 05 . Categorical data were analysed with Chi-Square tests , followed by Bonferroni multiple comparisons corrections . Numerical data were tested first for their type of distributions . If data were distributed normally , Student t-tests were used for comparisons between two samples types or genotypes; when more than two samples were being compared , equality of variances was tested with either a Levene’s or Barlett’s test , and if found to be equal , One-Way ANOVA was used to test for significance , followed by a post-doc Dunnett test for multiple comparisons to control . If quantitative data were not distributed normally , then Mann-Whitney U-test was used for comparisons between two genotypes , and Kruskal-Wallis for comparisons between more , followed by post-hoc Dunn’s test for multiple comparisons to control . Post-hoc Bonferroni multiple comparisons test was used to compare all samples with all samples in a given experiment . | Everything that you experience leaves its mark on your brain . When you learn something new , the neurons involved in the learning episode grow new projections and form new connections . Your brain may even produce new neurons . Physical exercise can induce similar changes , as can taking antidepressants . By contrast , stress , depression , ageing and disease can have the opposite effect , triggering neurons to break down and even die . The ability of the brain to change in response to experience is known as structural plasticity , and it is in a tug-of-war with processes that drive neurodegeneration . Structural plasticity occurs in other species too: for example , it was described in the fruit fly more than a quarter of a century ago . Yet , the molecular mechanisms underlying structural plasticity remain unclear . Li et al . now show that , in fruit flies , this plasticity involves Toll receptors , a family of proteins present in the brain but best known for their role in the immune system . Fruit flies have nine different Toll receptors , the most abundant being Toll-2 . When activated , these proteins can trigger a series of molecular events in a cell . Li et al . show that increasing the amount of Toll-2 in the fly brain makes the brain produce new neurons . Activating neurons in a brain region has the same effect , and this increase in neuron number also depends on Toll-2 . By contrast , reducing the amount of Toll-2 causes neurons to lose their projections and connections , and to die , and impairs fly behaviour . Li et al . also show that each Toll receptor has a unique distribution across the fly brain . Different types of experiences activate different brain regions , and therefore different Toll receptors . These go on to trigger a common molecular cascade , but they modulate it such as to result in distinct outcomes . By working together in different combinations , Toll receptors can promote either the death or survival of neurons , and they can also drive specific brain cells to remain dormant or to produce new neurons . By revealing how experience changes the brain , Li et al . provide clues to the way neurons work and form; these findings may also help to find new treatments for disorders that change brain structure , such as certain psychiatric conditions . Toll-like receptors in humans could thus represent a promising new target for drug discovery . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | A Toll-receptor map underlies structural brain plasticity |
Nutritional restriction leads to protein translation attenuation that results in the storage and degradation of free mRNAs in cytoplasmic assemblies . In this study , we show in Drosophila S2 cells that amino-acid starvation also leads to the inhibition of another major anabolic pathway , the protein transport through the secretory pathway , and to the formation of a novel reversible non-membrane bound stress assembly , the Sec body that incorporates components of the ER exit sites . Sec body formation does not depend on membrane traffic in the early secretory pathway , yet requires both Sec23 and Sec24AB . Sec bodies have liquid droplet-like properties , and they act as a protective reservoir for ERES components to rebuild a functional secretory pathway after re-addition of amino-acids acting as a part of a survival mechanism . Taken together , we propose that the formation of these structures is a novel stress response mechanism to provide cell viability during and after nutrient stress .
Cell response to nutritional restriction includes stimulation of degradation pathways , such as autophagy , as well as attenuating anabolic pathways , such as protein translation ( Castilho et al . , 2014 ) . Another key anabolic pathway is protein transport through the secretory pathway . In mammals , one third of the proteome encounters this pathway ( Stevens and Arkin , 2000; Almen et al . , 2009 ) , such as the proteins delivered to the extracellular medium , the plasma membrane , or other cellular membrane compartments with the exception of mitochondria and the nucleus . After their synthesis at the endoplasmic reticulum , proteins exit the ER at specialized ER exit sites ( ERES ) defined by a cup-shaped ER overlaying COPII-coated vesicles in which newly synthesized proteins are packaged . They then reach the Golgi apparatus where they are further modified , sorted , and dispatched to their correct final localization . The COPII coat assembly requires 6 proteins , including the transmembrane protein Sec12 that acts as a GEF for the small GTPase Sar1 . GTP-bound Sar1 recruits Sec23/Sec24 , the inner COPII coat that in turn recruits Sec13/31 , the outer coat ( Bard et al . , 2006; d'Enfert et al . , 1991; Oka et al . , 1991; Rothman and Wieland , 1996; Schekman and Orci , 1996 ) . In addition , the large hydrophilic protein Sec16 has been found to play a major role in the COPII assembly and regulation ( Ivan et al . , 2008; Hughes et al . , 2009 ) ( Connerly et al . , 2005; Kung et al . , 2012; Bharucha et al . , 2013 ) and Sec16 mutation or loss of function leads to a severe impairment in trafficking through the secretory pathway . Stress strongly affects the functional organization of the secretory pathway . For instance , energy deprivation and osmotic shock also block secretion at the level of the ER exit and the cis Golgi , a response mostly triggered by impaired dynamics of the COPI coat , which mediates retrograde transport ( Jamieson and Palade , 1968; Cluett et al . , 1993; Lee and Linstedt , 1999 ) . Interestingly , GBF1 , the GEF of Arf , the small GTPase required for COPI assembly , is phosphorylated and consequently inactivated by AMPK under conditions of nutrient starvation and energy depletion , leading to a block in secretion ( Miyamoto et al . , 2008 ) . Furthermore , biosynthesis of PI4P in yeast that was shown to play a key role in coordinating trafficking from the Golgi with cell growth seems to also be sensitive to nutrient conditions ( Piao et al . , 2012 ) . Last , ER stress that elicits the so-called ‘unfolded protein response’ ( Shamu et al . , 1994 ) directly impedes on the functional organization of ERES in Drosophila S2 cells ( Kondylis et al . , 2011 ) and reduces COPII subunits assembly in human cells ( Amodio et al . , 2009 ) . Furthermore , we have recently reported that serum starvation of Drosophila S2 cells also results in a distinct change in the ERES organization , namely Sec16 cytoplasmic dispersion away from ERES , in a conserved ERK7-dependent mechanism ( Zacharogianni et al . , 2011 ) that leads to protein secretion inhibition . In this study , we focus on amino-acid starvation that leads to the formation of a novel , non-membrane bound cytoplasmic stress assembly that contains ERES components and that we call ‘Sec bodies’ [Figure 1A and ( Zacharogianni et al . , 2011 ) ] . Sec bodies do not represent terminal aggregates . They are reversible , act as a reservoir for ERES components to reconstruct a functional secretory pathway upon re-feeding and are critical for cell survival during stress and upon stress relief . Furthermore , they display properties similar to those of Stress Granules , which place them in the rapidly growing class of cytoplasmic mesoscale assemblies and more specifically , the category of liquid droplets . 10 . 7554/eLife . 04132 . 003Figure 1 . Amino-acid starvation induces the formation of a novel stress assembly in Drosophila S2 cells . ( A ) Immunofluorescence ( IF ) visualization of Delta-myc ( using an anti-Delta antibody ) in S2 cells in Schneider's ( normal growth conditions ) or incubated with Krebs Ringer Bicarbonate buffer ( KRB ) for 4 hr ( amino-acid starvation ) . Note that in Schneider's Delta reaches the plasma membrane whereas it is retained intracellularly in starved cells . ( B ) IF visualization of endogenous Sec16 in Drosophila S2 cells grown in Schneider's and incubated in KRB for 4 hr . Note the formation of Sec bodies ( arrows ) . ( C ) IF visualization of Sec31 , Sec23 and co-visualization of GFP-Sec23 , mCherry-Sec24AB , Sec24CD-GFP , Sar1-GFP with Sec16 in S2 cells in Schneider's and KRB for 4 hr . ( D ) IF co-visualization of dGRASP/Sec24CD-GFP , GRIP/Sec16 , and Fringe-GFP/Sec16 in S2 cells grown in Schneider's and incubated in KRB for 4 hr . ( E ) Kinetics of Sec body formation in S2 cells incubated in KRB over indicated time ( up to 6 hr ) expressed as the percentage of cells exhibiting ERES , intermediates ( see ‘Materials and methods’ ) , and Sec bodies . Scale bars: 10 μm ( A–D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 00310 . 7554/eLife . 04132 . 004Figure 1—figure supplement 1 . Sec body formation and autophagy . ( A–A′ ) Quantification of Atg5 punctae formation in S2 cells incubated with rapamycin ( A ) and incubated with KRB with or without wortmannin ( A′ ) for indicated time points . Note that as expected , autophagy ( marked by Atg5 ) is stimulated by rapamycin and starvation ( KRB ) and inhibited by wortmannin . ( B–B′ ) Quantification of Sec body formation ( marked with Sec16 ) in cells incubated in KRB with and without wortmannin ( B ) and with and without bafilomycin ( B′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 00410 . 7554/eLife . 04132 . 005Figure 1—figure supplement 2 . Sec body formation and single amino-acids . Quantification of the prevention of Sec body formation by specific amino-acids upon incubation in KRB for 4 hr . This is expressed as the percentage of cells exhibiting the normal growth Sec16 localization at ERES . All amino-acids are added at 15 mM except for 3 ( gray bars ) that were also tested at their concentration in Schneider's medium . Note that at 15 mM , histidine , aspartate , and asparagine significantly decreases Sec body formation . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 00510 . 7554/eLife . 04132 . 006Figure 1—figure supplement 3 . Sec body formation in vivo . Projection of four equatorial confocal planes of Sec16 and Spectrin ( D ) in the follicular epithelium covering an egg-chamber from an ovary dissected from a virgin fly fattened for 3 days and incubated ex-vivo in KRB for 4 hr , and from an ovary dissected from a 36-hr starved virgin female . Note that in both cases , Sec16 is found in large punctae reminiscent of Sec bodies . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 00610 . 7554/eLife . 04132 . 007Figure 1—figure supplement 4 . Sec body formation and mammalian cells . IF visualization of Sec16 in immortalized MEFs incubated in growth medium ( DMEM ) or KRB plus bafilomycin for 7 hr . Note that Sec16 is remodeled into larger structures . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 00710 . 7554/eLife . 04132 . 008Figure 1—figure supplement 5 . Human Sec16A is incorporated to Sec bodies in starved Drosophila S2 cells . IF visualization of human Sec16A-V5 transfected in Drosophila S2 cells . Note that it partially localizes to ERES in fed cells but is efficiently incorporated in Sec bodies upon starvation . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 008
Amino-acid starvation of Drosophila S2 cells ( i . e . , cell incubation in Krebs Ringers Bicarbonate buffer , KRB , ( Gaccioli et al . , 2006 ) ) leads to inhibition of protein transport through the secretory pathway , as shown by monitoring the plasma membrane localization of the transmembrane reporter Delta ( Kondylis and Rabouille , 2003 ) . In cells grown in Schneider's , Delta reaches the plasma membrane , whereas in KRB , it is retained intracellularly ( Figure 1A ) . Amino-acid starvation also results in the formation of novel Sec16-positive spherical structures ( Figure 1B ) . In addition to Sec16 , these structures also contain COPII subunits Sec23 , the two Sec24 orthologs Sec24AB ( CG1472 , Hau ) ( Norum et al . , 2010 ) and Sec24CD ( CG10882 , Sten ) ( Forster et al . , 2010 ) , and Sec31 , and we therefore name them ‘Sec bodies’ . Conversely , Sec bodies do not contain Sar1 ( Figure 1C ) , COPI components and clathrin ( not shown ) . They also do not contain dGRASP ( that amino-acid starvation drives to complete dispersion in the cytoplasm ) , the TGN GRIP-domain protein dGCC185 , and the Golgi integral membrane protein , Fringe-GFP ( Figure 1D ) , although these two latter proteins are often found in close proximity to Sec bodies . The morphology of the ER , on the other hand , does not seem affected by amino-acid starvation ( not shown ) . Quantitation of this starvation phenotype reveals that although Sec bodies are present in 20% of cells after 1 hr of amino-acid starvation , it takes between 4 hr and 6 hr to get 90% of the cells displaying the typical Sec body pattern ( Figure 1E ) , that is 7 ± 3 Sec bodies/cell , including 1 to 5 with a diameter comprised between 0 . 6 and 0 . 8 μm . Interestingly , 4–6 hr corresponds to the end of the autophagy peak , a degradative pathway stimulated by starvation ( Klionsky et al . , 2012 ) ( Figure 1—figure supplement 1A′ ) as assessed by the appearance of Atg5 punctae ( Figure 1—figure supplement 1A , A′ ) . In this regard , pharmacological inhibition of autophagy ( by wortmanin or bafilomycin ) results in a premature formation of Sec bodies ( Figure 1—figure supplement 1B , B′ ) , consistent with the notion that Sec bodies form in response to a reduced level of intracellular amino-acid concentration . Given the Sec body content in COPII subunits , we used immuno-electron microscopy ( IEM ) to test whether they are not simply a collection of COPII vesicles . Sec bodies are electron dense structures and non-membrane bound , although small membrane profiles can occasionally be observed in their core and often ER is in close proximity ( Figure 2A , arrows ) . Sec body formation is associated with the loss of the typical early secretory pathway morphology ( Kondylis and Rabouille , 2009 ) . ERES and Golgi stacks are no longer visible . 10 . 7554/eLife . 04132 . 009Figure 2 . Sec bodies are non-membrane bound structures . ( A ) Immuno-electron microscopy ( IEM ) visualization of Sec16 ( 10 nm colloidal gold ) in Sec bodies in ultrathin sections of S2 cells incubated in KRB for 4 hr . Arrows point to membrane in close proximity of Sec bodies . E , endosomes; n , nucleus; m , mitochondria . ( B ) Visualization of Sec bodies ( Sec16 , green ) and lipid droplets ( marked by oil-red-O , red ) . Note that 95% of Sec bodies do not co-localize with lipid droplets . ( C–C′ ) Visualization of Sec bodies ( Sec16 , red ) and Atg5-GFP punctae ( C ) and GFP-Atg8 ( C′ ) after 4 hr starvation . Note that 82% of Sec bodies do not co-localize with Atg5 or Atg8 punctae . Scale bars: 500 nm ( A ) ; 10 μm ( B , C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 009 Sec body formation is specific for amino-acid starvation as heat shock , ER stress ( tunicamycin and DTT treatment ) , glucose starvation , oxidative stress ( arsenate treatment ) , hypoxia ( 1% O2 for 19 hr ) , and respiration uncoupling ( CCCP for 4 hr ) do not lead to this response ( not shown ) . Furthermore , to assess whether Sec bodies form in response to the withdrawal of specific amino-acids , cells were incubated in KRB in the presence of individual amino-acids . Histidine , aspartate , and asparagine ( at 15 mM ) strongly prevent Sec body formation but others also do so , albeit more mildly ( Figure 1—figure supplement 2 ) , suggesting that the signaling pathway is complex . Taken together , we demonstrate that amino-acid starvation leads to the remodeling of ERES components into Sec bodies . Sec bodies seemingly also form in vivo , for instance in ovaries starved either ex vivo or dissected from starved female flies ( Figure 1—figure supplement 3 ) . We also asked whether they form in mammalian cells . Although we do observe some degrees of remodeling of COPII components upon mammalian cell starvation , it remains unclear whether the resulting structures are Sec bodies ( Figure 1—figure supplement 4 ) . However , when human Sec16A is transfected in S2 cells , it is efficiently recruited to Sec bodies along with the endogenous components ( Figure 1—figure supplement 5 ) . This suggests that in mammalian cells , Sec body formation is perhaps less pronounced due to different signaling events and not to the properties of the ERES components themselves ( at least Sec16 ) . The IEM analysis shows that Sec bodies are not endosomes or lipid droplets , as their ultrastructure is very different from these organelles [see ( Teixeira et al . , 2003 ) , for lipid droplet ultrastructure in Drosophila] . This is confirmed by the fact that Sec bodies are negative for neutral lipids stained by oil red O that stains lipid droplet content ( Figure 2B ) . As mentioned above , amino-acid starvation is known to induce autophagy , but we demonstrate that Sec bodies are not autophagosomes , as they do not co-localize with Atg5 or Atg8 ( Figure 2C ) . Furthermore , as Sec bodies do not contain dGRASP , they are clearly different from the recently described yeast ‘compartment for unconventional protein secretion’ ( CUPS ) ( Bruns et al . , 2011 ) . Amino-acid starvation also results in protein translation inhibition/stalling that leads to the accumulation of untranslated mRNAs . Those are stored in Stress Granules ( Kedersha et al . , 1999; Anderson and Kedersha , 2008 ) , or degraded in Processing Bodies ( P-Bodies ) , both cytoplasmic ribonucleoprotein particles ( RNPs ) comprising mRNAs , RNA binding proteins , RNA processing machineries ( P-bodies ) , and translation initiation factors ( Stress Granules ) . We therefore tested whether Sec bodies are related to these structures . We visualized Stress Granules using endogenous FMR1 ( Fragile X mental retardation protein 1 ) , an RNA binding protein , and elF4E , a translation initiation factor , and P-bodies with Tral ( Trailer Hitch ) , a like-SM protein . In normal growth conditions , FMR1 , elF4E , and Tral are largely diffuse in the cytoplasm and Tral is also found in small punctae representing steady-state P-bodies ( Figure 3A ) ( Eulalio et al . , 2007 ) . Upon amino-acid starvation , as expected , Stress Granules form and P-Bodies enlarge ( Figure 3A ) ( Shimada et al . , 2011 ) in agreement with reported phenotypes in many cell types , ( Buchan et al . , 2008; Stoecklin and Kedersha , 2013 ) . In S2 cells , they form a dual structure , Stress Granule/P-Bodies ( SG/PB ) , in which FMR1 strictly co-localizes with Tral ( Figure 3A ) . 10 . 7554/eLife . 04132 . 010Figure 3 . Sec bodies are distinct from Stress Granules and P-bodies . ( A ) IF visualization of endogenous FMR1 , eIF4E ( green ) , and Tral ( red ) in cells growing in Schneider's and incubated in KRB for 4 hr . Note that upon starvation , Stress Granules ( FMR1 , eIF4E ) form and P-Bodies ( Tral ) enlarge to co-localize in SG/PBs . ( B ) IF visualization of Sec16 ( Sec bodies ) and FMR1 ( SG/PB ) in cells incubated in KRB for 4 hr . Sec bodies and SG/PBs are distinct structures but have a spatial relationship . ( C ) IEM visualization of FRM1 and Tral in ultrathin sections of S2 cell incubated with KRB for 4 hr . Note that the Tral and FMR1 positive SG/PBs ( asterisk ) are clearly different from Sec bodies ( Figure 2A ) . Scale bars: 10 μm ( A , B ) ; 500 nm ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 010 Sec bodies and SG/PBs form under the same conditions and in the same time frame , and although they have a spatial relationship and are often found adjacent to each other , Sec bodies are clearly distinct from SG/PBs ( Figure 3B ) . To confirm that they are indeed different structures , we also performed IEM of FMR1 and Tral in starved cells ( Figure 3C ) and compared them to Sec bodies ( Figure 2A ) . The FMR1/Tral positive structures are less electron dense , round and regular , and they appear to be often surrounded by mitochondria , which is not the case for Sec bodies . Taken together , these results show that Sec bodies are a novel stress assembly triggered by amino-acid starvation that is distinct from compartments and structures that are also formed upon this condition . We then asked how Sec bodies form . Time-lapse imaging of live cells using GFP-Sec23 reveals that once the cells sense amino-acid depletion , a pool of ERES rapidly disappears by releasing their components in the cytoplasm , and the remaining ones are rapidly transformed into smaller round structures . These small structures do not seem to efficiently fuse with one another . Instead , they seem to act as a seed and grow by recruiting ERES components from the cytoplasm where they were released to reach the typical Sec body size ( Figure 4A , and Video 1 ) . 10 . 7554/eLife . 04132 . 011Figure 4 . Sec bodies form at ERES but COPII- and COPI-coated vesicle formation is not required . ( A ) Stills of a GFP-Sec23 time-lapse video of a cell incubated in KRB ( t = 0 ) for 60 min showing Sec body formation . ( B–B″ ) IF visualization of miniSec16-V5 ( B , B′ ) and ΔNC2-3-Sec16-V5 ( B , B″ ) in S2 cells incubated in Schneider's ( B ) and KRB for 4 hr . Endogenous Sec16 is in red . Note that the cup-shaped ER forms a cradle for Sec bodies ( B′ , insert ) . ( C ) IF visualization of Sec16 and GFP-Sec23 in mock and Sar1-depleted S2 cells grown in Schneider's and incubated in KRB for 4 hr . Note that the ERES are enlarged in Sar1-depleted cells ( arrows ) and Sec bodies form in both conditions to the same extent . ( D ) IF visualizations of Sec16 and Delta-myc in S2 cells incubated with brefeldin A ( BFA ) in Schneider's and KRB for 3 hr . Note that Delta transport is inhibited in both cases as Delta is retained intracellularly . ( E ) IF visualization of Sec16 in S2 cells grown in Schneider's and incubated in KRB for 4 hr in the presence or absence of brefeldin A ( BFA ) . Note that pre-incubation with the drug does not affect Sec body formation during starvation . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 01110 . 7554/eLife . 04132 . 012Video 1 . Time-lapse movie of Sec body formation . GFP-Sec23 time-lapse video of two cells incubated in KRB ( t = 0 ) for 1 hr . One frame was taken every 3 min and the videos are displayed at 7 frame/s ( related to Figure 4A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 012 To assess , as suggested by the time-lapse , whether Sec bodies form at ERES , we used a truncated version of Sec16 , miniSec16 ( 690-1954 ) , which comprises the minimal Sec16 sequence required for ERES localization ( Ivan et al . , 2008 ) ( Figure 4B ) but is not incorporated into Sec bodies upon amino-acid starvation ( Figure 4B′ ) . Instead , miniSec16 remains associated to the cup-shaped ER of the ERES and seems to cradle the forming Sec bodies ( marked by endogenous Sec16; Figure 4B′ ) . This indicates that Sec bodies form where ERES were present , in line with their observed proximity to ER membrane ( arrows in Figure 2A ) . The time-lapse also suggested that the Sec body enlargement is mediated by recruitment of ERES components that have been dispersed in the cytoplasm . To test this further , we used a Sec16 deletion mutant ( ΔNC2-3 ) that does not localize to ERES and is mostly cytoplasmic ( Figure 4B ) because it lacks the region that mediates its recruitment to ERES ( NC2-3 ) ( Ivan et al . , 2008 ) . We found that it is efficiently recruited to Sec bodies , showing that Sec16 can be recruited from the cytoplasm and contributes to Sec body enlargement ( Figure 4B″ ) . Furthermore , this result indicates that the localization to ERES prior to starvation is not necessary for incorporation to Sec bodies . Importantly , the recruitment of this cytosolic mutant is not due to its interaction with endogenous Sec16 as the NC2-3 also contains the Sec16 oligomerization domain ( Ivan et al . , 2008 ) . This suggests that a distinct Sec16 domain responds to amino-acid starvation . Taken together , these results show that Sec bodies form ( at least initially ) where ERES were located and increase in size by recruiting ERES components dispersed in the cytoplasm . We then ask whether membrane traffic through the early secretory pathway is required for Sec body formation . To test if COPII vesicle formation is required , we depleted Sar1 by RNAi before starvation . Sar1 depletion is evidenced by a strong reduction ( 40 ± 5% ) in cell proliferation and ERES enlargement ( Ivan et al . , 2008 ) ( Figure 4C , arrows ) . However , Sec body formation was found to be as efficient as in control ( mock depleted ) cells ( Figure 4C ) . Second , we pharmacologically inhibited protein trafficking with Brefeldin A ( Figure 4D ) and found that this treatment before and during starvation does not affect Sec body formation ( Figure 4E ) . This demonstrates that transport in the early secretory pathway via COPI and COPII vesicle formation is not required for the formation of Sec bodies . Given the Sec body content in ERES components ( Figure 1 ) , we then tested using RNAi if they are necessary for Sec body formation . Depleting Sec16 did not yield satisfactory conclusions as we have shown that Sec16 is critical for the organization of ERES in Drosophila S2 cells ( Ivan et al . , 2008 ) and Sec16 depletion led to the aggregation of most of the COPII components even in cells grown in full medium ( see Figure 2A–C″ of Ivan et al . , 2008 ) . However , in the few depleted cells where an observation could be made , Sec bodies did not form ( not shown ) . We then depleted the two Sec24 gene products that are both expressed in S2 cells ( see DRSC , Drosophila RNAi screening center , http://www . flyrnai . org/ ) ( see ‘Materials and methods’ for depletion controls ) . When Sec24AB-depleted cells are starved , the normal Sec body formation ( marked by Sec16 , Figure 5A ) is impaired ( Figure 5B ) . The distribution of diameters of the resulting structures shows that they are twofold smaller and twice as many , when compared to Sec bodies in mock-depleted cells ( Figure 5E ) . In agreement with Sec24 forming a complex with Sec23 , Sec23 depletion also results in the same phenotype ( Figure 5C , E ) . These smaller structures are not classical Sec bodies . By IF , some of them appear to have a horseshoe shape . By IEM , they appear as a collection of Sec16 positive vesicular and tubular membrane profiles , some small , some large probably corresponding to ERES mixed with Golgi fragments , and a third category that we name ‘intermediates’ as they are reminiscent of Sec bodies by their round shape but that contain membrane and are smaller in size when compared to Sec bodies found in mock-depleted cells ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 04132 . 013Figure 5 . Sec23 and Sec24AB are key factors for Sec body formation . ( A–D ) IF visualization of Sec16 in mock ( A ) , Sec24AB ( B ) , Sec23 ( C ) , and Sec24CD ( D ) depleted S2 cells in Schneider's and incubated in KRB for 4 hr . Note that Sec body formation is inhibited upon Sec24AB and Sec23 , but not upon Sec24CD depletion . Boxed areas are shown at higher magnification . ( E ) Distribution of Sec body size ( shown as frequency of observed Sec body diameter ) in mock , Sec24AB , Sec24CD , and Sec23-depleted and starved cells ( as in A ) ( dsGFP: 45 cells , 341 Sec bodies; dsSec24AB: 43 cells , 585 Sec bodies; dsSec24CD: 35 cells , 245 Sec bodies; dsSec23: 36 cells , 504 Sec bodies ) . Note that the Sec body's mean diameter decreases by 1 . 8-fold upon Sec24AB and Sec23 depletion and that Sec bodies are twice as many . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 01310 . 7554/eLife . 04132 . 014Figure 5—figure supplement 1 . The smaller structures generated in starved Sec24AB-depleted cells are not Sec bodies . ( A ) Sec16 positive horseshoe shape structures observed by IF . B: Gallery of Sec16 positive IEM profiles: smaller clusters of membrane , larger ones probably corresponding to ERES mixed with Golgi fragments , intermediate structures ( round shape but with membrane in their core and of smaller size than typical Sec bodies ) . Scale bars: 2 μm ( A ) and 500 nm ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 014 Interestingly , Sec24CD depletion ( Figure 5D , E ) does not lead to the same phenotype as Sec24AB depletion and Sec bodies form seemingly normally , showing specificity for one Sec24 homologue . Taken together , this indicates a key and novel role for Sec23 , Sec24AB ( and perhaps Sec16 ) in Sec body formation , which is distinct from their classical role in ER exit via COPII vesicle formation that is not required ( Figure 4C–E ) . Given that the Sec bodies are non-membrane bound , we then asked how their components are prevented from freely diffusing in the cytoplasm . One class of stress related cytoplasmic structures are liquid droplets that are described to result from phase separation-induced liquid demixing in the cytoplasm ( Brangwynne et al . , 2009 , 2011; Hyman and Simons , 2012 ) . They are defined as non-membrane bound , spherical , reversible structures , the components of which diffuse easily within the droplet but are in slower exchange with the cytoplasm . Furthermore , liquid droplets are known to contain proteins that are prone to engage in weak protein–RNA or protein–protein interactions as their components display a high level of low complexity sequences ( LCS , defined as regions of low amino-acid diversity ) ( Kato et al . , 2012 ) ( Figure 6—Source data 1 ) . P-granules in C . elegans , nucleoli , but also Stress Granules and P-bodies have been shown to have liquid droplet properties ( Brangwynne et al . , 2009 , 2011; Hyman and Simons , 2012 ) . We therefore set out to assess whether Sec bodies are also liquid droplets . First , we used SEG , a bioinformatics tool that determines the LCS content of proteins recruited to the Sec bodies . Interestingly , we found that Sec16 and the two Sec24 gene products , Sec24AB , Sec24CD ( Figure 6A ) display a significantly higher LCS content when compared to other proteins related to the early secretory pathway and to the entire Drosophila proteome ( our analysis , see ‘Materials and methods’ , Figure 6A′; Figure 6—Source data 1 ) . Sec16 LCSs are situated throughout its sequence with the notable exception of its conserved central domain ( CCD , aa 1090–1590 ) ( Figure 6A ) . On the other hand , Sec24AB and Sec24CD LCSs are mostly situated at the N-terminus of the protein sequence in a manner that is partially conserved throughout evolution ( Figure 6—figure supplement 1 ) . Furthermore , as recently suggested ( Das et al . , 2014 ) , LCSs correspond to a high level of unstructured sequences and we show that it is indeed the case for Sec24AB , Sec24CD , and Sec16 ( using HHpred , http://toolkit . tuebingen . mpg . de/hhpred/ ) , Figure 6—figure supplement 2 and not shown , respectively ) . 10 . 7554/eLife . 04132 . 015Figure 6 . Sec body proteins contain low complexity sequences that are necessary for Sec body recruitment . ( A–A′ ) Schematic representation of the Low Complexity Sequences ( blue bars ) in Sec16 , Sec24AB , Sec24CD , and Sec23 ( A ) . The red bars mark the boundaries of the Sec16 domains . Genome wide analysis of Low Complexity Sequence ( LCS ) in the Drosophila proteome , in proteins related to the secretory pathway and proteins related to Stress Granules/P-bodies ( A′ ) . B: IF localization of sfGFP-tagged full-length Sec24B , Sec24AB LCS , and Sec24AB nonLCS in S2 cells in Schneider's and KRB for 4 hr , together with endogenous Sec16 ( red ) . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 01510 . 7554/eLife . 04132 . 016Figure 6—Source data 1 . Table showing the LCS content of proteins related to Stress Granules and P-bodies as well as the early secretory pathway ( related to Figure 6A , A′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 01610 . 7554/eLife . 04132 . 017Figure 6—figure supplement 1 . LCS content of Sec24 in different species . ( A ) A restricted phylogenic tree of the Sec24 sequences analyzed in ( B ) Note that S . pombe and S . cerevisiae Sec24 are similar to each other and to H . sapiens Sec24C and D , and D . melanogaster Sec24CD . However , they are distant to D . melanogaster Sec24AB , H . sapiens Sec24A and B . Furthermore , D . melanogaster Sec24AB is distant to D . melanogaster Sec24CD . ( B ) LCS analysis and schematics in Sec24 sequences of different organisms . Note that most sequences contain a significant percentage of LCS in the N-terminal third of the protein with the exception of S . pombe , A . melifera and G . gallus ( related to Figure 6A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 01710 . 7554/eLife . 04132 . 018Figure 6—figure supplement 2 . Secondary structure prediction of Drosophila Sec24AB using HH pred . C/c denotes the unstructured , H/h the alpha helices , and E/e , the beta sheets . Note their absence in the 405 amino-acids of the N-terminus corresponding to LCS . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 01810 . 7554/eLife . 04132 . 019Figure 6—figure supplement 3 . Sec24AB LCS is not sufficient to drive Sec body formation . S2 cells were depleted of endogenous Sec24AB . When starved ( KRB ) , this resulted in the formation of small structures ( as in Figure 5 ) . The transfection of Sec24AB LCS-sfGFP in these depleted cells did not rescue the formation of Sec bodies . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 019 Remarkably , two of the LCS enriched proteins Sec24AB and Sec16 are also required for Sec body assembly , suggesting that this feature might be necessary . However , not all ERES residing and LCS rich proteins are necessary for Sec body formation , as Sec24CD that contains the same amount of LCS , is not . We then tested whether LCSs were necessary for protein recruitment to Sec bodies and/or sufficient for their formation . We focused on Sec24AB as the LCSs are clustered to the first 415 amino-acids at the N-terminus ( Sec24AB LCS ) and compared their Sec body recruitment to this of its nonLCS region ( aa 416–1184 ) ( Figure 6B ) . LCS-sfGFP is largely recruited to ERES under normal growth conditions although not as efficiently as full-length Sec24AB . Under starvation conditions , it localizes to Sec bodies as full-length Sec24AB and seems to lead to their enlargement . This demonstrates that the LCS rich region of Sec24AB is sufficient to mediate recruitment to Sec bodies . Conversely , the nonLCS region is mostly cytoplasmic and remains largely so upon starvation , although a small pool is recruited to the Sec bodies . This shows that the LCS rich N-terminus region of Sec24AB plays a key role in recruitment of Sec24 to Sec bodies . We then tested whether the Sec24AB LCS was sufficient to drive Sec body formation . To do so , cells were depleted of endogenous Sec24AB ( resulting in the formation of Sec16 positive smaller structures ) followed by the expression of Sec24AB LCS . If this is sufficient , we expect that Sec bodies would form . However , although Sec24AB LCS is recruited to the smaller structures , Sec bodies did not significantly form ( Figure 6—figure supplement 3 ) . This suggests that either the nonLCS region of Sec24AB participates to Sec body formation , even though on its own , it is only slightly recruited , or that one or multiple other factors are involved in driving Sec body formation . Second , we assessed whether the FRAP properties of Sec bodies are compatible with liquid droplets , that is , assemblies made through phase separation . When a fraction of such an assembly ( GFP marked ) is photobleached , the recovery is quick as the molecules within mix instantaneously . However , when entirely photobleached , the recovery is slower as the exchange with the surrounding cytoplasm is not as efficient . We used Sec16-sfGFP and GFP-Sec23 that are efficiently incorporated to Sec bodies . When Sec bodies are partially bleached , the recovery is very fast for both GFP-Sec23 and Sec16-sfGFP , and the maximum intensity is approximately 50% of the original one . After complete photobleaching , however , Sec bodies recover more slowly and only to 10% of the initial fluorescence intensity , showing an inefficient exchange with the surrounding cytoplasm ( Figure 7A , A′; Video 2 and Video 3 ) . This is comparable to FRAP properties of Stress Granules that are well-documented liquid droplets . When entirely bleached , Stress Granules recover more than Sec bodies and when partially bleached , they recover slightly less ( Figure 7B , B′; Video 4 and Video 5 . This indicates that the Sec23 and Sec16 diffuse more quickly within Sec bodies than FMR1 in Stress Granules , but that their phase transition barrier is higher . 10 . 7554/eLife . 04132 . 020Figure 7 . Sec bodies have FRAP properties consistent with liquid droplets . ( A–A′ ) Percentage of fluorescence recovery after photobleaching ( FRAP ) over time of individual Sec bodies , marked by GFP-Sec23 ( Video 2 ) , and ΔNC1-Sec16-sfGFP ( Video 3 ) in S2 cells incubated in KRB for 4 hr . The triangles in B show the FRAP of Sec bodies that have been entirely bleached ( n = 3 , arrowheads in B′ ) . The circles show the FRAP of Sec bodies ( n = 3 , arrows in B′ ) that have been partially bleached . The dashed circles in B′ indicate the Sec bodies that have been entirely bleached and assessed in stills taken from Video 2 and 3 . ( B–B′ ) Percentage of fluorescence recovery after photobleaching ( FRAP ) over time of individual Stress Granules marked by FMR1-sfGFP ( Video 4 and 5 ) in S2 cells incubated in KRB for 4 hr . The triangles in C show the FRAP of Stress Granules that have been entirely bleached ( n = 3 , arrowheads in C′ ) . The circles show the FRAP of Stress Granules ( n = 3 , arrows in B′ ) that have been partially bleached . The dashed circles in B′ indicate the Stress Granules that have been bleached and assessed in stills taken from Video 4 and 5 . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02010 . 7554/eLife . 04132 . 021Video 2 . FRAP of one half-bleached Sec body and one fully bleached . FRAP video of two GFP-Sec23 positive Sec bodies , one photobleached partially and one entirely , recorded every 10 ms for at least 20 s , and then every minute . The video is displayed at 7 frame/s ( related to Figure 7B′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02110 . 7554/eLife . 04132 . 022Video 3 . FRAP of a fully bleached Sec body . FRAP video of one ΔNC1-Sec16-sfGFP positive Sec body entirely bleached , recorded every 10 ms for at least 20 s , and then every minute . The video is displayed at 7 frames/s ( related to Figure 7B′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02210 . 7554/eLife . 04132 . 023Video 4 . FRAP of a fully bleached Stress Granule . FRAP video of one FMP1-sfGFP positive Stress Granule entirely bleached , recorded every 10 ms for at least 20 s , and then every minute . The video is displayed at 7 frames/s ( related to Figure 7B′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02310 . 7554/eLife . 04132 . 024Video 5 . FRAP of a partially bleached Stress Granule . FRAP video of one FMP1-sfGFP positive Stress Granule partially bleached , recorded every 10 ms for at least 20 s , and then every minute . The video is displayed at 7 frames/s ( related to Figure 7B′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 024 Taken together , the spherical morphology , specific FRAP properties , and the presence of LCSs are features compatible with Sec bodies being liquid droplets . Third , we assessed the Sec body reversibility , a key feature of liquid droplets and we tested it using cells that were starved for 4 hr and further incubated in Schneider's . This results in the full recovery of their typical ERES pattern in less than 30 min ( Figure 8A; Video 6; Figure 8—figure supplement 1; Figure 8E ) . This convincingly shows that Sec bodies are not terminal aggregates . Furthermore , these ERES are functional as they support efficient transport in the secretory pathway ( Figure 8B ) to allow proliferation ( Figure 9B , solid dark blue line ) . 10 . 7554/eLife . 04132 . 025Figure 8 . Sec bodies are functionally reversible and act as reservoir for ERES components during starvation . ( A ) Stills of a GFP-Sec23 time-lapse video ( Video 6 ) of two cells recovering in Schneider's after 4 hr in KRB ( t = 0 , up to 60 min ) . Note that Sec bodies are reversed into ERES . ( B ) IF localization of Delta in cells that were starved ( KRB ) or not ( Schneider's ) followed further incubation in Schneider's for 2 hr . ( C ) Quantification of the percentage of cells with Delta at the plasma membrane in cells that were either starved ( KRB ) or not ( Schneider's ) followed by reversion in Schneider's . Delta was induced for 30 , 60 , 90 , and 120 min while cells were reverted in Schneider's . ( D–D′ ) IF visualization of Sec16 in cells starved in KRB supplemented or not with cycloheximide ( CHX , B ) , and in starved cells further incubated in Schneider's supplemented or not with CHX ( C′ ) . Note that neither Sec body formation nor reversal is affected by the presence of CHX . ( E ) Quantification of the Sec body reversal as described in B′ expressed as the percentage of cells exhibiting Sec bodies . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02510 . 7554/eLife . 04132 . 026Figure 8—figure supplement 1 . IF localization of Sec16 in cells recovering in Schneider's for 10–30 min after 4 hr in KRB . At 30 min , 90% of ERES are rebuilt . Note that the intensity of fluorescence of Sec16 is higher than in the cells that were maintained in Schneider's , reflecting the storage of the ERES components in Sec bodies . Scale bars: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02610 . 7554/eLife . 04132 . 027Video 6 . Sec bodies are reversible . GFP-Sec23 time-lapse video of two cells recovering in Schneider's after 4 hr in KRB ( t = 0 , up to 66 min ) . One frame was taken every 3 min and the video is displayed at 7 frame/s ( related to Figure 8A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 02710 . 7554/eLife . 04132 . 028Figure 9 . Sec body formation is a pro-survival mechanism . ( A ) Western blot of Sec31 , Sec23 , and tubulin ( loading control ) of lysates from GFP , Sec16 , Sec23 , and Sec24AB-depleted cells grown in Schneider's ( S ) and incubated with KRB for 4 hr ( K ) . ( B ) Graph of cell viability ( expressed as percentage of alive cells ) . The number of starting cells at t = 0 , either non- ( control , dark blue lines ) , mock- ( dsGFP , light blue lines ) , Sec24AB ( green lines ) , Sec24CD ( violet lines ) , and double Sec24 AB and CD ( orange lines ) depleted , is set at 100% . These cells are incubated in Schneider's ( dashed lines ) and KRB ( solid lines ) for 4 hr and further incubated in Schneider's up to 16 hr . Note that the mock depletion ( dsGFP , light blue dashed lines ) is slightly detrimental to cell survival upon amino-acid starvation when compared to non-depleted ( control , dark blue dashed line ) . ( C ) Quantification of the percentage of cells with Delta at the plasma membrane in mock- , Sec23 , and Sec24AB-depleted cells that were either starved ( KRB ) or not ( Schneider's ) followed by reversion in Schneider's . Delta was induced for 90 min while cells were reverted in Schneider's . Error bars in B represent standard error of the mean and in C standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 04132 . 028 Overall , although we have not been able to determine with certainty whether Sec bodies contain RNAs as all liquid droplets so far characterized do , we propose that amino-acid starvation leads to the formation of a novel stress assembly with liquid droplet features , the Sec bodies . Remarkably , addition of protein translation inhibitor cycloheximide during the reversal does not affect the ERES re-building ( Figure 8D′ , E ) . This suggests that Sec bodies act as a reservoir for ERES components allowing their re-mobilization upon stress relief to rebuild a functional secretory pathway . Of note , cycloheximide addition during starvation also does not affect Sec body formation ( Figure 8D′ ) . To test further whether Sec bodies act as a mechanism for ERES components during starvation preventing their degradation , we first monitored the level of ERES components during starvation . Remarkably , amino-acid starvation leads to an increased level of Sec16 , Sec23 , and Sec31 ( Figure 9A; Figure 9A′ , compare lanes 1 and 2 ) . We can rule out that this is due to an increase in protein translation during starvation as it is efficiently inhibited after 20 min ( not shown ) . It is therefore likely that Sec body formation leads to a stabilization of the ERES components and therefore protect them against degradation . We then asked whether this protection is inhibited when Sec bodies do not form . Upon Sec16 , Sec23 and Sec 24AB depletion , we observed two things: the first is that under normal growth conditions , the level of Sec31 is higher than in mock-depleted cells ( Figure 9A , compare lanes 3 , 5 and 7 to lane 1 ) . The second is that in depleted cells , this level is not maintained upon starvation . Sec31 level is decreased instead of being stabilized as in mock-depleted cells . Sec23 behaves similarly to Sec31 . In Sec24AB-depleted cells , however , its level is reduced even in cells incubated in full medium , suggesting that Sec23 turnover depends on the presence of Sec24AB . Altogether , this experiment suggests that starvation leads to ERES component stabilization that is inhibited when Sec bodies do not form . This supports the notion that Sec bodies act as a reservoir for ERES components to rebuild a functional secretory pathway upon stress relief . In this context , we investigated the relevance of Sec body formation in cell survival upon starvation and fitness upon stress relief . To do so , we exploited the fact that only one of the two Sec24 proteins is required for Sec body formation . Indeed , Sec bodies do not form in starved Sec24AB-depleted cells , whereas they do in Sec24CD-depleted cells ( see Figure 5 ) . As mentioned above , depletion of either of the Sec24 isoforms is effective but does not affect proliferation very much when cells are grown in normal medium ( Figure 9B , dashed lines and see ‘Materials and methods’ ) . However , upon amino-acid starvation , the number of Sec24AB-depleted cells decreases twice as much as the mock- ( ds GFP ) and Sec24CD-depleted cells ( Figure 9B , compare solid green to light blue , violet lines ) . As expected , the double depleted Sec24AB and CD cells also decline more quickly ( Figure 9B , solid orange line ) . When reverted to full medium , control and Sec24CD-depleted cells started to proliferate again , whereas the number of Sec24AB-depleted cells continues to diminish ( Figure 9B ) . As the absence of Sec bodies leads to a decreased stabilization of COPII components , one reason behind the cell lethality upon re-feeding could be the inefficient protein transport through the secretory pathway . To test this , we monitored the transport of Delta to the plasma membrane of starved mock- , and Sec24AB-depleted cells upon re-feeding for 90 min ( Figure 9C ) . Mock-depleted cells efficiently transport Delta whether starved or not ( Figure 9C; Figure 8C ) . Conversely , the efficiency of transport of Sec24AB-depleted cells after starvation is largely compromised , suggesting that formation of Sec bodies is critical for proper transport resumption . However , this result could simply be due to the depleted Sec24AB whose absence compromises transport even in cells kept in Schneider’s . Nevertheless , Delta transport in Sec24AB-depleted cells kept in Schneider's was found almost as efficient as mock-depleted cells , suggesting that Sec24CD compensates Sec24AB depletion . This indicates that the inhibition of Sec body formation , not the absence of Sec24AB , is detrimental to anterograde transport resumption . Taken together , these results show that Sec body formation is instrumental to efficient resumption of protein transport through the secretory pathway that contributes to cell survival and growth after re-feeding .
Here , we describe a novel , reversible , and non-membrane bound structure , the Sec body that forms in response to nutrient stress . Sec bodies comprise proteins that in normal growth conditions function as ERES components , including subunits of the COPII coat , namely Sec23 , Sec24AB , Sec24CD , and Sec31 as well as Sec16 , the upstream ERES organizer . A noticeable exception is the small GTPase Sar1 . One reason for this could be that Sar1 is devoid of LCS and this is currently under further investigation . Interestingly , the components of the other coats were seemingly not incorporated into Sec bodies but also did not form other structures , suggesting that remodeling of the ERES is sufficient to ensure inhibition of protein transport through the secretory pathway . In this regard , Sec body formation constitutes also a novel mechanism for attenuation/inhibition of protein transport through the secretory pathway . Cells can disperse ERES and Golgi components into the cytoplasm , as reported during mitosis the Golgi is fragmented ( Lucocq and Warren , 1987; Farhan et al . , 2010; Zacharogianni et al . , 2011 ) . The dramatic remodeling of the ERES that we describe here appears to be specific for amino-acid starvation , possibly underlining the acute and severe nature of this particular stress . It is also different from the response to serum starvation that requires ERK7 ( Zacharogianni et al . , 2011 ) . ERK7 appears to be involved to a small extent in the amino-acid starvation response , possibly facilitating the initial dispersion of a fraction of the ERES into the cytoplasm , as when depleted , it prevents Sec body formation upon amino-acid starvation by about 20% ( Zacharogianni et al . , 2011 ) . However , other signaling pathways , yet unidentified , are clearly at stake . Interestingly , one of Sec body components is Sec16 , a protein with a localization that is also modulated upon serum starvation . Furthermore , Sec16 is also phosphorylated in response to EGF signaling in human cells ( Farhan et al . , 2010 ) . It could therefore be emerging as one of platform integrating nutrient and growth factors availability . Sec bodies are novel stress structures and we have shown that they are not autophagosomes ( or substrates of autophagy ) , not lipid droplets and not CUPS as they are devoid of dGRASP that is found quantitatively re-distributed in the cytoplasm upon amino-acid starvation , suggesting a modification involving its lipid anchor or modification of its N-terminus . Sec bodies are also different from large reversible structures containing COPII components that have been described in yeast in a number of specific COPII mutants Sec12-4 and Sec16-2 ( Shindiapina and Barlowe , 2010 ) . These structures are thought to result from an imbalance between cargo incorporation in COPII-coated vesicles and the coat formation , and lowering the cargo load by inhibiting protein translation prevented their appearance . Sec body formation is , however , insensitive to translation inhibition by cycloheximide and is therefore different from these yeast structures . Last , we also show that Sec bodies are distinct from Stress Granules and P-bodies that also form upon amino-acid starvation . Therefore , Sec bodies are novel structures . Formation of mesoscale protein assemblies like Stress Granules , P-bodies or , more recently , Sec bodies is emerging as a general response to stress and especially nutrient stress , and is gaining increasing attention ( Hyman and Brangwynne , 2011; Wilson and Gitai , 2013 ) . For instance , in yeast under nutrient limiting conditions , metabolic enzymes and stress response proteins form reversible foci ( Narayanaswamy et al . , 2009 ) , such as purinosomes containing enzymes of the purine biosynthetic pathway ( An et al . , 2008; O'Connell et al . , 2012 ) , proteasome storage granules upon glucose restriction ( Laporte et al . , 2008; Peters et al . , 2013 ) , or , as recently described , glutamine synthetase filaments ( Petrovska et al . , 2014 ) . In challenging conditions , areas of localized biochemistry in the cytoplasm can be advantageous , as reagents and possibly energy can be focused to these specific areas . The reorganization of the cytoplasm through non-membrane bound protein assemblies could confer this rapid and spatio-temporally defined compartmentalization . In this regard , we have found that Sec bodies confer a fitness advantage to the cells under starvation ( see below ) . However , some stress assemblies ( especially cytoplasmic RNP granules ) can form dysfunctional RNA–protein assemblies that become irreversible and toxic for the cell . For instance , Stress Granule components have a strong relationship with degenerative diseases , such as ALS and laminopathies ( Ramaswami et al . , 2013 ) . Whether Sec body components could also form such deleterious aggregates remains to be established . Some of these mesoscale assemblies have liquid-like properties . These so-called liquid droplets are generally spherical and dynamic and form via phase separation ( liquid demixing ) of their components from the cytoplasm like a drop of oil in water . Their components display different rates of diffusion within the assembly and in the surrounding cytoplasm ( Hyman and Brangwynne , 2011 ) . They form via transient and weak protein–protein and protein–RNA interactions mediated by low amino-acid diversity stretches ( low complexity sequences , LCS ) , prone to engage in such interactions . Stress granules and P-bodies have been shown to be liquid droplets , and we show here that Sec bodies exhibit clear liquid droplet features as underlined by their spherical morphology , their reversibility , FRAP properties , and LCS content . In this regard , the presence of LCSs both in Sec16 and Sec24 is intriguing considering their role in cells under normal growth conditions where they act in sequence with many others to form the COPII coat . How the LCSs are shielded to make proteins competent for their function in COPII coat formation in growing cells remains to be investigated but the interaction with both cargo and Sec23 might be instrumental to their functioning as coat subunits . We show here that the LCS rich domain of Sec24AB is sufficient and necessary for Sec body incorporation upon amino-acid starvation , but not sufficient to induce Sec body formation . This has similarity to Tia1 , a key protein necessary for Stress Granule formation that has an LCS-prion like domain that is necessary to form stress granules ( Gilks et al . , 2004 ) . As mentioned above , the structures that fall in the liquid droplet category have been described to form through weak protein–RNA interactions . Although Sec bodies do not appear to contain RNAs , we propose that they are nonetheless liquid droplets . The absence of RNA might account for the very low and slow recovery we observed after complete photobleaching of whole Sec bodies when compared to Stress Granules that recovers to a higher degree . Shuttling of mRNA in and out of Stress Granules could drive more exchange between the structure and the surrounding cytoplasm , and this probably does not occur in Sec bodies . However , instead of protein–RNA interactions , Sec body components could establish weak protein–protein interactions helped by molecular modifications that could trigger conformational changes and perhaps exposure of their LCS . Importantly , one of the key features of a liquid droplet is to be efficiently reversible , and we show that Sec bodies are rapidly and completely reversible upon stress relief . This shows that they act as a reservoir of the ERES components that can be quickly remobilized to re-build a functional organelle , so that protein transport through the secretory pathway can resume once stress is relieved in order to support cell proliferation . Furthermore , Sec bodies have a role in protecting ERES components from degradation during starvation . That strengthens the fact that Sec bodies are neither autophagosomes nor a substrate of autophagy , unlike Stress Granules , which are reported to be cleared by autophagy ( Buchan et al . , 2013 ) . Last , we show that Sec body formation is critical for the cell viability during amino-acid starvation . It suggests a pro-survival mechanism , perhaps through the recruitment and inactivation of pro-apoptotic factors . This remains to be investigated . Taken together , amino-acid starvation inhibits both protein translation and protein transport through the secretory pathway and , similarly for both processes , results in the concomitant formation of cytoplasmic stress assemblies where key components necessary for cell survival are stored: untranslated mRNAs in Stress Granules and ERES components in Sec bodies .
Drosophila S2 cells were cultured in Schneider's medium supplemented with 10% insect tested fetal bovine serum ( referred to as Schneider's ) at 26°C as previously described ( Kondylis and Rabouille , 2003; Kondylis et al . , 2007 ) . Amino-acid starvation was carried out by incubating the cells for 4 hr ( or otherwise stated ) in Krebs Ringers' Bicarbonate buffer ( KRB , 10 mM glucose , 0 . 5 mM magnesium chloride , 4 . 53 mM potassium chloride , 120 . 7 mM sodium chloride , 0 . 7 mM dibasic sodium phosphate , 1 . 5 mM monobasic sodium phosphate , 15 mM sodium bicarbonate , 5 . 4 mM calcium chloride ) at pH 7 . 4 . We verified that simply adding 10% FBS to the buffer did not prevent the Sec body and Stress Granule formation . Single amino-acids were added at 15 mM ( unless otherwise indicated ) . Wild-type S2 cells or stably transfected were depleted by RNAi , as previously described ( Kondylis and Rabouille , 2003; Kondylis et al . , 2007 ) . Cells were analyzed after incubation with dsRNAs for 5 days ( or 4 days in the case of Sar1 depletion ) . Transient transfections were performed using the Effectene transfection reagent ( 301425; Qiagen , Germany ) according to the manufacturer's instructions . When cells were transfected with pMT constructs , expression was induced 48 hr after transfection with 1 mM CuSO4 for 1 . 5 hr . The newly synthesized proteins were allowed to localize for 1 hr after CuSO4 washout . When cells were transfected with the pUAS constructs , transfection was done 48 hr prior to the experiment . Drugs were used at the following concentrations: cycloheximide ( 10 μM ) , wortmanin ( 1 μM ) , bafilomycin ( 100 nM ) , rapamycin ( 2 µM ) , and brefeldin A ( 50 µM ) . When a drug treatment was followed by starvation , the cells were pretreated for 30 min in Schneider's following starvation in the presence of the drug . The following antibodies have been used in these experiments: rabbit polyclonal anti-Sec16 ( Ivan et al . , 2008 ) , 1:800 IF , 1:2000 WB; rabbit polyclonal anti-Sec23 ( Pierce PA1-069A ) , 1:200 IF , 1:1000 WB; rabbit polyclonal anti-Sec31 ( Bentley et al . , 2010 ) 1:200 IF; mouse monoclonal anti-V5 ( life technologies R960 ) , 1:500 IF; mouse monoclonal anti-FMR1 ( DSHB supernanant clone 5A11 ) , 1:10 IF; rabbit polyclonal anti-Tral , 1:200 IF; rabbit anti-dGRASP , 1:500 IF ( with methanol fixation ) ; mouse monoclonal anti-α-spectrin ( DSHB supernatant clone 3A9 ) , 1:20 IF; rat polyclonal anti-eIF4E , 1:200 IF ( with methanol fixation ) ; mouse monoclonal anti-Delta , ( DHSB clone C594 . 9B ) , 1:500 IF . The pRmeGFP-Sec23 , pRmSar1-eGFP , pMTmini-Sec16 ( NC2 . 3-CCD ) -V5 , pMTΔNC2 . 3-V5 constructs were described in the reference Ivan et al . ( 2008 ) . The pMTΔNC1-ΔCter-Sec16-V5 and the pMTΔNC1-Δ64-Sec16-V5 were described in the reference Zacharogianni et al . ( 2011 ) . The pMT-Atg5-V5 is a gift from Fulvio Reggiori . The pUASt-GFP-Sec24CD ( Sten ) and the pUASt-mCherry-Sec24AB were a kind gift from Stefan Luschnig . The Fringe-GFP construct is described in the reference Kondylis et al . ( 2007 ) . To generate ΔNC1Sec16sfGFP , sfGFP was amplified using the forward ( ggccgcggatggtgagcaagggcgagga ) and reverse ( gggtttaaacttacttgtacagctcgtccatg ) primers and cloned into PMTV5-B-ΔNC1Sec16 ( Zacharogianni et al . , 2011 ) using SacII and PmeI restriction sites . Human Sec16A-V5 was cloned in the pMT-V5-HisB vector using the primers forward ( tagccaccggtaccatgcctgggctcgaccga ) and reverse ( tacggaattcaagttcagcaccaggtgcttcctct ) to include the KpnI and EcoRI restriction sites . To generate FMR1-sfGFP , sfGFP was first amplified using the forward ( catgttcgaaatggtgagcaagggcgag ) and reverse ( catgaccggtcttgtacagctcgtccatgc ) primers containing the restriction sites for BstBI and AgeI , respectively and cloned into PMT-V5-His to replace the V5 tag , leading to PMT-sfGFP . To generate super folder GFP ( sfGFP ) tagged pMT-Sec24AB , pMT-Sec24AB LCS , and pMT-Sec24AB nonLCS , Sec24Ab LCS ( 1-415 aa ) were amplified from cDNA of Drosophila S2 cells using the forward ( gatggaattccaccatgtcgacttacaat ) and reverse ( gtcagggccctctggagcagtggttc ) , and Sec24AB nonLCS ( 416–1184 aa ) regions using forward ( cagtgaattcatgctaaacgtggctca ) and reverse ( gtcagggccctctttttgcaacacatt ) primers . Fragments were cloned into pMT-sfGFP using EcoRI and ApaI restriction sites . FMR1 cDNA was then amplified from the total cDNA of S2-cells using the forward ( catgggtacccaccatggaagatctcctcgtgga ) and the reverse ( catggaattcaaggacgtgc cattgaccag ) primers and inserted in pMT-sfGFP-His using KpnI and EcoRI restriction sites . The following primers were used to amplify cDNA templates using for RNAiForwardReverseSec24ABTaatacgactcactataggggccaaccggtttcaatcagtaatacgactcactatagggaggaggtagctggggttgacec24CDTaatacgactcactatagggccctagagtgctccggctattaatacgactcactatagggcgctctccttcgctgttc , Sec16ttaatacgactcactatagggagagccagaggatcagcatcttaatacgactcactatagggagagcgatcccacagcagtc , Sec23Ttaatacgactcactataggggtgcaggatatgctcggaatttaatacgactcactataggggtggagctgggattcaatgt , Sar1Ttaatacgactcactatagggatgttcacttgggactggttcttaatacgactcactatagggagaatctctcgagcccacttcaa The DNA fragments were used for in vitro transcription using the T7 Megascript kit ( AMBION ) to generate the dsRNAs used for RNAi . The efficiency of Sec24AB and CD depletion was estimated by transfecting mCherry-Sec24AB in Sec24AB-depleted cells and Sec24CD-GFP to Sec24CD-depleted cells and comparing the level of transfection ( number of cells expressing the fluorescent protein ) to this of non-depleted cells . In a typical experiment , 24 . 1 ± 2 . 0% non-depleted cells were transfected with mCherry-Sec24AB vs 2 . 5 ± 1% in Sec24AB-depleted cells , and 27 . 3 ± 2 . 4% non-depleted cells were transfected with GFP-Sec24CD vs 2 . 6 ± 0 . 7% in Sec24CD-depleted cells , showing that the 90% of the cells are depleted . The depletion of Sec23 and Sec16 were also 90% ( as measured by Western blot , not shown ) . S2 cells were plated on glass coverslips , treated , fixed in 4% PFA in PBS for and processed for Immunofluorescence as previously described ( Kondylis and Rabouille , 2003 ) . Alternatively ( as indicated for specific antibodies and dyes ) , the cells were fixed in ice-cold methanol for 5 min washed with PBS and processed for IF as above . Samples were viewed under a Leica SPE confocal microscope using a 63× lens and 1 . 5–3× zoom . 17–22 confocal planes are projected to image the whole cells . IEM was performed as described previously ( Kondylis and Rabouille , 2003; van Donselaar et al . , 2007 ) . To monitor Delta transport through the secretory pathway in starved Delta-S2 cells ( Kondylis and Rabouille , 2003 ) . Delta expression was induced by adding 1 mM CuSO4 to Schneider's medium for 30 min before incubating the cells in KRB or further in Schneider's ( control ) ( Figure 1A ) . To monitor Delta transport upon reversion , cells were incubated 4 hr in Schneider's or KRB . After 4 hr , the media were changed to Schneider's supplemented with 1 mM CuSO4 for 10–120 min ( Figure 8B , C ) . To monitor Delta transport in Delta S2-depleted cells , 0 . 75 million cells were mock- ( dsGFP ) , Sec24AB- , and Sec23-depleted for 5 days in a 6-well plate cell . Cells were then were split in 2 and plated on glass coverslips in a 12-well plate . They were allowed to attach for 1 hr and the media were changed for either Schneider's or KRB . After 4 hr , the media were changed to Schneider's supplemented with 1 mM CuSO4 for 90 min ( Figure 9C ) . Cells were fixed and processed for IF using Delta antibody . Transport efficiency was calculated as percentage of cells expressing Delta at the plasma membrane over the total number of cells expressing Delta . Between 30 and 60 cells were analyzed per condition . S2 cells transiently expressing Drosophila Atg5-V5 and mouse GFP-Atg8 were incubated for increasing length of time in Schneider's supplemented or not with rapamycin and in KRB supplemented or not with wortmannin and bafilomycin . The percentage of cells showing Atg5-V5 punctae was determined . Experiments were done in triplicate . Time lapse of Sec body formation and disassembly was performed on S2 cells stably expressing GFP-Sec23 . For Sec body formation , cells were incubated in KRB ( t = 0 ) at 26°C . For Sec body disassembly , cells were starved for 4 hr and further incubated in Schneider's ( t = 0 ) at 26°C . Cells were viewed with a Leica AF7000 Fluorescence microscope . 10 z planes with a z step of 0 . 7 μm of were recorded every 3 min . The FRAP experiments were performed on cells expressing GFP-Sec23 , ΔNC1-Sec16-sfGFP , or FMR1-sfGFP for 1 . 5 hr ( expression induced with CuSO4 ) followed by incubation in Schneider's for 1 hr and starvation in KRB for 4 hr . Sec bodies and Stress Granules were entirely or partially ( half ) photobleached using a 488 nm laser at 100% laser power for 750 msec . FRAP was recorded using a PerkinElmer Ultraview VoX spinning disk microscope with the volocity software . Fluorescence recovery was recorded every 10 ms for the first 14 s after bleaching , and thereafter every 10 s for 2 min . The amount of LCS was determined for each protein and isoform annotated in FlyBase release FB2014_02 using SEG ( ftp://ftp . ncbi . nih . gov/pub/seg/ ) ( Wootton and Federhen , 1996 ) . The LCS content of each protein was tested for enrichment by a hypergeometric test against the whole proteome . For the proteins that have multiple isoforms the longest isoform was chosen for comparison . Cells were lysed in 50 mM Tris–HCl pH 7 . 5 , 150 mM NaCl , 1% Triton X‐100 , 50 mM NaF , 1 mM Na3VO4 , 25 mM Na2‐β‐glycerophosphate supplemented with a protease inhibitors tablet ( Roche ) . The lysate were cleared by centrifugation 4°C for 15 min at 14 , 000 rpm , and proteins were separated on SDS-PAGE followed by western blot . HEK-293T , MEFs , and COS7 cells were cultured in standard DMEM and incubated in KRB supplemented with 100 nM bafilomycin for 7 hr ( up to 16 hr for Cos cells ) to mimic the optimal conditions found for S2 cells . 0 . 75 million cells were non- , mock- ( dsGFP ) , Sec24AB , Sec24CD , and Sec24AB and CD depleted for 5 days in a 6-well plate . They proliferated to reach 2 . 5 , 1 . 9 , 1 . 8 , 1 . 85 , and 1 . 9 million , respectively and this was set at 100% ( t = 0 ) . These cells were then either starved in KRB for 4 hr or further incubated in Schneider's , their number monitored and expressed as a percentage of t = 0 . The medium was changed to Schneider's and their proliferation monitored further up to 16 hr . Experiments were performed in triplicates . Oregon R+ virgin females were fattened on standard food supplemented with yeast for 3 days . They were subsequently either dissected to harvest the ovaries for the ex vivo treatments ( incubation in Schneider's and KRB for 4 hr ) or transferred to humidified empty vials for 36 hr before dissection . IF were performed as described in the reference Giuliani et al . ( 2014 ) . Three independent experiments were performed for quantification of the Sec body phenotype as scored by immunofluorescence . At least three fields were analyzed comprising at least 100 cells per condition . Averages and standard deviations reflect variation throughout the experiments . For Sec bodies we considered cells with at least one large , round ( >0 . 5 μm ) structure as exhibiting Sec bodies . Cells with smaller round structures and/or haze were considered intermediate . For all measurements p-values were calculated with Excel . Sec body diameter was measured using the Leica LAS software . At least 35 cells were analyzed per condition , in each of which all fluorescent foci ( at least 500 ) were measured . Distribution curves were made using Excel . | Proteins are needed by living cells to perform vital tasks and are made from building blocks called amino-acids . However , if a cell is starved of amino-acids , protein assembly comes to a halt , and if cells are deprived of amino acids for a long time , the cell may die . To survive short periods of amino-acid starvation , the cell has developed many protective mechanisms . For example , it can start to break down existing proteins , allowing the cell to scavenge and reuse the amino-acids to make other proteins that are more important for short-term survival . The cell may also temporarily halt certain processes: for example , newly constructed proteins may no longer be transported from the cell structure where they are made—called the endoplasmic reticulum—to their final destinations in the cell . However , the protein transport apparatus is also made of proteins and needs to be protected from being broken down so that once starvation ends , the cell can more quickly return to normal working order . Zacharogianni et al . identify a strategy cells use to store and protect part of their protein transport apparatus during times of stress . Starving fruit fly cells of amino-acids causes the cells to form protective stress assemblies incorporating the proteins associated with the ‘exit sites’ that release proteins from the endoplasmic reticulum . These assemblies are called Sec bodies , and when amino-acid starvation ends , these bodies release the exit site components unharmed . This allows the cell to quickly resume protein transport and so speeds the cell's recovery . If the Sec bodies do not form , the cells are more likely to die during amino-acid starvation . The Sec bodies are distinct from previously identified stress assemblies that form in the cell during stress , but they share features with them , such as being liquid droplets . Some of these assemblies have been linked to degenerative diseases like amyotrophic lateral sclerosis ( ALS ) . Further research will be necessary to determine if there are any similar harmful side effects associated with the formation of Sec bodies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | A stress assembly that confers cell viability by preserving ERES components during amino-acid starvation |
MicroRNAs ( miRNAs ) are important regulators of stem and progenitor cell functions . We previously reported that miR-142 and miR-150 are upregulated in human breast cancer stem cells ( BCSCs ) as compared to the non-tumorigenic breast cancer cells . In this study , we report that miR-142 efficiently recruits the APC mRNA to an RNA-induced silencing complex , activates the canonical WNT signaling pathway in an APC-suppression dependent manner , and activates the expression of miR-150 . Enforced expression of miR-142 or miR-150 in normal mouse mammary stem cells resulted in the regeneration of hyperproliferative mammary glands in vivo . Knockdown of endogenous miR-142 effectively suppressed organoid formation by BCSCs and slowed tumor growth initiated by human BCSCs in vivo . These results suggest that in some tumors , miR-142 regulates the properties of BCSCs at least in part by activating the WNT signaling pathway and miR-150 expression .
MicroRNAs ( miRNAs ) are evolutionally conserved small non-coding RNAs that regulate the translation of mRNAs . They are recruited to an RNA-induced silencing complex ( RISC ) and bind to the seed sequence within the 3′ untranslated region ( UTR ) of target mRNAs , leading to destabilization and/or translational suppression of the target mRNAs ( Bartel , 2009 ) . The immunopurification ( IP ) of Argonaute ( Ago ) , a central component of the RISC in the human and mouse , followed by microarray analyses ( Ago IP/microarray method ) makes it possible to isolate any Ago-associated miRNAs and mRNAs without relying on the mechanism of regulation ( i . e . mRNA decay or translational suppression ) , or sequence conservation , enabling a comprehensive identification of the miRNA-target genes in an unbiased manner . This provides quantitative information about the mRNAs that are regulated by miRNAs ( Hendrickson et al . , 2008 , 2009 ) . miRNAs are able to regulate the expression of hundreds of target mRNAs simultaneously and control a variety of cell functions including cell proliferation , stem cell maintenance , and differentiation ( Lewis et al . , 2005 ) . We previously identified a human breast cancer stem cell ( BCSC ) population ( a CD44+ CD24−/low lineage− population of human breast cancer cells ) that in many human breast tumors is enriched for the ability to drive tumor formation in a mouse xenograft model as compared to the remaining non-tumorigenic cancer cells ( NTG cells ) within the same breast tumor ( Al-Hajj et al . , 2003 ) . Comprehensive analyses of the expression profile of 466 miRNAs revealed that 37 miRNAs are differentially expressed between the human BCSCs and NTG cells ( Shimono et al . , 2009 ) . Among them , both miR-200c and miR-183 are downregulated in the human BCSCs and suppress the protein expression of the stem cell self-renewal gene , BMI1 , and miR-200c suppresses the protein expression of the EMT regulator ZEB1 ( Shimono et al . , 2009; Wellner et al . , 2009 ) . Enforced expression of miR-200c can strongly suppress the tumor formation driven by human BCSCs and the mammary ducts formation by normal mammary stem cells in vivo , suggesting that miR-200c is a regulator of normal mammary and BCSCs . On the other hand , the expression of miRNAs , such as miR-142 , miR-150 , and miR-155 , are upregulated in human BCSCs ( Shimono et al . , 2009 ) . Among them , miR-155 was originally identified as a product of the oncogenic BIC gene locus in B cell lymphoma ( Eis et al . , 2005 ) . Abnormal proliferation and myelodysplasia are seen when miR-155 expression is sustained in the blood system ( O'Connell et al . , 2008 ) . Furthermore , miR-155 functions as an oncogenic miRNA in various cancers , including leukemia and breast cancers ( Czyzyk-Krzeska and Zhang , 2013 ) . Dysregulation of miR-142 and miR-150 are reported in leukemia , gastric and lung cancers , but their roles in breast cancer or BCSCs are not elucidated . miR-142 is involved in hematopoiesis , immune responses , and T cell differentiation ( Chen and Lodish , 2005; Ramkissoon et al . , 2006; Wu et al . , 2007; Visone et al . , 2009 ) and is upregulated in bronchoalveolar stem cells ( Qian et al . , 2008 ) , suggesting it might have a role in the regulation of tissue stem cells . miR-142 is upregulated in human T-cell acute lymphoblastic leukemia ( Lv et al . , 2012 ) but is downregulated in acute myeloid leukemia ( Wang et al . , 2012 ) . miR-150 is upregulated in adult T-cell leukemia cells ( Bellon et al . , 2009 ) , gastric cancers ( Wu et al . , 2010 ) , and lung cancers ( Zhang et al . , 2013 ) . The canonical Wnt pathway signal is transduced by β-catenin and plays a critical role in many adult stem cells , including those of the breast and intestine ( Reya and Clevers , 2005; Zeng and Nusse , 2010 ) . The fact that some cancer cells share the extended self-renewal ability with normal stem cells , and that the canonical Wnt signaling pathway is implicated in both stem cell self-renewal and cancer suggests that normal physiological regulators of stem cell functions might be ‘hijacked’ in cancer ( Reya and Clevers , 2005 ) . A variety of putative transcriptional targets of the canonical Wnt signaling pathway , such as c-Myc and cyclin D1 , are identified ( He et al . , 1998; Shtutman et al . , 1999 ) . Adenomatous polyposis coli ( APC ) is a component of the destruction complex that destabilizes β-catenin and suppresses the activity of the canonical WNT signaling pathway . In human colon cancers , mutations in the APC gene are the most commonly known acquired genetic change for the aberrant activation of the canonical WNT signaling pathway during the tumor development and progression ( Kinzler et al . , 1991; Nishisho et al . , 1991; Cottrell et al . , 1992 ) . In a model for the stepwise progression of the colon tumorigenesis , APC gene mutations play an important role in the initiation step followed by successive mutations in other genes , including K-Ras and p53 ( Fearon and Vogelstein , 1990 ) . During clonal progression , dysregulation of downstream β-catenin targets , such as c-Myc , can relieve colon cancer cells of their dependence on β-catenin signaling ( van de Wetering et al . , 2002 ) . The expression of APC is not limited to the intestine but is widely observed in many other tissues , including lung , liver , kidney , and mammary tissue . However , the role of the suppression of APC and the activation of the canonical WNT signaling in the tumor initiation process of the tissues other than the colon largely remains unknown , because APC mutations are less frequent in tumors originating from these tissues . For example , recent data from the TCGA reveals an ∼2% incidence of APC mutations in breast cancer ( the TCGA Research Network: http://cancergenome . nih . gov/ ) . Dampening of APC inhibition of β-catenin appears to be more often due to promoter methylation ( 36–54% ) and loss of heterozygosity ( LOH ) ( 23% ) than to somatically acquired APC mutations ( ∼2% ) in human breast cancers ( Sarrio et al . , 2003; Jin et al . , 2001; Banerji et al . , 2012; The cancer genome atlas network 2012 ) , but the presence of the APC promoter methylation and LOH are independent of tumor size or stage ( Jin et al . , 2001; Sarrio et al . , 2003 ) . The APC mutations in breast cancers are typically found at sites distinct from the APC mutation cluster region in colorectal cancers and are much more frequently seen in advanced than early stage breast cancers ( Furuuchi et al . , 2000 ) . On the other hand , APC is targeted by the miRNAs , such as miR-27 , miR-155 , and miR-142 ( Lu et al . , 2009; Wang and Xu , 2010; Liu et al . , 2011; Hu et al . , 2013 ) . These observations suggest that in addition to LOH , promoter methylation , and the APC mutations , miRNAs that target APC may regulate the aberrant activation of the canonical WNT signaling pathway for the initiation of human breast cancers , the enhancement of niche independence , and the aberrant proliferation of the human BCSCs . In this study , we studied the roles of miR-142 and miR-150 , two of the miRNAs that are upregulated in the human BCSCs , in the enhancement of the canonical WNT signaling pathway and in the regulation of human BCSCs . We employed the Ago IP/microarray method to show the relevance of the targeting of APC by miR-142 , but not by miR-150 , and identified various target mRNAs that were efficiently recruited to RISC by these miRNAs . Upregulation of miR-142 enhanced the canonical WNT signaling pathway and transactivated the expression of miR-150 . Both microRNAs had the ability to induce hyperproliferation in mammary tissue . Finally , we show that knockdown of miR-142 reduced the clonogenicity of BCSCs in vitro and tumor growth in vivo . Our results provide the insights into the roles and mechanisms of two of the upregulated miRNAs in human BCSCs .
miRNAs suppress protein production by recruiting target mRNAs to the RISC where they associate with Ago , a core component of the RISC . To explore the possible target mRNAs of miR-142 and miR-150 which were more highly expressed in human BCSCs than in the NTG cells , we performed Ago IP/microarray to comprehensively identify mRNAs that were recruited to Ago by miR-142 and miR-150 . Lysates from human embryonal kidney ( HEK ) 293T cells transfected with or without the precursor for miR-142 or miR-150 were immunopurified using an anti-Ago antibody . Amplified RNAs from the Ago-immunopurified samples were labeled and hybridized to the Human Exonic Evidence Based Oligonucleotide ( HEEBO ) microarrays ( Klapholz-Brown et al . , 2007; Hendrickson et al . , 2008 ) . We identified dozens of potential targets of miR-142 and miR-150 ( Figure 1A and Figure 1A—source data 1–4 ) . The types of miRNA seed match ( i . e . the association of target mRNA with the region centered on the nucleotides 2–7 of the 5′ region of miRNAs ) correlate with the efficiency of mRNA targeting by miRNAs ( Bartel , 2009 ) . The efficiency of the miR142-dependent and miR-150-dependent recruitment of specific mRNAs to the RISC complex correlated well with the types of miRNA seed matches: mRNAs with the 8mer match were most efficiently enriched , followed by those with the 7mer-m8 , 7mer-1A , and 6mer sites ( Figure 1B and Figure 1—figure supplement 1 ) . Although both miR-142 and miR-150 were predicted to target the APC mRNA by a computer algorithm , the results of the Ago IP/microarray experiments showed that the APC mRNA was efficiently recruited to RISC by miR-142 , and that it was hardly recruited to RISC by miR-150 ( Figure 1A—source data 1–4 ) . Among the genes associated with the activity of the WNT signaling pathway , we found that the APC mRNA was strongly recruited to Ago by miR-142 ( Figure 1C ) . In the parallel experiments , amplified RNAs from the whole cell lysate from HEK293T cells transfected with or without the miR-142 precursor were labeled and hybridized to the HEEBO microarrays . The results showed that the abundance of the APC mRNA was consistently reduced in the miR-142-transfected HEK293T cells ( Supplementary file 1 ) . 10 . 7554/eLife . 01977 . 003Figure 1 . Recruitment of the APC mRNA to Ago by miR-142 or miR-150 . ( A ) Unsupervised hierarchal cluster of AgoIP/microarray from HEK293T cells transfected with mock , miR-142- or miR-150-precursor . The lysates of HEK293T cells transfected with mock , or 30 nM miR-142- or miR-150- precursor were immunopurified by an anti-Ago antibody . RNA was isolated from the immunopurified lysate and amplified for the HEEBO microarray analyses . Rows correspond to the putative miR-142 and miR-150 targets ( local false discovery rate ( FDR ) 1% ) , and columns represent individual experimental samples . ( B ) Cumulative distribution of the change for the Ago IP mRNAs classified by the types of the seed matches in the 3′ UTR ( Bartel , 2009 ) . Overall efficiency of Ago IP enrichment is 8mer > 7mer-m8 > 7mer-1A > 6mer-n2-7 ( nucleotides 2–7 ) > 6mer-n3-8 ( nucleotides 3–8 ) in the HEK293T cells transfected with miR-142 precursor . ( C ) Ago IP enrichment of mRNAs . The mRNA enrichment in the Ago IP samples of the miR-142-expressing HEK293T cells over those of the mock transfected cells . The results are derived from two independent transfections . The results for APC , AXIN1 , and GSK3B enrichment are presented . The data are mean ± standard deviation ( SD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 00310 . 7554/eLife . 01977 . 004Figure 1—source data 1 . The result of the Ago-IP microarray experiment using the miR-142-expressing cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 00410 . 7554/eLife . 01977 . 005Figure 1—source data 2 . The result of the second Ago-IP microarray experiment using the miR-142 expressing cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 00510 . 7554/eLife . 01977 . 006Figure 1—source data 3 . The result of the Ago-IP microarray experiment using the miR-150 expressing cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 00610 . 7554/eLife . 01977 . 007Figure 1—source data 4 . The result of the second Ago-IP microarray experiment using the miR-150 expressing cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 00710 . 7554/eLife . 01977 . 008Figure 1—figure supplement 1 . Cumulative distribution of the change for the Ago IP mRNAs classified by the types of the seed matches in the 3′ UTR ( Bartel , 2009 ) . Overall efficiency of Ago IP enrichment is 8mer > 7mer-m8 > 7mer-1A > 6mer-n2-7 ( nucleotides 2–7 ) > 6mer-n3-8 ( nucleotides 3–8 ) in the HEK293T cells transfected with miR-150 precursor . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 008 We tested the ability of miR-142 to regulate the 3' UTR of the APC mRNA using a luciferase reporter assay . The 3′ UTR of the APC mRNA , which contained a predicted target site for miR-142-3p was cloned into the pGL3-MC vector downstream of a luciferase minigene ( Figure 2A ) . HEK293T cells , which very weakly express miR-142 , were co-transfected with the pGL3 luciferase-APC 3′UTR vector , or the control luciferase vector , pRL-TK Renilla luciferase vector , and miRNA precursors . We observed that when the 3′ UTR of the APC mRNA was included in the luciferase transcript , co-transfection of the miR-142 precursor suppressed the luciferase activity by 37% ( Figure 2B ) . Mutations of the two nucleotides within the predicted target site for miR-142-3p in the 3′ UTR of the APC mRNA significantly weakened the ability of miR-142 to suppress the luciferase activity ( Figure 2A , B ) , suggesting that miR-142 specifically targets the predicted seed sequence in the 3′ UTR of the APC mRNA to suppress the translation of the APC mRNA . 10 . 7554/eLife . 01977 . 009Figure 2 . Targeting of APC by miR-142 . ( A ) Schematic representation of the predicted miR-142 target site within the 3′ UTR of APC . The predicted target site for miR-142-3p is located at the proximal portion of the APC 3′ UTR . Two nucleotides complementary to the seed sequence ( the nucleotides 2–7 of miRNA ) of miR-142-3p were mutated in the APC mutant plasmid . The number indicates the position of the nucleotides in the reference wild-type sequence of APC ( NM_000038 . 5 ) . ( B ) Activity of luciferase gene linked to the 3′ UTR of APC . The pGL3 luciferase reporter plasmids with the wild-type or mutated 3′ UTR of APC were transiently transfected into HEK293T cells along with 25 nM miR-142 precursor or negative control precursor . Co-transfected Renilla luciferase reporter was used for normalization . Luciferase activities were measured after 48 hr . The mean of the results from the cells transfected by pGL3-Control vector with control precursor was set at 100% . The data are mean ± SD ( n = 3 , ***p < 0 . 005 ) . ( C ) miR-142 suppressed endogenous APC expression . miR-142-expressing HEK293T cells were cultured for 6 days , and APC protein level was analyzed by Western blotting . The intensities of the bands for APC and β-actin were measured by ImageJ software . Difference in the APC protein level between the lysate of the control precursor transfected cells and that of miR-142 transfected cells was statistically significant ( n = 4 , **p < 0 . 01 ) . ( D ) Suppression of endogenous APC by miR-142 in the breast cancer cells . miR-142-expressing breast cancer MCF7 and MDA-MB-231 cells were cultured for 2 days , and the APC protein level was analyzed by Western blotting . ( E ) Elevation of the endogenous APC protein level by miR-142 knockdown . HEK293T cells and MDA-MB-231 cells were infected with the anti-miR-142-3p-expressing lentivirus , and GFP-positive cells were sorted by a cell-sorter . APC protein level was analyzed by Western blotting . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 009 Then we tested the ability of miR-142 to regulate expression of endogenous APC protein . Western blot analyses showed that , after 6 days , the protein level of APC was decreased in the HEK293T cells expressing miR-142 when compared to the control precursor transfected cells ( Figure 2C ) . To confirm our findings , breast cancer cells , such as MCF7 and MDA-MB-231 cells , were transfected with the miR-142 precursor . We found that the protein level of APC was reduced in both MCF7 and MDA-MB-231 cells ( Figure 2D ) . Next , the expression of miR-142-3p was knocked down using an anti-miR-142-3p-expressing lentivirus in the HEK293T , MDA-MB-231 , and MCF7 cells which expressed miR-142 at low levels . We found that the protein level of APC was elevated in HEK293T , MDA-MB-231 , and MCF7 cells transduced with the anti-miR-142-3p vector ( Figure 2E and data not shown ) . These results suggest that miR-142 effectively targets the APC mRNA and reduces the protein level of APC in the cells including the human breast cancer cells . Inhibition of APC expression by miR-142 could , in principle , diminish the dependence of the canonical WNT signaling pathway on extrinsic WNT signals . To test this possibility , we evaluated the effects of miR-142 on the β-catenin-dependent transcription of the luciferase gene using cells transfected with a TOPFlash luciferase plasmid , which contained the β-catenin/TCF binding sites in the promoter region of a luciferase gene . As a control , cells were transfected with an otherwise identical FOPFlash plasmid , in which the β-catenin/TCF binding sites were mutated . The ratio of the measured luciferase activity of the TOPFlash plasmid to that of the FOPFlash plasmid provides a measure of the activity of the WNT/β-catenin signaling in the cells ( van de Wetering et al . , 1997 ) . HEK293T cells were transfected with the miR-142 precursor and incubated with 20 mM lithium chloride ( LiCl ) for 6 hr to stimulate the canonical WNT signaling pathway ( by inhibiting the activity of the glycogen synthase kinase 3 ) ( van Noort et al . , 2002 ) . We found that the normalized TOPFlash/FOPFlash ratio of the luciferase activities increased by 364% in the miR-142 transfected HEK293T cells compared to that in the control precursor transfected cells ( Figure 3A ) . In accordance with this result , the elevation of the normalized TOPFlash/FOPFlash value in the miR-142 transfected cells was observed by the stimulation with Wnt3A ( Figure 3A ) . To further confirm that miR-142 activated the canonical WNT signaling pathway by targeting APC , we co-transfected the APC expression plasmid , pCMV-Neo-APC , that contains the complete coding sequence and the partial 3′ UTR sequence of the APC mRNA , along with the TOPFlash or FOPFlash vectors and the miRNA precursors . The miR-142-3p target site within the pCMV-Neo-APC plasmid sequence was mutated to produce the pCMV-APC mutant plasmid ( Figure 3B ) . The co-transfection of the control plasmid , pCMV-Neo-Control , or the pCMV-Neo-APC plasmid that contains miR-142-3p targeted sequence within the 3′ UTR region , did not affect the ability of miR-142 to elevate the normalized TOPFlash/FOPFlash value . On the contrary , co-transfection of pCMV-Neo-APC mutant plasmid significantly suppressed the normalized TOPFlash/FOPFlash value to an intensity comparable to that of the control miRNA precursor transfected cells ( Figure 3C ) . These results suggest that the activation of canonical WNT signaling pathway by miR-142 is mostly mediated by the ability of miR-142 to reduce the protein level of APC . 10 . 7554/eLife . 01977 . 010Figure 3 . Activation of the canonical WNT signaling pathway by miR-142 . ( A ) Activation of the canonical WNT signaling pathway by miR-142 . HEK293T cells were transfected with 25 nM miR-142 precursor along with TOPFlash or FOPFlash vector ( with or without TCF binding sites ) . Cells were stimulated with 20 mM LiCl ( left ) or with Wnt3A conditioned medium ( right ) for 6 hr . The canonical WNT signaling pathway activities were measured by dividing a normalized TOPFlash value by a normalized FOPFlash value . The mean of the results from the cells transfected with control precursor was set at 100% . The data are mean ± SD ( n = 3 , **p < 0 . 01 , ***p < 0 . 005 ) . ( B ) Schematic representation of the pCMV-Neo-APC expression vector that codes the full-length APC sequence together with wild-type or mutated miR-142 targeted site located at the 3′ UTR . Two nucleotides complementary to the seed sequence of miR-142-3p were mutated in the pCMV-Neo-APC mutant plasmid . The number indicates the position of the nucleotides in the reference wild-type sequence of APC ( NM_000038 . 5 ) . ( C ) miR-142 targets APC to activate the canonical WNT signaling pathway . HEK293T cells were transfected with 25 nM miR-142 precursor along with the TOPFlash or FOPFlash vector and the pCMV-Neo-APC expression vector or its mutant vector . Cells were stimulated with 20 mM LiCl for 6 hr . The activities of the canonical WNT signaling pathway were measured as described in ( A ) . The mean of the results from the cells transfected with control precursor was set at 100% . The data are mean ± SD ( n = 3 , *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 010 The organoid culture system of the mammary tumor cells allows the in vitro formation of the organoids that maintain cellular hierarchy of the original tissue from which it derived ( Sato et al . , 2009; Zeng and Nusse , 2010 ) . In this system , the tissue derived stem cells were embedded in Matrigel and cultured in the media that contained a cocktail of the growth factors , including ones that stimulate the WNT signaling pathways . We and others have previously shown that the MLTV-Wnt-1 tumor contains a CSC population ( Shackleton et al . , 2006; Cho et al . , 2008 ) . Because miR-142 is upregulated in the BCSCs and stimulates the canonical WNT signaling pathway that is an important regulator of the stem cell properties in mammary tissues , we tested the ability of miR-142 to regulate the organoid formation of BCSCs . The sorted murine mammary CSCs of the MLTV-Wnt-1 tumor were infected with the control lentivirus or the anti-miR-142-3p-expressing lentivirus that suppressed the expression of miR-142-3p and expressed GFP . The GFP-positive cells infected with the control lentivirus formed round-shaped organoids in 7 days ( Figure 4—figure supplement 1A ) . In contrast , most of the cells infected with the anti-miR-142-3p-expressing lentiviruses failed to form the round-shaped organoids ( Figure 4—figure supplement 1A ) . The number of the GFP positive organoids was significantly decreased when the mammary CSCs were infected with the anti-miR-142-3p-expressing lentivirus ( Figure 4—figure supplement 1B ) . These results suggest that miR-142-3p is an important regulator of the organoid formation ability in murine mammary CSCs . To confirm that these findings are applicable to the human BCSCs , we infected the human breast cancer cells with the anti-miR-142-3p-expressing lentiviruses and evaluated the ability to form the organoids derived from the human BCSCs . BCSCs of the human breast cancer xenograft tumors were infected with the control lentivirus or the anti-miR-142-3p-expressing lentiviruses . The GFP positive cells infected with the control lentivirus formed round-shaped organoids in 10 days ( Figure 4A ) . In contrast , most of the cells infected with the anti-miR-142-3p-expressing lentiviruses failed to form the round-shaped organoids ( Figure 4A ) . The number of the GFP positive organoids was significantly decreased when the human BCSCs were infected with the anti-miR-142-3p-expressing lentivirus ( Figure 4B ) . Then we analyzed the effects of miR-142 on the cell proliferation and apoptosis of the cancer cells in the organoids . The organoids were incubated with 5-ethynyl-2´-deoxyuridine ( EdU ) , and the percentage of the EdU positive cells was analyzed by flow cytometry . We found that the percentage of the EdU-positive cells in the GFP-positive organoids was significantly reduced when the human BCSCs were infected with the anti-miR-142-3p-expressing lentivirus ( Figure 4C ) . Then we analyzed the percentage of apoptotic cells in the GFP-positive organoids by flow cytometry . We found that the percentage of the annexin V-positive cells in the GFP positive organoids was significantly increased when the human BCSCs were infected with the anti-miR-142-3p-expressing lentivirus ( Figure 4D ) . These results suggest that miR-142 critically affects the clonogenic properties of the BCSCs by regulating the proliferation and apoptosis of the breast cancer cells . 10 . 7554/eLife . 01977 . 011Figure 4 . Suppression of the organid formation by the inhibition of miR-142 in BCSCs . ( A ) Representative images of the organoids from the human BCSCs infected with control or anti-miR-142-3p-expressing lentiviruses and cultured for 10 days . The upper panels are the phase-contrast images of the organoids , and the lower panels are the fluorescent microscopic images for the detection of GFP . Bars , 100 μm . ( B ) The number of organoids formed by the human BCSCs infected with control or anti-miR-142-3p-expressing lentiviruses . The data are mean ± SD ( n = 10 , *p < 0 . 05 ) . ( C ) Percentage of the EdU-positive cells among the GFP-positive breast cancer cells in the organoids formed by the human BCSCs infected with control or anti-miR-142-3p-expressing lentiviruses . The data are mean ± SD ( n = 4 , *p < 0 . 05 ) . ( D ) Percentage of the annexin V-positive cells among the GFP-positive breast cancer cells in the organoids formed by the human BCSCs infected with control or anti-miR-142-3p-expressing lentiviruses . The data are mean ± SD ( n = 4 , *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 01110 . 7554/eLife . 01977 . 012Figure 4—figure supplement 1 . ( A ) Representative images of the organoids from the mouse MMTV-Wnt1 breast cancer cells infected with control or anti-miR-142-3p-expressing lentiviruses and cultured for 7 days . The upper panels are the phase-contrast images of the organoids , and the lower panels are the fluorescent microscopic images for the detection of GFP . Bars , 100 μm . ( B ) The number of organoids formed by the mouse MMTV-Wnt1 breast cancer cells infected with control or anti-miR-142-3p-expressing lentiviruses . The data are mean ± SD ( n = 3 , *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 012 Both miR-142 and miR-150 are more highly expressed in BCSCs relative to the non-tumorigenic breast cancer cells ( Shimono et al . , 2009 ) . Because the promoter region of miR-150 contains a potential β-catenin/TCF transcription factor binding site ( Figure 5A ) , we investigated whether miR-150 expression is regulated by the canonical WNT signaling pathway . HEK293T cells were stimulated with 20 mM LiCl for 7 hr , then crosslinked with formaldehyde , and chromatin immunoprecipitation was performed either with an antibody against β-catenin or , as a control , with mouse IgG . The anti-β-catenin antibody selectively enriched DNA fragments containing the potential β-catenin/TCF binding site within the promoter of the miR-150 precursor ( Figure 5B ) . Then , using semi-quantitative real-time PCR , we found that the expression of miR-142 was able to upregulate the expression of miR-150 in human breast cancer MDA-MB-231 cells ( Figure 5C ) . Moreover , inhibition of the canonical WNT signaling pathway using the siRNA against β-catenin reduced the expression of miR-150 in the miR-142-expressing MDA-MB-231 cells ( Figure 5D ) . We further confirmed that the inhibition of miR-142 in human breast cancer organoids formed by human BCSCs infected with the anti-miR-142-3p-expressing lentiviruses decreased the expression of miR-150 ( Figure 5E ) . Taken together , these results indicate that the upregulation of miR-142 induces the expression of miR-150 at least partially through the canonical WNT signaling pathway . 10 . 7554/eLife . 01977 . 013Figure 5 . Enhancement of miR-150 transcription by the canonical WNT signaling pathway . ( A ) Schematic representation of the potential β-catenin/TCF binding site within the flanking genomic sequence of the miR-150 precursor . Box: potential β-catenin/TCF binding site ( WWCAAWG/CWTTGWW ) and its sequence . Box with an arrow: position of the miR-150 precursor and direction of transcription . Black arrows: relative positions of the PCR primers for the chromatin IP analyses . ( B ) Chromatin IP for a potential β-catenin/TCF binding site . Lysate of cross-linked HEK293T cells was immunoprecipitated by an anti-β-catenin antibody or mouse IgG . A putative β-catenin/TCF binding site was amplified by PCR . The template for input was purified from the 1% of total cell lysate . The β-catenin/TCF binding site for c-MYC was amplified as a positive control , and the genomic sequence for β-actin was amplified as a negative control . ( C ) miR-142 induced the transcription of miR-150 . The amount of miR-150 in the MDA-MB-231 cells transfected with a miR-142-expressing plasmid was analyzed by quantitative real-time PCR . Each circle represents one experiment . Bars indicate median . Differences of the amount of miR-150 between the miR-142-expressing and the control MDA-MB-231 cells were statistically significant ( *p < 0 . 05 ) . ( D ) Decrease of the miR-150 transcription by β-catenin knockdown . β-catenin was knocked down by the siRNA against CTNNB1 in MDA-MB-231 cells expressing miR-142 . Each circle represents one experiment . Bars indicate median . The amount of miR-150 was analyzed by quantitative real-time PCR . Difference in the amount of miR-150 between the cells transfected with a siRNA against CTNNB1 and those transfected with a control siRNA was statistically significant ( *p < 0 . 05 ) . ( E ) Decrease of miR-150 expression by the miR-142 inhibition in human BCSCs . The human BCSCs derived from the human breast cancer xenograft were infected with the anti-miR-142-expressing or control letiviruses and incubated in an organoid culture medium for 48 hr . The amount of miR-150 was analyzed by quantitative real-time PCR . Each circle represents one experiment . Bars indicate median ( *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 013 To clarify the effect of the upregulation of miR-142 and miR-150 on the mammary tissue in vivo , murine mammary epithelial cells were transduced with the lentiviruses expressing miR-142 or miR-150 and the regenerated mammary tissues were histologically analyzed . Expression of miR-142 or miR-150 in the lentivirus infected cells was confirmed by real-time PCR ( data not shown ) . We infected 5 × 104 lineage− murine mammary epithelial cells with the miR-142 or miR-150-expressing lentivirus and transplanted them into the cleared mammary fat pads of syngeneic mice . Non-infected and control lentivirus infected mammary cells were used as controls . Whole mount mammary tissue analyses and histological analyses were performed 8 weeks after transplantation . Transduction of the lentivirus into the mammary transplants was confirmed by the expression of ZsGreen . Overall , 7 out of 12 transplants with non-infected mammary cells and 7 out of 12 transplants with mammary cells infected with control lentivirus formed a mammary tree , suggesting that lentivirus infection was highly efficient and , by itself , did not perturb engraftment of mammary cells . Whole-mount staining and histological analysis of mammary fat pads injected with the control lentivirus-infected mammary cells showed the outgrowth of a normal-looking mammary tree structure consisting of single layers of myoepithelial and luminal cells ( Figure 6A , B ) . In contrast , mammary cells infected with the miR-142 lentivirus formed disorganized structures with multiple layers of cells ( Figure 6A , B ) . Histological analysis revealed that miR-142-expressing mammary cells formed clusters of cells with an abnormal appearance in which the normal mammary structure was disrupted because of the aberrant cell proliferation ( Figure 6B ) . Mammary cells infected with the miR-150-expressing lentivirus formed a hyperplastic mammary tree with extremely increased branching and thick mammary ducts ( Figure 6A , C ) . Histological analysis of the mammary trees showed that miR-150-expressing mammary cells formed thick hyperplastic ducts with multiple layers of mammary cells , with scattered small cavities among mammary cell layers ( Figure 6B , C ) . But unlike miR-142-expressing mammary cells , miR-150-expressing mammary cells had a normal cellular appearance . To confirm that both the dysplastic mammary tissue formed by the miR-142-expressing mammary cells and the hyperplastic mammary tissue formed by the miR-150-expressing mammary cells were both hyperproliferative , we immunostained these tissues with an anti-PCNA antibody . We found that the epithelial cells in the miR-142-expressing and miR-150-expressing mammary tissues were positively stained with an anti-PCNA antibody ( Figure 6D ) . In contrast , the mammary tissue regenerated by the control lentivirus transfected mammary cells was rarely stained with an anti-PCNA antibody . The results indicate that the enforced expression of either miR-142 or miR-150 induces hyperproliferation in the mammary tissues and that the phenotype of the miR-142-expressing mammary tissue is more severe and accompanied with abnormal morphology . 10 . 7554/eLife . 01977 . 014Figure 6 . Mammary dysplasia or hyperplasia formed by the mammary cells expressing miR-142 or miR-150 in vivo . ( A ) The whole-mount mammary fat pad staining . Murine mammary cells isolated from FVB/NJ mice were infected with the control lentivirus , or miR-142 or miR-150-expressing lentiviruses . The 5 × 104 infected mammary cells were transplanted into cleared mammary fat pads of the same strain weaning age female mice . Mammary duct outgrowth was analyzed 8 weeks later . The whole-mount mammary tissue was stained by a carmine alum staining solution . ( B ) Hematoxylin and Eosin staining of the mammary tissue formed by mammary cells infected with miR-142 or miR-150-expressing lentiviruses . Mammary duct outgrowth was analyzed 8 weeks after transplantation . ( C ) Degree of branching and number of mammary epithelial cell layers in the mammary tissue formed by mammary cells infected with miR-150 . Degree of branching was analyzed by counting the number of branches observed within a 50× field of the whole-mount mammary fat pad specimens . Number of mammary epithelial cell layers of the ducts was counted in the hematoxylin and eosin stained specimens ( **p < 0 . 01 , ***p < 0 . 005 ) . ( D ) Increase of the cell proliferation in the mammary tissues regenerated by the miR-142- or miR-150-expressing mammary cells . The tissues were stained with an anti-PCNA antibody followed by an Alexa Fluor 594-conjugated secondary antibody . Blue , DAPI; red , PCNA . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 01410 . 7554/eLife . 01977 . 015Figure 6—figure supplement 1 . Activation of β-catenin in the mammary tissue expressing miR-142 . Immunohistochemical staining of the mammary tissue formed by mammary cells expressing miR-142 . Tissue was stained with an anti-β-catenin or anti-active-β-catenin antibody followed by an Alexa Fluor 488-conjugated secondary antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 015 Next , we analyzed the expression and localization of β-catenin in the regenerated mammary tissues . While the signal for the β-catenin in mammary tissue formed by the control lentivirus-infected mammary cells was weakly detectable and its localization was mostly membranous , we observed cytoplasmic and nuclear localization of β-catenin and the positive staining of active β-catenin in the dysplastic cell clusters formed by the mammary cells infected with the miR-142-expressing lentivirus ( Figure 6—figure supplement 1 ) . Although the immunohistochemistry of β-catenin in general suffers from non-specific staining , we detected nuclear β-catenin in the miR-142-expressing mammary tissue . We speculate that nuclear β-catenin was detectable in these immunohistochemistry experiments because the WNT signaling pathway was highly overstimulated in the miR-142-expressing mammary tissue . These results suggest that miR-142 regulates the properties of mammary cells by activating the WNT signaling pathway and conferring the aberrant proliferative ability . To evaluate the role of miR-142 in the growth of human breast cancers initiated by the human BCSCs , we infected human BCSCs isolated from a patient-derived human breast cancer xenograft ( PDX ) with the anti-miR-142-3p-expressing lentivirus or the control lentivirus . Then , 1 × 104 of the infected BCSCs were injected into the mammary fat pad region of the NSG mice . The flow cytometric analyses revealed that 94% and 96% of the human BCSCs were infected with the control or the anti-miR-142-3p-expressing lentivirus , respectively , 2 days after the infection ( data not shown ) . The growth of the human breast cancer xenografts formed by the anti-miR-142-3p-expressing BCSCs was significantly slower than that of the control human breast cancer xenografts ( Figure 7A ) . We further confirmed that consistent with the results of the human breast cancer cell lines ( Figure 2 ) , the protein level of APC was elevated in the human breast cancer cells isolated from the anti-miR-142-3p-expressing breast cancer xenograft ( Figure 7B ) . These results suggest that the inhibition of miR-142-3p in human BCSCs from this patient elevated the expression of APC protein and suppressed the proliferation of the breast cancer cells in vivo . 10 . 7554/eLife . 01977 . 016Figure 7 . Suppression of the Tumor Growth Initiated by Human BCSCs Expressing the Anti-miR-142-3p in vivo . ( A ) CD44+ CD24-/low lineage− human BCSCs were isolated from an early passage human breast xenograft tumor and infected by the anti-miR-142-3p-expressing lentivirus or control lentivirus . Ten thousand infected cells were injected into the mammary fat pad region of immunodeficient NSG mice . Tumor growth was monitored for 2 months after injection . The data are mean ± SD ( n = 10 , * *p < 0 . 01 , ***p < 0 . 005 ) . ( B ) APC expression in the xenograft tumors was derived from the control or anti-miR-142-3p-expressing lentivirus infected BCSCs . Tumors were dissociated and APC expression was analyzed by western blotting . DOI: http://dx . doi . org/10 . 7554/eLife . 01977 . 016
Our results demonstrate that APC is indeed a relevant target of miR-142 . In both normal and malignant mammary cells , miR-142 activates the canonical WNT signaling pathway and regulates proliferation at least in part through upregulation of the WNT/β-catenin signaling . However , because each miRNA can have hundreds of targets in general , it is likely that other pathways also have roles in the regulation of normal and malignant breast epithelium by miR-142 . Expression of APC protein is modulated by miRNAs , such as miR-135 , miR-27 , miR-155 , and miR-142 ( Nagel et al . , 2008; Lu et al . , 2009; Wang and Xu , 2010; Hu et al . , 2013 ) . Elevated expression of miR-135 , which targets the APC mRNA , has been observed in colorectal adenomas and adenocarcinomas with/without biallelic APC mutations , suggesting that this miRNA could play a primary or synergistic role in activation of the canonical WNT signaling pathway during colon cancer development ( Nagel et al . , 2008 ) . In a model of osteoblast differentiation , miR-27 and miR-142 were shown to regulate APC expression and to positively modulate the canonical WNT signaling pathway ( Wang and Xu , 2010; Hu et al . , 2013 ) . miR-155 targets APC and inhibition of miR-155 using the anti-miR-155 increases the protein level of APC ( Lu et al . , 2009 ) . It is noteworthy that both miR-142-3p and miR-155 are more highly expressed in the human BCSCs than in the NTG cells ( Shimono et al . , 2009 ) . In this study , we show that miR-142 reduced the protein level of APC and enhanced the canonical WNT signaling pathway in normal and malignant mammary epithelium . This is important for the stem cells in the mammary gland ( Haegebarth and Clevers , 2009; Zeng and Nusse , 2010 ) . The dysregulated WNT signaling can enhance niche independence and promote aberrant proliferation of stem cells . Taken together , upregulation of miR-142 , miR-150 , and miR-155 may not only be hallmarks of BCSCs but could actually contribute directly to aberrant proliferation of these cells . In this report , we show that miR-150 is a target of miR-142 via its regulation of the canonical WNT/β-catenin signaling . Notably , miR-150 plays a role in the proliferation induced by miR-142 and the canonical WNT/β-catenin signaling . Our results suggest that miR-150's regulation of proliferation is downstream of the canonical WNT/β-catenin signaling . In addition , the APC mRNA has a predicted target site for miR-150 . We constructed the pGL3 luciferase expression plasmid in which the miR-150 target site within the 3′ UTR of the APC mRNA was cloned downstream of a luciferase minigene and found that the activity of this luciferase plasmid was reduced by 32% when the miR-150 precursor was co-transfected ( data not shown ) . However , the results of the Ago IP/microarray experiments revealed that APC mRNA was much less efficiently recruited to RISC by miR-150 ( Figure 1A—source data 3–4 ) . Because each miRNA has multiple target genes , we speculate that the presence of many other target genes with higher affinity to the miR-150-containing RISC will perturb the ability of miR-150 to regulate the APC mRNA and that miR-150 has much weaker effect , if any , on the WNT signaling pathway in vivo . Some miRNAs are known to be regulated by transcriptional factors which have critical roles in regulating development , proliferation , and cell survival ( Bracken et al . , 2008; O'Donnell et al . , 2005; Raver-Shapira et al . , 2007; Yamakuchi et al . , 2010; Chang et al . , 2007; Ma et al . , 2007 ) . For example , the transcriptional repressors , ZEB1 and ZEB2 , which can play a positive role in EMT , suppress transcription of the miR-200a-miR-200b-miR-429 cluster , which in turn represses expression of ZEB1 and ZEB2 , thereby forming a double negative feedback loop ( Bracken et al . , 2008 ) . c-Myc enhances expression of the miR-17-92 cluster , which in turn represses E2F1 expression and thereby suppresses cell-cycle entry ( O'Donnell et al . , 2005 ) . The tumor suppressor Tp53 regulates expression of miRNAs , including miR-34a , miR-107 , and miR-200 ( Chang et al . , 2007; Raver-Shapira et al . , 2007; Yamakuchi et al . , 2010; Chang et al . , 2011 ) . Expression of miR-10b , a miRNA that promotes breast cancer cell metastasis by inhibiting HoxD10 , is regulated by the transcription factor Twist , a major regulator of the epithelial-to-mesenchymal transition ( Ma et al . , 2007 ) . We identified a TCF/β-catenin binding site in the promoter region of miR-150 and showed that miR-142 stimulates the canonical WNT signaling pathway in vivo and elevates the expression of miR-150 . It has recently been reported that miR-150 induces proliferation and suppresses apoptosis of breast cancer cells ( Huang et al . , 2013 ) . A simple model to explain the upregulation of miR-142 and miR-150 in human BCSCs is that suppression of the APC protein expression by miR-142 increases the activity of the canonical WNT signaling pathway and thereby enhances the miR-150 expression . We note that our results show that miR-142 upregulates the expression of miR-150 when miR-142 is highly expressed and/or the WNT signaling pathway is strongly stimulated in mammary epithelial cells . Considering that miR-142 is highly expressed in human BCSCs but weakly expressed or undetectable in the stem/progenitor population of the mammary epithelial cells , our result suggests that the upregulation of miR-142 and the increase of the miR-150 expression by miR-142 could be especially relevant in the initiation and/or progression of human breast cancers driven by human BCSCs in vivo . The results of the mouse models suggest that APC dysregulation is associated with initiation and progression of at least some breast cancers . Mice heterozygous for a germline mutation in Apc ( Apc+/850T , ApcMin mice ) spontaneously develop mammary tumors , although at significantly lower frequency than intestinal tumors ( Moser et al . , 1993 ) . Mice heterozygous for a C-terminal truncation mutation , Apc+/1572T , develop multifocal mammary tumors with pulmonary metastasis at much higher frequency ( Gaspar et al . , 2009 ) . MMTV-promoter-Wnt1 transgenic mice develop breast tumors composed of both luminal and myoepithelial component and expanded mammary stem cell pools ( Cho et al . , 2008 ) . Finally , heterozygous mutation of APC in immature , but not mature , mammary epithelium induces breast tumors ( Kuraguchi et al . , 2009 ) . These results suggest that activation of the canonical WNT signaling pathway in mammary epithelium , by mutation of APC , excessive ectopic WNT expression , and/or stabilization of β-catenin contribute to mammary tumor initiation in the same way as colon cancer at least in mouse models of the mammary tumors . However , although these mouse models that upregulate the canonical WNT signaling activity induce mammary gland tumors ( Moser et al . , 1993; Gaspar et al . , 2009; Kuraguchi et al . , 2009 ) , APC mutations are less frequently found in human breast cancers ( Furuuchi et al . , 2000; Jin et al . , 2001; Sarrio et al . , 2003 ) . APC is a key tumor suppressor that regulates the canonical WNT signaling pathway and is involved in development and homeostasis of a variety of cells including stem cells ( Reya and Clevers , 2005 ) . In addition , APC has additional important cellular functions , including roles in cell adhesion , migration , organization of actin and microtubule networks , spindle formation , and chromosome segregation ( Aoki and Taketo , 2007 ) . Thus , dysregulation of these processes by mutations in the APC gene is frequently implicated in tumor initiation and progression . Our results show that upregulation of miR-142 can significantly enhance the canonical WNT signaling pathway through the suppression of APC including in the mammary cells . Therefore , miR-142 , a miRNA frequently upregulated in human BCSCs than in the NTG cells ( Shimono et al . , 2009 ) , could provide at least a part of the shared molecular mechanism for aberrant activation of the canonical WNT signaling pathway in BCSCs and for the initiation and progression of breast cancers .
Primary breast cancer specimens and normal breast tissues were obtained from the consented patients as approved by the Research Ethics Boards at Stanford University and at the City of Hope Cancer Center in California . All animal experiments were carried out under the approval of the Administrative Panel on Laboratory Animal Care of Stanford University . Human embryonic kidney ( HEK ) 293T cells , and MCF7 , and MDA-MB-231 breast carcinoma cells were maintained in Dulbecco's modified Eagle's medium ( DMEM ) with 10% FBS , 100 U/ml penicillin , 100 μg/ml streptomycin with or without 250 ng/ml amphotericin B ( Invitrogen , Carlsbad , CA ) and incubated at 5% CO2 at 37°C . HEK293T cells were seeded at 6 × 106 cells in 10-cm tissue culture plates 12 hr prior to transfection . All transfections were carried out with Lipofectamine 2000 ( Invitrogen ) , according to the manufacturer's instructions . A set of four plates was transfected with same miRNA precursors and each plate was transfected with 300 pmol miRNA precursor ( Ambion , Austin , TX ) . Plates were washed with 1× ice-cold phosphate-buffered saline ( PBS ) at 48 hr post-transfection and 700 µl of an ice-cold lysis buffer ( 150 mM KCl , 25 mM Tris–HCl , pH 7 . 4 , 5 mM EDTA , 0 . 5% Nonidet P-40 , 0 . 5 mM DTT , 100 U/ml SUPERase In ( Ambion ) with Complete proteinase inhibitor ( Roche , Germany ) ) was added to a 10-cm plate . After incubating for 30 min at 4°C , plates were scraped and the lysates were combined and spun at 14 , 000 rpm at 4°C for 30 min . The supernatant was collected and filtered through a 0 . 45 µm syringe filter . The biotinylated anti-Argonaute antibody ( 12 . 5 μg; biotin was conjugated to an anti-human Ago antibody ( #01-2203 , Wako , Japan ) ) ( Hendrickson et al . , 2009 ) was mixed with 250 μl Dynal M-280 Streptavidin coated magnetic beads ( Invitrogen ) , which were equilibrated by washing three times with lysis buffer and twice with PBS . The beads were incubated with the lysate at 4°C for 4 hr and washed twice with 10× volume of lysis buffer for 5 min . Five percent of the beads were frozen for SDS-PAGE analysis after the second wash . RNA was extracted directly from the remaining beads with 25:24:1 phenol:chloroform:isoamyl alcohol ( Invitrogen ) . Trace amounts of phenol were removed by chloroform extraction and RNA was precipitated using Glycogen ( Invitrogen ) as a carrier . RNA pellets were resuspended in 30 µl of RNase free water and stored at −80°C . Detailed methods for microarray experiments are available at the Brown lab website ( http://cmgm . stanford . edu/pbrown/protocols/index . html ) . HEEBO oligonucleotide microarrays were produced by Stanford Functional Genomic Facility . The HEEBO microarrays contain ∼45 , 000 70-mer oligonucleotide probes , representing ∼30 , 000 unique genes . A detailed description of this probe set can be found at http://microarray . org/sfgf/heebo . do . RNA from IP experiments was hybridized to microarrays printed on epoxysilane glass ( Hendrickson et al . , 2009 ) . For HEEBO microarray experiments , poly-adenylated RNAs were amplified in the presence of aminoallyl-UTP with Amino Allyl MessageAmp II aRNA kit ( Ambion ) . For expression experiments , universal reference RNA was used as an internal standard to enable reliable comparison of relative transcript levels in multiple samples ( Stratagene , La Jolla , CA ) . Amplified RNA ( 3–10 µg ) was fluorescently labeled with NHS-monoester Cy5 or Cy3 ( GE Healthcare , United Kingdom ) . Dye-labeled RNA was fragmented , then diluted in a 50 µl solution containing 3× SSC , 25 mM HEPES-NaOH , pH 7 . 0 , 20 µg human Cot-1 DNA ( Invitrogen ) , 20 µg poly ( A ) RNA ( Sigma-Aldrich , St . Louis , MO ) , 25 μg yeast tRNA ( Invitrogen ) , and 0 . 3% SDS . The sample was incubated at 70°C for 5 min , spun at 14 , 000 rpm for 10 min in a microcentrifuge , then hybridized at 65°C for 12–16 hr ( Hendrickson et al . , 2009 ) . Following hybridization , microarrays were washed in a series of four solutions containing 400 ml of 2× SSC with 0 . 05% SDS , 2× SSC , 1× SSC , and 0 . 2× SSC , respectively . The first wash was performed at 65°C for 5 min . The subsequent washes were performed at room temperatures for 2 min each . Following the last wash , the microarrays were dried by centrifugation in a low-ozone environment ( <5 ppb ) to prevent the destruction of Cy dyes ( Fare et al . , 2003 ) . Once dry , the microarrays were kept in a low-ozone environment during storage and scanning ( see http://cmgm . stanford . edu/pbrown/protocols/index . html ) . Microarrays were scanned using either AxonScanner 4200 or 4000B ( Molecular Devices , Sunnyvale , CA ) . PMT levels were auto-adjusted to achieve 0 . 1–0 . 25% pixel saturation . Each element was located and analyzed using GenePix Pro 5 . 0 ( Molecular Devices ) . Data were filtered to exclude elements that did not have a regression correlation of ≥0 . 6 between Cy5 and Cy3 signal over the pixels compromising the array element of and intensity/background ratio of ≥2 . 5 in at least one channel , for 60% of the arrays . For cluster and SAM analysis of Ago+/− miRNA IPs vs mock IPs , measurements corresponding to oligonucleotides that map to the same EntrezID were treated separately and the data were globally normalized per array , such that the median log2 ratio was 0 after normalization . A 525 bp fragment of the APC 3′ UTR ( corresponding to the positions of 8589–9113 of the NM_000038 . 5 ) was amplified by PCR using the cDNA of HEK293T cells as a template and cloned into the pGEM-T-Easy vector . Then the APC 3′ UTR product was cloned at the 3′ of the luciferase gene of the pGL3-MC vector ( Shimono et al . , 2009 ) . All products were sequenced . The pCMV-Neo-Bam APC plasmid was obtained from Addgene ( Cambridge , MA ) . To produce the control pCMV-Neo-Control plasmid , the pCMV-Neo-Bam-APC plasmid was digested by BamHI and ligated . Mutation of the putative miR-142-3p target sequences within the 3′ UTR of APC and within the pCMV-Neo-Bam-APC plasmid was generated using the QuikChange Site-Directed Mutagenesis kit ( Stratagene ) . The sequences of hsa-miR-142 including stem loop structure and 200–300 base pairs of up-stream and down-stream flanking regions were cloned by PCR using genomic DNA of HEK293 cells as a template . The products were cloned into multicloning sites of pEIZ-HIV-ZsGreen vector ( kind gift from Dr . Zena Werb , UCSF ) ( Welm et al . , 2008 ) . Lentiviruses were produced as described ( Tiscornia et al . , 2006 ) . For knockdown of miR-142-3p , the miRNA Zip ( anti-miRNA ) plasmid specific for miR-142-3p ( MZIP142-3p-PA-1 ) together with the scrambled control RNA-expressing plasmid was purchased from System Bioscience ( Mountain View , CA ) and lentiviruses were produced , according to the manufacturer's instructions . HEK293T cells were seeded at 1 × 105 cells per well in 48-well plates the day prior to transfection . All transfections were carried out with Lipofectamine 2000 ( Invitrogen ) , according to the manufacturer's instructions . Cells were transfected with 320 ng pGL3 luciferase expression construct containing the 3′ UTR of human APC , 40 ng pGL4 . 74 hRLuc/TK Renilla luciferase vector ( Promega , Fitchburg , WI ) , and 25 nM hsa-miR-142 precursor or negative control precursor ( Ambion ) . Forty-eight hr after transfection , luciferase activities were measured using the Dual-Luciferase Reporter Assay System ( Promega ) and normalized to Renilla luciferase activity . For the TOPFlash/FOPFlash Reporter assays , cells were transfected with 320 ng TOPFlash or FOPFlash luciferase expression construct containing TCF/β-catenin binding sites , 40 ng pRL-TK Renilla luciferase vector ( Promega ) , and 25 nM hsa-miR-142 precursor or negative control precursor ( Ambion ) . To co-transfect pCMV-Neo-APC plasmids , cells were transfected with 200 ng TOPFlash or FOPFlash luciferase expression construct , 20 ng pRL-TK Renilla luciferase vector ( Promega ) , 300 ng pCMV-Neo-APC plasmids , and 25 nM hsa-miR-142 precursor or negative control precursor ( Ambion ) . 48 h after transfection , cells were stimulated by 20 mM LiCl or 15% Wnt3A conditioned medium for 6 hr . Luciferase activities were measured using the Dual-Luciferase Reporter Assay System ( Promega ) and normalized to Renilla luciferase activity . All experiments were performed in triplicate . The cells were seeded at 7 × 105 cells in a 6-well plate , transfected with miR-142-precursor , and cultured for 2 days . To analyze the effect of anti-miR-142-3p in breast cancer cells and human breast cancer xenograft cells , the cells were infected with the anti-miR-142-3p-expressing lentivirus and the GFP-expressing cells were collected using a cell sorter . Cells were washed with PBS twice and soaked in hypotonic buffer ( 10 mM Tris–HCl pH 7 . 5 , 5 mM MgCl2 ) with Complete proteinase inhibitor ( Roche ) on ice for 20 min and sonicated . Then an SDS sample buffer ( 50 mM Tris–HCl , pH 6 . 8 , 2% SDS , 10% glycerol 5 mM EDTA , 0 . 02% bromophenol blue , 3% β-mercaptoethanol ) was added to the total cell lysate . Samples were separated on SDS-4-15% gradient polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride filters ( Amersham ) . After blocking with 5% skim milk in 0 . 05% Tween 20/PBS , filters were incubated with 1:100 diluted anti-APC polyclonal antibody ( C-20 , sc-896 , Santa Cruz Biotechnology , Dallas , TX ) or 1:500 diluted anti-β-actin monoclonal antibody ( C4 , sc-47778 , Santa Cruz Biotechnology ) . Then 1:5000-10 , 000 diluted peroxidase-conjugated sheep anti-rabbit or mouse IgG antibody ( Amersham , United Kingdom ) was added and developed using the SuperSignal West Dura Substrate ( Thermo Scientific , Waltham , MA ) . The experiments were repeated four times and the intensity of the band was measured using the ImageJ software . Mouse MMTV-Wnt1 breast tumors were digested and CD45-Thy1+CD24+ murine breast cancer cells were isolated using a cell sorter as described previously ( Cho et al . , 2008 ) . Three thousand CD45-Thy1+CD24+ murine breast cancer cells were sorted directly into 150 μl of the organoid medium ( Advanced DMEM/F12 medium ( Invitrogen ) supplemented with 2 mM GlutaMax ( Invitrogen ) , 10 mM HEPES ( Sigma-Aldrich ) , 1 mM sodium pyruvate ( Lonza , Switzerland ) , 10% FBS , 500 ng/ml R-spondin1 ( R&D systems , Minneapolis , MN ) , 100 ng/ml Noggin ( Peprotech , Rocky Hill , NJ ) , ITES media supplement ( Lonza ) , 50 ng/ml EGF ( Peprotech ) , 10 μM Y-27632 ( Calbiochem , Gibbstown , NJ ) , 100 U/ml penicillin , 100 μg/ml streptomycin , 250 ng/ml amphotericin B ( Invitrogen ) ) . The cells were infected with 25 multiplicity of infection ( moi ) of control or anti-miR-142-3p-expressing lentiviruses and cultured in the ultra-low attachment 96-well plate for 12 hr . To prepare a feeder layer , 2 × 103 irradiated 3T3-L1 cells ( Sigma-Aldrich ) were seeded in a 96-well plate and cultured with a feeder medium ( Advanced DMEM/F12 medium supplemented with 2 mM GlutaMax , 10 mM HEPES , 1 mM sodium pyruvate , 10% FBS , 100 U/ml penicillin , 100 μg/ml streptomycin , and 250 ng/ml amphotericin B ( Invitrogen ) ) at 5% CO2 at 37°C for 16 hr . Then feeder media were replaced by 40 μl of growth factor reduced Matrigel ( BD Bioscience , Franklin Lakes , NJ ) . Three thousand lentivirus-infected MMTV-Wnt1 tumor cells in 150 μl of the organoid media were plated on the Matrigel and cultured at 5% CO2 at 37°C . The CD44+ CD24−/low human BCSCs were collected from the dissociated single cell suspension of the early-passaged human breast xenograft ( COH69 ) . The cells were infected with 20 moi of control or anti-miR-142-3p-expressing lentiviruses and cultured in the ultra-low attachment 96-well plate for 12 hr . To prepare a feeder layer , 1 × 104 irradiated L Wnt-3A cells ( ATCC , Manassas , VA ) were seeded in a 96-well plate and cultured with a feeder medium at 5% CO2 at 37°C for 16 hr . Then feeder media were replaced by 50 μl of growth factor reduced Matrigel ( BD Bioscience ) . Five thousand lentivirus-infected human BCSCs in 100 μl of the organoid media were plated on the Matrigel and cultured at 5% CO2 at 37°C . The organoids were observed using a microscope ( Leica DMI 6000 B , Leica , Germany ) . The organoid cells were incubated with 10 μM EdU for 2 hr and the proliferation of the dissociated organoid cells was analyzed using the Click-iT Plus EdU Alexa Flour 647 Flow Cytometry Assay Kit ( Life Technologies , Carlsbad , CA ) , according to the manufacturer's instructions . The flow cytometric analyses of apoptotic cells in the organoids were performed using an APC-conjugated Annexin V ( BioLegend , San Diego , CA ) . Chromatin IP was performed using the Magna ChIP G kit ( Upstate , Lake Placid , NY ) , according to the manufacturer's protocol . HEK293T cells were seeded at 2 × 107 cells in 15-cm tissue culture plates and cells were treated with 20 mM LiCl for 7 hr . Cells were crosslinked by adding formaldehyde directly to culture medium to a final concentration of 1% and incubated for 10 min , followed by incubation with glycine for 5 min . After washing twice with ice-cold PBS , cells were scraped and collected in a microcentrifuge tube . Cell were precipitated by spinning at 3000 rpm at 4°C for 5 min and incubated with the lysis buffer with proteinase inhibitor on ice for 15 min . Tubes were spun , the nuclear lysis buffer with proteinase inhibitor was added and sonicated to form shared chromatin of about 200–1000 base pairs in length ( data not shown ) . The product was diluted 10 times by ChIP dilution buffer , and 2 μg of an anti-β-catenin antibody ( clone 14 , BD Bioscience ) or control normal mouse IgG ( Santa Cruz Biotechnology ) together with 20 μl of fully suspended protein G magnetic beads were added . Lysates were incubated overnight with rotation at 4°C and washed with low salt buffer , high salt buffer , LiCl buffer , and TE buffer . DNA was purified by using the column provided in the kit . The primer set was designed for the human genomic sequence flanking the putative TCF/β-catenin binding site ( Figure 5A ) . The sequences of primers are: miR-150 forward , GTGTGCAGTTTCTGCGACTCAG; reverse , CACTGGTACAAGGGTTGGGAGAC; c-MYC forward , GCACGGAAGTAATACTCCTCTCCTC; reverse , CAGAAGAGACAAATCCCCTTTGCGC; β-actin forward , GTGTCTAAGACAGTGTTGTGGGTGTAG; reverse , CTGGGGTGTTGAAGGTCTCAAACATG . The amount of TCF/β-catenin biding sequence enriched by chromatin IP was evaluated by PCR and subsequent agarose gel electrophoresis . RT , pre-PCR , and the real-time PCR for miRNA expression analyses were performed by the real-time PCR method as described previously ( Tang et al . , 2006; Shimono et al . , 2009 ) . The abundance of each miRNA was measured individually by using the 7900HT Fast Real-Time PCR System ( Applied Biosystems , Foster City , CA ) . Results were normalized by the amount of small nuclear RNA expression , C/D box 96A , C/D box84 , or U6 snRNA . Cells were transfected with miR-142-precursor and cultured for 2 days . To analyze the effect of anti-miR-142-3p in human breast cancer xenograft cells , the cells were infected with the anti-miR-142-3p-expressing lentivirus and the GFP-expressing cells were collected using a cell sorter . To knockdown β-catenin , the cells were transfected with siRNA against CTNNB1 ( Life Technologies ) or negative control siRNA ( Life Technologies ) using the Lipofectamine RNAiMAX reagent ( Life Technologies ) . The cells were lysed with TRIzol ( Life Technologies ) , and total RNAs were extracted following the manufacture's protocol . SuperScript VILO ( Life Technologies ) and TaqMan MicroRNA Reverse Transcription kit ( Life Technologiies ) were used for the mRNA and miRNA measurement experiments , respectively . Then , RT products were amplified with TaqMan PreAmp master Mix ( Life Technologies ) , when necessary . The abundance of each mRNA or miRNA was measured using a 7900HT Fast Real-Time PCR system ( Applied Biosystems ) . Mammary epithelial cells were obtained as described previously ( Stingl et al . , 2006 ) . Briefly , mammary glands from 8- to 14-week old virgin female FVB were digested with collagenase/hyaluronidase ( StemCell Technologies , Canada ) . Single cell suspension was obtained by dissociation of the fragments with 0 . 25% trypsin , and dispase and DNase I ( StemCell Technologies ) . Lineage− mammary epithelial cells were obtained after removing the CD45+ , Ter119+ , and CD31+ cells using the EasySep mouse epithelial cell enrichment kit ( StemCell Technologies ) . The isolated cells were mixed with 5 moi of lentivirus and incubated for 16 hr at 5% CO2 at 37°C . Our flow cytometry analysis using the lentivirus-infected mammary cells revealed that ∼50% of the mammary cells were infected 1 day after infection . Fifty-thousand lentivirus infected cells were injected into cleared mammary fat pad of weaning age FVB/NJ female mouse . After 8 weeks , ZsGreen expression of the transplanted mammary tissue was checked under the fluorescent microscope ( Leica DMI 6000 B ) . Whole-mount mammary tissue was fixed in Carnoy's fixative ( 60% ethanol , 30% chloroform , 10% glacial acetic acid ) for 4 hr , and whole-mount staining was performed using carmine alum staining solution . For hematoxylin and eosin staining , the mammary tissue with ZsGreen expression was fixed in formalin , embedded in paraffin , and stained by hematoxylin and eosin staining method . The stained tissue was observed using a microscope ( Leica DM 4000 B ) . Formalin-fixed paraffin-embedded tissue was cut in serial sections and mounted on glass slides . The tissue sections were deparaffinized and rehydrated . Antigen retrieval was performed using a 0 . 01 M citrate buffer ( pH 6 . 0 ) by heating the sample in a microwave for 20 min . The slides were incubated with an anti-PCNA antibody ( clone D3H8P , Cell Signaling , Danvers , MA ) , an anti-β-catenin antibody ( 1:200 , clone 14 , BD Bioscience ) , or an anti-active-β-catenin antibody ( 1:100 , clone 8E7 , Millipore , Billerica , MA ) at 4°C overnight . Then tissue was stained by an Alexa Fluor594-conjugated secondary antibody ( 1:200 , Jackson Laboratory , Bar Harbor , ME ) or an Alexa Fluor 488-conjugated secondary antibody ( 1:200 , Invitrogen ) at room temperature for 1 hr . The stained tissue was observed using a fluorescent microscope ( Leica DM 4000 B ) . The CD44+ CD24-/low lineage− human BCSCs were isolated by a cell sorter . BCSCs were infected by 20 moi of anti-miR-142-3p-expressing lentivirus or control lentivirus by spin infection for 2 hr . Infected cells were washed with PBS and were mixed with Matrigel ( BD Biosciences ) . Ten thousand infected cells were injected into mammary fat pad of female NSG mouse . Tumors were measured twice a week , and their volume was estimated using the formula: volume = ab2/2 ( a , length; b , width ) ( Tanaka et al . , 1990 ) . All experiments were carried out under the approval of the Administrative Panel on Laboratory Animal Care of Stanford University . Data from experiments were statically analyzed using T-test . For the results of real-time PCR , we employed Mann–Whitney U-test . | Messenger RNA molecules take the information encoded in a gene's DNA sequence and turn it into instructions for building a protein . However , if certain smaller molecules of RNA—called microRNAs—bind to a messenger RNA molecule , they ‘silence’ it , which prevents the information in the messenger RNA from being translated to make a protein . Despite their small size , microRNAs are very powerful . These molecules are able to simultaneously inhibit the translation of hundreds of messenger RNAs and perform many roles , including controlling cell growth and maintaining populations of stem cells . Furthermore , microRNAs have been linked to different aspects of the growth of cancerous cells . Certain microRNAs appear to suppress tumors by regulating the growth of the stem cells found there , while others appear to be ‘hyperactive’ in cancers—including breast cancer , colon cancer , and blood cancer . In 2009 , researchers compared the amount of microRNA in breast cancer stem cells that are highly capable of forming tumors with the amount in other cancer cells within the same tumor . Amongst other differences , two microRNAs ( called miR-142 and miR-150 ) were found to be hyperactive in human breast cancer stem cells . One of them , miR-142 , is known to target a gene called APC that inhibits the renewal of normal stem cells . Mutations in the APC gene have been linked to colon cancer , and scientists have suggested that the mutations inactivate APC in cancer cells to promote unregulated cell growth . Breast tumors rarely have mutations in the APC gene , but Isobe et al . wondered whether microRNAs that target this gene might also promote the growth of these tumor cells . Isobe et al . —including several of the researchers involved in the 2009 work—show that miR-142 does target the APC gene in human breast cancer stem cells , and silences it . With the gene silenced , a cancer-promoting pathway turns on and more miR-150 is made . Increasing the amount of either miR-142 or miR-150 causes excessive cell growth in breast tissue and can form abnormal breast tissue in mice . Reducing the amount of miR-142 in human breast cancer stem cells slows the growth of breast tumors . Although they only make up a small population of human breast cancer cells , focusing on breast cancer stem cells could uncover the cancer-promoting pathways that are activated in human breast cancers . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"cancer",
"biology"
] | 2014 | miR-142 regulates the tumorigenicity of human breast cancer stem cells through the canonical WNT signaling pathway |
In high-income countries , one’s relative socio-economic position and economic inequality may affect health and well-being , arguably via psychosocial stress . We tested this in a small-scale subsistence society , the Tsimane , by associating relative household wealth ( n = 871 ) and community-level wealth inequality ( n = 40 , Gini = 0 . 15–0 . 53 ) with a range of psychological variables , stressors , and health outcomes ( depressive symptoms [n = 670] , social conflicts [n = 401] , non-social problems [n = 398] , social support [n = 399] , cortisol [n = 811] , body mass index [n = 9 , 926] , blood pressure [n = 3 , 195] , self-rated health [n = 2523] , morbidities [n = 1542] ) controlling for community-average wealth , age , sex , household size , community size , and distance to markets . Wealthier people largely had better outcomes while inequality associated with more respiratory disease , a leading cause of mortality . Greater inequality and lower wealth were associated with higher blood pressure . Psychosocial factors did not mediate wealth-health associations . Thus , relative socio-economic position and inequality may affect health across diverse societies , though this is likely exacerbated in high-income countries .
The Tsimane are a population of >16 , 000 Indigenous Amerindians living in >90 communities at the edge of the Amazon basin in lowland Bolivia . Tsimane communities consist of dispersed household clusters tied together by networks of kinship , cooperative production and consumption ( Hooper et al . , 2015; Jaeggi et al . , 2016 ) , as well as usually a school and soccer field . Community meetings convene to discuss and resolve important matters , including conflicts within the community . As such , we treat the community as the salient scale of status competition ( Alami et al . , 2020; von Rueden et al . , 2018; von Rueden et al . , 2008; von Rueden et al . , 2019; von Rueden et al . , 2014 ) and calculated relative wealth and inequality at this level . The Tsimane remained relatively isolated from the larger Bolivian economy until the 1970s and still widely practice traditional subsistence ( swidden horticulture , hunting , and fishing ) , which contributes >90% of their calories ( Gurven et al . , 2017; Kraft et al . , 2018 ) . Cattle , introduced by missionaries and ranchers , are owned by a small minority of Tsimane . Over the past few decades , wage labor opportunities with loggers or ranchers and produce sales in the local market towns of San Borja and Yucumo have been increasing , as have formal schooling , Spanish fluency , and access to modern amenities such as electricity and health care . The population thus exhibits quantifiable gradients of modernization ( see Figure 1 ) . In terms of morbidity and mortality , the Tsimane are characterized by high infectious disease burden , with respiratory infections as the leading cause of death at all ages ( Gurven et al . , 2007 ) . Additionally , parasites such as helminths and giardia are highly prevalent ( Blackwell et al . , 2013; see also Table 2 ) . These conditions result in frequent , acute immune responses ( Blackwell et al . , 2016a ) but still a low incidence of chronic conditions such as hypertension or atherosclerosis , due to high levels of physical activity and other protective factors ( Gurven et al . , 2012; Gurven et al . , 2016; Gurven et al . , 2009; Kaplan et al . , 2017 ) .
Because wealth varied considerably by age ( Figure 1A ) , we used an age-corrected measure of relative wealth that reflects one’s wealth relative to this age trajectory ( see Materials and methods ) . This corrects for random variation in the age structure of sampled communities and arguably better captures the essence of relative socio-economic rank: what matters is how one compares to others , relative to general trends such as wealth ( status , influence , etc . ) accumulating with age . At the high end of the wealth distribution ( Figure 1B ) , much of the variation was driven by livestock , especially cattle . Figure 1C–D illustrates variation in mean wealth and wealth inequality among the study communities . Mean wealth was generally lowest in communities located in the interior forest ( Figure 1C , bottom right ) , which are remote and inaccessible by road for much of the year ( due to washed-out bridges ) ; and in those communities downriver from San Borja ( Figure 1C , top ) , which experience frequent flooding and are within or adjacent to a protected bioreserve that limits resource extraction . Somewhat unexpectedly , mean wealth was higher further from the market town of San Borja ( correlation between mean wealth and distance to market r = 0 . 36 , df = 38 , p=0 . 02 ) . We operationalized inequality by calculating community-level Gini coefficients for age-corrected wealth ( see Materials and methods ) . Wealth inequality was generally higher in communities closer to the market towns of San Borja and Yucumo , where Tsimane can sell produce and purchase market goods , though some villages near towns also show low inequality ( Figure 1D ) ( correlation between Gini and distance to market r = –0 . 38 , df = 38 , p=0 . 01 ) . Inequality was marginally lower in richer communities ( r = –0 . 22 , df = 38 , p=0 . 17 ) . Community size was not significantly related to distance ( r = –0 . 18 , p=0 . 26 ) , mean wealth ( r = 0 . 11 , p=0 . 50 ) , or inequality ( r = 0 . 00 , p=0 . 99 ) . In sum , villages near towns had both higher inequality and lower mean wealth due to both more wealthy individuals and more very poor individuals in these communities . To examine the effects of household wealth and community wealth inequality on psychosocial or health outcomes , we used Bayesian multilevel models with appropriate controls and random effects at the individual , household , and community level ( see Materials and methods ) . Wealth was divided into relative wealth , centered on the community mean , and mean community wealth . Operationalizing wealth this way means we are in principle able to tease apart within-community wealth differentials , that is , one’s position in the local socio-economic hierarchy , from community-level differences in access to resources , that is , mean community wealth ( Kreft et al . , 1995 ) . However , in practice , models with wealth centered on the village produced virtually identical estimates to models with wealth centered on the sample as a whole ( see Supplementary file 1a-1m ) , largely because villages did not differ strongly in mean wealth ( median –0 . 03 , range: –1 . 0–0 . 66 Z-scores , 80% between –0 . 43 and 0 . 37 ) . Thus , community-relative and population-relative wealth were highly correlated ( r = 0 . 92 ) . Bayesian models produce a posterior distribution of parameter estimates that can be summarized in various ways ( McElreath , 2020 ) . Here , we provide coefficient plots ( Figures 2—5 ) showing posterior medians , as well as 75% and 95% highest posterior density intervals; we also provide prediction plots as supplements to these figures . In the text , we report results as standardized coefficients ( β ) for Gaussian models or as log odds ( β ) and odds ratios ( ORs ) for logistic models , both represented by the posterior mean , as well as the proportion of the posterior above zero ( P>0 ) , that is , the likelihood of a positive association . Higher or lower values of this number represent stronger certainty for a non-zero effect , while values near 0 . 5 indicate complete uncertainty about the direction of an association , if any . In addition , we report Cohen’s d as a standardized measure of effect size to allow comparison between continuous and binary variables; d is reported as the posterior median and the median absolute deviation ( MAD ) ( a more robust measure of dispersion than the standard deviation ) . For simplicity , we refer to effect sizes of d > 0 . 2 as ‘strong , ’ those >0 . 1 as ‘moderate , ’ and consider the rest to be ‘weak’ though potentially still suggestive of a general pattern . Similarly , we refer to posterior support of >0 . 975 ( or <0 . 025 , if negative ) as ‘high certainty’ and those with support >0 . 875/<0 . 125 as ‘moderate certainty , ’ corresponding to the entire 95% or 75% highest posterior density intervals respectively not overlapping with 0 , and we consider the rest to be ‘uncertain . ’ However , we encourage readers to use the full information on the posteriors to inform their own inference . Means and 95% credible intervals for all parameters are reported in Supplementary file 1a-1o . These tables also provide Bayesian R2 ( Gelman et al . , 2019 ) as a goodness-of-fit measure , indicating that in most models the predictors and random effects jointly explained about 20–40% of the variance in the data ( R2 range: 0 . 16–0 . 91 ) . Overall , for adults , household wealth was associated , with various effect sizes and degrees of confidence , with beneficial health outcomes except gastrointestinal illness , which showed no association ( Figure 2; Supplementary file 1f-1m ) . Community mean wealth had more mixed associations with health outcomes . Specifically , household wealth was associated with lower systolic blood pressure ( β = −0 . 01 , P>0=0 . 37 , Cohen’s d = −0 . 01 [0 . 02] ) and lower diastolic blood pressure ( β = −0 . 04 , P>0=0 . 02 , d = −0 . 05 [0 . 02] ) , though both effect sizes were small and only the latter association had high certainty . Community mean wealth was strongly and with high certainty associated with lower systolic ( β = −0 . 29 , P>0=0 . 00 , d = −0 . 34 [0 . 09] ) and diastolic ( β = −0 . 21 , P>0=0 . 01 , d = −0 . 24 [0 . 11] ) blood pressure . Household wealth also associated with better self-rated health ( reverse coded β = −0 . 02 , P>0=0 . 14 , d = −0 . 03 [0 . 02] ) , lower odds of infectious ( β = −0 . 06 , P>0=0 . 26 , d = −0 . 02 [0 . 05] , OR = 0 . 94 ) and respiratory ( β = −0 . 04 , P>0=0 . 69 , d = −0 . 03 [0 . 05] , OR = 0 . 96 ) illness , and lower total morbidity ( β = −0 . 02 , P>0=0 . 23 , d = −0 . 02 [0 . 04] ) , though again most effect sizes were small and there was high uncertainty . There was no evidence for an association with gastrointestinal infection . However , there was a moderate though uncertain association between community mean wealth and lower gastrointestinal illnesses ( β = −0 . 32 , P>0=0 . 21 , d = −0 . 16 [0 . 21] , OR = 0 . 72 ) . Household wealth was weakly and uncertainly associated with lower BMI ( β = −0 . 01 , P>0=0 . 24 , d = −0 . 02 [0 . 04] ) , but community mean wealth was weakly associated with higher BMI ( β = 0 . 06 , P>0=0 . 74 , d = 0 . 12 [0 . 20] ) . Using population-relative wealth , rather than community-relative wealth had little effect on these associations ( Supplementary file 1a-1m ) . In sum , despite mostly small effect sizes and high uncertainty , the general pattern was for wealthier adults to have better outcomes . For juveniles ≤ 15 years of age ( Figure 3; Supplementary file 1n , o ) , household wealth was weakly associated with lower total morbidity ( β = −0 . 04 , P>0=0 . 06 , d = −0 . 06 [0 . 04] ) , and in particular , moderately lower odds of respiratory illness ( β = −0 . 24 , P>0<0 . 01 , d = −0 . 13 [0 . 05] , OR = 0 . 79 ) . However , both household and community mean wealth were associated with higher odds of gastrointestinal illness ( β = 0 . 13 , P>0=0 . 95 , d = 0 . 07 [0 . 04] , OR = 1 . 14; β = 0 . 49 , P>0=0 . 81 , d = 0 . 27 [0 . 30] , OR = 1 . 63 ) and community mean wealth was associated with other infections ( β = 0 . 81 , P>0=0 . 87 , d = 0 . 44 [0 . 37] , OR = 2 . 25 ) and higher total morbidity ( β = 0 . 05 , P>0=0 . 82 , d = 0 . 31 [0 . 33] ) with mostly strong effect sizes but high uncertainty . In sum , for juveniles , wealth was moderately associated with reduced risk of respiratory illness , while community wealth was strongly associated with several negative health outcomes . For adults , inequality was associated with higher levels of three morbidity-related outcomes and lower levels of four outcomes ( Figure 2; Supplementary file 1f , m ) . Consistent with predictions of worse health with inequality ( P2a ) , greater inequality was weakly associated with higher blood pressure ( systolic: β = 0 . 05 , P>0=0 . 98 , d = 0 . 06 [0 . 03]; diastolic: β = 0 . 02 , P>0=0 . 75 , d = 0 . 03 [0 . 04] ) , and strongly with a greater likelihood of respiratory illness ( β = 0 . 35 , P>0=0 . 93 , d = 0 . 20 [0 . 13] , OR = 1 . 36 ) . Despite these harmful associations with inequality , people in more unequal communities had a strongly lower likelihood of other infections ( β = −0 . 62 , P>0=0 . 02 , d = −0 . 33 [0 . 16] , OR = 0 . 54 ) and to a more uncertain degree , total morbidity ( β = −0 . 07 , P>0=0 . 25 , d = −0 . 07 [0 . 13] ) , and gastrointestinal infections ( β = −0 . 12 , P>0=0 . 22 , d = −0 . 06 [0 . 09] , OR = 0 . 89 ) . Associations with BMI were negligible ( β = 0 . 01 , P>0=0 . 62 , d = 0 . 03 [0 . 08] ) . In contrast , for juveniles ( Figure 3; Supplementary file 1m , o ) , BMI was lower in more unequal communities ( β = −0 . 06 , P>0=0 . 05 , d = −0 . 08 [0 . 05] ) . Inequality had little effect on total morbidity and was moderately associated with less infectious illness ( β = −0 . 23 , P>0=0 . 24 , d = −0 . 13 [0 . 19] , OR = 0 . 79 ) , but greater respiratory illness ( β = 0 . 21 , P>0=0 . 81 , d = 0 . 11 [0 . 13] , OR = 1 . 23 ) and gastrointestinal illness ( β = 0 . 17 , P>0=0 . 74 , d = 0 . 10 [0 . 13] , OR = 1 . 19 ) , both of which are highly prevalent among juveniles . For adults , greater household wealth was associated with better outcomes in four of five psychological and social measures , with no association for the fifth ( Figure 2; Supplementary file 1a-1e ) . Household wealth was strongly and with high certainty associated with having more labor partners ( reverse coded β = –0 . 13 , P>0=0 . 01 , d = −0 . 49 [0 . 20] ) , and weakly and uncertainly , with fewer depressive symptoms ( β = −0 . 04 , P>0=0 . 14 , d = −0 . 05 [0 . 05] ) , fewer non-social problems ( i . e . , self-reported concerns over food insecurity , debt , and illness; β = −0 . 06 , P>0=0 . 12 , d = −0 . 08 [0 . 07] ) , and lower urinary cortisol ( β = −0 . 02 , P>0=0 . 27 , d = −0 . 02 [0 . 04] ) . There was no support for an association with social conflicts . Unlike household wealth , community mean wealth was not clearly associated with any psychosocial outcome , though there were strong but uncertain associations with more labor partners ( reverse coded β = −0 . 16 , P>0=0 . 29 , d = −0 . 77 [0 . 94] ) but also more non-social problems ( β = 0 . 28 , P>0=0 . 77 , d = 0 . 33 [0 . 45] ) . Contrary to predictions , inequality was largely associated with fewer stressors and psychological or social problems ( Figure 2; Supplementary file 1a-1e ) . The strongest evidence was for fewer non-social problems in more unequal communities ( β = −0 . 15 , P>0=0 . 07 , d = −0 . 17 [0 . 13] ) , with weak evidence for fewer conflicts ( β = −0 . 04 , P>0=0 . 33 , d = −0 . 01 [0 . 09] ) , and more labor partners ( β = −0 . 05 , P>0=0 . 32 , d = −0 . 29 [0 . 38] ) with more inequality . We tested the prediction ( P3 ) that the effects of wealth or inequality on health were mediated via psychosocial pathways using formal mediation analysis ( Baron and Kenny , 1986; MacKinnon et al . , 2007 ) . Specifically , this involves estimating the association between wealth/inequality and psychosocial variables ( ‘path a’ ) , as well as between psychosocial variables and health outcomes ( ‘path b’ ) ; if both are statistically significant and the association between wealth/inequality and health outcomes ( ‘path c’ , or direct effect ) is weaker , then there is evidence that there is an indirect effect of wealth/inequality on health via psychosocial variables ( i . e . , the psychosocial variable is a mediator ) . As reported above , paths a were mostly supported for household wealth , that is , household wealth was associated with four of the five psychosocial variables , but not for community wealth or inequality . Supplementary file 1q-1s presents mediation analyses with each health outcome variable and each psychosocial variable as a potential mediator , including estimates of the direct ( path c , as reported above ) and indirect effects , the mediator effects ( path b ) , and the proportion mediated ( indirect effect/total effect ) . See Appendix 1 for a discussion and graphical depiction of the causal relationships assumed by this mediation approach . The only convincing evidence for mediation was found for depression and non-social problems mediating the effect of household wealth on diastolic blood pressure; specifically , household wealth was negatively associated with diastolic blood pressure ( path c ) as well as with depression and non-social problems ( paths a; see above , Figure 2 ) , and both depression ( β = −0 . 03 , P>0=0 . 20 ) and non-social problems ( β = −0 . 08 , P>0=0 . 05 ) were themselves negatively associated with diastolic blood pressure ( paths b ) . However , there were no other cases where both paths a and b were well supported , the indirect effects of household wealth , community wealth , or inequality were virtually always zero for any mediator ( including depression and non-social problems ) , and the proportion mediated was generally small or highly uncertain ( Supplementary file 1q-1s ) . Overall , there was little evidence of mediation . Of the included covariates , many were associated with outcomes . For adults ( Figure 4; Supplementary file 1a-1m ) , age was positively associated with all negative health outcomes except respiratory illness as well as depression and social conflict . Male sex was associated with increased blood pressure but lower depression , conflicts , non-social problems , urinary cortisol , infection illness , and total morbidity , and with better self-rated health . Increasing distance from the market town was associated with increased blood pressure , more conflicts , respiratory illness , and gastrointestinal illness , as well as lower BMI . However , it was also associated with lower depression and urinary cortisol . Community size was generally associated with more positive psychological and social variables , but also higher blood pressure and infection . Household size was associated with worse psychological and social condition , with the exception of labor partners , which were higher for large households . Results for juveniles largely reflect similar associations ( Figure 5; Supplementary file 1n and o ) . In some cases , the inclusion of covariates improved model R2 statistics , though in many models changes in fit were negligible ( Supplementary file 1a-1o ) . In general , the inclusion of covariates reduced the variance attributable to random effects for individual , household , and community . Posterior distributions for wealth and inequality associations were all similar whether covariates were included or excluded ( i . e . , the posteriors overlap substantially ) , though there were some minor differences between the posterior means that were largely inconsequential for inference . Finally , we conducted several post-hoc tests to examine whether wealth-health associations were contingent on sex or whether relative wealth effects were contingent on levels of inequality and vice versa . For example , inequality could trigger increased stress and competitiveness only in men given a history of higher reproductive skew in males ( Daly , 2016 ) and inequality might affect the wealthier and poorer differently ( P2b ) , that is , poorer individuals may fare even worse in more unequal contexts . For this reason , we included wealth-by-inequality , wealth-by-sex , or inequality-by-sex interactions in models . A number of models favored interactions though there was little consistency across outcomes ( Figure 6; Supplementary file 1p ) . For depression , systolic and diastolic blood pressure , and self-rated health , poorer men showed worse outcomes than wealthier men , though there was little effect of wealth for women . In contrast , poorer women reported more non-social problems . Poor individuals showed both increased conflicts and reduced labor partners in unequal places , while wealthier individuals reported more conflicts and fewer labor partners in equal communities . In unequal communities , wealth had little effect on respiratory illness , while in more equal places , wealthier individuals were less likely to be diagnosed with respiratory illness . Contra P2b , there was no consistent indication that inequality was worse for poorer individuals , while males were somewhat more affected by being poor .
We tested whether within-community relative wealth , community wealth , and community-level wealth inequality were associated with a broad range of psychological , social , and health outcomes in a large sample of households and communities in a relatively egalitarian small-scale subsistence society . Overall , our results showed substantial heterogeneity in terms of the direction and magnitude of associations between wealth , wealth inequality , and health , which contrasts with the more consistent socio-economic health gradients in high-income countries . Nevertheless , some findings supported an association between wealth or inequality and health outcomes , though these associations were not mediated by psychosocial factors . Consistent with the prediction that higher relative position in a socio-economic hierarchy improves outcomes ( P1 ) , we found that household wealth relative to others in the community , capturing one’s rank within the local socio-economic hierarchy , was associated with lower blood pressure , and for juveniles , lower total morbidity and fewer respiratory infections . Relative household wealth was also generally associated with better health and psychosocial outcomes , but with more uncertainty in the posterior estimates , and for juveniles , relative household wealth was associated with increased gastrointestinal illness . Community mean wealth , capturing the absolute access to resources of households within that community , was also strongly associated with lower blood pressure for adults , but there was high uncertainty in estimates for other outcomes . Conversely , in support of P2a inequality was associated with higher blood pressure in adults and more respiratory disease in both adults and juveniles . It was also associated with lower BMI in juveniles , which in this energetically limited population likely represents a negative outcome . However , contra P2a inequality was also associated with lower levels of other infections ( mostly fungal and yeast infections and lice ) as well as fewer non-social problems , and there were several null results ( Figures 2 and 3 ) . Although most effect sizes were weak to moderate ( most Cohen’s d < 0 . 2 ) , these statistically weak results could still have significant biological and clinical impacts , as elaborated below . The finding that higher inequality associated with greater likelihood of respiratory disease is perhaps the most significant in terms of well-being and biological fitness . Respiratory illness is the leading cause of mortality at all ages in this population ( Gurven et al . , 2007 ) and continues to be a major source of morbidity . The likelihood of being diagnosed with respiratory illness was predicted to differ greater than threefold , 8–28% , between the least and most unequal communities indicating substantial fitness costs to inequality . However , this effect of inequality appears to primarily affect wealthy individuals , bringing their prevalence up to the level of poorer individuals ( Figure 6 ) . In one Tsimane community ( with relatively high average income compared to other communities ) , von Rueden et al . , 2014 found lower risk of respiratory infection among influential men but no effect on respiratory infection ( though trending in direction of higher risk ) for men with higher income . With current data , we cannot determine the mechanism responsible for this association between inequality and respiratory disease . The association could reflect differences in immune function as suggested by other research on psychosocial influences on infectious disease ( Aiello et al . , 2018; Chen and Miller , 2013; McDade et al . , 2016 ) , despite a lack of evidence for psychosocial mediation here . The association between inequality and respiratory disease could also be spurious , despite our best efforts to control for relevant covariates , or it could reflect differences in exposure not captured by distance to town or community size ( such as population density or frequency of contact with outsiders ) ; in this context , it should be noted that effects of inequality on health are arguably only expected for outcomes for which there is a socio-economic gradient in the first place ( Pickett and Wilkinson , 2015 ) , which was not the case for respiratory disease here . One of the strongest , most certain and most consistent associations of wealth ( both household and community level ) and inequality was with blood pressure , a major contributor to chronic disease in high-income countries . There was a clear socio-economic gradient in blood pressure within and between communities , and blood pressure was higher in more unequal communities . These effects were observed primarily in men . While most Tsimane are not hypertensive and do not have heart disease ( Gurven et al . , 2012; Kaplan et al . , 2017 ) , the predicted effects of wealth and inequality on blood pressure were substantial: systolic blood pressure was predicted to increase by 0 . 32 SD ( i . e . , 4 . 0 mmHg ) and diastolic blood pressure by 0 . 40 SD ( 3 . 7 mmHg ) in the most unequal compared to the most equal communities; conversely , wealth was protective such that the lowest blood pressures were predicted for people in the richest communities ( 7 . 4 mmHg systolic and 4 . 2 mmHg diastolic lower ) and the richest households within communities ( 1 . 0 mmHg systolic and 3 . 1 mmHg diastolic lower ) . In high-income countries , such changes in blood pressure correspond to as much as a 10% change in the risk of major cardiovascular disease events ( see Figure 2 in Ettehad et al . , 2016 ) . Among the Tsimane , it corresponds to as much as 40 years of age-related increases in blood pressure ( Gurven et al . , 2012 ) . As novel , obesogenic foods enter the Tsimane diet ( Kraft et al . , 2018 ) , market integration increases stress ( Konečná and Urlacher , 2017; von Rueden et al . , 2014 ) , sanitation improves ( Dinkel et al . , 2020 ) , and protective lifestyle factors like physical activity and helminth infections are changing ( Gurven et al . , 2013; Gurven et al . , 2016 ) , people in unequal communities , especially the poor ( see Figure 6 ) , may be at increasingly greater risk of chronic disease . Increases in blood pressure with modernization have also been reported in many other subsistence populations , and may partly stem from stress caused by integrating into a dominant culture ( Dressler , 1999; Konečná and Urlacher , 2017 ) . In this context , it is also worth noting that while the range of our village-level Gini values ( 0 . 15–0 . 43 ) was similar to that of income inequality among high-income countries ( e . g . , Denmark: 0 . 24; USA: 0 . 45 ) , it was considerably lower than the range of wealth inequality in these countries ( e . g . , Japan: 0 . 55; USA: 0 . 81 [Nowatzki , 2012] ) . Thus , the reported associations between wealth/inequality and blood pressure may still be relatively harmless for the Tsimane , but lay the foundation for chronic disease under more mismatched conditions . An alternative interpretation for some of these associations may be that causality is reversed , with poor health leading to less wealth or exacerbated inequality . On the face of it , this seems plausible for respiratory illness , which reduces work productivity . However , the fact that we see no direct association with wealth for adults , and only an association with inequality , seems to argue against such a mechanism . We did find an association between wealth and respiratory disease for juveniles – perhaps having sicker children puts some strain on wealth accumulation . For blood pressure , it is harder to imagine how reverse causality might occur since the blood pressure changes we observed are unlikely to affect wealth . Regardless , a limitation of our data is that we cannot determine the direction of causation given our cross-sectional design . Other confounds might also be possible , for example , if people preferentially assort by health or wealth by moving between villages . Beyond respiratory disease and blood pressure , many associations were inconclusive . This heterogeneous picture may seem surprising given robust directional findings from studies in high-income countries , especially for SES health gradients . One possibility for this difference is that hierarchy-stress associations produce more consistent health effects in an epidemiological context characterized by chronic , rather than infectious , disease . As argued above , our finding that one of the most consistent wealth-health associations was with blood pressure would support this argument since hypertension is a risk factor for most chronic diseases and consistently associated with socio-economic position and inequality in high-income countries ( Kim et al . , 2008; Shahu et al . , 2019 ) , but unlikely to be harmful for most Tsimane ( Gurven et al . , 2012; Kaplan et al . , 2017 ) . However , there are also consistent associations between socio-economic position and infectious disease in high-income countries ( Aiello et al . , 2018; Snyder-Mackler et al . , 2020 ) , suggesting that epidemiological context alone does not account for inconsistent results . Another possible source of heterogeneity is the scale at which relative wealth and inequality are measured . Literature reviews suggest that at an international scale as many as 83% of studies find associations , while in studies of areas the size of neighborhoods , only 45% find associations ( Kondo et al . , 2012; Pickett and Wilkinson , 2015; Wilkinson and Pickett , 2006 ) . Pickett and Wilkinson , 2015 suggest that this heterogeneity reflects the scale at which inequality is perceived as most salient . Here , we assessed relative wealth and inequality at the scale of the residential community , a salient arena of daily cooperation and competition ( Alami et al . , 2020; Gurven et al . , 2015; Gurven et al . , 2008b; Jaeggi et al . , 2016; von Rueden et al . , 2008; von Rueden et al . , 2014 ) . This local level is also similar in scale to group-level hierarchies in other social species that show hierarchy-associated stress responses ( Sapolsky , 2005; Snyder-Mackler et al . , 2016; Tung et al . , 2012 ) . Furthermore , substituting community-relative wealth with wealth relative to the whole Tsimane population made little difference for results ( Supplementary file 1a-1m ) , suggesting that the choice of scale within this relatively small-scale society did not matter . Modern technologies , such as television , may upset these comparisons and the functioning of hierarchy-related adaptations by making the global seem local; however , few Tsimane have regular access to television and other media . Nevertheless , it is possible that at least some Tsimane perceive inequality in reference to the local non-Tsimane population , or other regions of Bolivia , which was not captured by our study . Interacting with members of the dominant culture can be a source of stress ( Dressler , 1999; Konečná and Urlacher , 2017; von Rueden et al . , 2014 ) , even if the Tsimane are arguably doing fairly well financially compared to other rural Bolivians ( Godoy et al . , 2007 ) . Thus , we might not have been able to capture a relevant scale of comparison for some people , which could explain why associations at other scales were less consistent . However , this argument also applies to studies in high-income countries – where the relevant scales could be anything from neighborhoods to countries – and does not necessarily explain why results were inconsistent ( as opposed to simply weaker ) when measured at a less salient scale . Finally , another explanation for heterogeneous associations is that our measure of household wealth may capture several distinct dimensions of socio-economic status , with partly orthogonal effects on health . On the one hand , greater wealth affords more respect and influence within communities , which is associated with lower cortisol and better health among the Tsimane ( von Rueden et al . , 2014 ) and elsewhere ( Decker , 2000 ) ; this is likely the dimension captured by our subjective status data . On the other hand , household wealth is accumulated through participation in the market economy , which is associated with greater stress – higher cortisol , blood pressure – among the Tsimane ( von Rueden et al . , 2014 ) and elsewhere ( Dressler , 1999; Konečná and Urlacher , 2017 ) . The risks of different infectious diseases may also vary along these dimensions , with people who more frequently visit town and interact with outsiders possibly being more exposed to respiratory pathogens ( Kaplan et al . , 2020 ) . Thus , household wealth may in part be inconsistently associated with health because of these opposing processes . Several psychosocial variables were directly associated with health . Conflicts and depression were associated with lower BMI and blood pressure , perhaps indicating the effects of stress or lack of access to resources ( depression is associated with low productivity among the Tsimane; Stieglitz et al . , 2014 ) . Depression and non-social problems were associated with worse self-rated health , again possibly via stress or direct effects of resource availability . However , associations between wealth or inequality and health outcomes were not mediated when including psychosocial variables in models ( P3 ) , and there was almost no evidence for indirect effects proceeding through these pathways . An obvious limitation is that our sample sizes for the mediation analysis were smaller than for other analyses ( Supplementary file 1q-1s ) , though most were still large enough to capture any meaningful effect . It is also possible that our measures of psychosocial stress were inadequate , for example , a single urinary cortisol measure likely captures overall differences in cortisol excretion ( Yehuda et al . , 2003 ) , but does not capture changes in diurnal cortisol patterns that are typically associated with chronic stress ( García et al . , 2017; Miller et al . , 2007 ) . But the lack of mediation found here may also point to more nuanced mechanisms such as changes in physical activity related to different subsistence strategies or other lifestyle factors not accounted for here . For subsistence societies experiencing socio-economic change , whether relative status increases , decreases , or has no effect on stress and health may depend on the status measure and its association with social support . A study of four Tsimane communities found that influential men with greater social support had lower cortisol ( von Rueden et al . , 2014 ) , but higher cash income associated with higher cortisol ( von Rueden et al . , 2014; see also Konečná and Urlacher , 2017 ) . In another study of the Tsimane , higher incomes predicted lower BMIs , unless individuals had relatively more social support ( Brabec et al . , 2007 ) . It therefore remains unclear what mechanisms were responsible for the wealth-health associations found here , though hierarchy is known to affect immune function , and thereby infectious disease morbidity independently of stress and associated HPA activity ( Aiello et al . , 2018; Miller et al . , 2011; Snyder-Mackler et al . , 2020; Snyder-Mackler et al . , 2016 ) . In sum , we present the most comprehensive test of hierarchy-health associations in a subsistence society to date . In support of an evolutionary argument that conceptualizes hierarchy-health effects as stemming from evolved reaction norms adjusting people’s behavior and physiology to the rank and local competitive regime they find themselves in ( Daly and Wilson , 1997; Griskevicius et al . , 2011; Pepper and Nettle , 2014 ) , we found that wealth and inequality were associated with several health outcomes , though other associations were negligible or in the opposite direction to that predicted . In support of the argument that most hierarchy-health effects in high-income countries are caused by evolutionary mismatch ( Sapolsky , 2004 ) , we found that inequality was associated with blood pressure but in a range unlikely to affect health; however , this association could lead to hypertension , cardiovascular and metabolic disease as inequality further increases due to increased market integration and/or as novel foods and lifestyle factors enter the population ( Gurven et al . , 2012; Gurven et al . , 2016; Kaplan et al . , 2017; Kraft et al . , 2018 ) . Our study thus contributes to an evolutionary approach to public health that considers tradeoffs and mismatch as important links between socio-ecology , lifestyle , and health ( Eaton et al . , 2002; Wells et al . , 2017 ) .
All data were collected under the auspices of the Tsimane Health and Life History Project ( THLHP ) ( Gurven et al . , 2017 ) by a team of Bolivian medical professionals and Tsimane researchers . Wealth data were collected in 2006–2007 and 2013 . Here , we only included wealth data collected prior to a rare catastrophic flood in February 2014 that destroyed crops and household goods in the vast majority of Tsimane communities ( Trumble et al . , 2018 ) . Figure 1—figure supplement 1 summarizes how many individuals were included in the sample , out of all individuals ever sampled by the THLHP . Household wealth was assessed through an inventory of commonly owned items including traditional goods , that is , items manufactured from local organic materials ( e . g . , canoes , bows and arrows ) , market goods , that is , industrially produced items obtained through trade or purchase ( e . g . , bicycles , motorbikes ) , and livestock ( e . g . , pigs , cows ) , which were subsequently converted into their local market value in Bolivianos and summed ( Figure 1 ) . Objective household wealth arguably provides only an indirect measure of people’s subjective wealth and status ( Norton , 2013 ) , but these data were most widely available for this study . Furthermore , household wealth correlated significantly , albeit weakly , with subjective status ( Amir et al . , 2019; Woolard et al . , 2019; r = 0 . 17 , df = 147 , p<0 . 05 ) and subjective wealth rank ( r = 0 . 29 , df = 150 , p<0 . 001 ) . Previous work among the Tsimane ( Undurraga et al . , 2016 ) has also shown that more visible forms of wealth , such as the household items counted here , influenced subjective health more than less visible forms of wealth , such as the size of cultivated fields . To prevent differences in age sampling between villages from affecting wealth and inequality estimates , we followed Borgerhoff Mulder et al . , 2009 and adjusted wealth values for the age of the head of household by fitting generalized additive models for location scale and shape ( GAMLSS ) to the distribution of wealth-by-age to obtain wealth-by-age Z-scores . Wealth Z-scores derived from GAMLSS , representing centile values , were used in all analyses in part because wealth was skewed in distribution , and also expected to have diminishing returns at higher values ( i . e . , 100 Bolivianos are worth more to a poor individual than a wealthy one ) . However , to determine whether Z-scoring with GAMLSS altered results by normalizing the shape of the wealth distribution , we also repeated analyses with standardized wealth ( i . e . , [household wealth – population-average wealth]/standard deviation of population-average wealth ) , which preserves the skew . There were no qualitative differences in inference between the two methods , largely because Z-scoring with GAMLSS primarily affects outliers on the far high end of the distribution . Note that ‘Z-score’ can have two slightly different meanings; for wealth and BMI ( see below ) , we generally mean centile values from GAMLSS unless otherwise noted , for all other variables Z-scores refer simply to standardized values ( i . e . , [x – mean ( x ) ]/sd ( x ) ) . Mean wealth and wealth inequality at the community level ( for communities with ≥ 9 households ) were calculated after converting wealth Z-scores back into equivalent values in Bolivianos at age 50 ( see Figure 1 ) . We used the Gini coefficient to measure inequality; other inequality measures ( e . g . , median share , 90/10 ratio ) generally correlate highly ( r > 0 . 94 ) with Gini ( Kondo et al . , 2009 ) and were therefore not considered . In other studies , local scales of measuring inequality , such as at the community level used here , tend to produce smaller effects on health than those at larger scales , such as states or countries ( Kondo et al . , 2009; Wilkinson and Pickett , 2006 ) . In the Tsimane context , it is unclear whether that will be the case given low residential mobility and concentration of work and socializing within communities . However , Tsimane visit other communities and sporadically engage in market-based interaction with non-Tsimane , and comparisons with wealthier neighbors can contribute to Tsimane status aspirations ( Schultz , 2019 ) . Nevertheless , as mentioned above ( Study population ) , we consider the community to be the most relevant arena for status competition among Tsimane ( though substituting community-relative wealth with population-relative wealth made little difference; see Supplementary file 1a-1m ) . Note that most studies on health effects of inequality use income inequality ( but see Nowatzki , 2012 ) , which is less unequally distributed than wealth . Cash income among the Tsimane during this study period was sporadic and many households may have no income in a given sampling period , which leads to overestimated Ginis . We therefore preferred wealth and wealth inequality as a more reliable measure of households’ long-term access to resources and its distribution . The THLHP has been recording biomedical and anthropological data during roughly annual medical examinations and interviews by THLHP physicians and research assistants on an increasing number of communities since 2002 . Here , we included any data collected within 2 years of an individual’s wealth data , that is , the potential range of data was 2004–2009 and 2011–2015 . Table 1 summarizes how many individuals out of all the ones with wealth data ( see also Figure 1—figure supplement 1 ) were included for each outcome variable . Depressive symptoms were measured using an adapted 18-item questionnaire ( Stieglitz et al . , 2014 ) , the responses to which were summed to yield an overall depression score . The same interview also asked whether participants experienced conflicts with several kinds of social partners as well as non-social problems , such as food insecurity , illness , or debt; affirmative answers were summed to yield a composite measure of social conflicts and non-social problems , respectively . A household’s cooperation network was measured as the number of people from different households who helped in that household’s fields in a given year . Cortisol was measured in first-morning urine using enzyme-linked immunosorbent assays and corrected for specific gravity ( see von Rueden et al . , 2014 for details ) . BMI Z-scores were calculated by GAMLSS using Tsimane-specific growth curves ( Blackwell et al . , 2016b ) ( R package at: https://github . com/adblackwell/localgrowth ( Blackwell , 2021 ) copy archived at swh:1:rev:81ce799bdc6da90d48e5ad8afd6ad0f3b19494d2 ) as well as the total distribution of Tsimane adult BMIs , representing deviations from the local population average for a given age and sex . Diastolic and systolic blood pressures were measured by THLHP physicians using an aneroid sphygmomanometer . Self-rated general health was measured using a five-point scale from ( ‘very bad’ [1] to ‘excellent’ [5] ) . Morbidity at the time of the medical check-up was assessed by physicians using the International Classification of Disease , 10th edition ( ICD-10 classification ) and then grouped into 18 clinically meaningful categories following the Clinical Classifications System ( CCS ) ( https://www . hcup-us . ahrq . gov/toolssoftware/ccs/ccsfactsheet . jsp ) ; morbidities in any of these categories were summed to give a total morbidity score potentially ranging from 0 ( no morbidities ) to 18 ( at least one morbidity in each category ) . In addition , we also examined the presence/absence of infectious and parasitic diseases ( CCS 1 , hereafter ‘infections’ ) , diseases of the respiratory system ( CCS 8 , ‘respiratory illness’ ) , and diseases of the digestive system ( CCS 9 , ‘gastrointestinal illness’ ) , which represent the most common causes of morbidity and mortality in this population ( Gurven et al . , 2020; Gurven et al . , 2007 ) . See Table 2 for examples of the six most common diagnoses in these three categories . Distance to the town of San Borja was measured as nearest route ( whether by river or road ) from the center of the community and provides a proxy for access to modern amenities . Community size and household size were summarized from complete population censuses conducted regularly by the THLHP . Thus , they include all individuals , not just those sampled for wealth or other covariates . Prior to analysis , all variables except binary variables were standardized into Z-scores . Urinary cortisol was log transformed prior to standardization to reduce skew , as is common practice . All outcomes were modeled as Gaussian , except the presence/absence of specific morbidities ( Bernoulli ) . Each analysis modeled an individual-level outcome as a function of individual- , household- , and community-level characteristics ( Table 1 ) . Thus , we fit the following base model for each outcome: Outcomeijkl ~ β0 + ( β1 * Sexj ) + ( β2 * Agej ) + ( β3 * relative household wealthk ) + ( β4 * Community-level Ginil ) + ( β5 * Community-level mean wealthl ) + ( β6 * Community Sizel ) + ( β7 * Distance of community to market townl ) + ( β8 * Household Sizel ) uj+ uk + ul+ eijkl wherein the subscripts denote measurement i , individual j , household k , and community l , respectively . β0 is the intercept , all other βs are slopes , us are random intercepts , and e is the residual error ( not available for Bernoulli responses ) . Variance inflation factors ( VIFs ) indicated virtually no collinearity among predictors ( all VIFs < 3 ) . In order to test whether potential wealth-health associations were mediated by psychosocial stress , we reran all health models ( blood pressure , self-rated health , total morbidity , infections , respiratory and gastrointestinal illness ) with pertinent psychosocial variables as covariates and used the mediation function in the sjstats package ( Lüdecke , 2021 ) to estimate direct and indirect effect . In addition , we also ran a series of exploratory analyses in which we added interaction terms . We used Bayesian multilevel models fit with the brms package v . 2 . 13 . 5 ( Bürkner , 2017 ) in R 4 . 0 . 2 for all analyses . All models used regularizing priors ( fixed effects: normal , mean = 0 , SD = 1; random effects: half-Cauchy , location = 0 , scale = 2 ) , which imposes conservatism on parameter estimates and reduces the risk of inferential errors ( Gelman et al . , 2013; McElreath , 2020 ) . All models converged well as assessed by inspecting trace plots and standard diagnostics ( all Rhat < 1 . 01 ) . All data and R code are available at https://doi . org/10 . 5281/zenodo . 4567498 with any updates at https://github . com/adblackwell/wealthinequality ( Jaeggi et al . , 2021 , copy archived at swh:1:rev:da16ac6b20732fe1939478450d81ac32fdcce202 ) . | Poverty is bad for health . People living in poverty are more likely to struggle to afford nutritious food , lack access to health care , or be overworked or stressed . This may make them susceptible to chronic diseases , contribute to faster aging , and shorten their lifespans . In high-income countries , there is growing evidence to suggest that a person’s ‘rank’ in society also impacts their health . For example , individuals who have a lower position in the social hierarchy report worse health outcomes , regardless of their incomes . But it is unclear why living in an unequal society or having a lower social status contributes to poorer health . One possibility is that inequalities in society are creating a stressful environment that leads to worse physical and mental outcomes . It is thought that this stress largely comes from how humans evolved to prioritize reaching a higher social status over having a long and healthy life . If this is the case , this would mean that the link between social status and health would also be present in non-industrialized communities where social hierarchies tend to be less pronounced . To test this , Jaeggi , Blackwell et al . studied the Indigenous Tsimane population in Bolivia who live in small communities and forage and farm their own food . The income and relative wealth of 870 households from 40 Tsimane communities were compared against various outcomes , including symptoms associated with depression , stress hormone levels , blood pressure , self-rated health and several diseases . Jaeggi , Blackwell et al . found poverty and inequality did not negatively impact all of the health outcomes measured as has been previously reported for industrialized societies . However , blood pressure was higher among people with lower incomes or those who lived in more unequal communities . But because the Tsimane people generally have low blood pressure , the differences were too small to have much effect on their health . People who lived in more unequal communities were also three times more likely to have respiratory infections , but the reason for this was unclear . This shows that social determinants such as a person’s wealth or inequality can affect health , even in communities with less rigid social hierarchies . In industrial societies the effect may be worse in part because they are compounded by lifestyle factors , such as diets rich in fat and sugar , and physical inactivity which can also increase blood pressure . This information may help policy makers reduce health disparities by addressing some of the social determinants of health and the lifestyle factors that cause them . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"medicine"
] | 2021 | Do wealth and inequality associate with health in a small-scale subsistence society? |
Predicting evolutionary change poses numerous challenges . Here we take advantage of the model bacterium Pseudomonas fluorescens in which the genotype-to-phenotype map determining evolution of the adaptive ‘wrinkly spreader’ ( WS ) type is known . We present mathematical descriptions of three necessary regulatory pathways and use these to predict both the rate at which each mutational route is used and the expected mutational targets . To test predictions , mutation rates and targets were determined for each pathway . Unanticipated mutational hotspots caused experimental observations to depart from predictions but additional data led to refined models . A mismatch was observed between the spectra of WS-causing mutations obtained with and without selection due to low fitness of previously undetected WS-causing mutations . Our findings contribute toward the development of mechanistic models for forecasting evolution , highlight current limitations , and draw attention to challenges in predicting locus-specific mutational biases and fitness effects .
Adaptation requires the realization of beneficial mutations . As self-evident as this may be , predicting the occurrence of beneficial mutations and their trajectories to improved fitness is fraught with challenges ( Lässig et al . , 2017 ) . Nonetheless progress has been made for phenotypically diverse asexual populations subject to strong selection . Effective approaches have drawn upon densely sampled sequence data and equilibrium models of molecular evolution to predict amino acid preferences at specific loci ( Luksza and Lässig , 2014 ) . Predictive strategies have also been developed based on selection inferred from the shape of coalescent trees ( Neher et al . , 2014 ) . In both instances the models are coarse-grained and sidestep specific molecular and mutational details . There is reason to by-pass molecular details: mutation , being a stochastic process , means that for the most part details are likely to be idiosyncratic and unpredictable . But an increasing number of investigations give reason to think otherwise – that adaptive molecular evolution might follow rules ( Yampolsky and Stoltzfus , 2001; Pigliucci , 2010; Stern , 2013; Laland et al . , 2015 ) . This is particularly apparent from studies of parallel molecular evolution ( Colosimo et al . , 2005; Woods et al . , 2006; Ostrowski et al . , 2008; Flowers et al . , 2009; Meyer et al . , 2012; Tenaillon et al . , 2012; Zhen et al . , 2012; Herron and Doebeli , 2013; Galen et al . , 2015; Bailey et al . , 2017; Kram et al . , 2017; Stoltzfus and McCandlish , 2017 ) , and particularly from studies that show parallel evolution to be attributable – at least in part – to factors other than selection ( McDonald et al . , 2009; Lind et al . , 2015; Bailey et al . , 2017 ) . A standard starting position for predicting adaptive evolution recognises the importance of population genetic parameters including mutation rate , generation time , population size , selection and more recently information on the distribution of beneficial fitness effects . An often used approach appeals to ‘origin-fixation’ models that estimate the probability that selection will realise particular mutational trajectories based on the assumption that the rate of evolution is a function of the rate at which mutations originate , multiplied by their fixation probability ( for review , see ( McCandlish and Stoltzfus , 2014 ) , for application , see for example ( Weinreich et al . , 2006 ) ) . Leaving aside a number of simplifying and restrictive assumptions , population genetic approaches such as those encompassed by origin-fixation models require pre-existing knowledge of ( or assumptions concerning ) mutation rate to a particular phenotype and selection coefficients ( obtained in the absence of frequency dependent effects ) and thus have predictive power only in specific situations and with the benefit of hindsight ( McCandlish and Stoltzfus , 2014 ) . Looking to the future , there is hope that one day it might be possible to predict the course of evolutionary change in response to specific environmental challenges in much the same way as gene function , regulation and interactions can be predicted today based on knowledge of nucleotide sequence data . A central issue is to define the necessary information . Mechanistic understanding of the connection between genotype and phenotype combined with knowledge of the genomic bases of mutational bias offer opportunities for progress . Mutations arise randomly with respect to utility , but genetic architecture can influence the translation of mutation into phenotypic variation: the likelihood that a given mutation generates phenotypic effects depends on the genotype-to-phenotype map ( Alberch , 1991; Gompel and Prud'homme , 2009; Stern and Orgogozo , 2009; Rainey et al . , 2017 ) . Thus , the functions of gene products and their regulatory interactions provide information on likely mutational targets underpinning particular phenotypes . For example , consider a hypothetical structural determinant subject to both positive and negative regulation and whose over-expression generates a given adaptive phenotype . Assuming a uniform distribution of mutational events , mutations in the negative regulator ( and not the positive activator ) will be the primary cause of the adaptive phenotype . This follows from the fact that loss-of-function mutations are more common than gain-of-function mutations . Indeed , an emerging rule indicates that phenotypes determined by genetic pathways that are themselves subject to negative regulation are most likely to arise by loss-of-function mutations in negative regulatory components ( McDonald et al . , 2009; Tenaillon et al . , 2012; Lind et al . , 2015; Fraebel et al . , 2017 ) . Mutation is not equally likely at each nucleotide of a given genome ( Lind and Andersson , 2008; Lynch , 2010; Seier et al . , 2011; Foster et al . , 2015; Reijns et al . , 2015; Sankar et al . , 2016; Stoltzfus and McCandlish , 2017 ) . Numerous instances of mutational bias have been reported . Prime examples are simple sequence repeats such as homopolymeric nucleotide tracts or di- , tri- and tetrameric repeats that mutate at high frequency via slipped strand mispairing ( Levinson and Gutman , 1987 ) . These readily identifiable sequences define contingency loci in obligate human pathogens and commensals ( Moxon et al . , 1994 ) and are widespread in eukaryotic genomes ( Tautz and Renz , 1984 ) . The behaviour of contingency loci can be further modulated by defects in components of methyl-directed mismatch repair systems ( Richardson and Stojiljkovic , 2001; Martin et al . , 2004; Hammerschmidt et al . , 2014; Heilbron et al . , 2014 ) . Certain palindromic structures also lead to mutational bias ( Viswanathan et al . , 2000; Lovett , 2004 ) and promote amplification events that can increase target size for additional mutations ( Roth et al . , 1996; Kugelberg et al . , 2010; Reams and Roth , 2015 ) Transition-transversion bias ( Stoltzfus and McCandlish , 2017 ) and elevated mutation rates at CpG sites ( Galen et al . , 2015 ) can further skew the distributions of mutational effects . Additional bias arises from the chromosomal neighbourhood of genes under selection ( Steinrueck and Guet , 2017 ) , the location of genes with regard to interactions with DNA replication/transcription machineries ( Sankar et al . , 2016 ) , and environmental factors that affect not only mutation rate but also the spectra of mutational events ( Krašovec et al . , 2017; Maharjan and Ferenci , 2017; Shewaramani et al . , 2017 ) . Beyond the genotype-to-phenotype map and mutational biases , predicting adaptive evolution requires ability to know a priori the fitness effects of specific mutations . At the present time there is much theoretical and empirical interest in the distribution of fitness effects ( DFE ) ( Eyre-Walker and Keightley , 2007 ) — and particularly the DFE of beneficial mutations ( Orr , 2005 ) — because of implications for predicting the rate of adaption and likelihood of parallel evolution ( de Visser and Krug , 2014 ) , but knowledge of the shape of the distribution is insufficient to connect specific mutations to their specific effects , or to their likelihood of occurrence . Such connections require a means of knowing the connection between mutations and their environment-specific fitness effects . This is a tall order . A starting point is to understand the relationship between all possible mutational routes to a particular phenotype and the set that are realised by selection . Here we take a bacterial system in which the genetic pathways underpinning evolution of the adaptive ‘wrinkly spreader’ ( WS ) type are known and use this to explore the current limits on evolutionary forecasting . Pseudomonas fluorescens SBW25 growing in static broth microcosms rapidly depletes available oxygen establishing selective conditions that favour mutants able to colonise the air-liquid interface . The most successful mutant-class encompasses the WS types ( Ferguson et al . , 2013; Lind et al . , 2017b ) . These types arise from mutational activation of diguanylate cyclases ( DGCs ) that cause over-production of the second messenger c-di-GMP ( Goymer et al . , 2006; McDonald et al . , 2009 ) , over-production of an acetylated cellulose polymer ( Spiers et al . , 2002; Spiers et al . , 2003 ) and ultimately formation of a self-supporting microbial mat ( Figure 1A ) . McDonald et al . ( McDonald et al . , 2009 ) showed that each time the tape of WS evolution is re-run mutations generating the adaptive type arise in one of three DGC-encoding pathways ( Wsp , Aws , or Mws ) ( Figure 1A ) . Subsequent work revealed that when these three pathways are eliminated from the ancestral type evolution proceeds along multiple new pathways ( Lind et al . , 2015 ) resulting in WS types whose fitnesses are indistinguishable from those arising via mutations in Wsp , Aws , or Mws . Parallel evolution of WS involving preferential usage of the Wsp , Aws and Mws pathways is therefore not explained by selection: repeated use of Wsp , Aws and Mws stems from the fact that these pathways are subject to negative regulation and thus , relative to pathways subject to positive regulation , or requiring promoter-activating mutations , gene fusion events , or other rare mutations , present a large mutational target ( Lind et al . , 2015 ) . Given repeatability of WS evolution , knowledge of the Wsp/Aws/Mws pathways , plus genetic tools for mechanistic investigation — including capacity to obtain WS mutants in the absence of selection — the WS system offers a rare opportunity to explore the extent to which knowledge of the genotype-to-phenotype map can lead mechanistic models for evolutionary forecasting . Our findings show that short-term mechanistic-level predictions of mutational pathways are possible , but also draw attention to challenges that stem from current inability to a priori predict locus-specific mutational biases and environment-specific fitness effects .
Knowledge of the rate at which mutation generates WS types via each of the Wsp , Aws and Mws pathways — unbiased by the effects of selection — provides a benchmark against which the predictive power of null models can be appraised . To achieve such measures we firstly constructed a set of genotypes containing just one of the three focal pathways: PBR721 carries the Wsp pathway but is devoid of Aws and Mws , PBR713 carries the Aws pathway but is devoid of Wsp and Mws , while PBR712 harbours the Mws pathway but is devoid of Wsp and Aws . Into each of these genotypes a promoterless kanamycin resistance gene was incorporated immediately downstream of the promoter of the cellulose-encoding wss operon and transcriptionally fused to an otherwise unaffected wss operon ( Figure 1B ) . In the ancestral SM genotype the cellulose promoter is inactive in shaken King’s Medium B ( KB ) broth ( Spiers et al . , 2002 ) and thus the strain is sensitive to kanamycin . When a WS-causing mutation occurs , the wss promoter becomes active resulting in a kanamycin-resistant WS type ( Fukami et al . , 2007; McDonald et al . , 2011 ) . Individual growth of this set of three genotypes in shaken KB , combined with plating to detect kanamycin-resistant mutants , makes possible a fluctuation assay ( Luria and Delbrück , 1943; Hall et al . , 2009 ) from which a direct measure of the rate at which WS mutants arise can be obtained . Importantly , because WS types are maladapted in shaken broth culture , the screen for kanamycin-resistant clones allows the pathway-specific mutation rate to WS to be obtained without the biasing effects of selection for growth at the air-liquid interface ( Figure 1B ) . The results are shown in Figure 2 . The mutation rate was highest for the Aws pathway ( 6 . 5 × 10−9 ) ; approximately double that of Wsp ( 3 . 7 × 10−9 ) and an order of magnitude higher than that of the Mws pathway ( 0 . 74 × 10−9 ) ( Figure 2 ) . The rate at which WS mutants arose from the ancestral genotype in which the three pathways are intact ( 11 . 2 × 10−9 ) was approximately the sum of the rates for the three pathways ( 11 . 0 × 10−9 ) confirming that the Wsp , Aws and Mws pathways are the primary routes by which WS types evolve ( Lind et al . , 2015 ) . That the Aws pathway has the greatest capacity to generate WS is surprising given the smaller target size ( three genes and 2 . 3 kb compared to seven genes ( 8 . 4 kb ) in the Wsp pathway ) . Much is known about the function and interactions among components of each of the three focal pathways . This knowledge allows development of models that capture the dynamic nature of each pathway and thus allow predictions as to the likelihood that evolution will precede via each of the three mutational routes . An unresolved issue is the extent to which these models match experimental findings . Following a brief description of each pathway we describe the models . The 8 . 4 kb Wsp pathway is a chemotaxis-like system ( Goymer et al . , 2006; Güvener and Harwood , 2007; Römling et al . , 2013; Micali and Endres , 2016 ) composed of seven genes with the first six genes ( wspA-wspF ) being transcribed as a single unit and the last ( wspR from its own promoter ( Bantinaki et al . , 2007 ) . WspA ( PFLU1219 ) is a methyl-accepting chemotaxis ( MCP ) protein that forms a complex with the CheW-like scaffolding proteins WspB ( PFLU1220 ) and WspD ( PFLU1222 ) . WspA senses environmental stimuli and transmits the information via conformational changes in the WspA/WspB/WspD complex to effect activity of WspE ( PFLU1223 ) , a CheA/Y hybrid histidine kinase response regulator . WspE activates both the WspR ( PFLU1225 ) diguanylate cyclase ( DGC ) and the CheB-like methylesterase WspF ( PFLU1224 ) following transference of an active phosphoryl group . The activity of WspA is modulated by methylation by the constitutively active CheR-like methyltransferase WspC ( PFLU1221 ) that transfers methyl groups to conserved glutamine residues on WspA . The demethylase WspF serves to remove these groups when in the phosphorylated active form . WS mutants are known to arise by mutations in the WspF negative regulator and also in the WspE kinase ( McDonald et al . , 2009 ) . In vitro manipulations of WspR that abolish repression of the DGC domain by the response regulator domain are known , but have never been observed to occur in experimental populations ( Goymer et al . , 2006 ) . The 2 . 3 kb aws operon contains three genes transcribed from a single promoter ( awsXRO ) . Homologous genes in Pseudomonas aeruginosa ( yfiRNB , PA1121-1119 ) have been characterised in detail ( Malone et al . , 2010; Malone et al . , 2012; Xu et al . , 2016 ) . The outer membrane lipoprotein AwsO ( PFLU5209 ) has an OmpA domain , a signal peptide and binds to peptidoglycan . AwsO is thought to be the sensor whose activity is modulated in response to envelope stress ( Malone et al . , 2012 ) . AwsO sequesters the periplasmic protein AwsX ( PFLU5211 ) at the outer membrane . AwsX functions as a negative regulator of the DGC AwsR ( PFLU5210 ) in the inner membrane . Both increased binding of AwsX to AwsO or loss of negative regulation by inactivation of the interaction between AwsX and AwsR can lead to WS ( McDonald et al . , 2009; Malone et al . , 2010; Malone et al . , 2012 ) . The 3 . 9 kb mwsR gene ( PFLU5329 ) is known as morA ( PA4601 ) in Pseudomonas aeruginosa , and encodes a predicted membrane protein with both a DGC domain that produces c-di-GMP and a phosphodiesterase ( PDE ) domain that degrades c-di-GMP . Little is known of the molecular details determining its function , but both catalytic domains appear to be active ( Phippen et al . , 2014 ) . Deletion of the PDE domain results in a WS phenotype with activity being dependent on a functional DGC domain ( McDonald et al . , 2009 ) . If the specific effects of changing each nucleotide ( and sets of nucleotides ) were known then models for each pathway would not be required . Here , we show how knowledge of genetic architecture can be used to build models that predict the likelihood that mutations generating WS types will arise in a given pathway – and even in specific genes . In the following section we present four null models that incorporate increasing levels of information concerning the genotype-to-phenotype map . The goal of these models is two-fold: firstly to demonstrate that incorporation of knowledge of genetic architecture allows development of models with explanatory value and secondly , to define minimal necessary information for reliable forecasting . The results are summarised in Figure 3 , which displays the experimental data from Figure 2 ( Figure 3A ) , along with predictions from each of the null models . Null Model I is intentionally naïve . It uses sequence length as a proxy for mutational target size , but ignores genetic organisation , function of predicted proteins and interaction among proteins . The model assumes that mutational target size is proportional to the number of nucleotides at a given locus and thus the probability that a given pathway is used to generate WS relative to another is simply the ratio of the probability of generating WS for the two focal pathways . For any one pathway the probability that a mutation generates WS is given by 1- ( 1-p ) n , where p is the probability of a mutation at a nucleotide and n is the number of nucleotides in the pathway . If the mutation probability is low such that the expected number of mutations in a pathway is below 1 , that is , np <<1 then the binomial approximation can be used: 1- ( 1-p ) n = np . Thus the probability that evolution follows the Wsp pathway over the Aws pathway is: 8400 p1/ 2300 p2 = 3 . 65 p1/p2 , where p1 and p2 are the mutation rates for each pathway . Assuming equivalency of mutation rate , p1 = p2 , evolution is predicted to proceed via the Wsp pathway 3 . 65 times more often than via Aws , with evolution predicted to proceed via Mws 1 . 65 times more often than Aws . Comparison with experimental data shows a departure both in terms of the priority of pathways used by evolution and the frequency of pathway usage ( Figure 3A versus 3B ) . Null Model II builds on Model I but only in a marginal sense . It recognises that nucleotides defining loci of interest are organised into genes , and therefore adopts gene number as a proxy for mutational target size . As above , the probability that a pathway is used is 1- ( 1-p ) n = np , but in this instance n is the number of genes . Model II predicts that mutations in Wsp generate WS types 2 . 33 times more often than mutations in Aws , with mutations in Mws being 3-fold less likely to generate WS compared to mutations in Aws ( Figure 3C ) . This marginal adjustment makes little difference to the fit between experimental data and predictions . Past work has shown the explanatory value of information that comes from knowledge of gene ( protein ) function and interactions ( McDonald et al . , 2009; Lind et al . , 2015 ) . These form the basis of Null Model III . The relevant functions and interactions are depicted in Figure 4 as reaction diagrams that reflect how changes in different interactions affect the production of WS types . Organizing interactions within each pathway according to reactions has the advantage that it allows for a standard mathematical description of the biochemical dynamics using differential equations ( see Figure 4 and Figure 4—figure supplement 1 to 3 , ) . An additional advantage is that such an approach allows ready incorporation of new experimental data including interaction type and interaction strength . One immediate consequence of this approach is that production of WS types is described entirely through biochemical reactions . Knowing whether a WS-type is generated by a particular mutation amounts to determining which reaction rates are altered by the mutation and whether those changes affect the likelihood of producing WS types . This leads to an approach where mutations in components can be classified according to their effects on reaction rates: enabling mutations increase reaction rates , whereas disabling mutations decrease reaction rates . Based on the mathematical models for the pathways , each reaction rate can be altered ( either increased or decreased ) to generate higher levels of the key WS effector ( the active DGC ) . For example , in the Wsp pathway , if reactions 2 or 6 experience a disabling mutation , or alternatively , if any of the reactions 1 , 3 , 4 , or 5 experience an enabling mutation , then the level of activated ( phosphorylated ) WspR is increased . For now , it is assumed that increased activity of the relevant DGC generates a WS type . Reactions whose enabling/disabling mutations increase the amount of WS effector are termed enabling/disabling reactions . By comparing the number of disabling and enabling reactions in different pathways it is possible to calculate the relative likelihood that evolution uses each mutational pathway . The Wsp pathway encompasses two disabling and four enabling reactions , whereas the Aws pathway is defined by one disabling and three enabling reactions . Mws encompasses a single disabling reaction and two enabling reactions . If disabling mutations are as likely as enabling mutations , then the likelihood that a pathway will be used by evolution is simply the ratio of the total number of enabling and disabling reactions: 6:4:3 ( Wsp:Aws:Mws ) . If instead , disabling mutations are much more likely than enabling mutations , then enabling reactions can be ignored and the likelihood that a pathway will be used by evolution becomes simply the ratio of the number of disabling reactions: 2:1:1 ( Wsp:Aws:Mws ) ( Figure 3D left hand panel ) . If the reverse is true , then the likelihood of use is a ratio of the number of enabling reactions: 4:3:2 ( Wsp:Aws:Mws ) ( Figure 3D right hand panel ) . In all cases , mutations in the Wsp pathway are predicted to be 1 . 3–2 times more likely to generate WS than mutations in the Aws pathway , with the Aws pathway being the target of evolution 1–1 . 5 times more often than the Mws pathway . Although still inflating the importance of the Wsp pathway relative to Aws , Mws is more prominent than in null Model II . Null Model IV takes a further step toward mechanistic accuracy by endowing interactions with pleiotropic and continuous effects . This removes two simplifying assumptions of null Model III likely to limit predictive power . In Model III mutations affect one reaction at a time , but mutations in certain components , for example wspA , can affect more than a single reaction ( WspA appears in reactions 1–4 ) . Additionally , in Model III , changes in reaction rate are assumed to be binary , however reactions may have a range of effects on the evolution of WS types . To accommodate pleiotropic effects , null Model IV systematically considers all combinations of enabling and disabling changes to reaction rates and determines the likelihood that a WS type is generated . An example of one set of the possible mutations ( mi ) in Wsp is 1 , –1 , 0 , 0 , 0 , 0 ( an increase in r1 , a decrease in r2 , but no change in r3 , r4 , r5 , or r6 ( Figure 3A ) ) . Since the Wsp pathway has six reaction rates this amounts to 36 or 729 total combinations . However , note that reaction 3 does not share any reactants or products with reactions 5 or 6 . Thus , mutations such as 0 , 0 , 1 , 0 , 1 , 0 or 0 , 0 , 1 , 0 , 0 , 1 are not considered because they require mutations in two separate genetic components . To accommodate a range of effects null Model IV simulates enabling/disabling changes of different magnitudes and determines the resulting effect on the respective effector DGC ( see Materials and methods ) . Briefly , the approach addresses the lack of information concerning biochemical reaction rates and molecular concentrations in the mathematical models describing WS-producing pathway dynamics . By repeatedly sampling from the space of all possible reaction rates , initial concentrations , and magnitudes of effects , this approach computes the probability that a particular set of mutations ( mi ) , for example 1 , –1 , 0 , 0 , 0 , 0 , results in a wrinkly spreader . This probability is represented as the conditional probability P ( WS |mi ∈ Wsp ) , which motivates a Bayesian formulation to compare the relative probability that the different pathways produce WS . To this end , the probability that a particular pathway will be used is decomposed into two terms: the probability that a particular set of mutations ( mi ) occurs in Wsp ( or Aws , or Mws ) represented as P ( mi ∈ Wsp ) and the probability that those mutations give rise to a wrinkly spreader represented as P ( WS |mi ∈ Wsp ) ( or Aws , or Mws ) . ( 1 ) P ( WS∩m∩Wsp ) =∑iP ( WS|mi∈Wsp ) P ( mi∈Wsp ) To estimate P ( mi ∈ Wsp ) we assume fixed probabilities of enabling and disabling changes and compute the product . Thus , the probability of mi = 1 , –1 , 0 , 0 , 0 , 0 is pepd ( 1 − pe − pd ) 4 , where pe is the probability of a mutation with an enabling effect and pd is the probability of a mutation with a disabling effect . Recognising the value of accommodating the possibility of localised mutational bias we note that pe and pd can be adjusted for the affected reactants . The second term , P ( WS |mi ∈ Wsp ) , relies on our sampling methodology and describes the probability that a set of disabling/enabling changes of different magnitudes will yield a WS type ( see Materials and methods ) . Despite the mechanistic advances incorporated into null Model IV the Wsp pathway is still predicted to be the pathway most commonly used by evolution . The extreme cases in which disabling mutations are more probable than enabling mutations ( and vice versa ) are shown in Figure 3E . The results of simulations for the full range of pd and pe values are shown in Figure 5 . Figure 5A shows that the Wsp pathway is predicted to be the target of mutation 1 . 5–1 . 9 times more often than the Aws pathway while Figure 5B shows that the Mws pathway is predicted to be the target of mutation 0 . 4–0 . 5 times less often that the Aws pathway . While these results agree with the experimental data showing Mws to be least likely pathway to be followed , the predictions are at odds with the data showing WS types to be twice as likely to arise from mutation in Aws , versus Wsp . The causes of this discrepancy are accounted for in the next section ‘analysis of mutants reveals sources of mutational bias’ . The Bayesian approach of null Model IV has additional utility: it allows prediction of genes likely to be affected by mutation . Predictions as to specific genetic targets come from appraisal of the relative importance of each reaction in terms of generating WS types ( Figure 5—figure supplements 1 and 2 ) . While it is recognised that a majority of WS mutations arise from defects in negative regulators of DGCs , such as WspF and AwsX ( McDonald et al . , 2009; Lind et al . , 2015 ) , further predictions are possible based on impacts of alterations in gene function on reaction rates . For example , with reference to the Wsp pathway ( Figure 5—figure supplement 1 ) , there are two reactions ( 2 and 6 ) that are affected by WspF function: r2 describes the rate of removal of methyl groups from the WspA signalling complex and r6 the rate at which WspF is activated by transfer of active phosphoryl groups from the WspE kinase . Loss-of-function ( disabling ) mutations being much more common than gain-of-function ( enabling ) mutations means that WspF , WspA and WspE are all likely targets . The null model predicts that in the area of parameter space in which Wsp is most likely compared to Aws , 45% of the time WS will be generated when the second reaction , r2 , is altered ( Figure 5—figure supplement 1 ) . The same is true for reaction r6 . Thus the presence of a negative regulator is predicted to extend the mutational target size well beyond the gene itself . This is also true for Aws , where r3 is the main contributor to the WS phenotype in the case where disabling change is more common than enabling change . Here mutations are predicted not only in the negative regulator AwsX , but also in the interacting region of the DGC AwsR ( Figure 5—figure supplement 2 ) . Loss-of-function mutations in negative regulators and their interacting partners are not the only predicted targets . For Wsp r1 , r3 , r4 , and r5 are altered approximately 5% of the time in the parameter region where disabling mutations are more common than enabling mutations , but contribute more when the rate of enabling mutations is increased ( Figure 5—figure supplement 1 ) . Enabling mutations based on null Model IV are likely to be found in WspC ( increasing r1 ) , WspABD ( increasing r3 ) , WspABD/WspE ( increasing r4 ) and WspR ( increasing r5 ) ( Figure 4A ) . For Aws , enabling mutations are predicted to increase reaction r1 by mutations causing constitutive activation of AwsO , r2 increasing binding of AwsO and AwsX and r4 increasing formation of the active AwsR dimer ( Figure 4B , Figure 5—figure supplement 2 ) . In summary , high rates of WS mutations are predicted for wspF , wspE , wspA , awsX and awsR with lower rates for wspC , wspR and awsO . Several of these predictions sit in accord with previous experimental observations , however , notable are predictions that evolution might also target wspA and wspR , two genes that have not previously been identified as mutational causes of WS types ( McDonald et al . , 2009 ) . There are several reasons why predictions from the models might be out of kilter with experimental data on mutation rates . We firstly looked to the distribution of WS generating mutations among the 109 mutants collected during the course of the fluctuation assays . Of the 109 mutants , 105 harboured a mutation in wsp ( 46 mutants ) , aws ( 41 mutants ) or mws ( 18 mutants ) ( Figure 6A , Figure 6—source data 1 ) . The remaining four had mutations in previously described rare pathways ( PFLU0085 , PFLU0183 ) , again confirming that these non-focal pathways produce just a fraction of the total set of WS mutants ( Lind et al . , 2015 ) . The distribution of mutations for each of the three pathways is indicative of bias . As shown in Figure 6B , almost 29% of all WS-causing mutations ( adjusted for differences in mutation rates between the three pathways ) were due to an identical 33 base pair in-frame deletion in awsX ( Δt229-g261 , ΔY77-Q87 ) , while a further 13% were due to an identical mutation ( 79 a->c , T27P ) in awsR . At least 41 different mutations in Aws lead to WS: if mutation rates were equal for these sites the probability of observing 20 identical mutations would be extremely small . In fact 10 million random samplings from the observed distribution of mutations failed to recover this bias . While the Wsp pathway also contains sites that were mutated more than once ( six positions were mutated twice , one site three times and one five times ) , sources of mutational bias in Wsp were less evident than in Aws ( Figure 6B ) . The mathematical models presented above assumed no mutational bias , but the null models can be modified to incorporate such bias . With focus on the AwsX hotspot , we show this for models II-IV . Null Model IV of the Wsp , Aws , and Mws pathways allow estimation of the relative probability that a WS is produced by enabling or disabling changes in a certain reaction rate . This means that in order to evaluate the success of our predictions , the 105 WS mutations must be connected to their likely effects on reaction rates . For some mutations this is straightforward . If a mutation completely disables the negative regulator WspF , this will reduce r2 and r6 , thereby producing a wrinkly spreader . In other cases interpretations of likely effects are more difficult and requires knowledge of the molecular functions of the proteins and their interactions , which needs to be obtained from literature , bioinformatics and structural modelling . For example , mutations in WspA can potentially produce WS by disruption of the interaction with WspF ( reducing r2 ) , increase methylation by WspC ( increasing r1 ) , increase signalling rate ( r3 ) or increase phosphorylation of WspE ( r4 ) . A summary of the interpretation of the molecular effects of mutations and how they are connected to reaction rates is available in Table 1 with a more detailed description in Appendix 1 . Mutations were found in five genes in the Wsp pathway . The majority of the mutations were found in the negative regulator WspF or its interacting partners WspE and WspA . These mutations likely reduced reaction rates of r2 and r6 in accord with predictions of null Model IV . The remaining mutations in WspC and WspR were less common , in accord with predictions and likely due to increases in reaction rates r1 and r5 , respectively . Results for the Aws pathway were also in line with predictions with the majority of the mutations arising in the negative regulator AwsX or the interacting part of AwsR ( both decreasing r3 ) , with rare mutations in AwsO and other parts of AwsR . For MwsR , loss-of-function mutations were expected primarily in the phosphodiesterase ( PDE ) domain , but mutations were found in both diguanylate cyclase ( DGC ) and PDE domains . A structural analysis revealed that mutations were clustered in the interface between the domains and unlikely to disrupt PDE function . These mutations most likely change the dynamics between the DGC and EAL domains in a way that increases production of c-di-GMP . The null models – and especially null Model IV – successfully predicted the mutational targets and explained mutation rates to WS when knowledge of mutational hot spots was included . To what degree can such knowledge allow prediction of the outcome of the original experimental evolution under selection ( McDonald et al . , 2009 ) ? A comparison between the frequencies of mutations isolated here ( without selection ) with those isolated under selection , reveals several notable differences ( Figure 8 ) . The most obvious difference is in use of the Wsp pathway , which is most commonly used ( 15/24 ) under selection and yet produces WS types at a lower rate than the Aws pathway . Differences are also apparent in the spectrum of wsp mutations , with no wspA mutations being found under selection despite being the most commonly mutated gene without selection ( 15/46 ) , and the previous failure to detect wspR mutants in a screen of 53 WS mutants ( Goymer et al . , 2006 ) . The most obvious explanation for the differences in mutational spectra between WS isolated with or without selection ( Figure 8 ) is that certain mutants have a lower fitness and thus their relative frequencies will be lower in the original evolution experiment ( McDonald et al . , 2009 ) . We measured the fitness of representative WS types with mutations in each of the mutated genes ( wspA , wspC/D , wspE , wspF , wspR , awsX , awsR , awsO , mwsR ) in 1:1 competitions against a reference WspF ΔT226-G275 deletion mutant marked with GFP ( Figure 9 ) . This type of fitness data should be interpreted with caution because the fitness of WS mutants are frequency-dependent ( Rainey and Travisano , 1998 ) and some WS mutants are superior in early phase attachment as opposed to growth at the air-liquid interface ( Lind et al . , 2015 ) . Nevertheless , the competition experiments provide an estimate of fitness when several different WS mutants compete at the air-liquid interface ( a likely situation given a ~ 10−8 mutation rate to WS and a final population size of >1010 ) . The fitness data account for the over- or under-representation of some WS mutants when grown under selection ( McDonald et al . , 2009 ) compared to those uncovered without selection ( as reported here ) . The three wspF mutants , the wspC-wspD fusion , and the wspE mutants have similar fitness ( p>0 . 38 , except for WspE K734E that has slightly higher fitness ( p=0 . 027 ) , two-tailed t-tests ) . In contrast , both wspA mutants are slightly less fit ( p<0 . 0214 , two-tailed t-test ) and both wspR mutants are severely impaired ( p<0 . 00007 , two-tailed t-test ) ( Figure 9 ) . This sits in accord with previous work in which mutations generating WS obtained with selection have been detected in wspF and wspE , but not wspA or wspR ( Goymer et al . , 2006; McDonald et al . , 2009 ) . All awsXRO mutants have similar lower fitness ( p<10−6 , two-tailed t-test ) compared to the wspF reference strain ( Figure 9 ) , which explain why under selection these are found at lower frequencies compared to mutations in the wsp pathway ( McDonald et al . , 2009 ) despite a roughly two-fold higher mutation rate to WS .
The issue of evolutionary predictability and the relative importance of stochastic events compared to deterministic processes have a long history in evolutionary biology ( Darwin , 1872; Simpson , 1949; Jacob , 1977; Gould , 1989; Conway Morris , 2003; Orgogozo , 2015 ) . Recent interest has been sparked by an increasing number of observations that both phenotypic and molecular evolution , under certain circumstances , can be remarkably repeatable ( Colosimo et al . , 2005; Shindo et al . , 2005; Jost et al . , 2008; Barrick et al . , 2009; Lee and Marx , 2012; Meyer et al . , 2012; Zhen et al . , 2012; Herron and Doebeli , 2013 ) , but whether these cases are representative for evolutionary processes in general remains to be determined . A related question , with greater potential for practical applications , is whether it is possible to forecast short-term evolutionary events and if so , then the challenge is to stipulate the data necessary to make successful predictions . Our uniquely detailed knowledge of the WS experimental evolution system has provided a rare opportunity to determine the contributions of mutational bias and genetic architecture to the generation of new adaptive phenotypes , and consequently explore the limits of evolutionary forecasting . A thorough understanding of the function of the molecular species and their interactions allowed development of null models ( see especially Model IV ) that capture essential features of the genotype-to-phenotype map sufficient to predict the relative likelihood that evolution will follow each of the three principle pathways , along with specific mutational ( genetic ) targets . Even the most sophisticated of these models failed to forecast outcomes that matched the experimental data shown in Figure 2 , however , the reason became apparent upon characterisation of the set of WS mutants obtained without selection: the presence of mutational hotspots . Armed with knowledge of sources of mutational bias it was a simple matter to refine Model IV leading to predictions matching those observed by experiment . Problematic at the current time is inability to a priori detect all causes of mutational bias , however , it is likely that this will improve as understanding of the biochemical causes of bias improves and algorithms trained to recognise and detect nucleotide patterns indicative of bias are implemented . One specific deletion ( ΔY77-Q87 ) in awsX was found to account for nearly half ( 20/41 ) of the mutations in the Aws pathway . Thus , despite the existence of hundreds of possible mutations leading to WS ( this work and ( McDonald et al . , 2009; McDonald et al . , 2011; Lind et al . , 2015 ) ) one single mutation accounts for more than one quarter of all WS mutations . While the six base pair direct repeat flanking the deletion provides a convincing explanation for its increased rate , it is not clear why this deletion would be ten times more common than the ΔP34-A46 deletion in the same gene that is flanked by ten base pair repeats and contains five base pairs identical to those from the ΔY77-Q87 deletion ( Figure 6—source data 1 ) . There are also instances where single base pair substitutions are overrepresented: the AwsR T27P mutation is found in nine cases , while eight other single pair substitutions in Aws were found only once . Consider further the fact that wspE ( a gene of ~2 . 3 kb ) , where changes to only four specific amino acids repeatedly cause WS , and wspF ( a gene of ~1 kb ) where any mutation that disrupts function results in WS ( Figure 6A ) contribute equally to the rates of at which new WS types arise . It is evident from these findings and from related studies ( Pollock and Larkin , 2004 ) that there is need for detailed experimental measurement of local mutation rates in specific systems . Such investigations stand to contribute to understanding of the causes of mutational bias and the extent to which biases might be conserved among related or even unrelated organisms . If local nucleotide sequence is the major determinant , an estimate of mutation rate will apply strictly to very closely related species , but if the dynamics of molecular processes , such as transcription and replication ( Sankar et al . , 2016 ) , are major influences then estimates might be applicable to a wider range of species . Forecasting mutational routes to new phenotypes is one component of a comprehensive forecasting strategy . The second requirement is ability to a priori predict fitness effects . Solving this problem is the Holy Grail for predicting evolution , but at the current time this is not possible . That it matters is made clear by the observation that a subset of all possible WS mutants was found after experimental evolution ( Figure 8 ) due to inability of some to successfully compete with high fitness WS types ( Figure 9 ) . Direct measurement of the fitness effects of large numbers of mutations is difficult , time-consuming , typically only possible for microbial species and against the spirit of identifying a priori predictors of fitness effects . Future success likely rests on ability to infer fitness from parameters such as estimated effects of mutations on thermodynamic stability ( Capriotti et al . , 2005; Dean and Thornton , 2007; Rodrigues et al . , 2016 ) , molecular networks , evolutionary conservation of amino acid residues ( Glaser et al . , 2003; Ng and Henikoff , 2003; Landau et al . , 2005; Kumar et al . , 2009; Celniker et al . , 2013; Choi and Chan , 2015 ) or machine-learning methods combining several types of information ( Bromberg and Rost , 2007; Li et al . , 2009; Capriotti et al . , 2013; Yates et al . , 2014; Hecht et al . , 2015 ) . Possibilities likely also exist to extrapolate findings from a small number of mutations that are either directly constructed and assayed in the laboratory or through fitness estimates of polymorphism data from natural populations . Recent work on the prediction of the fitness effects of random mutations in several genes suggests that in many cases large effect mutations can be predicted using methods based on evolutionary conservation ( Lind et al . , 2017a ) . On a less ambitious scale it may be possible to take advantage of the fact that the distribution of fitness effects associated with mutations in single genes is often bimodal – a consequence of many mutations causing complete loss-of-function rather than intermediate deleterious; effects ( Sanjuan , 2010; Hietpas et al . , 2011; Jacquier et al . , 2013; Firnberg et al . , 2014; Lind et al . , 2017a; Sarkisyan et al . , 2016; Lundin et al . , 2017 ) . As such it may be sufficient to know gene function plus consequences arising from loss of function mutations and thus treat mutations within single genes as having equivalent fitness ( Sanjuan , 2010; Jacquier et al . , 2013; Sarkisyan et al . , 2016; Lind et al . , 2017a ) . Interestingly WS mutations occurring in the same gene typically show similar fitness effects ( Figure 9 ) . While still requiring experimental data , this simplification may fuel exploration of the relationship between large numbers of mutations and their fitness effects and thus understanding of the extent to which fitness effects are transferable between strains with different genetic backgrounds or closely related species ( Ungerer et al . , 2003; Pearson et al . , 2012; Wang et al . , 2014 ) . Taken at face value , our findings give reason to question the value of aspiring to forecast evolution from first principles based on mechanistic understanding . But we argue against such pessimism and point firstly to the value that stems from a clear understanding of current limitations on forecasting: defining what is known , what is not known , and what needs to be known . That it is possible to take knowledge of the genotype-to-phenotype map and forecast with good accuracy targets and rates is an important advance . There is good reason to suspect that the principles outlined here and previously ( Lind et al . , 2015 ) are transferable to other systems and even generalizable ( Lind , 2018 ) . Evolutionary forecasting is in its infancy . In the short term it is likely to be most successful for biological systems where there are experimental data on a large number of independent evolutionary events , such as is the case for influenza , HIV and cancer ( Kouyos et al . , 2012; Fraser et al . , 2014; Lawrence et al . , 2014; Luksza and Lässig , 2014; Neher et al . , 2014; Eirew et al . , 2015 ) . Evolution might appear idiosyncratic indicating that every specific system requires detailed investigation , but our hope is that deeper knowledge of the genotype-to-phenotype map , distribution of fitness effects and mutational biases will allow short term forecasts to be produced using modelling without the need for large-scale experimental studies . A major boost to further refinement of evolutionary forecasting is likely to come from combining coarse ( top down ) and fine-grained ( bottom-up ) approaches . Our demonstration that simple null models of functional networks can produce quantitative predictions is a step forward allowing predictions to be directly tested in other experimental systems ( Lind , 2018 ) .
The strains used in the study are all Pseudomonas fluorescens SBW25 ( Silby et al . , 2009 ) or derivatives thereof . The reporter construct ( pMSC ) , used for isolation of WS mutants before selection , fused the Pwss promoter to a kanamycin resistance marker ( nptII ) ( Fukami et al . , 2007; McDonald et al . , 2011 ) . P . fluorescens strains with deletions of the wsp ( PFLU1219-1225 ) , aws ( PFLU5209-5211 ) and mws ( PFLU5329 ) operons were previously constructed as described by McDonald et al . ( McDonald et al . , 2011 ) . All experiments used King’s medium B ( KB ) ( King et al . , 1954 ) , solidified with 1 . 5% agar and incubation was at 28°C . All strains were stored in glycerol saline KB at −80°C . Strains with the pMSC reporter construct and either wild type genetic background or double deletions of aws/mws , wsp/mws or wsp/aws were used to estimate mutation rates to WS before selection . Overnight cultures were diluted to approximately 103 cfu/ml and 60 independent 110 μl cultures were grown for 16–19 hr ( OD600 = 0 . 9–1 . 0 ) with shaking ( 200 rpm ) in 96-well plates before plating on KB plates with 30 mg/l kanamycin . Viable counts were estimated by serial dilution and plating on KB agar . One randomly chosen colony per independent culture with WS colony morphology was restreaked once on KB agar . The assay was repeated at least four times for the double deletion mutants and twice for the wild type strain in order to obtain enough mutants to allow estimation of mutation rates . Mutations rates and confidence intervals were estimated using the Ma-Sandri-Sarkar Maximum Likelihood Estimator ( Hall et al . , 2009 ) available at www . keshavsingh . org/protocols/FALCOR . html . The mutation rates between the different strains were statistically evaluated using a t-test as previously described ( Rosche and Foster , 2000 ) , but this method has only been shown to be valid in cases where total population size is not significantly different for the strains used . In our assay this was not the case , as determined by ANOVA , and therefore the results of the statistical analysis should be interpreted with caution . As the estimated number of mutants per well was <0 . 5 for all strains , the biasing effect of differences in fitness between WS mutants is minimal . Mutations causing the WS phenotype were identified by Sanger sequencing of candidate genes in the remaining common pathway to WS , for example the wsp operon for the aws/mws deletion strain . In a few cases where no mutations were identified in the previously established WS target genes , we used genome sequencing ( Illumina HiSeq , performed by Macrogen Korea ) . Competition assays were performed as previously described ( Lind et al . , 2015 ) by mixing the WS mutant 1:1 with a reference strain labelled with green fluorescent protein and measuring the ratio of each strain before and after static growth for 24 hr using flow cytometry ( BD FACS Canto ) . We used a WspF ΔT226-G275 deletion mutant as the reference strain because WspF mutants are the most commonly found WS type when grown under selective conditions ( McDonald et al . , 2009 ) and the in frame deletion of 50 amino acids most likely represents a complete loss-of-function mutation with minimal polar effects on the downstream wspR . Selection coefficients per generation were calculated as s = [ln ( R ( t ) /R ( 0 ) ) ]/[t] , as previously described ( Dykhuizen , 1990 ) where R is the ratio of alternative WS mutant to WspF ΔT226-G275 GFP and t the number of generations . Viable counts on KB plates of initial and final populations were performed to calculate the number of generations . Stability of colony morphologies was confirmed and data from microcosms with >5% smooth colonies were excluded ( two cases ) . Control competition experiments with isogenic WspF ΔT226-G275 reference strains with and without GFP were used to correct for the cost of the GFP marker . Control competitions were also used to determine the cost of the double deletions and the reporter construct relative to a wild type genetic background , for example an AwsX ΔY77-Q87 mutant in Δwsp/Δmws background with pMSC was competed with a GFP labeled AwsX ΔY77-Q87 mutant in wild type background . Competitions were performed in independently inoculated quadruplicates for each strain with the number of replicates based on previous work ( Lind et al . , 2015 ) . Homology models of the structure of WspA , WspE , WspR , AwsR , AwsO and MwsR were made using Phyre2 in intensive mode ( http://www . sbg . bio . ic . ac . uk/phyre2 ) ( Kelley et al . , 2015 ) . The differential equation models describe the interactions between proteins in each of the three WS pathways . In order to solve the differential equations , two pieces of information are required: ( i ) the initial concentrations of the molecular species and ( ii ) the reaction rates . Although this information is unavailable a random-sampling approach was used to generate different random sets of initial concentrations and reaction rates . Each random set was used to establish a baseline of potential WS expression making it possible to evaluate whether a set of mutations results in a WS type . Effectively , this approach allows sampling of the probability distribution P ( WS |mi ∈ Wsp ) used in our Bayesian model . We randomly sample 1000 different sets of reaction rates and initial concentrations from uniform priors: reaction rates were sampled randomly from a uniform distribution on log space ( i . e . 10U[−2 , 2] ) and initial concentrations of reactants were sampled from a uniform distribution U[0 , 10] . For each set , the appropriate differential equation model was integrated and the steady state concentration of the compounds that correspond to a wrinkly spreader ( RR in Aws , R* in Wsp and D* for Mws ) computed . This served as a baseline for the non-WS phenotype that was used for comparison to determine whether combinations of mutations result in increased WS expression . After obtaining the baseline , we implemented particular combinations of enabling/disabling mutations ( a mi ) . Ideally , a distribution linking enabling/disabling mutations to a fold change in reaction rates would be used , but this information is unavailable . In order to progress the effect sizes for enabling and disabling mutations were sampled from 10U[0 , 2] and 10U[−2 , 0] , respectively , and then multiplied by the reaction rates . The differential equations were then solved for the same time that it took the baseline to reach steady state . The final concentration of R* ( Figure 4A ) , RR ( Figure 4B ) and D* ( Figure 4C ) was then compared to the baseline and the number of times out of 1000 that the WS-inducing compound increased served as an estimate of P ( WS|mi ∈ Wsp ) . The probability distribution stabilized by 500 random samples and additional sampling did not produce significant changes ( data not shown ) . The absence of empirical data on reaction rates , initial concentrations , and expected mutation effect size meant using a random sampling approach requiring estimates for parameter ranges . Parameter ranges were chosen to be broad enough to capture differences spanning several orders of magnitudes while allowing numerical computations for solving the differential equations . To assess the effect of these ranges on the results , the sampling procedure was repeated for WSP for three different parameter regimes ( i ) an expanded range for initial concentrations [0–50] , ( ii ) an expanded range for reaction rates 10[-3 , 3] , ( iii ) a compressed range for mutational effect size 10^[-1 , 1] . This analysis shows that qualitative results are robust to these changes ( see Figure 5—figure supplement 3 ) . Source code and equations are available in Supplementary file 1 . | Predicting evolution might sound like an impossible task . The immense complexity of biological systems and their interactions with the environment has meant that many biologists have abandoned the idea as a lost cause . But despite this , evolution often repeats itself . This repeatability offers hope for being able to spot in advance how evolution will happen . To make general predictions , it is necessary to understand the mechanisms underlying evolutionary pathways , and studying microbes in the laboratory allows for real-time experiments in evolution . One of the best studied microbes for experimental evolution is Pseudomonas fluorescens , which repeatedly evolves flattened wrinkled colonies instead of round smooth ones when there is limited oxygen . The underlying molecular pathways that lead to this change have been studied in detail . Lind et al . developed mathematical models to predict how often the three most common pathways would be used and which genes were most likely to be mutated . After controlling for the effects of natural selection and refining the models to take into account mutation hotspots , Lind et al . were able to accurately predict the genes that would be targeted by mutations . The findings suggest that biologists need not lose hope when it comes to the goal of predicting evolution . A deep understanding of the molecular mechanisms of evolutionary changes are essential to predicting the mutations that lead to adaptive change . The results are an important first step towards forecasting organisms’ responses to changing conditions in the future . In the short term , this is important for medical issues , including antibiotic resistance , cancer and immune receptors . In the long term , predicting the course of evolution could be essential for survival of life on the planet . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"genetics",
"and",
"genomics"
] | 2019 | Predicting mutational routes to new adaptive phenotypes |
Tumors often co-exist with T cells that recognize somatically mutated peptides presented by cancer cells on major histocompatibility complex I ( MHC-I ) . However , it is unknown why the immune system fails to eliminate immune-recognizable neoplasms before they manifest as frank disease . To understand the determinants of MHC-I peptide immunogenicity in nascent tumors , we tested the ability of thousands of MHC-I ligands to cause tumor subclone rejection in immunocompetent mice by use of a new ‘PresentER’ antigen presentation platform . Surprisingly , we show that immunogenic tumor antigens do not lead to immune-mediated cell rejection when the fraction of cells bearing each antigen ( ‘clonal fraction’ ) is low . Moreover , the clonal fraction necessary to lead to rejection of immunogenic tumor subclones depends on the antigen . These data indicate that tumor neoantigen heterogeneity has an underappreciated impact on immune elimination of cancer cells and has implications for the design of immunotherapeutics such as cancer vaccines .
Human cancers bear uniquely distinguishable features on the surface of their cells in the form of neoantigens , which are peptides derived from mutated , foreign or oncofetal proteins that are presented in complex with major histocompatibility complex I ( MHC-I ) molecules . These short , 8–11 amino acid fragments specifically mark cancer cells and activate potent immune responses that can lead to effective anti-cancer therapy ( Tran et al . , 2014; Rosenberg and Restifo , 2015; Zacharakis et al . , 2018 ) . Yet , the existence of T cells that recognize neoantigens is often not sufficient to eliminate tumors . Karl Hellström first described the coexistence of tumor-specific lymphocytes together with cancer cells in human solid tumors as a paradox 50 years ago ( Hellström et al . , 1968 ) . In mice , an analogously enigmatic observation has been made that sporadic tumors occurring in aged or carcinogen-treated mice induce strong T cell responses only when transferred into new hosts , thereby preventing engraftment ( Heike et al . , 1994; Dubey et al . , 1997; Shankaran et al . , 2001 ) . Thus , tumor-specific T cells can lead to tumor rejection in a new host or when used as a cancer therapy , but somehow immune surveillance is evaded during early tumorigenesis in the host that originally developed an immunogenic tumor . The increased rate of tumor formation in immunocompromised individuals has led to the hypothesis that the immune system can and does eliminate some tumors , particularly virally induced tumors , before they become clinically apparent ( Schulz , 2009 ) . We hypothesized that if the immune system can eliminate some early tumors , but not others , perhaps it is because some antigens are more potent at inducing effective T cell responses during early tumorigenesis . Identification and characterization of neoantigens that can induce an effective immune response and clear cancer cells is critical to understanding why and how immunogenic tumors develop in immunocompetent hosts . Immunogenic peptides have been discovered in animal studies by injection of thousands or millions of tumor cells bearing neoantigens into animals and observing tumor rejection . However , the robust immune activation and tumor rejection in these cases is not analogous to the events of early tumorigenesis in humans , when the number of transformed cells is miniscule . Thus , even though some tumors are immunogenic , it is not clear why the host cannot eliminate them when the tumors first arise . If the biochemical features of neoantigens that lead to effective T cell responses were known , it might be possible to identify which tumors bear immunogenic antigens . Some reports have linked peptide immunogenicity to the biochemical characteristics of amino acid residues at certain positions along the MHC-I ligand ( Calis et al . , 2013 ) , while others have focused on the difference between the affinity of a wild-type ligand and a mutated ligand ( Duan et al . , 2014 ) . However , the absence of a large , unbiased data set of known immunogenic and non-immunogenic peptides has stymied the validation of these approaches . Indeed , most known immunogenic antigens are derived from viral proteins; few mutationally-derived neoantigens are confirmed as bonafide immunogenic peptides in mice or humans . Here , we have developed a novel genetic method to express libraries of precisely defined MHC-I ligands in mammalian tumor cells and have used this method to ask questions about MHC peptide immunogenicity in immunocompetent animals during early tumorigenesis . Using libraries of genetically encoded MHC-I ligands , we tracked the dynamic growth and depletion of thousands of tumor subclones in vivo and noted a striking failure of cancer immunosurveillance that is potentially analogous to the failure of immune surveillance in humans during early tumorigenesis . We demonstrate for the first time that the ability of the naive immune system to surveille a nascent tumor and reject immunogenic subclones is limited by the fraction of cells expressing each unique antigen . Furthermore , we show that these rejection thresholds vary among antigens . Our data are consistent with the observation in humans that patients whose tumors have high numbers of subclones—and thus more subclonal neoantigens—have increased levels of relapse and worse survival than patients with more homogenous tumors ( McGranahan et al . , 2016; Reuben et al . , 2017; Turajlic et al . , 2018a ) . Thus , our data provide an antigen-specific rationale for the impact that tumor heterogeneity has on survival of human patients . According to our findings , antigen-specific immune effects are limited during early tumorigenesis , which has implications for the emergence and outgrowth of immunogenic tumors .
We have developed a reductionist method for encoding diverse peptide/MHC ( pMHC ) ligands in mammalian cells ( termed ‘PresentER’ ) , which was purposefully designed to enable us to ask questions about the immunogenicity of individual peptide epitopes presented by cancer cells at relatively physiologic levels . We used this method to perform high-throughput , pooled screening of MHC-I ligand immunogenicity in wild-type mice ( Figure 1A ) . The PresentER antigen minigene is comprised of an endoplasmic reticulum ( ER ) signal sequence followed by a short peptide/epitope . Expression of the peptide and its display on MHC-I does not require proteasomal degradation or peptide processing , thus enabling precise definition of the exact epitope displayed to the immune system . As previously described ( Gejman , 2018a ) , Transporter associated with antigen presentation ( Tap ) deficient cell lines expressing PresentER antigen minigenes lead to surface presentation of the encoded MHC-I peptide , detectable by multiple modalities , including fluorescently labeled antibodies directed to specific MHC-I ligands , mass spectrometry based immunopeptidomics and antigen-specific T cell reactivity . To demonstrate the applicability of PresentER antigen minigenes to study MHC-I ligand immunogenicity , we first asked if cancer cells encoding known immunogenic ( mouse Tyrp1/gp75 TAYRYHLL W223A , H224Y ( Dyall et al . , 1998 ) ; mouse Ddx5/p68 SNFVFAGI S551F ( Dubey et al . , 1997 ) ; synthetic SIYRYYGL ( Udaka et al . , 1996 ) ; synthetic VTFVFAGL ( Dubey et al . , 1997 ) ; chicken ovalbumin SIINFEKL ) or non-immunogenic ( mouse Serpinf1/Pedf MSIIFFLPL ( Wang et al . , 2006 ) ; scrambled chicken ovalbumin FEKIILSN; mouse Ndufa4/dEV8 EQYKFYSV ( Holler et al . , 2003 ) ; mouse Ddx5/p68 SNFVSAGI ( Dubey et al . , 1997 ) ; mouse Tyrp1/gp75 TWHRYHLL ( Dyall et al . , 1998 ) ; Mouse Trp2 SVYDFFVWL ) MHC-I ligands would be rejected by wild-type ( WT ) animals . The C57BL/6 syngeneic , Tap deficient mouse cell line RMA/S was transduced with these antigen minigenes and we first checked that the antigens were expressed at physiologically normal levels on the cell surface . Previously we , and others , had shown that endogenously expressed MHC-I tumor antigens are presented on the cell surface at several hundred to several thousand sites per cell ( Dao et al . , 2013; Sergeeva et al . , 2011; Mathias et al . , 2017 ) . Using a radiolabeled TCR mimic antibody ( 25-D1 . 16 ) reactive with SIINFEKL/H-2Kb , we demonstrated that there were on average 3500 binding sites per RMA/S cell expressing PresentER-SIINFEKL , compared with ~90 sites on cells expressing PresentER-MSIIFFLPL , suggesting that the PresentER system allows display of the epitopes in a relatively physiologic range . We injected 5 × 106 cells subcutaneously into WT C57BL6/N mice . Tumors expressing known non-immunogenic peptides grew , while tumors expressing known immunogenic peptides were rejected in some or most animals ( Figure 1B ) . Tap deficient cells were used because peptides derived from proteins that are not directed to the ER ( e . g . eGFP ) are not presented on MHC as these cells lack the ability to transport peptide from the cytoplasm into the endoplasmic reticulum . At 7 days after tumor injection , T cells specific for the chicken ovalbumin peptide SIINFEKL could be detected in tumor draining lymph nodes of animals injected with PresentER-SIINFEKL expressing cells ( Figure 1C ) . Haemotoxylin and Eosin ( H and E ) staining of regressing PresentER-SIINFEKL tumors showed lymphocytic infiltration with hyalin rich fibrin deposits , indicating cell death . By contrast , tumors with PresentER-FEKIILSN ( a non-MHC-I binding peptide ) were well-vascularized , highly cellular and with trace lymphocytic infiltration ( Figure 1D ) . In order to verify that the mechanism of tumor rejection was indeed T cell dependent , we injected Rag-/- animals with SIINFEKL or FEKIILSN positive tumors and confirmed that SIINFEKL tumors were not rejected ( Figure 1E ) . Taken together , these results indicate that RMA/S tumors expressing PresentER antigen minigenes can recapitulate the known immunogenicity of individual mouse MHC-I ligands in a T cell dependent manner . To study immunogenicity in vivo at high throughput and complexity in this model , we first wanted to understand if immune responses directed at immunogenic antigens lead to rejection of cells presenting non-immunogenic antigens . If so , the ability to study immunogenicity in a pooled in vivo setting might be compromised . Mice were injected with pairs of tumors expressing an immunogenic and a non-immunogenic antigen minigenes ( one minigene-expressing tumor on each flank ) : SIINFEKL/FEKIILSN , SIYRYYGL/EQYKFYSV and TAYRYHLL/TWHRYHLL . The immunogenic SIINFEKL , SIYRYYGL and TAYRYHLL expressing tumors were rejected , but the non-immunogenic FEKIILSN , EQYKFYSV and TWHRYHLL tumors found on the contralateral flank were not ( Figure 2A ) . This suggests that tumor rejection is local and that an effective immune response to an immunogenic tumor does not affect the growth of a non-immunogenic tumor . Next , to test if the immune system could identify and kill immunogenic subclones within a largely non-immunogenic tumor , we cultured cells with varying ratios of immunogenic ( SIINFEKL ) and non-immunogenic ( FEKIILSN ) cells . These cells were injected subcutaneously into mice and tumor size was monitored . In tumors where PresentER-SIINFEKL cells were greater than 25% of a tumor , tumors were smaller and tumor rejection occurred more frequently , especially when tumors were comprised of ≥50% immunogenic cells ( Figure 2B ) . Next , we wanted to test if immunogenic subclones within a largely non-immunogenic tumor could be eliminated . We cloned mCherry into the PresentER vector and mixed PresentER-SIINFEKL ( mCherry ) with non-immunogenic PresentER-MSIIFFLPL ( eGFP ) cells at varying ratios and injected them into congenically marked CD45 . 1 mice . On day 16 , the tumors were harvested and flow cytometry was performed to identify which ( CD45 . 2 positive ) tumor cells remained . Remarkably , within a non-immunogenic ( eGFP labeled ) tumor , immunogenic ( mCherry labeled ) sub-populations were eliminated ( Figure 2C ) . Thus , in this model , the immune system is capable of recognizing and selectively depleting immunogenic tumor subclones within the context of a largely non-immunogenic tumor . We hypothesized that a pooled screen in vivo might reveal the determinants of immunogenic MHC-I ligands if a tumor bearing a library of MHC-I antigen were depleted of cells bearing immunogenic peptide-MHC while tumor cells bearing non-immunogenic pMHC were spared ( Figure 3A ) . Using the PresentER system , such a screen could be done on large scale and identify hundreds or thousands of immunogenic antigens at once , in contrast to identifying immunogenic antigens one-by-one . Immunoediting in vivo leading to loss of tumor clones with immunogenic neoantigen-encoding mutations has previously been observed in syngeneic mouse tumor models ( Matsushita et al . , 2012; DuPage et al . , 2012; DuPage et al . , 2011 ) , suggesting that this approach might be viable . We searched the mouse proteome in silico and randomly selected 5 , 000 8-mer peptides that were predicted by NetMHCPan to bind to the B6 mouse MHC-I allele H-2Kb . We also designed a library of mutated peptides by selecting single amino acid substituents of each peptide in the wild-type peptide library that did not eliminate MHC-I binding ( Figure 3B ) . The MHC-I affinities and properties of the mutated peptides are further described in Figure 3—figure supplement 1 and Supplementary file 2 . On average , the peptides in the mutated library have slightly higher affinity for MHC-I than the peptides in the wild type library , but the plurality of mutated peptides are within 100 nM of their non-mutated counterparts ( Figure 3—figure supplement 1A–B ) . The residues that are changed from the wild-type to the mutant library tend to be at positions 4 , 6 and 7 ( Figure 3—figure supplement 1C ) . The majority of mutations are isomorphic ( polar >polar , hydrophobic >hydrophobic and charged >charged ) , but ~⅓ of peptides feature a hydrophobic >other substitution ( Figure 3—figure supplement 1D ) . Several known immunogenic and known non-immunogenic control antigen minigenes were included in each library , including some that were described in Figure 1 . The libraries of wild type and mutant libraries were separately cloned and introduced into RMA/S cells by transduction at multiplicity of infection <0 . 3 , thereby ensuring that few cells received more than one minigene . Naive C57BL6/N mice were injected with either ( a ) 5 × 106 cells expressing the wild type library , ( b ) 5 × 106 cells expressing the mutant library , ( c ) 5 × 106 mutant library cells plus 1 × 106 wild type RMA/S ( ‘padded’ ) or ( d ) 5 × 106 wild type library cells mixed with 5 × 106 mutant library cells ( n = 5 per group ) . The tumors were allowed to grow for 17 days and then harvested ( Figure 3C and Figure 3—figure supplement 2A ) . Genomic DNA was extracted from all of the tumors , as well as RMA/S library cells frozen on the day of injection ( ‘pretumor’ samples ) and the minigenes encoded by each cell were amplified by PCR and sequenced by Illumina next generation sequencing . Comparison of minigene abundance in the tumor outgrowth with minigene abundance in the pre-tumor samples led to the surprising finding that no minigenes were robustly depleted during growth in vivo ( Figure 3D and Figure 3—figure supplement 2 and Figure 3—source data 1 ) . This result stood in stark contrast to the experiments described above , in which mutant peptide expressing clones were efficiently depleted in immunocompetent mice . Some minigenes that were not abundant in the library ( <1/10 , 000 ) at the time of injection were depleted or dropped out entirely in the tumor due simply to stochastic drop-out; however , minigenes that were abundant in the library at time of injection maintained their abundance despite in vivo growth of the tumor . Surprisingly , even positive control minigenes encoding strongly immunogenic peptides ( orange points ) were not depleted in this context . Analysis of minigene abundance in individual animal tumors ( as opposed to the average of several tumors ) yielded the same conclusion that few , if any , minigenes were reliably depleted ( Figure 3—figure supplement 3 ) . The absolute number of cells at time of injection can be estimated from the relative abundance of each minigene in the pre-tumor samples . The least abundant minigenes were found at 1 per 5 × 106 cells ( 0 . 00002% ) and the most abundant at ~23×103 per 5 × 106 ( ~0 . 5% ) . The immunogenic controls SIINFEKL , VTFVFAGL and SNFVFAGI were found at 0 . 005–0 . 013% of cells before injection into mice , which correspond to between 250 and 650 cells injected out of 5 × 106 . The number of cells expressing the immunogenic antigens upon injection is very low and thus analogous to the number of cells that are present during early tumor development , when immunogenic transformed cells first arise . The surprising inability of the immune system to eliminate these demonstrably immunogenic cells may shed light on the limits of immune cell activation and killing during the early stages of tumorigenesis that enable outgrowth and escape of immunogenic cancers . A possible explanation for failure of immune surveillance in our RMA/S MHC-I minigene library model might be insufficient antigen available in the growing tumor to activate an initially productive anti-tumor immune response in naive mice . If this were true , we hypothesized that T cells from antigen experienced mice might be able to detect and kill immunogenic cells in a highly heterogeneous tumor . Although vaccination with soluble peptides has been shown to generate robust T cell immunity ( Ott et al . , 2017; Sahin et al . , 2017 ) , this is not cost effective at scale . Alternatively , there is precedent for the idea that vaccination with a library of mutated antigens can lead to immunogenic T cell responses that lead to slower tumor growth or clearance ( Engelhorn et al . , 2006 ) . We decided to vaccinate mice with irradiated tumor cells bearing the library of MHC-I peptides . In order to avoid confounding immunity to the RMA/S cells themselves ( Van Hall et al . , 2006 ) , we used CRISPR/Cas9 to generate a B6-syngeic Tap2-/- MCA205 fibrosarcoma cell line . A single cell clone of Tap2 knockout MCA205 was selected and Tap2 knockout was validated by RT-PCR and next generation sequencing . Decreased surface MHC-I staining was expected and observed , because the Tap complex is a key chaperone of peptide/MHC-I formation ( Figure 4—figure supplement 1 ) . WT B6 mice were vaccinated three times , once every 6 days , with 1 × 107 irradiated MCA205∆Tap2 cells bearing the wild-type library minigenes ( Figure 4A ) . Splenocytes and draining lymph nodes from three vaccinated and three non-vaccinated mice were harvested at day 18 after the final vaccination and analyzed for the presence of antigen experienced T cells . Five control peptide tetramers were used , three of which are immunogenic and were present in the library ( SIINFEKL , SNFVFAGI , VTFVFAGL ) , one which is not immunogenic but was present in the library ( MSIIFFLPL ) and one which is immunogenic but not found in the library ( SIYRYYGL ) . Only the immunogenic peptides found in the library showed an increased number of CD44+/tetramer+ CD8 T cells , while the other two peptides did not show significant changes ( Figure 4B ) . Therefore , vaccination with the library yielded detectable T cell populations specific to the immunogenic peptides . Vaccinated and non-vaccinated mice were then challenged with 5 × 106 RMA/S cells bearing the wild type peptide library . Slower tumor growth was noted in the vaccinated as compared to the non-vaccinated mice , suggesting that a vaccine-related anti-tumor effect may have occurred ( Figure 4C ) . However , neither vaccinated nor unvaccinated animals showed depletion of immunogenic control minigenes in relation to the non-immunogenic control minigenes ( Figure 4D–F and Figure 4—source data 1 ) . Thus , although slower tumor growth was noted , antigen-specific immunity was not observed in response to prophylactic vaccination in the library setting , suggesting a possible response to some other , broadly expressed , cellular antigens not represented in the peptide library . While we have demonstrated that PresentER minigenes peptides can generate effective antigen-specific immunity in bulk RMA/S tumor assays , the same response does not occur in tumors bearing libraries of MHC-I ligands . In order to test if there is a threshold level of tumor cell clonality necessary to effectively activate the immune system , CD45 . 1 mice were injected with mixtures of immunogenic ( labeled with eGFP ) and non-immunogenic ( labeled with mCherry ) RMA/S cells and flow cytometry was performed on reisolated tumors 17 days later ( Figure 5A ) . Relative to their proportion upon engraftment , tumor cells bearing the immunogenic peptides SIINFEKL and TAYRYHLL were depleted when they comprised as little as 1% of the tumor . Below 1% , depletion of cells bearing these two minigenes could not be detected . Depletion of cells bearing the SNFVSAGI peptide could be reliably detected at 50% and in some tumors at 10% , however depletion could not be detected when the SNFVSAGI cells were found at less than 10% of the tumor ( Figure 5B ) . These findings are surprising , as they indicate that immunogenic tumor subclones can persist within a tumor and that rejection or persistence is dependent on tumor cell percentage of the total tumor mass during tumorigenesis , and not immunogenicity of the cell alone . Failure of the immune system to eliminate immunogenic subclones present at low fractional abundance could either be due to the low quantity of total antigen present in the tumor or to the low percentage of cells bearing each antigen . In order to discriminate between these two possibilities , we increased the amount of tumor injected , thus increasing the total amount of antigen the immune system sees while keeping the percentage of each antigen within the tumor the same . We grew RMA/S library tumors in Rag-/- mice , harvested the tumors at 20 days and retained a portion of the material for sequencing . The rest of the tumor material was transferred into the flank of WT B6 or Rag-/- mice . Each animal received approximately 1 milliliter of tumor fragments ( ~2 . 5×108 cells ) , which represents a 40–100 fold increase in cells expressing each antigen ( Figure 5C ) . After 17 days , the transferred tumors were harvested and sequenced . Once again , as we observed in both naive and vaccinated mice , robust depletion of cells bearing immunogenic peptide minigenes did not occur . Overall minigene abundance was highly correlated between the tumor material transferred and the tumors harvested 17 days later in both WT and Rag-/- mice and in both the wild type and mutant libraries ( Figure 5D and Figure 5—source data 1 ) . These results reveal that it is the percentage of tumor cells bearing each antigen within a tumor and not the total quantity of antigen that determines if an effective immune response occurs .
The immune system is capable of recognizing cancer cells as foreign based on the presentation of altered or unusual MHC-I ligands on the surface of tumor cells—a phenomenon that has been leveraged for cancer therapies such as immune checkpoint blockade and adoptive cell transfer ( Tran et al . , 2014; Rosenberg and Restifo , 2015; Zacharakis et al . , 2018 ) . Despite this , the immune system fails to clear immunogenic , clinically detected tumors on its own , as has been paradoxically noted for decades ( Hellström et al . , 1968 ) . It is not clear at what point in tumorigenesis the immune system begins to initiate ( ineffectively ) a response to immunogenic MHC-I ligands presented on cancer cells . There is epidemiologic evidence that immunocompromised patients develop more tumors , suggesting that the immune system may prevent some tumors from ever manifesting clinically . However , recent data also suggests that mutations accumulate at high rates in sun-damaged , but otherwise normal , tissue at levels comparable to cancer cells ( Martincorena et al . , 2015 ) and that tumors face overall little negative selective pressure ( Martincorena et al . , 2017 ) . Indeed , immune escape by loss of MHC ( Masuda et al . , 2007; Cabrera et al . , 2000; Gudmundsdóttir et al . , 2000; So et al . , 2005; Shukla et al . , 2015 ) , when it occurs , is a late and subclonal event ( McGranahan et al . , 2017; Turajlic et al . , 2018b ) . If T cell surveillance were highly effective during early tumor development , negative selective pressure would be notable in the evolutionary trajectories of tumors and HLA loss would be expected to occur early , frequently and in a clonal manner . We interpret the data from human and animal experiments to suggest that routine T cell immunosurveillance of nascent cancer cells does occur , but is limited by unknown factors , and that recognition of tumors as foreign occurs mostly later in tumor development . If tumor immunosurveillance is sometimes effective in the early growth of a tumor , we hypothesized that it might be due to potent neoantigens expressed by cancer cells . Discovery of the biochemical characteristics of mutationally-derived neoantigens that are immunogenic would be important to clarify why some neoantigens are tolerated by the immune system and others are not . The significance of this question is underscored by the major challenge currently facing cancer immunotherapy: the identification of patients whose tumors bear immunogenic neoantigens—and are thus likely to respond well to immune checkpoint blockade or other immunotherapies—vs those patients who will be unnecessarily exposed to potentially toxic therapies without a chance for efficacy . Here , we have developed a reductionist approach to determine which MHC-I ligands are immunogenic in immunocompetent mice . We demonstrate that tumors are rejected when an immunogenic pMHC is expressed on all or most injected cancer cells . To our surprise , we have discovered that effective T cell responses are not mounted against highly immunogenic peptides when cells expressing these peptides are a minor fraction of the tumor , which is the case in early developing tumors in humans . Furthermore , the threshold percentage of tumor cells necessary to yield an antigen-specific immunogenic T cell response varies with the antigen . We propose that ineffective T cell responses may be a consequence of intratumoral ( or intra-tissue ) heterogeneity and that neoantigen clonal fraction is an important , overlooked aspect of MHC-I antigen immunogenicity . There are many mechanisms by which immune mediated killing of immunogenic cancer cells can fail . Here we report the first evidence that tumor heterogeneity is an explicit factor leading to failure of effective immunity in a tumor and that the maximum level of heterogeneity tolerable before immune escape occurs is dependent on the antigen . Immune checkpoints or active tumor suppression do not explain the observed immune evasion because such mechanisms would be expected to suppress T cell killing irrespective of the fraction of tumor cells that expresses an immunogenic epitope . This is evidenced by robust depletion of the highly immunogenic peptides SIINFEKL and TAYRYHLL when present at ≥1% of the tumor . The mechanism by which T cell responses are restrained when immunogenic MHC-I antigens are present at low frequency is not yet clear . Low levels of antigen presentation may lead to ineffective cross-presentation of antigen in the tumor draining lymph node , thus limiting T cell activation ( Spiotto et al . , 2002 ) . In animal models , low levels of immunogenic epitopes of oncogenic drivers presented on transformed cells early during tumorigenesis ( at the premalignant stage ) were shown to induce a program of cellular hypo-responsiveness in tumor-specific ( oncogene-specific ) CD8 T cells ( Willimsky and Blankenstein , 2005; Schietinger et al . , 2016 ) . After prophylactic vaccination with irradiated library cells we observed smaller tumors and increased numbers of antigen-specific T cells in splenocytes and draining lymph nodes , suggesting that some antigen-specific T cell proliferation does occur . While we demonstrated sufficient levels of the high affinity , highly immunogenic antigens on the cell surface , tumor associated epitopes of lower affinity , immunogenicity and expression levels should produce an even more pronounced deficit in T cell recognition . We speculate that the mechanism of immune escape in this model is either ineffective T cell activation or failure of activated T cells to identify and kill antigen-positive cells present at low abundance within a sea of cells displaying irrelevant antigens . The observation that animals prophylactically vaccinated with MCA205∆Tap2 cells are partially protected from outgrowth of RMA/S tumors may be due to several possibilities . For instance , the vaccinations may have activated T cells that recognize some immunogenic peptides in the library , or T cells that recognize mutated peptides expressed by both cell types , or , most intriguingly , T cells specific to T cell epitopes associated with impaired peptide processing ( TEIPPs ) . TEIPPs are T cell epitopes derived from wild type proteins that are not normally presented , unless antigen presentation is altered by loss of Tap , which yields a relative deficit of peptides translocated into the ER . TEIPPs in RMA/S cells can be immunogenic ( Van Hall et al . , 2006; Doorduijn et al . , 2016 ) . Although RMA/S and MCA205∆Tap2 are unlikely to share neoantigens derived from somatic mutations , they may share TEIPPs that contribute to protection against RMA/S tumor growth . A caveat to the vaccination experiments is that differences in the antigen presentation machinery of MCA205∆Tap2 and RMA/S cells may lead to incomplete congruence in the MHC-I peptidome of these two cell lines . These differences may lead to some peptides encoded by the PresentER library ( or by the endogenous genome of the cell ) to be poorly presented in one cell line but well presented in another cell line . Although we have focused our efforts on understanding the role of CD8 T cells , other infiltrating cell types are also important for anti-tumor effects . In particular , CD4 T cells have been shown to play important roles in the CD8 T cell responses to tumors ( Borst et al . , 2018 ) and to target neoantigens ( Ott et al . , 2017; Sahin et al . , 2017 ) . However , our experimental design does not allow us to assess the role of CD4 T cells in CD8 T cell mediated anti-tumor responses . Thus , while it may be the case that the magnitude or quality of the immune response may differ when both CD8 and CD4 cells recognize a related immunogenic antigen , we are unable to study this with the tools we have developed to perform pooled library experiments because we cannot encode both a specific MHC II and MHC I ligand in a single minigene . Moreover , there is the confounding possibility that some CD4 help is coming from unrelated antigens , such as mutated proteins found in RMA/S or the exogenous proteins introduced into the cells by the PresentER vector . More broadly , the heterogeneity in actual human tumors is characterized by differential infiltration with antigen presenting cells , which can change the sensitivity of the immune system to detect immunogenic antigens present at low clonal fractions , and thus could impact our findings . Although we could not modulate the infiltrating immune cells in our experimental tumor models , we appreciate that this may influence the potency of a subpopulation of immunogenic cells . It would be interesting to understand how various levels of antigen presenting cells contribute to dynamic thresholds of T cell immunogenicity . The PresentER system we have employed does have some potential biochemical caveats that may impact the interpretation of the data . While we have not observed alternative cleavage patterns in individual peptides we have studied ( Gejman , 2018b ) , in a library setting we cannot test if every minigene is properly yielding its encoded MHC-I ligand . For instance , some encoded peptides may not bind well to MHC-I or may only bind in the presence of an MHC loading chaperone . Some peptides may lead to improper or no cleavage of the signal peptide whereas other peptides may be shortened ( e . g . by ERAP1 , an ER associated endopeptidase ) , altered or destroyed in the endoplasmic reticulum . New therapeutic modalities such as immune checkpoint blockade have shown clinical efficacy in the treatment of human tumors , but the toxicity of the drug regimens has led to many efforts to find biomarkers that predict patient response . Tumors with high mutation burdens—which are more likely to have immunogenic neoantigens ( Van Allen et al . , 2015; Yarchoan et al . , 2017; Snyder et al . , 2014 ) —and tumors that have mismatch repair ( MMR ) deficiency or microsatellite instability ( MSI-H ) respond well to checkpoint blockade ( Le et al . , 2015; Le et al . , 2017 ) . In mouse models , syngeneic tumors with MMR gene knock outs accumulate mutations over time and grow more slowly in wild type mice than do tumors without MMR deficiency . Cell lines derived by sub-cloning of MMR deficient lines—thus increasing the clonal fractions of each neoantigen—grow more slowly or are rejected entirely . In all cases , MMR inactivation in tumors leads to better responses to checkpoint blockade than the parental tumors ( Germano et al . , 2017 ) . Moreover , survival is inversely related to tumor neoantigen clonality in human patients with lung adenocarcinoma ( McGranahan et al . , 2016; Reuben et al . , 2017 ) . Patients whose tumors bear high numbers of subclonal ( or branched ) neoantigens have increased levels of relapse and worse survival than those patients with more homogenous tumors ( McGranahan et al . , 2016; Reuben et al . , 2017 ) . Response to checkpoint blockade in lung and skin tumors is also associated with lower levels of intratumoral heterogeneity ( McGranahan et al . , 2016 ) . The combination of these data are highly suggestive that while total neoantigen burden is important for long term survival and response to checkpoint blockade , neoantigen clonality is an additional important factor in mediating tumor regression . The discovery that intratumoral heterogeneity prevents effective T cell responses has implications for the development of therapeutic cancer vaccines , checkpoint blockade , adoptive T cell therapies , and studies of tumor immunogenicity . In addition , our data may provide a new understanding of mechanisms of immune surveillance and its failure that allows growth and evolution of tumors and subclones . For instance , recent data shows that subclones with very low clonal fraction yet distinct intratumoral functions may be important to tumor survival and growth ( Vinci et al . , 2018 ) —suggesting that even low abundance tumor subclones may be important targets for immunotherapy . In general , we may speculate on one mechanism for cancer escape and progression , in which small early cancers generally do not bear immunogenic epitopes derived from their limited number of driver oncogenic proteins and few passenger mutations . Then as the tumors evolve , the neoantigens appearing in subclones do not reach a clonal fraction high enough to breach their antigen-specific immunogenicity thresholds , thereby allowing escape of these otherwise immunogenic clones . This model may help to explain why sun-damaged and aged healthy tissue is replete with mutations , sometimes reaching frequencies seen in human cancers ( Martincorena et al . , 2015; Risques and Kennedy , 2018 ) . Cumulatively , these findings suggest that effective T cell immunity is restrained in the context of healthy tissue and growing tumors and paint a picture of T cell immunogenicity that is poorly captured by existing models of tissue immunosurveillance .
6-8 week old C57BL6/N mice were purchased from Envigo or Taconic Biosciences . 6–8 week old B6 . SJL-Ptprca/BoyAiTac ( known as CD45 . 1 mice ) and B6 . 129S6-Rag2tm1Fwa N12 Mice ( known as Rag2 KO ) were purchased from Taconic Biosciences . Mice were shaved before subcutaneous engraftment of the indicated number of RMA/S cells in 100 uL PBS . Tumor volumes were calculated using caliper measurements and the standard modified ellipsoid formula: tumor volume = ( LxW2 x 0 . 52 every 2–3 days . Animals were euthanized when tumor volume exceeded 2000 mm3 or if ulceration was noted . Vaccination of animals was performed by subcutaneous injection of 10 × 106 irradiated ( 20 Gy ) MCA205-ΔTap2 cells expressing libraries of minigenes . For cell surface staining , cells were incubated with appropriate fluorophore-conjugated mAbs for 30 min on ice and washed twice before resuspension in the viability dye DAPI at 1 μg/mL . Flow cytometry data were collected on a LSRfortessa ( BD ) or an Accuri C6 ( BD ) and analyzed with FlowJo V10 software . The antibodies used in this study were anti-H-2Kb-APC clone AF6-88 . 5 ( Biolegend 116517 ) , anti-SIINFEKL/H2kb-APC clone 25-D1 . 16 ( Biolegend 141605 ) , anti-CD45 . 2-APC clone 104 ( eBioscience 17-0454-81 ) , anti-CD8a-FITC clone Ly-2 ( BD Pharmingen 553030 ) , anti-CD3-PerCP clone 145–2 C11 ( BD Pharmingen 553067 ) . The following fluorescently labeled H-2Kb tetramers were obtained through the NIH Tetramer Core Facility: SIINFEKL , SIYRYYGL , MSIIFFLPL , SNFVFAGI and VTFVFAGL . Radioimmuneassay to determine the number of SIINFEKL/H-2Kb molecules on the surface of RMA/S PresentER-SIINFEKL cells was performed as previously described ( Dao et al . , 2013; Chang et al . , 2017 ) . A guide RNA sequence targeting murine Tap2 ( ATGGGGCTGTTGCGCTGAGC ) was cloned into the LentiCRISPRv2 ( Sanjana et al . , 2014 ) plasmid ( Addgene plasmid 52961 ) , a gift from Feng Zhang ( Broad Institute , Cambridge , Massachusetts , USA ) . MCA205 fibrosarcoma cells were transiently transfected using Lipofectamine 2000 ( Thermo Fisher Scientific 11668027 ) following standard manufacturer’s protocols . 24 hr later , successful transfectants were selected using 5 ug/mL Puromycin for 3 days before expansion and single-cell subculture . Genetic ablation of Tap2 was verified by next generation sequencing of the Tap2 loci confirming a frameshift deletion in both alleles , and RT-PCR analysis . Primers for RT-PCR analysis are listed in Supplementary file 1 . Reduced cell-surface H-2Kb expression was also verified by flow cytometry . RMA/S cells expressing PresentER antigen #1 ( eGFP ) were mixed with RMA/S cells expressing PresentER antigen #2 ( mCherry ) at defined ratios ( e . g . 1:10 , 1:100 , 1:1000 , etc ) and validated by flow cytometry immediately before injection into CD45 . 1 mice . After 17 days , tumors were harvested , cut into small pieces and disaggregated by incubation at 37°C with Liberase TL ( Sigma-Aldrich 5401020001 ) , DNAse I ( Worthington Biochemical LS002139 ) and ACK lysis buffer ( Thermo Fisher Scientific A1049201 ) . Single cell suspensions of tumor cells were stained with CD45 . 2 and DAPI and collected on the same flow cytometer , using the same settings and gates as on day 0 . The number of eGFP and mCherry cells was calculated based on the same gates used on the day of injection . Non-fluorescent cells were ignored for the purposes of analysis and the percentage of eGFP and mCherry cells in the tumor were normalized to sum to 100% . The fold change in cells expressing each antigen was calculated as: ( percentage of normalized eGFP cells in tumor ) / ( percentage of eGFP cells at D0 ) . The individual PresentER minigenes specified above were cloned into the PresentER backbone as previously described ( Tran et al . , 2014 ) . Oligonucleotides were amplified using T3_SfiI and T7_SfiI primers ( Supplementary file 1 ) . HEK293T Phoenix-Ampho cells were transfected with each plasmid and , after 24 hr , viral supernatant was harvested every 12 hr . RMA/S were transduced with limiting amounts of viral supernatant at 1000xg for 2 hr at 37°C in six well non-tissue culture treated plates . PresentER minigenes vectors are available on Addgene ( 102942 , 102943 , 102944 , 102945 , 102946 ) . The peptides included in the libraries were found in Uniprot database of canonical mouse protein sequences ( UP000000589 ) . Substrings of unique eight amino acid sequences were collected and affinity to H-2Kb was calculated using NetMHCPan v4 . 0 . 5000 randomly selected peptides with predicted ic50 <500 nM were selected and constitute the ‘wild type peptide library . ’ A single random amino acid substitution was made to each member of the wild type library to generate the ‘mutant peptide library . ’ Substitutions which generated another wild type peptide were excluded . Libraries of PresentER minigenes were cloned as previously described ( Tran et al . , 2014 ) . Library metadata is provided in Supplementary file 2 . Oligonucleotide libraries were ordered from CustomArray and amplified with Phusion polymerase using WT and Mutant minigene library-specific primers ( Supplementary file 1 ) . Amplicons were digested with SfiI and passed through a MinElute column . The PresentER cassette vector was also digested with SfiI , treated with calf intestinal phosphatase and ligated to the oligonucleotides with T4 ligase . Ligation products were phenol extracted and electroporated into DH5 electrocompetent cells . Electroporated cells were plated and counted . At least 1000x fold more transformants than minigene library members were required to proceed to plasmid DNA extraction ( >5×106 colonies ) . The colonies were scraped off the plate and grown for 3 . 5 hr in TB +ampicillin at 37°C at 225 rpm . The bacteria were maxiprepped using the Qiagen maxiprep kit and library representation was checked by Illumina sequencing . Library containing retrovirus was produced by transfection of 15 cm plates of HEK293T Phoenix-AMPHO cells with library plasmid DNA . Viral supernatant fluid was collected beginning at 24 hr after transfection and continuing every 12 hr until 72 hr post transfection . Viral supernatant fluid was pooled , concentrated with Clontech Retro-X concentrator and frozen . Concentrated viral supernatant fluid titers were determined by transduction of RMA/S cells . Libraries of minigene expressing RMA/S cells were generated by transduction at an MOI <0 . 3 ( 1 , 000xg for 2 hr at 37°C in six well non-tissue culture treated plates ) and selection with puromycin . Library expressing cells were maintained in cultures of >1000 x cells per number of minigenes in the library ( i . e . at least 5 × 106 ) . To verify that minigene representation was not compromised during cloning , all libraries were sequenced from plasmid prior to transduction into mammalian cells . Minigenes were either directly amplified with P5 and barcoded P7 primers and sequenced using a custom , or they were amplified with a nested PCR protocol followed by Illumina library preparation and sequencing at the Integrated Genomics Operation at MSKCC ( Supplementary file 1 ) . Genomic DNA was extracted from cultured cells or mechanically disaggregated tumors with the Gentra Puregene kit and minigenes were amplified from genomic DNA . In order to avoid amplification bias due to high PCR cycle numbers or multiple integration events , PCR was performed with the minimum number of cycles possible . The minimum number of PCR cycles was empirically-determined for each batch of samples , but was typically 18–22 cycles when the template was genomic DNA . All steps were performed in PCR hoods or other clean environments with precautions to avoid cross contamination between samples . In order to avoid under-sampling low abundance minigenes , each sample of genomic DNA was amplified using DNA equivalent to >1 , 000 fold the number of minigenes in the library . Reads were mapped to the PresentER minigene libraries with Bowtie2 using default settings . Reads that did not map to the minigenes in the library were discarded . The number of reads aligning to each minigene was divided by the total number of reads aligning to the library to obtain the relative abundance of each minigene . Quality control to ensure library representation and absence of contamination were performed on every sample . All data analysis was performed in R . Some PresentER minigenes are available from Addgene ( e . g . , #102942 , #102943 , #102945 , #102946 , #102944 ) and others are available by request directly from the investigators . Data are available in the following repositories: DOI:10 . 5281/zenodo . 1310902 , DOI:10 . 5281/zenodo . 1309836 and DOI:10 . 5281/zenodo . 1308909 . | T cells are specialized agents of the immune system that can detect and attack tumors . They spot their target by identifying small pieces of proteins – or antigens – at the surface of diseased cells . In particular , they can recognize the new and abnormal antigens that a cancer cell often displays . Yet , cells that become cancerous and start displaying suspicious antigens can manage to escape T cells and grow into full tumors . Why does the immune system not recognize and kill these early cancers before they get out of control ? One possibility is that T cells do not identify certain antigens carried by cancer cells . To test this , researchers have conducted experiments where they inject a mouse with cancer cells that display a single new antigen . If the animal develops a tumor , then this antigen does not trigger an immune response . However , this method is slow and laborious , because only one antigen can be tested at the time . Instead , Gejman , Chang et al . developed a new technique , PresentER , where a rodent gets injected with a mix of millions cancer cells that each displays a different antigen . This way , many thousands of new antigens can be studied in one go . The tumors are left to grow for several weeks before they are removed and analyzed to see which cells survived and which have been killed by the immune system . Unexpectedly , the nature of the antigen did not make a big difference . Instead , cancer cells with new antigens could go undetected if they were rare and made up only a small proportion of all the different cancer cells in a tumor . However , the immune system would eliminate the exact same cancer cells when they were the major component of a cancerous lump . Future research now has to explore exactly how rare cancer cells can hide amongst other cells , and remain invisible to the body . Armed with this knowledge , it might be possible to improve cancer therapy by prompting the immune system to target these emerging threats earlier . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2018 | Rejection of immunogenic tumor clones is limited by clonal fraction |
A considerable proportion of mammalian gene expression undergoes circadian oscillations . Post-transcriptional mechanisms likely make important contributions to mRNA abundance rhythms . We have investigated how microRNAs ( miRNAs ) contribute to core clock and clock-controlled gene expression using mice in which miRNA biogenesis can be inactivated in the liver . While the hepatic core clock was surprisingly resilient to miRNA loss , whole transcriptome sequencing uncovered widespread effects on clock output gene expression . Cyclic transcription paired with miRNA-mediated regulation was thus identified as a frequent phenomenon that affected up to 30% of the rhythmic transcriptome and served to post-transcriptionally adjust the phases and amplitudes of rhythmic mRNA accumulation . However , only few mRNA rhythms were actually generated by miRNAs . Overall , our study suggests that miRNAs function to adapt clock-driven gene expression to tissue-specific requirements . Finally , we pinpoint several miRNAs predicted to act as modulators of rhythmic transcripts , and identify rhythmic pathways particularly prone to miRNA regulation .
Circadian clocks orchestrate daily oscillations in mammalian behaviour , physiology , and gene expression . In mammals , a master pacemaker in the brain's suprachiasmatic nucleus ( SCN ) synchronises subsidiary oscillators present in most peripheral cell types ( reviewed in Dibner et al . , 2010; Mohawk et al . , 2012 ) . Timekeeping by peripheral and SCN clocks relies on negative transcriptional feedback loops that engender oscillatory gene expression . In the core loop , BMAL1:CLOCK transcription factor heterodimers drive the expression of repressors , encoded by the Period ( Per1 , 2 ) and Cryptochrome ( Cry1 , 2 ) genes . PER:CRY protein complexes subsequently accumulate in the nucleus and repress BMAL1:CLOCK-mediated transcription . Due to their instability , repressor protein and mRNA abundance rapidly drops below the threshold required for autorepression , clearing the way for a new cycle . The mechanisms that control the stability of clock proteins have been studied to considerable extent and frequently involve post-translational protein modifications that control proteasomal degradation ( e . g . , Yagita et al . , 2002; Eide et al . , 2005; Shirogane et al . , 2005; reviewed in Mehra et al . , 2009; Chong et al . , 2012 ) . It is less well understood how the decay of core clock mRNAs is controlled ( reviewed in Kojima et al . , 2011; Lim and Allada , 2013 ) . Cyclically expressed transcription factors such as BMAL1:CLOCK ( Panda et al . , 2002; Rey et al . , 2011 ) or REV-ERBα/β ( Ueda et al . , 2002; Le Martelot et al . , 2009; Bugge et al . , 2012; Cho et al . , 2012 ) relay the timing information from the core clock to clock output pathways by driving the rhythmic expression of clock-controlled genes ( CCGs ) , many of which are tissue-specific ( Panda et al . , 2002; Storch et al . , 2002 ) . In mouse liver , up to 15% of expressed mRNAs accumulate in a rhythmic fashion ( Vollmers et al . , 2009; Mohawk et al . , 2012 ) . At least in part , the synergistic activation of genes by circadian and tissue-specific transcription factors may account for the rhythmic expression of cell type-specific transcripts . Conceivably , tissue-specific post-transcriptional regulation could participate in this endeavour as well . Recent studies in liver have indeed suggested that a substantial proportion of mRNA abundance rhythms is generated post-transcriptionally ( Koike et al . , 2012; Le Martelot et al . , 2012; Menet et al . , 2012 ) and have speculated on the involvement of miRNAs in this process ( Menet et al . , 2012 ) . MicroRNAs are short , non-coding RNA molecules that inhibit the translation and promote the destabilisation of mRNAs by base-pairing with sequence elements that are typically located in the 3′ untranslated regions ( 3′ UTRs ) of their target transcripts ( reviewed in Krol et al . , 2010; Fabian and Sonenberg , 2012; Yates et al . , 2013 ) . Mammalian genomes encode >1000 miRNAs ( Bentwich et al . , 2005 ) and each may have hundreds of targets . It has thus been estimated that up to 60% of mammalian protein-coding transcripts can be regulated by miRNAs ( Lewis et al . , 2005; Friedman et al . , 2009 ) . It is therefore likely that these regulatory molecules carry functions in circadian gene expression as well , both at the level of the core clock and clock output genes . The antisense-inactivation of miR-219 and miR-132 in the SCN has indeed been reported to result in mild lengthening of period and in defective light-induced clock resetting , respectively ( Cheng et al . , 2007 ) . Moreover , an elegant study by Chen et al . ( 2013 ) has recently reported dramatic period shortening ( ≈2 hr ) of free-running rhythms in miRNA-deficient mouse embryonic fibroblasts ( MEFs ) , likely caused by the lack of three miRNAs ( miR-24 , miR-29a , miR-30a ) targeting Per1 and Per2 mRNAs . Other miRNAs have been noted for their capacity to regulate core clock transcripts as well ( Kiriakidou et al . , 2004; Meng et al . , 2006; Nagel et al . , 2009; Tan et al . , 2012; Lee et al . , 2013; Shende et al . , 2013 ) , but their circadian functions in vivo are still unclear . Similarly , the functions that miRNAs assume in clock output pathways have barely been investigated in mammals , except in a few cases . The mir122 locus thus encodes a highly abundant , hepatocyte-specific miRNA involved in the regulation of cholesterol and lipid metabolism ( Krutzfeldt et al . , 2005; Esau et al . , 2006 ) that is clock-controlled in mouse liver ( Gatfield et al . , 2009; Menet et al . , 2012 ) . Mature miR-122 appears to act as a modulator of circadian output pathways , despite being non-rhythmic due to its high metabolic stability ( Gatfield et al . , 2009; Kojima et al . , 2010 ) . Conceivably , other miRNAs could target rhythmic transcripts as well . We have now comprehensively analysed the contribution that miRNAs make to the regulation of hepatic circadian gene expression by genetically inactivating miRNA biogenesis in the livers of adult mice . Around-the-clock profiling of mRNAs and pre-mRNAs from Dicer-deficient ( Harfe et al . , 2005 ) and control livers thus allowed the global detection of post-transcriptional and miRNA-dependent regulation of hepatic gene expression . We have addressed the following main questions . What functions do miRNAs have in the regulation of the hepatic core clock ? To what extent do miRNAs contribute to the post-transcriptional generation of mRNA rhythms ? And third , for transcripts subject to both miRNA regulation and cyclic transcription , how do these mechanisms integrate and cooperate to yield the net rhythmic output ? Finally , the comprehensive atlas of miRNA-dependent ( circadian and non-circadian ) gene expression established in this study exhaustively charts the regulatory space that miRNAs occupy in liver gene expression and may thus serve as an important resource beyond chronobiology .
To comprehensively uncover miRNA-regulated gene expression in the liver , we generated a mouse model in which miRNA biogenesis could be inactivated in hepatocytes , which constitute around 80% of liver mass ( Weibel et al . , 1969 ) . We used mice carrying conditional knockout alleles for the Dicer1 gene ( termed Dicerflox in the following ) , which encodes the ribonuclease that converts pre-miRNAs to mature miRNAs ( Harfe et al . , 2005 ) , and a Cre-ERT2 recombinase expressed as an internal ribosome entry site ( IRES ) -controlled transgene in the 3′ UTR of the endogenous Albumin locus ( AlbCre-ERT2 in the following ) ( Schuler et al . , 2004 ) . The Albumin locus conferred hepatocyte-specificity and the Cre-ERT2 fusion allowed for induction of knockout in adult animals using tamoxifen . Previous publications have reported constitutive hepatocyte-specific knockouts of Dicer ( Hand et al . , 2009; Sekine et al . , 2009a , 2009b ) and to our knowledge the alleles we employed have not been used in combination before . We thus first confirmed their suitability for efficient miRNA depletion in the livers of adult animals . To this end , we analysed the kinetics of recombination at the Dicer locus at different timepoints after intraperitoneal tamoxifen ( tx ) injection . Recombination was undetectable before but highly efficient from 2 to 3 weeks after tx treatment ( Figure 1—figure supplement 1A ) . We assessed the consequences on miRNA expression by northern blot analysis probing for the highly abundant , hepatocyte-specific miR-122 , which has been previously noted for its long half-life ( Gatfield et al . , 2009 ) and thus served as a proxy for overall miRNA depletion from hepatocytes . One month after tx treatment , miR-122 was virtually undetectable ( Figure 1—figure supplement 1B , C ) . Moreover , loss of mature miR-122 was accompanied by increased levels of its precursor , pre-miR-122 ( Figure 1—figure supplement 1B , C ) , as expected ( Harfe et al . , 2005 ) . Interestingly , we observed that at later timepoints beyond 2 months after tx treatment , miR-122 expression began to recover ( Figure 1—figure supplement 1D , E ) . The most likely explanation for this observation was that after miRNA loss the liver eventually renewed its hepatocytes from a cell population that had escaped recombination . In summary , we concluded that our experimental system appeared suitable for the efficient knockout of Dicer and , at least judging by miR-122 , for the depletion of miRNAs . Approximately 1 month after tx treatment appeared to represent a suitable timepoint for experiments aiming at a comparison of knockout with control animals . We induced the knockout in more than 40 mice ( male , aged 3–6 months ) , entrained them to light–dark cycles for 1 month ( LD 12:12; ad libitum feeding ) and sacrificed them at 4-hr intervals around-the-clock ( Figure 1A ) . A littermate control group was treated identically . We combined both wild-type and heterozygous males for controls because a single functional Dicer allele was sufficient to ensure miRNA processing to wild-type levels ( Figure 1—figure supplement 2A ) , as expected from previous studies ( Harfe et al . , 2005; Kanellopoulou et al . , 2005; Murchison et al . , 2005; Chen et al . , 2008; Frezzetti et al . , 2011 ) . Pools of liver total RNA ( 3–4 animals ) were assembled into two independent full time series around-the-clock ( ZT0-20 ) for both knockouts and controls . Northern blot analysis confirmed loss of miR-122 in the knockouts across all timepoints ( Figure 1B , C ) . Microarray ( Figure 1—figure supplement 2B ) and northern blot analyses ( Figure 1—figure supplement 2C ) showed that beyond miR-122 , miRNA levels were globally decreased . Other hepatocyte-enriched miRNAs ( e . g . , miR-148a , miR-192 and miR-194 [Sun et al . , 2004; Landgraf et al . , 2007; Tang et al . , 2007; Farid et al . , 2012] ) were thus virtually undetectable in knockouts . Probably owing to their expression also in non-hepatocyte cells , known ubiquitous miRNAs ( e . g . , miR-21 [Landgraf et al . , 2007] , miR-26a , or let-7 family members [Lagos-Quintana et al . , 2003] ) decreased less strongly , and miRNAs with a particularly strong expression in non-hepatocytes ( e . g . , miR-126 in hepatic stellate cells [Guo et al . , 2013] ) were almost unchanged , as expected ( Figure 1—figure supplement 2B , C ) . In all cases , reduced miRNA levels correlated with increased pre-miRNA abundance ( Figure 1B , D , Figure 1—figure supplement 1C ) . Moreover , even in the knockout , pre-miR-122 still showed rhythmic accumulation with a trough at ZT12 ( Figure 1B , D ) , which was a consequence of clock-driven rhythmic transcription at the mir122 gene locus ( Gatfield et al . , 2009; Menet et al . , 2012 ) . Other clock-controlled genes such as Dbp ( Ripperger et al . , 2000 ) still oscillated as well ( Figure 1E ) . We thus concluded that cyclic gene expression per se was still functional in Dicer knockouts and that the RNA time series collected from the two genotypes was suitable to uncover the effects of miRNA loss on the rhythmic expression of core clock and clock output genes in hepatocytes . 10 . 7554/eLife . 02510 . 003Figure 1 . Analysis of hepatic Dicer knockout using miR-122 as a diagnostic marker . ( A ) Schematic of the Dicer knockout protocol used throughout the study . Conditional knockout and control littermates ( heterozygotes and wild-type ) carrying the AlbCre–ERT2 allele that ensures hepatocyte-specific expression were injected with tamoxifen , entrained to 12-hr light/12-hr dark cycles for a month , and sacrificed at the indicated Zeitgeber Times . ( B ) Northern blot analysis demonstrating that miR-122 is virtually undetectable in knockout animals . In contrast , the Dicer substrate , pre-miR-122 , accumulates to higher levels due the lack in turnover by Dicer . Each loaded sample is a mix of RNAs prepared from 3 to 4 independent animals . The two series are the same as used for the RNA-seq analysis . tRNAThr served as a loading control . ( C ) Quantification of miR-122 abundance from northern blot shown in ( B ) after normalisation to tRNAThr . In the knockouts , mean miR-122 levels only reached 1 . 4% of control levels . ( D ) Quantification of pre-miR-122 expression ( normalised to tRNAThr ) from northern blot shown in ( B ) . On average , pre-miR-122 accumulates to >sixfold higher levels than in controls due to the absence of Dicer processing . Note that even in the knockout , pre-miR-122 still shows the rhythmic accumulation that is the result of cyclic transcription at the mir122 locus . ( E ) Quantitative real-time PCR analysis of Dbp , a typical core clock output gene , clearly indicates that rhythmic gene expression per se still occurs in Dicer knockout livers . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 00310 . 7554/eLife . 02510 . 004Figure 1—figure supplement 1 . Kinetics of Dicer knockout in liver . ( A ) PCR analysis of recombination kinetics at the Dicer locus . Male and female mice homozygous for the Dicerflox allele ( also carrying AlbCre-ERT2 ) were injected with tamoxifen on five consecutive days and sacrificed 0 , 8 , 13 , 22 days after the first injection . Cre-mediated deletion of exon 24 of the Dicer locus was detected using PCR with three primers that produce different products for floxed and floxed-deleted alleles ( the location of the primers relative to the floxed exon 24 is schematically represented at the right of the figure ) . High efficiency of deletion thus occurred from 2 to 3 weeks after tamoxifen treatment . ( B ) Northern blot analysis confirms that 2–3 weeks after tamoxifen treatment miR-122 abundance strongly decreased . After around 1 month , miR-122 was virtually undetectable , whereas pre-miR-122 accumulated to higher levels due to the lack in Dicer processing . ( C ) Quantification of the northern blot shown in ( B ) . miR-122 levels ( normalised to tRNAThr ) on day 0 were set to 100% . ( D ) Male mice heterozygous ( control ) and homozygous ( knockout ) for the Dicerflox allele were injected with tamoxifen as in ( A–C ) and sacrificed 30 days and 71 day after the first injection . As in ( B ) , 1 month after injection miR-122 was undetectable in the knockouts . Interestingly , 2–3 months after tamoxifen treatment , miR-122 expression recovered . This effect was independent of whether the animals carried a single or two copies of the AlbCre-ERT2 allele ( data not shown ) . ( E ) Quantification of the northern blot shown in ( D ) . miR-122 levels ( normalised to tRNAThr ) in controls at day 30 were set to 100% . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 00410 . 7554/eLife . 02510 . 005Figure 1—figure supplement 2 . Global analysis of miRNA depletion in Dicer knockout livers . ( A ) miRNA microarray analysis of wild-type vs heterozygote knockout liver RNA shows that a single functional copy of Dicer is sufficient to process miRNAs to wild-type levels . No significantly different miRNA species was detected ( FDR-corrected for multiple testing ) . The seemingly highly expressed miRNAs labelled in grey correspond to probes that consistently give signal on the microarrays , but whose expression is dubious as they are not detectable in liver small RNA-seq and thus may rather represent false positive signals from the array . ( B ) miRNA microarray analysis of knockout vs control liver RNA shows that miRNAs are globally downregulated in the knockout . Pools over all time points from the RNA-seq series were used . The apparently highly expressed and unchanged miRNAs labelled in grey correspond to probes that consistently give signal on the microarrays , but whose expression is doubtful as they are not detectable in liver small RNA-seq and thus may rather represent false positive signals from the array . miRNAs in blue were confirmed by northern blot in ( C ) , miRNAs in green are those proposed by Chen et al . ( 2013 ) to be regulators of Per1 and Per2 . miR-451 ( red ) has been reported for its Dicer-independent processing ( ‘Discussion’ ) . Known hepatocyte-enriched miRNAs ( e . g . , miR-122 , miR-148a , miR-192 and miR-194 ) were thus virtually undetectable in knockouts . Known ubiquitous miRNAs ( e . g . , miR-21 , miR-26a , or let-7 family members ) decreased less strongly , probably owing to their expression also in non-hepatocyte cells , which make up ca . 20% of the liver . miRNAs with a particularly strong expression in non-hepatocytes ( e . g . , miR-126 in hepatic stellate cells ) were almost unchanged ( see main text ) . Quantile normalisation used for microarrays in ( A ) and invariant normalisation for ( B ) . FC , fold-change difference . ( C ) Northern blots using the same RNA pools as in ( B ) confirm the downregulation of miRNAs , which is accompanied by higher pre-miRNA levels . Values below the northern blots show how their quantification compares to the microarrays . See ( B ) for comments on liver cell-type specific expression . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 005 MicroRNAs inhibit the translation and promote the decay of their mRNA targets . How these events are elicited , and in which order they occur , is still subject of ongoing debate . However , given that in most cases mRNA degradation appears to be an endpoint in miRNA activity , changes in mRNA levels are the most commonly used readout for regulation by miRNAs ( reviewed in Fabian and Sonenberg , 2012 ) . To analyse how rhythmic RNA accumulation was affected by the loss of miRNAs , we employed RNA high-throughput sequencing ( RNA-seq ) . We chose a protocol for library generation that was based on the random priming of rRNA-depleted ( ‘Ribo-Zero’ ) total RNA , which permitted the simultaneous quantification of mRNAs ( exon-mapping reads ) and pre-mRNAs ( intron-mapping reads ) . Given that splicing occurs co-transcriptionally and is relatively fast , the intron-containing pre-mRNAs are short-lived and their abundance can serve as a proxy for gene transcription rates ( Koike et al . , 2012 ) . This strategy thus identifies post-transcriptional ( mRNA ) gene expression changes that occur in the Dicer knockouts and likely represent bona fide miRNA targets , as opposed to transcriptional changes ( mRNA and pre-mRNA affected ) , which likely represent secondary effects of miRNA loss . Conceivably , any transcription factor that is a direct miRNA target would engender numerous such secondary effects . We sequenced knockout and control time series with comparably high coverage . Reads mapped to different RNA classes , as expected ( Figure 2A , Figure 2—figure supplement 1 , Figure 2—source data 1 ) . We did not detect gross distortions in RNA populations that would compromise the comparability of knockout and control samples , although some differences reached statistical significance ( Figure 2—figure supplement 1 , Figure 2—source data 1 ) . We then examined transcripts from protein-coding loci , which contributed around half of all obtained sequences ( 51 . 6 ± 1 . 0% in knockout , 47 . 9 ± 1 . 3% in control; Figure 2A , B , Figure 2—figure supplement 1 , Figure 2—source data 1 ) . Reads mapping to exonic sequences were categorised as ‘mRNA’ and those mapping to introns or intron–exon boundaries as ‘pre-mRNA’ ( ‘Materials and methods’ ) . Pre-mRNAs are of generally low abundance , but genome-wide there is an estimated 20-fold more transcribed intronic than exonic sequence space ( Sakharkar et al . , 2005 ) . In accordance , pre-mRNAs were readily detectable transcriptome-wide , representing around a quarter of protein-coding reads ( 25 . 5 ± 4 . 1% in knockout , 21 . 5 ± 3 . 1% in control; Figure 2A , B , Figure 2—source data 1 ) . Finally , the RNA-seq data also provided an additional confirmation of the high efficiency of the Dicer knockout ( Figure 2—figure supplement 2 ) . 10 . 7554/eLife . 02510 . 006Figure 2 . Genome-wide quantification of pre-mRNA and mRNA abundance by RNA-seq . ( A ) Summary of RNA-seq results from Dicer KO ( left ) and control ( right ) . Percentages are relative to total reads across all time points . ( B ) Time-resolved analysis of intron- ( yellow ) and exon-mapping ( green ) RNA-seq reads in the two series of knockout and control mice . ( C ) Venn diagram indicating the number of genes whose expression was detectable in Dicer knockout and control animals on the mRNA ( exon ) and pre-mRNA ( intron ) levels ( threshold: 0 . 1 RPKM for mRNA , 0 . 01 RPKM for pre-mRNA in at least 1/3 of samples ) . ( D ) Comparison of mRNA expression ( RPKM ) of protein coding genes in Dicer knockouts vs controls ( averaged over all time points ) . Green dots correspond to transcripts whose levels are statistically significantly >1 . 5-fold different between genotypes and considered as differentially expressed . The red mRNAs are highly different on the transcriptional ( pre-mRNA ) level and were excluded from circadian analyses; see ( E ) . ( E ) Comparison of pre-mRNA expression ( RPKM ) of protein coding genes in Dicer knockouts vs controls . Green dots correspond to transcripts whose levels are statistically significantly >1 . 5-fold different between genotypes and considered as differentially expressed . Red dots correspond to genes with higher pre-mRNA fold-change than the 90% quantile of differentially expressed pre-mRNAs ( corresponding to >4 . 1-fold expression differences between knockout and control ) ; due to the highly different transcription rates , they were excluded from circadian analyses . ( F ) Analysis of transcriptome-wide mRNA/pre-mRNA ratios ( as a measure of mRNA stability ) between Dicer knockouts and controls . Note that in the Dicer KO , mRNAs become globally more stable ( modes of distributions indicated in red ) . ( G ) mRNA/pre-mRNA ratio changes are predictive for direct miRNA targets . Known targets of liver-specific miR-122 ( red ) thus have overall increased mRNA/pre-mRNA ratios in the knockout , in contrast to targets of miR-124 ( dark blue ) , which is not expressed in liver . Transcripts encoding ribosomal proteins ( pale blue ) seem to be overall excluded from miRNA regulation , as expected . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 00610 . 7554/eLife . 02510 . 007Figure 2—source data 1 . RNA-seq read statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 00710 . 7554/eLife . 02510 . 008Figure 2—figure supplement 1 . Analysis of RNA-seq data by RNA classes . Distribution of RNA-seq reads in Dicer knockout ( red ) and control mice ( blue ) to different RNA classes . Percentages represent mean ± standard deviation across the 12 samples per genotype . Significant differences ( p<0 . 05 , student's t test ) are indicated . See Figure 2—source data 1 for exact numbers . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 00810 . 7554/eLife . 02510 . 009Figure 2—figure supplement 2 . Reads mapping to the Dicer locus confirm high knockout efficiency . ( A ) Distribution of RNA-seq reads in knockouts ( red ) and controls ( blue ) across the portion of the Dicer locus that carries the loxP-flanked exon 24 , which contains part of the catalytic domain of Dicer ( Harfe et al . , 2005 ) . The quantification of reads shows that in Dicer knockout livers , exon 24 has 14 . 6 ± 5 . 7% of the read level of control . Given that hepatocytes make up ca . 80% of the liver , this means that the knockout in these cells must be close to 100% . ( B ) Expression plots of Dicer mRNA ( left panel ) and pre-mRNA ( right panel ) in Dicer knockouts ( red ) and control ( blue ) . The two time series are plotted together in the same graph ( vertical lines connect the data points from the two series ) . Abscissas show Zeitgeber Time , left ordinates RPKM values , right ordinates raw read number only normalised to sequencing depth in the samples . Since exon 24 that is deleted after knockout induction is relatively short , reads accumulated over the whole locus on the pre-mRNA level barely change and reads on the mRNA level drop to ca . 50% . Note that Dicer expression is non-circadian in liver . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 00910 . 7554/eLife . 02510 . 010Figure 2—figure supplement 3 . Several genes show extreme transcriptional changes in Dicer knockouts . ( A ) A number of genes showed strikingly different transcription in Dicer knockouts , including several imprinted loci such as Igf2 or Cdkn1c . ( B ) Strong up-regulation was observed for a group of genes known for their development-specificity and sexual dimorphism with usually low expression in adult male livers , such as certain members of the cytochrome P450 ( Cyp ) superfamily ( Wiwi et al . , 2004; Buckley and Klaassen , 2009 ) , for example Cyp2b9 , Cyp2b13 , Cyp3a41a and Cyp3a41b . ( C ) Conversely , several male-specific transcripts were strongly downregulated , for example Cyp2d40 , Cyp4a12a , Ugt2b38 . These gene expression signatures have previously been reported from Dicer knockouts and could indicate that hepatocytes turn to a more fetal-like state upon miRNA loss ( Hand et al . , 2009; Sekine et al . , 2009a ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 010 In the complete data set , we detected 13 , 408 individual protein-coding loci ( Figure 2C ) . The intersecting set ( mRNA and pre-mRNA detected in knockout and control ) contained 11 , 990 genes ( 89 . 4% ) ; another 361 transcripts ( 2 . 7% ) common to both genotypes corresponded to intronless genes . To assess the changes engendered by miRNA loss , we first compared mRNA and pre-mRNA levels separately between the genotypes ( averaging over timepoints ) . It was apparent that Dicer deficiency caused large changes in mRNA and pre-mRNA abundance ( Figure 2D , E ) . Applying a 1 . 5-fold cut-off on differential expression , we found that 30 . 3% of mRNAs and 15 . 5% of pre-mRNAs were significantly changed in the Dicer knockout . These changes were skewed towards increased levels for mRNAs ( 19 . 8% increased , 10 . 5% decreased ) but slightly less so for pre-mRNAs ( 9 . 8% increased , 5 . 7% decreased ) . In summary , these findings confirmed our expectation that global miRNA loss would lead to the widespread de-repression of targets that affected a substantial part of the transcriptome . Inevitably , these primary effects would provoke numerous secondary responses involving altered transcription . Striking transcriptional effects were detected for several imprinted loci and for a group of known development-specific and sexually dimorphic genes ( see Figure 2—figure supplement 3 for examples ) , as previously reported from Dicer knockouts ( Hand et al . , 2009; Sekine et al . , 2009a ) . We excluded the group of transcriptionally most altered genes in subsequent circadian analyses , as vastly different transcription rates would likely confound comparative analyses of rhythmic properties , in particular regarding the precise post-transcriptional contributions ( marked in red in Figure 2D , E ) . We next calculated transcriptome-wide mRNA/pre-mRNA ratios , which are a measure of mRNA stability and thus an indicator of post-transcriptional regulation ( Zeisel et al . , 2011 ) . The distribution of mRNA/pre-mRNA ratios was globally shifted to higher values in the knockout ( Figure 2F ) and we thus assessed whether an increased mRNA/pre-mRNA ratio in the Dicer knockout could be predictive of miRNA activity . Using experimentally validated targets ( Hsu et al . , 2011 ) of miR-122 as a positive control and of miR-124 as a negative control ( this miRNA is not expressed in liver , [Landgraf et al . , 2007] ) , we indeed observed that the majority of miR-122 targets showed increased mRNA/pre-mRNA ratios in Dicer knockouts , whereas most miR-124 targets did not ( Figure 2G ) . Ribosomal protein mRNAs , whose 3′ UTRs are particularly short ( Caldarola et al . , 2009 ) and thus probably not regulated by miRNAs , did not show increased mRNA/pre-mRNA ratios either ( Figure 2G ) . We thus concluded that incorporating mRNA/pre-mRNA ratios in our analyses could enrich for likely direct miRNA targets in the dataset . We next assessed the expression of transcripts encoding core clock components and clock regulators ( listed in Figure 3—source data 1 ) . With regard to mRNA/pre-mRNA ratios , the transcripts of the Period gene family ( Per1 , Per2 , Per3 ) showed a moderate ( average 1 . 6-fold ) , but significant increase in Dicer knockouts ( Figure 3A ) . The time-resolved RNA-seq data ( Figure 3B ) confirmed that Per1 and Per2 mRNAs accumulated to higher levels throughout the day despite similar rhythmic pre-mRNA abundance . For Per3 , the higher mRNA/pre-mRNA ratio mainly resulted from decreased transcription ( Figure 3—figure supplement 1 ) . Moreover , we noticed that Cry2 showed post-transcriptional up-regulation in the Dicer knockout and was thus a potential miRNA target , although the case was less clear than for Per1 or Per2 since Cry2 transcription was increased at certain timepoints as well . Beyond these genes , the expression of most other core clock components , including Cry1 , Clock , Bmal1 ( Figure 3B ) and others ( Figure 3—figure supplement 1 ) only showed minor changes , if at all . Consistent with the mRNA profiles , western blot analysis confirmed an estimated twofold increase in the levels of PER2 protein and 1 . 5-fold higher CRY2 ( Figure 3C ) . Surprisingly , PER1 levels were barely increased ( 1 . 1-fold higher ) . A number of mechanisms could account for the observed uncoupling of protein from mRNA levels , including secondary effects of miRNA loss acting on Per1 mRNA translation or the protein degradation machinery . While we cannot pinpoint the exact mechanism that is operative in counter-regulating Per1 mRNA increase , it is interesting to note that the RNA binding protein Syncrip ( also known as hnrnp q ) is a positive regulator of Per1 translation ( Lee et al . , 2012 ) and showed decreased expression in Dicer knockouts ( Figure 3—figure supplement 1 ) . Importantly , however , it was clear that core clock functionality was overall unimpaired , as also exemplified by the virtually unchanged expression profiles of genes that are directly regulated by CLOCK:BMAL1 , PERs and CRYs , such as Dbp ( Figure 3—figure supplement 1 , Figure 1E ) . 10 . 7554/eLife . 02510 . 011Figure 3 . The hepatic core clock is remarkably resilient to miRNA loss . ( A ) mRNA/pre-mRNA ratio analysis of core clock transcripts indicates that Per1 , Per2 , and Per3 have a significantly increased ratio in Dicer knockouts . ( B ) Time-resolved RNA-seq data for selected core clock transcripts with Dicer knockouts ( red ) , controls ( blue ) , mRNA ( left panels ) , pre-mRNA ( right panels ) , and the two time series plotted together in the same graph ( vertical lines connect the data points from the two series ) . Abscissas show Zeitgeber Time , left ordinates RPKM values , right ordinates raw read numbers only normalised to sequencing depth in the samples . Per1 , Per2 , and Cry2 thus show post-transcriptional upregulation in the Dicer KO , but other core clock transcripts such as Clock , Bmal1 and Cry1 do not . ( C ) Western blot analysis of core clock protein expression in liver nuclear extracts . For each sample , liver nuclear extracts from three mice were pooled . U2AF65 served as a loading control . Quantfication revealed that in the knockout the fold-change ( FC ) of up-regulation was ca . 2x for PER2 protein , 1 . 5x for CRY2 , and 1 . 1x for PER1 . Overall , the core clock appears to be fully functional . ( D ) Example of free-running rhythms measured from liver explants of Dicer knockout and control animals carrying the mPer2Luc reporter gene . In these mice , a PER2-LUCIFERASE fusion protein is expressed from the endogenous Per2 locus and allows for the real-time recording of circadian bioluminescence rhythms; importantly , the mPer2Luc 3′ UTR is identical to that of wild-type Per2 . Raw bioluminescence was detrended using a 24-hr moving average . In this example , the knockout has a clearly longer period than the control . ( E ) Summary of several experiments as in ( D ) from a total of 9 control and 13 knockout mice . Although there is a trend to period lengthening upon miRNA loss , this effect is statistically not significant ( p=0 . 197; student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01110 . 7554/eLife . 02510 . 012Figure 3—source data 1 . Core clock genes . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01210 . 7554/eLife . 02510 . 013Figure 3—figure supplement 1 . Core clock gene expression in Dicer knockouts . Apart from Per1 , Per2 , and Cry2 , core clock gene expression shows very little post-transcriptional changes upon miRNA loss , indicating that these transcripts are not miRNA targets in the liver . Note that some transcripts show transcriptional alterations , for example Nr1d1/Rev-erbα or Dec1 . Syncrip/Hnrnpq encodes an RNA binding protein involved in Per1 translation whose down-regulation may contribute to the discrepancy between Per1 mRNA and protein ( see main text for details ) . Dotted lines respresent cosine curve fits to the data . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 013 In our knockout model , miRNA loss occurred in the hepatocytes of a liver that was under constant entrainment by a genetically intact SCN clock . It was thus conceivable that a more severe phenotype was masked and would only become apparent under non-entrained , free-running conditions . It has indeed recently been reported that the period length of free-running rhythms in Dicer–deficient MEFs is dramatically shortened by ≈2 hr ( Chen et al . , 2013 ) . Moreover , the authors identified the lack of miRNA-mediated repression of Per1 and Per2 as the underlying mechanism . We thus measured free-running rhythms in liver explants from Dicer knockout and control mice carrying the mPer2Luc reporter allele ( Yoo et al . , 2004 ) . Interestingly , our experiments showed a trend towards period lengthening in knockout liver explants with a period that was on average 41 min longer than in controls . The effect , however , did not reach statistical significance ( p=0 . 197 ) ( Figure 3D , E ) . Overall , we concluded that Per1 , Per2 and Cry2 were likely miRNA targets in liver , but that the hepatocyte clock was relatively resilient to miRNA loss . The difference in phenotypes of liver and MEF Dicer knockouts could point to tissue-specific miRNA functions in clock regulation ( ‘Discussion’ ) . We next explored how circadian gene expression was affected beyond the core clock . We first wished to identify transcripts whose rhythms were post-transcriptionally driven by miRNAs and assess whether this mechanism could underlie the discrepancies between cyclic transcription and cyclic mRNA accumulation that have been reported ( Koike et al . , 2012; Le Martelot et al . , 2012; Menet et al . , 2012 ) . To identify oscillating gene expression transcriptome-wide and as comprehensively as possible , we adapted previously published methods for rhythmicity detection that we used at relatively low stringency ( ‘Materials and methods’ ) . Especially for pre-mRNAs , which are naturally noisier due to their low abundance , we wanted to avoid applying overly stringent criteria , as this could lead to the under-detection of transcriptionally and the overestimation of post-transcriptionally generated rhythms . Indeed , a previous study had noted that many genes that had failed to pass the imposed rhythmicity criteria on the transcriptional level visibly still exhibited transcriptional patterns that resembled those of the mRNA , albeit with much higher variability ( Menet et al . , 2012 ) . In control and Dicer knockout mice , 1630 and 1902 mRNAs , respectively , cycled with at least 1 . 5-fold peak-to-trough amplitudes ( ≈13% and 14% , respectively , of detected mRNAs; around half were shared between both groups; Figure 4A , Figure 4—source data 1 ) , consistent with previous estimates that up to 15% of mRNAs are rhythmic in wild-type mice ( Lowrey and Takahashi , 2004; Vollmers et al . , 2009 ) . On the pre-mRNA level , around 17% of transcripts were rhythmic ( 2083 in control and 2101 in knockouts; >50% shared ) . In both knockouts and controls , >60% of oscillating mRNAs were also rhythmic at the pre-mRNA level . These values are within the range of what has been reported previously from wild-type mice ( Koike et al . , 2012; Le Martelot et al . , 2012; Menet et al . , 2012 ) . Interestingly , a sizeable number of transcripts were rhythmic specifically in either controls or Dicer knockouts ( Figure 4A , Figure 4—figure supplement 1 ) . We were particularly intrigued by the relatively high number of transcripts that appeared to become rhythmic in the Dicer knockouts ( Figure 4A , sectors B–D ) and investigated whether this was caused as a side effect of low stringency rhythm detection . However , the overlap of rhythmic transcripts between genotypes did not increase with higher thresholds despite reducing the overall number of detected rhythmic events ( Figure 4—figure supplement 2 ) . Further analyses revealed that Dicer KO-specific rhythmic transcripts were strongly enriched for certain Gene Ontology ( GO ) terms , of which cell cycle-specific genes ( especially those involved in DNA replication , such as the members of the minichromosome maintenance complex , MCM ) were particularly noteworthy ( Figure 4—source data 2 , see Figure 4—figure supplement 3 for examples ) . These findings are in line with increased liver regeneration after miRNA loss ( Figure 1—figure supplement 1D , E ) and the circadian gating of the cell cycle in the regenerating liver ( Matsuo et al . , 2003 ) . 10 . 7554/eLife . 02510 . 014Figure 4 . miRNAs may drive the rhythmic accumulation of a small set of transcripts . ( A ) Venn diagram summarising the extent of rhythmicity detected on all levels , that is mRNA and pre-mRNA in Dicer KO and control . A 1 . 5-fold threshold on peak-to-trough ratio amplitudes estimated from cosine fits was imposed . Sectors are named B–I for further reference . Sector A ( not shown ) corresponds to transcripts that were not rhythmic on any level . ( B ) mRNA/pre-mRNA ratio analysis for all transcripts that were detected as rhythmic in control ( N = 1630 ) . Globally , rhythmic transcripts thus appear to distribute similarly to all transcripts . ( C ) The distribution of the change in mRNA/pre-mRNA ratio between Dicer knockout and control is globally similar for rhythmic transcripts as compared to all transcripts . Although there is a slight statistically significant upshift in the cyclic transcripts ( p=0 . 001 , Welch two-sample t test ) , rhythmic transcripts do not seem to be specifically enriched for or excluded from miRNA regulation . ( D ) Heatmap of sector I transcripts , for which rhythmicity is only detected on the level of the mRNA in control animals . Note that although rhythms are clearly most pronounced for control mRNAs , as expected , the visual impression is that there are still underlying , but noisier rhythms present on the pre-mRNA level and in Dicer knockouts . pre-mRNAs ( knockout and control ) are on a common scale , as are the mRNAs . ( E–I ) Ddx17 , Slc1a5 , Stx2 , Uba6 and Zpf697 are examples of transcripts that are transcriptionally non-rhythmic , but show mRNA rhythms that are Dicer-dependent , indicating that miRNAs could be involved in driving their post-transcriptional cycling . Note that all examples have relatively shallow rhythmic amplitudes . Interestingly , their rhythms fall into a similar phase , which could indicate that a common ( rhythmic ) miRNA could be involved . Dicer knockouts are shown in red , control in blue , mRNAs are on panels to the left , pre-mRNAs to the right; the two time series are plotted together in the same graph ( vertical lines connect the data points from the two series ) . Abscissas show Zeitgeber Time , left ordinates RPKM values , right ordinates raw read number only normalised to sequencing depth in the samples . Dotted lines are the cosine fits to the data . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01410 . 7554/eLife . 02510 . 015Figure 4—source data 1 . Rhythm parameters transcriptome-wide . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01510 . 7554/eLife . 02510 . 016Figure 4—source data 2 . GO term analysis Dicer rhythmic transcripts . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01610 . 7554/eLife . 02510 . 017Figure 4—figure supplement 1 . Transcriptome-wide rhythmicity detection in Dicer knockouts and controls . ( A ) Venn diagram summarising the extent of rhythmicity detected at all levels , that is mRNA and pre-mRNA in Dicer KO and control . A 1 . 5-fold threshold on peak-to-trough ratio amplitude estimated from cosine fits was imposed . Numbers in brackets and italics refer to how many of the transcripts found in the sector also have significantly increased mRNA/pre-mRNA ratios in Dicer knockouts . ( B–P ) Heatmaps for transcripts found in sectors B to P . Scales are independent for each panel; however , for each individual heatmap , pre-mRNAs ( knockout and control ) are on a common scale , as are the mRNAs , and can thus be directly compared . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01710 . 7554/eLife . 02510 . 018Figure 4—figure supplement 2 . Differences in detected rhythmic transcripts between Dicer knockout and control are not caused by too low stringency rhythm detection . ( A–H ) Venn diagrams summarising the extent of rhythmicity detected on all levels , that is mRNA and pre-mRNA in Dicer KO and control , using different parameters for rhythm detection . Variables were: FDR ( 0 . 05 vs 0 . 01 ) , amplitude ratio ( 1 . 5- vs 2-fold ) and the combination of two methods , JTK_cycle and harmonic fit as an AND or OR operation . The conditions shown under ( A ) are those used throughout the manuscript . The numbers in the Venn diagram segments and those in the summarising table below indicate that the relatively poor overlap between rhythmic events in the two genotypes is not caused by the comparably low stringency of rhythm detection applied in the study—on the contrary , higher stringency ( i . e . , decreasing FDR , imposing higher amplitudes , or only taking into account transcripts whose rhythm is detected by both methods , JTK_cylce AND the harmonic fit ) leads in some cases even to lower percentages in shared rhythmic transcripts . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01810 . 7554/eLife . 02510 . 019Figure 4—figure supplement 3 . The expression of genes linked to the cell cycle/DNA replication becomes rhythmic in Dicer knockout livers . Likely due to the higher cellular turnover and regeneration rate in Dicer knockout livers , the expression of many genes involved in the regulation of the cell cycle is increased and their rhythms are more pronounced . ( A ) Many genes encoding proteins involved in the initiation of DNA replication , especially members of the MCM complex , are up-regulated and become rhythmic in Dicer knockouts . ( B ) Components of the metaphase checkpoint show rhythms in Dicer knockouts . ( C ) Several genes that are required for chromosome condensation have rhythmic transcripts in Dicer knockouts . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 01910 . 7554/eLife . 02510 . 020Figure 4—figure supplement 4 . miRNAs as potential drivers of rhythmic mRNA accumulation . Shown are further examples ( next to those in Figure 4 ) of transcripts that are transcriptionally non-rhythmic , but show mRNA rhythms that are Dicer-dependent , indicating that miRNAs could be involved in driving their post-transcriptional cycling . Note that these transcripts have relatively shallow rhythmic amplitudes . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 02010 . 7554/eLife . 02510 . 021Figure 4—figure supplement 5 . Quantitative real-time PCR analysis of several transcripts shown in main Figures 4 and 5 . Shown are mean values of two time series ( filled circle and triangle ) , and the individual data points ( open circle and triangle ) plotted together in the same graph . ( A ) Stx2 and Uba6 are candidates for transcripts whose post-transcriptional rhythms are miRNA-dependent ( see main Figure 4G , H ) . ( B ) Aco2 , Fbxo21 , Ppppde1 and Sgpl1 show post-transcriptional phase delays in Dicer knockouts ( main Figure 5E ) . ( C ) Dnmt3b and Net1 show post-transcriptionally altered amplitudes in Dicer knockouts ( main Figure 5F ) . ( D ) Lipa becomes essentially non-rhythmic in the absence of miRNAs ( main Figure 5G ) . ( E ) Nr1d1/Rev-erbα pre-mRNA served as a control example for a rhythmically transcribed gene . cDNA were synthesised from 3 μg of DNase-treated total RNA using random hexamers and PCR-amplified using a SYBR green method as described in ‘Materials and methods’ . Mean levels were calculated from triplicate PCR assays for each sample and normalised to those obtained for the five control transcripts Csnk1a1 , Ctsd , Nudt4 , Smg1 and Trip12 . Oligonucleotides used for qPCR are listed in Supplementary file 2 ( Du et al . , 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 021 To explore the relationship between transcript rhythmicity and miRNA regulation , we first consulted the mRNA/pre-mRNA ratios as an indicator of post-transcriptional regulation and of mRNA stability changes that could be indicative of direct miRNA activity ( Figure 4B , C ) . For 30% of rhythmic transcripts in control mice ( 483 out of 1630 ) , mRNA/pre-mRNA ratios were significantly increased in Dicer knockouts ( Figure 4—figure supplement 1A , sectors I–P ) . Globally , rhythmic transcripts showed a slightly stronger shift to higher mRNA/pre-mRNA ratios than average ( Figure 4C; p=0 . 001; Welch's two-sample t test ) . We concluded that several hundred circadian mRNAs were potentially miRNA-regulated but that there was no striking enrichment or depletion for miRNA regulation among the circadian transcriptome . To identify candidates for post-transcriptionally driven rhythmicity involving miRNAs , we investigated the group of 290 transcripts whose cyclic accumulation occurred exclusively at the mRNA level in control animals ( sector I in Figure 4A , see Figure 4—source data 1 ) . A heatmap representation confirmed pronounced rhythmicity mainly at the control mRNA level , as expected , but also suggested that a considerable number of transcripts possibly still possessed underlying , yet noisier rhythms at the level of control pre-mRNA and/or Dicer knockout mRNA ( Figure 4D ) . We thus examined all profiles individually , also taking into account mRNA/pre-mRNA ratio changes occurring in the Dicer knockouts ( in total 93 of the 290 transcripts in sector I thus had increased mRNA/pre-mRNA ratios in knockouts , see Figure 4—figure supplement 1 ) . The visual inspection confirmed that only a minority of transcripts showed the features expected for true miRNA-driven post-transcriptional rhythmicity and that , instead , many of the non-rhythmic assignments were probably a result of noisier data ( Supplementary file 1 in Du et al . , 2014 and Figure 4—source data 1 for whole transcriptome gene expression plots and rhythmicity parameters ) . Visual selection resulted in no more than 20 transcripts for which we would confidently postulate that their rhythmic accumulation could indeed be miRNA-driven . As shown for Ddx17 , Slc1a5 , Stx2 , Uba6 , Zfp697 ( Figure 4E–I ) and others ( Figure 4—figure supplement 4 ) , these transcripts were all characterised by relatively low amplitude rhythms . As a validation with an independent technique , we confirmed the circadian expression detected by RNA-seq for two of the genes , Stx2 and Uba6 , also by quantitative real-time PCR ( qPCR ) ( Figure 4—figure supplement 5A ) . In summary , these findings suggested that miRNAs are only of marginal importance as drivers of circadian rhythmicity . Although we were able to identify several transcripts whose cyclic accumulation was Dicer-dependent and occurred post-transcriptionally , miRNA-mediated mechanisms can probably not account for the previously observed large discrepancy between mRNA and transcription rhythms ( Koike et al . , 2012; Le Martelot et al . , 2012; Menet et al . , 2012 ) . Hundreds of rhythmic mRNAs displayed increased stability ( as judged from mRNA/pre-mRNA ratios ) in Dicer knockouts ( Figure 4B , C ) and many of these transcripts likely represented direct miRNA targets . Circadian transcription paired with miRNA-mediated post-transcriptional control is thus probably a widespread phenomenon . We therefore investigated how such dual regulation added up at the level of mRNA cycling . Kinetic models of how mRNA stability influences rhythmic mRNA accumulation show that for a cyclically transcribed gene , the more stable the transcript , the later the phase and the lower the amplitude of its cycling ( Le Martelot et al . , 2012 ) , as schematically shown in Figure 5A . 10 . 7554/eLife . 02510 . 022Figure 5 . miRNAs adjust the mRNA phases and amplitudes of rhythmically transcribed genes . ( A ) Schematic representation of how mRNA stability affects circadian transcript accumulation . mRNAs with short half-lives ( blue ) will thus peak relatively early after the transcriptional peak ( orange ) and show higher peak-to-trough amplitudes than mRNAs with long half-lives ( red ) . ( B ) Heatmap representation of analysed transcripts that all show rhythmic transcription ( >1 . 5-fold amplitude ) and rhythmic mRNA accumulation ( no amplitude cut-off , in order to avoid biasing against transcripts whose amplitudes are regulated by miRNAs ) in Dicer knockouts and controls . The top and lower panels show the transcripts with ( N = 167 ) and without ( N = 505 ) , respectively , significantly increased mRNA/pre-mRNA ratios in Dicer knockouts . ( C ) Analysis of how the Dicer knockout changes the phases of rhythmic mRNA accumulation . Rhythmic transcripts from ( B ) are plotted according to the difference of the phase delay between mRNA and pre-mRNA peak in Dicer knockout vs control ( abscissa; the further to the right a transcript is located , the later its mRNA is shifted in Dicer knockouts ) and according to the change in mRNA/pre-mRNA ratio ( ordinate; transcripts at the top of the panel thus become more stable in Dicer knockouts ) . Transcripts with significantly higher mRNA/pre-mRNA ratios in Dicer knockouts are marked in red . Red and black vertical lines correspond to mean phase shifts for the two groups of transcripts . The blue line shows the correlation of phase delay and mRNA/pre-mRNA ratio over all transcripts ( Pearson's r ( 670 ) = 0 . 17; p=9 . 24e−06 ) . Difference in phase delays between the two groups of transcripts was tested using Wilcoxon rank sum test ( W ( 672 ) = 30 , 799; p=1 . 72e−07 ) . At the top of the graph , the cumulative density plot shows that the two groups of transcripts are clearly separated . ( D ) Analysis of how the Dicer knockout changes the amplitudes of rhythmic mRNA accumulation . Rhythmic transcripts from ( B ) are plotted according to how the ratio of the mRNA amplitude divided by the pre-mRNA amplitude is changed between Dicer knockout and control ( abscissa; the further to the left a transcript is located , the lower the amplitude becomes in Dicer knockouts ) and according to the change in mRNA/pre-mRNA ratio on the ordinate , as in ( C ) . Red and black vertical lines correspond to mean amplitude changes of transcripts with and without increased mRNA/pre-mRNA ratios . The blue line shows the correlation of amplitude change over all transcripts ( Pearson's r ( 670 ) = −0 . 0965; p=1 . 23e−02 ) . Difference in amplitudes between the two groups of transcripts was tested using Wilcoxon rank sum test ( W ( 672 ) = 36 , 518; p=9 . 39e−03 ) . ( E ) Expression plots for six examples of transcripts whose phases are post-transcriptionally delayed in the absence of miRNAs . Aco2 , Fbxo21 , Nampt , Ngfr , Pppde1 and Sgpl1 thus all have similar phases of rhythmic transcription ( pre-mRNA ) in Dicer knockout ( red ) and control ( blue ) ( right panels ) , but their mRNA accumulation is shifted to later times in the knockouts ( left panels ) . Dotted lines represent the cosine curve fits . ( F ) Examples of transcripts showing amplitude effects . Dnmt3a , Dnmt3b and Net1 thus all have similar transcriptional amplitudes in Dicer knockout ( red ) and control ( blue ) ( right panels ) , but the loss of miRNAs reduces the mRNA amplitudes ( left panels ) . ( G ) For rhythmically transcribed Lipa , the loss of miRNAs prevents rhythmic mRNA accumulation altogether . ( H ) Cpeb1 mRNA shows the expected phase effect ( later in Dicer knockout ) , but contrary to expectation , the mRNA amplitude is increased in the knockout . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 02210 . 7554/eLife . 02510 . 023Figure 5—source data 1 . Rhythmic transcripts used in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 02310 . 7554/eLife . 02510 . 024Figure 5—figure supplement 1 . Analysis of phase and amplitude changes occurring upon Dicer knockout . Similar to Figure 5C , D ( which shows the differences in mRNA-to-pre-mRNA phase delay and mRNA/pre-mRNA amplitude change , respectively , in Dicer knockout vs control ) , these graphs analyse mRNAs only ( A and C ) and pre-mRNAs only ( B and D ) . The figures use the same set of rhythmic transcripts as in Figure 5B–D . Overall , the data confirm that the effects of later phases and shallower amplitudes in Dicer knockouts are indeed generated on the mRNA , and not on the pre-mRNA level . ( A ) Changes in mRNA phases ( Dicer knockout vs control ) on the abscissa are plotted against the change in mRNA/pre-mRNA ratio ( Dicer knockout vs control ) on the ordinate . ( B ) Changes in pre-mRNA phases ( Dicer knockout vs control ) on the abscissa are plotted against the change in mRNA/pre-mRNA ratio ( Dicer knockout vs control ) on the ordinate . ( C ) The ratios of the mRNA amplitude in the Dicer knockout vs the amplitude in control on the abscissa are plotted against the change in mRNA/pre-mRNA ratio ( Dicer knockout vs control ) on the ordinate . ( D ) The ratios of the pre-mRNA amplitude in the Dicer knockout vs the amplitude in control on the abscissa are plotted against the change in mRNA/pre-mRNA ratio ( Dicer knockout vs control ) on the ordinate . ( E ) Plot of the effect on phase against the effect on amplitude ( i . e . , both abscissas of Figure 5C , D plotted against each other ) . The crosses show the centre of the data for transcripts with ( red ) and without ( black ) higher mRNA/pre-mRNA ratio in the Dicer knockouts . The circles show the standard deviation of the data points . Overall there is no statistically significant correlation between later phases and lower amplitudes . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 02410 . 7554/eLife . 02510 . 025Figure 5—figure supplement 2 . Permutation of the RNA-seq data set confirmed that the effects observed on the phases and amplitudes ( main Figure 5C , D ) are highly significant and specific to the Dicer knockout . Two different permutation analyses were performed . First ( panels A and B ) , independently for each gene the 12 timepoints were randomly permutated 10 , 130 times . For each individual permutation , the same newly shuffled order was applied genewise identically across control/knockout and mRNA/pre-mRNA . In the newly created data sets , rhythm detection , and analysis ( phase and amplitude determination ) were performed in an identical fashion to the analysis of the orginal data shown in the manuscript . Red triangles on the abscissa represent where the original data is located in the distribution of the random data sets . Collectively , these analyses show that there is not per se a trend in the data or its distribution that would lead to similar findings ( later phases , shallower amplitudes upon miRNA loss ) just by chance . ( A ) Left panel shows the distribution of phase delay differences between ‘higher mRNA/pre-mRNA ratio in Dicer KO’ and ‘other’ rhythmic transcripts from the permutated data and in the original data ( reported values correspond to difference between vertical black and red lines in Figure 5C ) . Only in 2 . 8% of cases did random permutations give a global phase delay difference effect that was as distant from 0 as the experimental data . Moreover ( middle panel ) , none of the permutations had a similarly high Wilcoxon statistics ( reported in top left corner of Figure 5C ) as the experimental data; Wilcoxon test statistics were converted to z-scores in order to ensure comparability between experimental and permutated data sets , which differ in the number of rhythmic transcripts ( N = 182 . 1 ± 10 . 6; mean ± SD ) . Finally ( right panel ) , similarly good Pearson's coefficients of the correlation between phase delay difference and mRNA/pre-mRNA ratios ( shown in blue in Figure 5C ) only occurred in 4 . 7% of permutations . ( B ) Analogous analysis as in ( A ) , but for amplitudes . Since the Dicer KO and control data sets kept their genotype identities in the reshuffling of time points , amplitude effects were overall conserved even after permutation , as expected . In a second permutation analysis ( panels C and D ) , time points were kept in the original order , but the assignment to ‘Dicer KO’ or ‘control’ was switched randomly . Downstream analyses as in ( A ) and ( B ) . ( C ) None of the 10 , 010 permutations had a comparable or stronger phase difference ( left panel ) analogous to that shown by vertical black and red lines in Figure 5C or a comparable Wilcoxon test z-score ( middle panel ) . Similar or stronger Pearson's correlation coefficients were only found in 0 . 03% of permutated data sets . ( D ) Amplitude analysis showing that only 3% of permutations led to a similar effect on the amplitude as the original data . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 025 We examined the circadian properties of transcripts that were rhythmic across all conditions , that is in knockouts and controls at the pre-mRNA ( >1 . 5-fold amplitude ) and mRNA levels ( without amplitude cut-off ) . Moreover , we grouped the transcripts according to whether their mRNA/pre-mRNA ratios were significantly increased in the Dicer knockout ( i . e . , transcripts more likely to represent direct miRNA targets; N = 167 ) or not ( i . e . , transcripts less likely to represent direct miRNA targets; N = 505 ) . A heatmap representation of the transcripts considered is shown in Figure 5B ( see also Figure 5—source data 1 for gene list ) . We first determined how miRNA loss affected the phases of rhythmic mRNA accumulation ( Figure 5C ) . We thus calculated the time lag between the transcripts' pre-mRNA and mRNA peaks for the Dicer knockouts , and subtracted the equivalent values calculated from the control mice . Positive and negative ‘phase delay differences’ thus signified that in the Dicer knockout , mRNA accumulation peaked later and earlier , respectively , than in controls ( an additional analysis of absolute pre-mRNA and mRNA phases can be found in Figure 5—figure supplement 1A , B ) . Phase delay differences were plotted against the mRNA/pre-mRNA ratio change ( Figure 5C; the 167 transcripts with a significantly higher ratio in the knockout are marked in red ) . This analysis revealed a shift to later phases of mRNA accumulation in the Dicer knockout , which correlated with increased mRNA/pre-mRNA ratios ( blue line in Figure 5C; Pearson's r ( 670 ) = 0 . 17; p=9 . 24e−06 ) . In the absence of miRNAs , mRNAs with significantly increased mRNA/pre-mRNA ratios thus peaked on average 48 min later ( vertical red line in Figure 5C ) , whereas the average for all other rhythmic transcripts ( vertical black line ) lay at 5 min . Both groups of transcripts , that is those that showed higher mRNA/pre-mRNA ratios and those that did not , were significantly different ( p-value 1 . 72e−07; Wilcoxon rank sum test; cumulative density plot at the top of Figure 5C ) . Although the globally detectable effect of miRNA loss on the phases of mRNA rhythms was relatively modest ( less than 1 hr ) , individual transcripts were affected more dramatically ( Figure 5E ) . For genes such as Aconitase 2 ( Aco2 ) , F-box protein 21 ( Fbxo21 ) , Nicotinamide phosphoribosyltransferase ( Nampt ) or Nerve growth factor receptor ( Ngfr ) , the phase-shifts amounted to several hours ( Figure 5E ) . We also confirmed the observed effects for selected genes by qPCR ( Figure 4—figure supplement 5B ) . Using a similar approach , we next assessed how the loss of miRNAs influenced the amplitudes of mRNA rhythms . For each transcript , we calculated the ratio of the mRNA amplitude to the pre-mRNA amplitude in the Dicer knockout and divided this value by the corresponding ratio in control mice . We thus found that miRNA loss led to globally shallower amplitudes in the Dicer knockout ( Figure 5D ) , although the mean effect was small and amounted only to a 7% decrease in fold-amplitude ( see difference between red and black vertical lines in Figure 5D; p-value for difference between transcripts with and without increased mRNA/pre-mRNA ratio: 9 . 39e−03; Wilcoxon rank sum test ) . Nevertheless , individual examples clearly suggested that miRNAs assume important functions in ensuring that rhythmically transcribed genes give rise to rhythmic mRNAs that oscillate with the desired amplitudes and magnitudes , as shown in Figure 5F ( qPCR validations in Figure 4—figure supplement 5C , D ) for the expression of DNA methyltransferases 3A ( Dnmt3a ) , 3B ( Dnmt3b ) and Neuroepithelial cell transforming gene ( Net1 ) , as well as for several other genes in Figure 5E . In some cases , such as Lysosomal acid lipase A ( Lipa ) , rhythmicity generated at the pre-mRNA level was completely lost at the mRNA level when miRNAs were absent ( Figure 5G ) . Here , miRNA-mediated decay could thus represent the mechanism ensuring that after cyclic transcription , mRNAs are sufficiently unstable to show rhythmic accumulation at all . Interestingly , the protein encoded by Lipa , lysosomal acid lipase ( LAL ) , is the key enzyme hydrolysing cholesteryl esters and triglycerides stored in lysosomes after LDL receptor-mediated endocytosis ( Fouchier and Defesche , 2013 ) and has been reported to be rhythmic in liver , but non-rhythmic in other organs ( Tanaka et al . , 1985 ) . Conceivably , miRNAs could be involved in rendering Lipa rhythms tissue-specific . Altogether the effects of miRNA loss on amplitudes were less uniform than those on phases ( compare Figure 5C , D ) . Several reasons could account for this difference . First , we noticed that amplitude estimations appeared technically more error-prone than phase estimations . Both parameters were read from cosine curves fitted to the data , but the peak-trough symmetry imposed by the cosine function frequently resulted in the underestimation of amplitudes; this was especially the case for high amplitude rhythms with pronounced ‘spiky’ appearance ( i . e . , with rapid rising and declining phases; see cosine fits of Nr1d1/Rev-erbα and Dbp in Figure 3—figure supplement 1 ) . Second , miRNAs likely only represent one of several mechanisms operative in amplitude modulation; direct consequences of miRNA loss on target mRNA amplitudes might thus be partially masked by secondary effects occurring in the Dicer knockouts . Third , the model of how phases and amplitudes should ideally correlate ( Figure 5A ) is almost certainly an oversimplification . Indeed , for transcripts such as Cpeb1 ( Figure 5H ) , later phases were even accompanied by higher amplitudes . Globally , there was indeed no significant correlation between phase delays and amplitude decreases ( Figure 5—figure supplement 1E ) . An uncoupling of amplitude and phase effects could occur , for example , through miRNA activity that is not constant over the day but confined to specific timepoints ( ‘Discussion’ ) . Overall , we concluded that the phase delays seemed to be the dominant , consistent consequence of miRNA loss in liver . Nevertheless , both the amplitude decrease and the phase delay were individually highly significant , as also shown by permutation tests in which we randomly reshuffled either the timepoints of the original data or the Dicer KO/control assignments on a genewise basis , followed by analyses of rhythm parameters . Among >10 , 000 datasets with permuted timepoints , the likelihood of finding similar or stronger delays in phase difference ( in both size and significance ) or correlations ( between mRNA/pre-mRNA ratios and phase difference ) than those in the original data were very low ( Figure 5—figure supplement 2A ) . The observed shift in phase differences was therefore not an intrinsic data set property . By contrast , because the expression level differences ( Dicer KO vs control ) were untouched by the reshuffling of timepoints , a sizeable number of permutations still showed amplitude effects comparable to the original data ( Figure 5—figure supplement 2B ) . However , when reshuffling occurred at the level of the genotypes , the permuted data no longer retained any trend reminiscent of the original data , neither for phases nor for amplitudes ( Figure 5—figure supplement 2C , D ) . We concluded that the correlations uncovered between miRNA loss and rhythmic gene expression parameters were not likely to have arisen by chance , for example as a result of a specific predisposition of the data structure or distribution that could have occurred in our particular time series and the downstream analyses . The correct timing and extent of rhythmicity is important for the daily execution of clock-regulated physiological functions . It was thus likely that the observed phase and amplitude changes had an impact on liver functions . To identify pathways that were particularly affected by miRNA loss , we investigated whether the post-transcriptionally up-regulated circadian transcripts were associated with specific GO terms ( Figure 6A ) . Using the same data set of circadian transcripts as before ( Figure 5 ) , several over-represented GO terms could be identified with statistical significance ( FDR-adjusted p-value<0 . 05 ) in the group of rhythmic transcripts with increased mRNA/pre-mRNA ratios in Dicer knockouts , but not among the remaining rhythmic transcripts . However , the enrichment was relatively low , indicating that the function of miRNAs in regulating circadian gene expression was broad rather than confined to specific rhythmic functions . The individual GO terms affected three major fields: cell adhesion , apoptosis/development , and ( lipid ) metabolism . In particular , the latter caught our attention because serum analyses of Dicer knockout animals indicated metabolic phenotypes as well , especially for cholesterol , triglyceride , and glucose levels , which were all reduced ( Figure 6—figure supplement 1 ) . At least in part , these effects were likely mediated by the loss of miR-122 , whose inactivation has previously been shown to impact lipid metabolism ( Krutzfeldt et al . , 2005; Esau et al . , 2006; Gatfield et al . , 2009; Wen and Friedman , 2012 ) . 10 . 7554/eLife . 02510 . 026Figure 6 . Regulation of circadian output pathways by specific miRNAs . ( A ) GO term analysis identifies specific pathways enriched in the group of circadian transcripts that are likely miRNA-regulated ( higher mRNA/pre-mRNA ratio in Dicer knockouts ) but not so in the remainder of rhythmic mRNAs; p-values are corrected for FDR due to multiple testing . Deviations from transcript numbers in Figure 5 are due to genes without associated GO term . ( B ) List of miRNAs for which more predicted targets than expected are found in the group of circadian transcripts with higher mRNA/pre-mRNA ratios in Dicer knockout . The miRNAs marked in red are expressed at readily detectable levels in liver; those in black were identified in liver by small RNA-seq , but only at very low levels . For miRNAs in grey , we found no evidence for expression in liver by RNA-seq . miRNA predictions performed with targetscan ( ‘Materials and methods’ ) . ( C ) Reporter construct to study the function of 3′ UTRs sequences . Within a lentiviral expression cassette ( defined by the long terminal repeats , LTRs , in blue ) , two different luciferase mRNAs are transcribed from the bidirectional Pgk1 promoter . Firefly luciferase serves as the reporter gene to test the effect of a particular 3′ UTRs ( pink ) , whereas renilla luciferase serves as a control for internal normalisation . ( D ) Effect of Per2 , Per1 and Cry2 3′ UTRs on reporter expression in control ( blue ) and Dicer knockout ( red ) primary hepatocytes . Lentivirally delivered full-length Per2 3′ UTRs-containing reporter ( nt 1–2129 ) is thus de-repressed in Dicer knockouts; this effect is mediated by the second half of the UTR ( 1017–2129 ) , probably in synergy through fragment 1425–1758 , which contains the predicted miR-24 and miR-30 sites from Chen et al . ( 2013 ) , and fragment 1739–2129 , which contains a predicted site for the miR-25/92 family . Per1 UTR is only slightly de-repressed in Dicer knockout hepatocytes . Cry2 appears to be miRNA-regulated by sites located in the 3′-terminal portion of the UTR , which contains predicted sites for miR-24 , miR-340 and let-7 . The miRNAs listed below the graph represent targetscan predictions filtered for those detected in liver by small RNA-seq . ( A ) and ( B ) identify miRNA regulation reported by Nagel et al . ( 2009 ) and Chen et al . ( 2013 ) , respectively . Data correspond to mean ± standard deviation from triplicate assays from independent lentiviral transductions using hepatocytes from the same mice . Each experiment was confirmed at least twice using hepatocytes from independent animals . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 02610 . 7554/eLife . 02510 . 027Figure 6—figure supplement 1 . Serum analyses in Dicer knockouts indicate metabolic defects . Analysis of ( A ) total cholesterol , ( B ) HDL cholesterol , ( C ) LDL cholesterol , ( D ) triglycerides , ( E ) glucose , and ( F ) serum lipase activity in the serum of 12 control and 13 Dicer knockout animals . Similar phenotypes on lipid metabolism have already been described for the single inactivation of miR-122 ( Krutzfeldt et al . , 2005; Esau et al . , 2006; Gatfield et al . , 2009 ) , and are thus probably due to the lack of this particular miRNA . ( G ) Summary of data shown in ( A–F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 027 The breadth of miRNA loss ( including potential secondary effects ) poses an obvious limit to the suitability of Dicer knockouts for further functional studies . Ideally , the targeted inactivation of single miRNAs that fulfil prospective functions in circadian regulation would allow for more straightforward investigations . To identify such miRNAs , we analysed whether the circadian transcripts that became more stable in Dicer knockouts were enriched for predicted miRNA binding sites ( Lewis et al . , 2005 ) . Interestingly , this analysis revealed nine miRNA seed families with more circadian targets than expected by chance ( Figure 6B ) specifically in the data set with higher mRNA/pre-mRNA ratios . Of these , miR-25 , miR-92 , miR-122 , miR-22 , and miR-29 are expressed at readily detectable levels in liver , as shown previously ( Landgraf et al . , 2007; Gatfield et al . , 2009; Vollmers et al . , 2009 ) and confirmed by northern blot ( Figure 1—figure supplement 2 ) and small RNA-seq ( data not shown ) . Given that miR-122 already has a history as a circadian output modulator ( Gatfield et al . , 2009 ) , it is tempting to speculate that the other miRNAs may represent specialised regulators of the rhythmic transcriptome as well . Interestingly , miR-29 is also one of the miRNAs that was reported to regulate Per1 in MEFs ( Chen et al . , 2013 ) . In the future , it would thus be exciting to explore the role of these miRNAs in targeted experiment in the context of clock-regulated biological pathways . The combination of increased mRNA/pre-mRNA ratios and miRNA target predictions is suggestive of bona fide miRNA-mediated regulation . Nevertheless , before embarking on follow-up experiments for individual transcripts , validation experiments to map miRNA binding sites in the targets' 3′ UTRs and to identify the responsible miRNAs will be necessary . To this end , we designed a luciferase reporter assay based on the lentiviral transduction of Dicer knockout and control primary hepatocytes ( Figure 6C ) . We chose the core clock transcripts to demonstrate the utility of the assay , as we were particularly intrigued by the difference in phenotypes between the Dicer knockout in liver ( i . e . , mild period lengthening , Figure 3E ) and the previously reported strong period shortening in MEFs ( ≈2 hr ) , which the authors attributed to miRNA-regulation of Per1 ( by miR-24 and miR-29 ) and Per2 ( by miR-24 and miR-30 ) ( Chen et al . , 2013 ) . We tested whether the firefly luciferase reporter gene carrying the full-length Per1 , Per2 or Cry2 3′ UTRs ( or fragments thereof ) was subject to de-repression in Dicer knockout hepatocytes , as compared to control hepatocytes . Consistent with the RNA-seq and western blot analyses ( Figure 3B , C ) , this assay indeed identified Per2 as the main miRNA-regulated core clock component ( Figure 6D ) . Moreover , truncated versions of the 3′ UTR that contained the miR-24 and miR-30 sites identified by Chen et al . recapitulated some of the miRNA regulation observed for the full-length Per2 3′ UTR sequence . However , additional regulation was also conferred by a portion of the 3′ UTR containing a predicted target site for the miR-25/92 seed family ( fragment 1739–2129 , Figure 6D ) and full regulation may thus require the synergistic activity of several sites . Per1 and Cry2 full-length 3′ UTRs showed 1 . 2-fold and 1 . 8-fold up-regulation , respectively ( Figure 6D ) . This was in the same range as what we had observed for the endogenous proteins by western blot ( Figure 3C ) . In particular , the lack of strong Per1 regulation in liver was a striking difference to the MEF data and could be responsible for phenotypic differences , as Chen et al . have proposed that the faster translation and higher accumulation of PERs ( approximately twofold higher PER1 and PER2 peak levels in Dicer knockout ) shortens the time delay within the oscillator's main feedback loop and speeds up the clock . It is possible that tissue-specific miRNA expression contributes to the observed differences; in line with a previous study ( Vollmers et al . , 2012 ) , we have thus observed that miR-30 is among the most abundant miRNA species in liver ( within the top 10 of miRNAs detected by small RNA-seq ) , but that miR-29 and miR-24 are expressed at lower levels ( not in the top 50 ) ( data not shown ) , which could thus partially explain the comparably weak effect of the Dicer knockout on PER1 in liver . In summary , in addition to shedding light on miRNA regulation of core clock transcripts , we concluded that the reporter assay in hepatocytes represented a useful tool for future studies aimed at verifying direct miRNA targets and at unravelling the contribution of specific miRNAs .
Recent publications have suggested widespread contributions of post-transcriptional regulation to circadian mRNA cycling and have challenged the assumption that rhythmic transcription is the main driver of rhythmic gene expression ( Koike et al . , 2012; Le Martelot et al . , 2012; Menet et al . , 2012 ) . MicroRNAs are important post-transcriptional regulators ( Bushati and Cohen , 2007; Krol et al . , 2010; Fabian and Sonenberg , 2012 ) whose functions in mammalian circadian biology are only beginning to emerge ( Lim and Allada , 2013; Mehta and Cheng , 2013 ) . We have used a drastic approach , that is the knockout of the miRNA biogenesis factor Dicer ( Harfe et al . , 2005 ) , to assess how the rhythmic transcriptome in mouse liver is altered in the absence of miRNA-mediated regulation ( Figure 7 ) . Considering the severity of the genetic model , our finding that the circadian system was globally very resilient to miRNA loss was surprising and somewhat reminiscent of the stability of circadian oscillators with regard to other perturbations such as large fluctuations in general transcription rates and temperature ( Dibner et al . , 2009 ) . Moreover , our findings are in line with the notion that miRNAs frequently function in the fine-tuning and modulation of gene expression ( Krol et al . , 2010; Fabian and Sonenberg , 2012; Yates et al . , 2013 ) . 10 . 7554/eLife . 02510 . 028Figure 7 . Model , summary , and speculations . In the liver , we propose that miRNAs play three distinct roles in the regulation of rhythmic gene expression . First , around 30% of rhythmically transcribed genes appear to be also regulated by miRNAs , which tunes the phases and amplitudes of mRNA accumulation . This group of transcripts is enriched for predicted binding sites for several miRNAs ( Figure 6B ) , such as miR-122 , miR-22 , miR-25/92 , and others . Second , for a very small group of transcripts ( <2% of all rhythmic mRNAs ) rhythms may be driven by miRNAs , but miRNA activity is unlikely to underlie major discrepancies between the rhythmic transcriptome and rhythmic transcription . Finally , miRNA activity seems to be dispensable for a functional hepatic core clock . Nevertheless , it is conceivable that under conditions where the clock is brought out of equilibrium and has to readjust ( e . g . , jet lag , food shifting ) the identified miRNA-mediated regulation in particular of Per2 ( but also of Per1 and Cry2 ) would be of functional importance . DOI: http://dx . doi . org/10 . 7554/eLife . 02510 . 028 The hepatic core clock remained fully functional and showed only modest period lengthening of free-running rhythms ( on average by 40 min ) in Dicer knockout liver explants . This phenotype is a plausible consequence of the approximately twofold increase in expression of PER2 , which we identified as the main miRNA-regulated core clock component in hepatocytes . Mice in which Per2 levels are increased due to additional Per2 transgenes thus show a similar period lengthening ( Gu et al . , 2012 ) , while Per2 knockout mice have short periods before becoming arrhythmic ( Zheng et al . , 1999; Bae et al . , 2001 ) . Conceivably , the longer period seen in Dicer knockout liver explants is masked in the animal due to constant entrainment by the SCN and other cues . It is well possible that a phenotype will manifest in knockouts only under conditions that bring the steady-state relationship between entrainment cues and the liver clock out of equilibrium , such as in jet lag or food shifting experiments ( Damiola et al . , 2000 ) ; this situation would be consistent with models stating that miRNAs confer robustness to oscillatory networks and denoise negative feedback loops ( Cohen et al . , 2006; Gerard and Novak , 2013 ) . Intriguingly , Per2 has already been noted to act as a link between systemic entrainment signals and local liver clocks ( Kornmann et al . , 2007 ) . It would thus be exciting to investigate clock readjustment kinetics in Dicer knockouts , for example using a novel method for the real-time recording of liver rhythms ( Saini et al . , 2013 ) . Interestingly , the previously reported dramatic period shorting ( ≈2 hr ) in Dicer knockout MEFs ( Chen et al . , 2013 ) is in stark contrast to what we observed in liver . Very likely , tissue-specific differences in miRNA activity form the basis of these differences and could explain that PER1 and PER2 are equally strongly affected in Dicer knockout MEFs , whereas in liver PER1 appears to be less strongly regulated . Interestingly , an earlier study on the SCN clock has revealed brain-specific miR-219 and miR-132 as regulators of period length and light-induced clock resetting , respectively ( Cheng et al . , 2007 ) . It is thus tempting to speculate that miRNAs generally function to post-transcriptionally tune the core clock in a cell type-specific fashion . At least in part , miRNA activity could thus also underlie known tissue-specific differences in phases and free-running period lengths ( Yoo et al . , 2004 ) . The similarity of core clocks proved advantageous for the comparison of clock-controlled gene expression between Dicer knockouts and controls animals . To this end , we found that miRNAs contributed only marginally to generating transcript rhythms at the post-transcriptional level . Less than 2% of all rhythmic mRNAs thus fulfilled our criteria for miRNA-driven rhythmicity and the identified examples ( e . g . , Figure 4E–I ) showed low amplitude cycling . This finding is consistent with the notion that miRNAs are usually highly stable molecules and generally not expected to show pronounced daily variations in abundance ( see below ) . By contrast , we have found that miRNAs exert important functions in refining the rhythmic mRNA accumulation profiles of cyclically transcribed genes . We thus estimate that for about 30% of the rhythmic transcriptome , miRNA-mediated regulation adds an additional layer of post-transcriptional control . Concretely , the absence of miRNAs caused a global shift of mRNA accumulation to later phases during the day . Moreover , peak-to-trough amplitudes of mRNA accumulation were overall reduced in Dicer knockouts , although this effect was globally smaller and more variable across the transcriptome . These results establish that miRNAs are in charge of an important regulatory control level sandwiched between rhythmic transcription and the final rhythmic mRNA and , eventually , protein output . Because many miRNAs are expressed in a tissue-specific fashion , they may thus be key to converting rhythmic transcriptional information into the desired tissue-specific rhythmic outcome . Our miRNA target prediction analysis would suggest that a discrete group of miRNAs ( miR-25/92 , miR-122 , miR-22 and miR-29 ) is specifically involved in the post-transcriptional tuning of circadian transcripts in the liver . Using the lentiviral reporter system , it will be possible to confirm some of the predicted miRNA-target interactions in hepatocytes , which could then form the basis for targeted experiments in which the expression of individual miRNAs is manipulated , for example by genetic or antisense loss-of-function techniques . Constantly expressed miRNAs can explain many of the observed phase and amplitude effects ( Figure 5A ) . For transcripts such as Lipa , miRNA-mediated regulation may thus merely represent a convenient mechanism to keep mRNAs sufficiently unstable to ensure that they cycle at all . However , the post-transcriptional profiles of transcripts such as Ddx17 , Slc1a15 ( Figure 4E , F ) , or Cpeb1 ( Figure 5H ) , could be suggestive of rhythmic miRNA activity . Several miRNAs have been previously reported as potentially rhythmic in mouse liver ( Na et al . , 2009; Vollmers et al . , 2012 ) , but our small RNA-seq profiling ( data not shown ) , as well as previous analyses of individual miRNAs ( Gatfield et al . , 2009 ) could not confirm high-cycling miRNA species . Similarly , the recently reported circadian expression of Dicer itself ( Yan et al . , 2013 ) was not evident from our data ( Figure 2—figure supplement 2B ) . Collectively , these findings suggest that the potential for rhythmic miRNA activity in mouse liver is relatively low . However , our approach , which analyses miRNA activity almost exclusively from the perspective of miRNA targets , cannot resolve this issue; a dedicated study would thus be informative . The Dicer knockout allele that we have used ( Harfe et al . , 2005 ) has previously served as an entry point to define the role of miRNAs in many other fields ( e . g . , Harris et al . , 2006; Chong et al . , 2008; Sheehy et al . , 2010 ) . In spite of the insinuated universality of miRNA loss , there are some limitations to the approach . First , in a few exceptional cases , miRNA biogenesis can be independent of canonical Dicer processing , for example for miR-451 ( Yang and Lai , 2011 ) . This miRNA is indeed not depleted from Dicer knockout livers ( Figure 1—figure supplement 2B and data not shown ) . Moreover , similar to most other studies , we deduced miRNA activity from target mRNA abundance . This approach seems justified overall because genome-wide measurements of miRNA action on mRNA and protein levels have shown that these correlate generally well ( Baek et al . , 2008; Selbach et al . , 2008 ) . Nevertheless , it should be kept in mind that miRNAs may affect mRNA levels less strongly than protein levels ( Selbach et al . , 2008; Yang et al . , 2010 ) , and that mRNA changes can even be absent altogether ( e . g . , Bhattacharyya et al . , 2006 ) . Finally , distinguishing direct from indirect effects in Dicer knockout data is challenging . We have used a measure of mRNA stability ( mRNA/pre-mRNA ratios ) to deplete our dataset of gene expression changes that involve altered transcription and that are probably mostly indirect . However , in specific cases miRNAs have been reported to directly interfere with transcription through promoter complementarity ( reviewed in Huang and Li , 2012 ) . Obviously , our analyses miss direct effects of miRNAs that are not post-transcriptional . On the other hand , the use of increased mRNA stability ( mRNA/pre-mRNA ratios ) to enrich for likely direct miRNA targets will inevitably result in false-positives that are post-transcriptionally regulated due to indirect effects , for example because components of the mRNA decay machinery are direct miRNA targets . In spite of these caveats , our study represents a comparatively complete analysis of miRNA activity in liver and should prove a valuable resource for further investigations of circadian and non-circadian functions that these regulatory molecules exert in hepatic gene expression , metabolism , and physiology .
Animal studies were conducted in accordance with the regulations of the veterinary office of the Canton of Vaud ( authorization VD2376 ) . All alleles used in the study have been published before that is , Dicerflox ( Harfe et al . , 2005 ) , AlbCre-ERT2 ( Schuler et al . , 2004 ) , and mPer2Luc ( Yoo et al . , 2004 ) , and were kindly provided by the Tabin , Metzger , and Takahashi labs , respectively . The genetic background of the animals used in this study was mixed . Male littermate knockout ( Dicerflox/flox;AlbCre-ERT2 ) and control ( Dicerflox/+ or Dicer+/+;AlbCre-ERT2 ) animals aged 3–6 months received tamoxifen treatment by intraperitoneal injections over 5 days ( total 2 mg ) essentially as described ( Schuler et al . , 2004 ) . We combined both wild-type and heterozygous male littermates for the controls because microarray analyses indicated that a single functional Dicer allele was sufficient to ensure miRNA processing to wild-type levels ( Figure 1—figure supplement 2A ) , as expected from previous studies ( Harfe et al . , 2005; Kanellopoulou et al . , 2005; Murchison et al . , 2005; Chen et al . , 2008; Frezzetti et al . , 2011 ) . Haploinsufficiency has indeed so far only been reported for tumour suppressor functions of Dicer in certain non-liver tumorigenesis models ( Kumar et al . , 2009; Arrate et al . , 2010; Lambertz et al . , 2010; Nittner et al . , 2012; Yoshikawa et al . , 2013 ) . After the last injection , mice were entrained to a 12-hr light:12-hr dark photoperiod with free access to food and water for 1 month , sacrificed at the respective Zeitgeber times ( ZT 0 , 4 , 8 , 12 , 16 , 20 ) , and the liver was snap-frozen in liquid nitrogen . Oligos used to genotype Dicer knockout efficiency from genomic DNA ( Figure 1—figure supplement 1A ) were DicerF1 , DicerR1 , and DicerDel ( see Supplementary file 2 in Du et al . , 2014 for sequences ) . Total RNA was prepared essentially as described previously ( Gatfield et al . , 2009 ) . The efficiency of the Dicer knockout was confirmed for each liver sample by northern blot probing for miR-122 . The protocol for miRNA northern blots has been described ( Gatfield et al . , 2009 ) . Total RNA pools for the two time series in knockouts and controls were assembled using identical amounts of RNA from 3 to 4 animals per pool . All pools were DNase-digested ( RQ1 DNase , Promega , Madison , WI ) in order to eliminate potential contamination from genomic DNA that would have rendered pre-mRNA quantifications impossible . The absence of genomic DNA contamination was confirmed in all pools by the comparison of RT− and RT+ reactions in quantitative real-time PCR ( qPCR ) using genomic probes . For qPCR analysis , cDNA was synthesised from 6 μg of DNase-treated total RNA using random hexamers and SuperScript II reverse transcriptase ( Invitrogen , Carlsbad , CA ) according to the supplier's instructions . cDNA was PCR-amplified using FastStart Universal SYBR Green Master ( Roche , Basel , Switzerland ) . Mean levels were calculated from triplicate PCR assays for each sample and normalised to those obtained for the control transcripts GusB and Eef1a1 ( Figure 1E ) and Csnk1a1 , Ctsd , Nudt4 , Smg1 and Trip12 ( Figure 4—figure supplement 5 ) . All oligonucleotide sequences are listed in Supplementary file 2 ( Du et al . , 2014 ) . Total RNAs ( 100 ng; same samples as for the RNA-seq series ) were hybridised to Mouse miRNA Microarray , Release 18 . 0 , 8 × 60K ( Agilent Technologies , Santa Clara , CA ) using miRNA Complete Labelling and Hyb Kit ( Agilent Technologies ) according to the supplier's instructions . An invariant normalisation method using five internal spikes ( hur_1 , hur_2 , hur_4 , hur_6 , mr_1 ) was used to normalise Dicer KO vs control samples in Figure 1—figure supplement 2B . Quantile normalisation was used for heterozygote knockout vs wild-type samples in Figure 1—figure supplement 2A . The dual luciferase reporter cassette in which the bidirectional phosphoglycerate kinase 1 ( PGK1 ) promoter ( Amendola et al . , 2005 ) drives the expression of firefly luciferase ( FL , used for 3′ UTR cloning ) and renilla luciferase ( RL , for normalisation ) was designed in silico and purchased as a synthetic clone ( GenScript , Piscataway , NJ ) in pUC57 plasmid . After excision from the plasmid via flanking SalI sites , the cassette was used to replace the XhoI cassette in lentiviral plasmid pWPT-GFP ( Addgene , Cambridge , MA ) , resulting in plasmid prLV1 that was used to clone 3′ UTRs of interest via XhoI/NotI restriction sites downstream of the firefly luciferase coding sequence . 3′ UTR sequences were amplified by PCR from liver cDNA or genomic DNA with specific oligonucleotides that carry NotI sites in the forward ( F ) and XhoI sites in the reverse ( R ) oligos , as listed in Supplementary file 2 ( Du et al . , 2014 ) . The identity of the cloned UTRs was verified by sequencing . From the prLV1 vectors with cloned UTRs , lentiviral particles were produced in 293T cells using envelope vector pMD2 . G and packaging plasmid psPAX2 as previously described ( Salmon and Trono , 2007 ) . Viral supernatant was spun 2 hr at 25 , 000 rpm , 4°C using Optima L-90K Ultracentrifuge ( SW32Ti rotor; Beckman , Brea , CA ) , then viral particles were resuspended with primary hepatocyte medium . Mice were anesthetised and livers were perfused through the inferior vena cava with 50 ml of washing buffer ( 137 mM NaCl , 2 . 7 mM KCl , 0 . 5 mM Na2HPO4 , 10 mM HEPES , pH 7 . 65 ) supplemented with 0 . 5 mM EDTA , then 50 ml of digestion buffer ( washing buffer supplemented with 7 mM CaCl2 , 0 . 4 mg/ml collagenase [C5138; Sigma-Aldrich , St . Louis , MO] ) at flow rate of 5 ml/min and at 37°C . Isolated cells were filtered with a cell strainer ( 100 μm , BD Falcon , Franklin Lakes , NJ ) and washed with 40 ml of primary hepatocyte medium ( Medium 199 , GlutaMAX Supplement , supplemented with 1% penicillin/streptomycin/glutamine ( PSG ) , 0 . 1% BSA , 10% FCS [all Gibco/Life Technologies , Carlsbad , CA] ) . The cells were resuspended with 10 ml of primary hepatocyte medium , counted , and plated in 12-well plates ( coated with 0 . 2% gelatin ) at a density of 2 × 105 cells/well . Medium was changed 4 hr later and viral supernatant was added . The cells were maintained at 37°C and 5% CO2 until harvested . For each lentiviral construct , transductions were performed in triplicates . Each experiment was repeated at least twice with cell preparations from independent mice . 6 days after lentiviral transduction , cells were collected using 5x Passive Lysis Buffer ( Promega ) and luciferase activity was measured using the Dual-Glo Luciferase Assay System ( Promega ) according to the manufacturer's protocol . Firefly luciferase signals were normalised to Renilla luciferase , and for each 3′ UTR construct this signal was then normalised to that of lentivector-control plasmid ( without cloned 3′ UTR , thus only containing generic vector 3′ UTR ) . Signals in control mice were set to 1 . Mice were anesthetised with isoflurane and sacrificed by decapitation . The liver was excised and immediately placed in cold Hank's balanced salt solution ( Invitrogen ) . Then , the liver was sliced into small pieces and cultured separately on Millicell culture membranes ( PICMORG50; Millipore , Billerica , MA ) with 1 . 1 ml of HEPES-buffered phenol red-free DMEM ( Gibco ) supplemented with 2% B27 ( Invitrogen ) , 1% PSG ( Gibco ) , 4 . 2 mM Na2HCO3 and 0 . 1 mM luciferin . Cultures were maintained at 37°C and 5% CO2 in a light-tight incubator , and bioluminescence was monitored continuously using LumiCycle 32 ( Actimetrics , Wilmette , IL ) . Proteins from mouse liver nuclei were prepared according to the NUN procedure ( Lavery and Schibler , 1993 ) . Each sample analysed by SDS-PAGE was a pool of protein extracts from 3 to 5 mice . SDS-PAGE and immunoblot analysis were performed according to standard protocols . Antibodies used were rabbit CRY1 , CRY2 , PER1 , PER2 , BMAL1 , and CLOCK ( kindly provided by S Brown and J Ripperger ) and U2AF65 ( Sigma ) . Western blots were quantified using ImageQuant TL 8 . 1 software ( Life Sciences ) . All blood chemistry was performed as described ( Le Martelot et al . , 2009 ) . DNase-treated total RNAs were subjected to rRNA depletion using Ribo-Zero Magnetic Kit ( Human/Mouse/Rat , Epicentre , Madison , WI ) following the supplier's instructions . The resulting rRNA-depleted RNA samples were then used to prepare random-primed cDNA libraries using the Illumina Tru-Seq RNA Sample Preparation Kit ( Illumina , San Diego , CA ) according to the manufacturer's recommendations . Multiplexed libraries were sequenced following the supplier's protocol ( 100 bp single-end reads ) on the Illumina HiSeq 2500 at the Lausanne Genomics Technologies Facility . Quality of the sequencing reads was initially assessed based on the quality values produced by the Illumina pipeline Casava 1 . 8 . 2 . Sequencing runs were only then allowed into our data analysis work flow , if their statistics related to quality of base calling and to preliminary alignment against mouse genome were within three standard deviations of the mean of all runs . The five quality related statistics used were the percentage of clusters passed filtering ( %PF clusters ) , mean quality score ( PF clusters ) , percentage of reads aligned , mean alignment score and percentage of alignment error , and had the following population means and standard deviations , respectively , 93 . 39 ± 1 . 85 , 36 . 34 ± 0 . 5 , 72 . 24 ± 3 . 0 , 290 . 6 ± 15 . 13 and 0 . 98 ± 0 . 15 . Presence of adapter sequences in the datasets was checked with cutadapt utility ( Martin , 2011 ) . Less than 0 . 65% of the reads were estimated to include an adapter sequence longer than 10 bps; and more than 95% of such reads were left unmapped after the alignment of the reads to the mouse genome . Therefore , it was not necessary to trim reads to remove the adapter sequences . Sequences were first mapped to local databases of mouse rRNA ( Ensembl and NCBI ) and rodent-specific repeat sequences ( RepBase Update version 18 . 10 [Jurka et al . , 2005] ) using bowtie2 version 2 . 1 . 0 ( Langmead and Salzberg , 2012 ) with default alignment parameters . Then reads were mapped to Genome Reference Consortium GRCm38 ( mm10 ) version of mouse reference genome sequence using tophat version 1 . 4 . 1 with known mm10 transcripts provided via the—transcriptome-index option ( Trapnell et al . , 2009 ) . For each read all mapping outcomes were considered and only non-rRNA , non-repeat reads that mapped uniquely to the mouse genome were selected for further analysis ( Figure 2—source data 1 ) . Expression levels for mRNA and pre-mRNA were estimated per locus rather then per isoform . To this end , gene annotations from Ensembl mouse database release 68 ( Flicek et al . , 2013 ) were flattened similarly as described earlier ( Anders et al . , 2012 ) with the following modifications . Prior to flattening , transcript isoform models whose expression was not evidenced by count data or splice junction coverage were removed from the annotation database . This was achieved first by estimating the expression level of each isoform contained in Ensembl database by cufflinks software version 2 . 0 . 2 ( Trapnell et al . , 2010 ) with following parameters: -G Mus_musculus . GRCm38 . 68 . gtf -u -j 0 . 8 -m 160 -s 50 -N . Isoform models which did not comprise at least 5% of the total expression of the locus or which were only expressed in fewer than three samples were excluded from the gene annotation database . Furthermore , spliced isoform models , whose splice junctions were not covered on average by 6 reads , or 40% or more of whose splice junctions were not covered at least by 6 reads were also removed from the annotation database . In addition to the removal of non-supported isoform models , novel exons overlapping with reference on the opposite strand were identified and added to the annotation database to improve the unambiguity of read counting . Briefly , cufflinks was run in reference annotation based transcript ( RABT ) mode ( Roberts et al . , 2011 ) and isoform models were assembled with cuffmerge software . Novel exon models with a class code ‘X’ , which overlap with a reference exon on the opposite strand , were then validated by splice junction coverage as outlined before and added to the annotation database . Finally , the gene models were flattened ( Anders et al . , 2012 ) . An in-house Python script was used to count the reads mapped within each annotation feature in a similar way as implemented in htseq-count utility software ( HTSeq: analysing high-throughput sequencing data with Python . [http://www-huber . embl . de/users/anders/HTSeq/] ) . Only reads which can be unambiguously identified as either exonic ( continuous or spliced ) or intronic for a single locus were counted towards the mRNA or pre-mRNA counts of that locus , respectively . Mappable and countable mRNA and pre-mRNA lengths ( in bps ) for each locus were calculated by means of generating all possible 100-bp long reads in silico ( faux reads ) for each transcript type and counting the faux reads through identical mapping and counting work flow used for real experimental reads . Preliminary inspection of the extent of differential expression and the presence of highly expressed genes was carried out by cumulative percentage plots of raw counts; and accordingly read counts of mRNA and pre-mRNA datasets were normalised with upper quantile ( Bullard et al . , 2010; Dillies et al . , 2013 ) and TMM ( Robinson and Oshlack , 2010 ) normalisation methods , respectively . Prior to normalisation , transcripts which did not have at least 10 counts in at least one third of the samples were removed from the datasets . Differential expression ( DE ) analysis was performed using the DESeq analysis tool version 1 . 12 . 1 ( Anders and Huber , 2010 ) . Briefly , normalisation factors calculated earlier were used to create DESeq count datasets for mRNA and pre-mRNA counts annotated with full experimental design ( treatment × time ) and dispersion values were calculated with default settings . Calculated p-values for DE of all transcripts between control and Dicer KO conditions were then adjusted for false discovery rate ( FDR ) by Benjamini and Hochberg ( BH ) method ( Benjamini and Hochberg , 1995 ) . A transcript was considered differentially expressed if it had an adjusted p-value smaller than 0 . 05 and had a fold-change greater than 1 . 5 . For further filtering of lowly expressed transcripts and for better comparability between datasets , RPKM values were calculated as the number of counted reads per 1000 mappable and countable bases per geometric mean of normalised read counts per million . The geometric mean of normalised read counts was 53 , 692 , 658 and 16 , 285 , 588 for mRNA and pre-mRNA datasets , respectively . Transcripts which did not have a RPKM value greater than 0 . 1 or 0 . 01 in at least one third of the samples were considered as ‘not expressed’ in the mRNA or the pre-mRNA dataset , respectively . As an estimate of the mRNA stability , we have calculated the log2 mRNA to pre-mRNA ratios of RPKM values separately for each sample . The distribution of log-ratios was inspected to assess normality . To test if a transcript's mean stability over all timepoints was increased in the absence of miRNAs , we applied a one-sided t test between control and KO conditions ( n = 12 ) . The p-values were then adjusted for FDR using BH method . We used the combined power of a parametric test based on harmonic regression similar to those implemented in cosiner based algorithms ( Cugini , 1993; Yang and Su , 2010 ) and a non-parametric test , JTK_CYCLE ( Hughes et al . , 2010 ) . To adapt the harmonic regression to sequencing count data , several modifications were introduced . First , a negative binomial generalised linear model ( GLM ) , with a log link function was used to regress the rounded normalised counts on a cycle with a 24-hr period . Count data were not detrended or smoothened before the regression . Dispersion values that were already fitted for each gene during DE analysis were used as an estimate of the shape parameter theta . From the fitted parameter values , then , amplitude and phase estimates were calculated for each gene . The goodness-of-fit for the harmonic model ( sinusoidal expression ) was tested with a likelihood ratio test against the updated null model including only the intercept parameter ( constant expression ) . The analysis was performed using the MASS package ( Rindskopf , 1997 ) in statistical computing environment R . For both tests , the two biological replicates per timepoint were treated as independent replicates and were not combined into a false 48-hr data-series . The p-values calculated by each test were then adjusted separately for FDR by BH method ( Benjamini and Hochberg , 1995 ) . For all further calculations , amplitude and phase estimates obtained from the harmonic fit were used . We categorised mRNAs and pre-mRNAs as rhythmic when they passed at least one of the statistical tests and had an amplitude-ratio ( peak/trough ) ≥1 . 5 . Combining two complementary tests in this way provided for relatively low stringency in rhythm detection , which we deemed important in particular for the pre-mRNA data , which is naturally noisier due to the low abundance of short-lived pre-mRNA molecules . To assess the significance of the effects of the treatment on amplitude and phase estimates , we performed two permutation tests ( N > 10 , 000 ) . For both tests , we applied the permutation on the normalised read counts per gene , always maintaining the mRNA and pre-mRNA attributes . Each permuted data set was then analysed by exactly the same steps used on the experimental data . Finally , the distribution of the difference in means , significance of the effects , and correlation coefficients under the null hypothesis were inspected for both amplitude and phase effects . To enable the comparison of the significance of the effects between samples of different sizes , the Wilcoxon test statistics were converted into a standardised z-score . In the first test , we permuted the timepoints while maintaining the treatment effects ( Dicer KO vs control ) to test the hypothesis that the observed phase delay differences in the stabilised circadian transcripts was not merely an outcome of altered expression levels ( in particular lower amplitudes ) in the Dicer KO ( in which case it would be observable rather frequently on randomly generated cyclic transcripts ) . In the second test , the genotype ( Dicer KO/control ) was permuted to test the hypothesis that the observed effects on both amplitudes and phase differences were not random . Statistical testing of enrichment of Gene Ontology terms within lists of genes was performed with MetaCore version 6 . 17 software suite ( Thomson Reuters , New York City , NY ) . For enrichment analysis of miRNA targets within lists of genes , the ‘Conserved Family Info’ table from the TargetScanMouse database version 6 . 2 ( Lewis et al . , 2005 ) was used to create an association database between miRNA families and the genes from the current study . Using an in-house Python script , raw counts of targets and non-targets were extracted from this database for the genes included in list . Background list included all 11 , 841 genes from Figure 2F . The statistical significance of enrichment of targets for each miRNA family in the test list against the background list was tested via Fisher's exact test on the contingency tables created from the raw counts . Calculated p-values were adjusted for FDR using BH method ( Benjamini and Hochberg , 1995 ) . Analysis was carried out in R environment ( R Core Team , 2013 ) . The RNA-seq data set produced in this study has been deposited in the Gene Expression Omnibus ( accession number GSE57313 ) . Supplementary files 1 and 2 are available at http://www . unil . ch/cig/page102471 . html and have been deposited in the Dryad Digital Repository ( Du et al . , 2014 ) . | The rising and setting of the sun have long driven the schedules of humans and other mammals . This 24-hr cycle influences many behavioural and physiological changes , including alertness , body temperature , and sleep . A region in the brain acts as a master clock that regulates these daily cycles , which are called circadian rhythms . Signals from the brain's master clock turn on and off ‘core clock genes’ in cells , which trigger cycles that cause some proteins to be produced in a circadian rhythm . The rhythm is specialized to a particular tissue or organ , and may help them to carry out their designated daily tasks . However , circadian rhythms might also be produced in other ways that do not involve these genes . Messenger RNA ( mRNA ) molecules have a central role in the production of proteins , and in the mouse liver , up to 15% of mRNA molecules are produced in circadian cycles . The liver performs essential tasks that control metabolism–including that of carbohydrates , fats , and cholesterol . Precisely timing when certain mRNAs and proteins reach peaks and troughs in their activities to coincide with mealtimes is important for nutrients to be properly processed . Other RNA molecules called microRNAs influence how mRNA molecules are translated into proteins . Now Du , Arpat et al . have looked at the influence of microRNAs on circadian rhythms in the mouse liver in greater detail . These experiments , which involved ‘knocking out’ a gene that is essential for the production of microRNAs , show that rather than setting the mRNA rhythms , the microRNAs appear to adjust them to meet the specific needs of the liver . Targeting specific microRNA molecules may reveal new strategies to tweak these rhythms , which could help to improve conditions when metabolic functions go wrong . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2014 | MicroRNAs shape circadian hepatic gene expression on a transcriptome-wide scale |
Spinal and bulbar muscular atrophy ( SBMA ) is a progressive neuromuscular disease caused by polyglutamine expansion in the androgen receptor ( AR ) protein . Despite extensive research , the exact pathogenic mechanisms underlying SBMA remain elusive . In this study , we present evidence that Nemo-like kinase ( NLK ) promotes disease pathogenesis across multiple SBMA model systems . Most remarkably , loss of one copy of Nlk rescues SBMA phenotypes in mice , including extending lifespan . We also investigated the molecular mechanisms by which NLK exerts its effects in SBMA . Specifically , we have found that NLK can phosphorylate the mutant polyglutamine-expanded AR , enhance its aggregation , and promote AR-dependent gene transcription by regulating AR-cofactor interactions . Furthermore , NLK modulates the toxicity of a mutant AR fragment via a mechanism that is independent of AR-mediated gene transcription . Our findings uncover a crucial role for NLK in controlling SBMA toxicity and reveal a novel avenue for therapy development in SBMA .
Spinal and bulbar muscular atrophy ( SBMA; MIM #313200 ) is an X-linked progressive neuromuscular disease ( Kennedy et al . , 1968 ) . Patients present in midlife with weakness of the limb and facial muscles , the latter of which often progress to dysarthria and dysphagia , occasionally leading to fatality . SBMA patients also commonly suffer from mild androgen insensitivity , presenting with gynecomastia , testicular atrophy , and decreased fertility ( Katsuno et al . , 2012 ) . SBMA was originally defined as a neurodegenerative disease affecting the proximal spinal and bulbar motoneurons , and muscle atrophy was considered secondary to motoneuron degeneration . Current opinion in the field of SBMA research , however , now favors a model in which SBMA also directly affects the skeletal muscles ( Yu et al . , 2006; Jordan and Lieberman , 2008; Monks et al . , 2008; Boyer et al . , 2013; Malena et al . , 2013; Oki et al . , 2015 ) , and , in fact , recent studies have shown that removing or decreasing the expression of the mutant protein within skeletal muscle is sufficient to rescue SBMA phenotypes in vivo ( Cortes et al . , 2014; Lieberman et al . , 2014 ) . This model of disease is supported by the finding that , in conjunction with neuronal loss , patients also show elevated creatine kinase levels and evidence of myopathic changes on biopsy ( Sorarù et al . , 2008; Chahin and Sorenson , 2009 ) . SBMA is caused by the expansion of a polymorphic CAG trinucleotide repeat located in the first exon of the Androgen Receptor ( AR ) gene ( La Spada et al . , 1991 ) . In wild-type AR , this repeat region encodes a stretch of 6–36 glutamines ( Q ) . In SBMA patients , in contrast , the region is expanded to 37 to 70Q , resulting in pathogenesis via a gain-of-function and partial loss-of-function mechanism ( Katsuno et al . , 2012 ) . SBMA is therefore one of nine identified polyglutamine ( polyQ ) repeat diseases , along with Huntington's disease , dentatorubral-pallidoluysian atrophy , and spinocerebellar ataxia ( SCA ) types 1 , 2 , 3 , 6 , 7 , and 17 . PolyQ expansion renders the host protein toxic , resulting in the formation of mutant protein aggregates and cell death; and the commonalities in the nature of the mutation and the presentation of the different polyQ disorders suggest the presence of a common pathogenic mechanism ( Orr , 2001 ) . Nonetheless , this mechanism has remained elusive and to date there are no cures or even effective therapies for most of these diseases . AR is a well-studied steroid hormone receptor that also plays a crucial role in additional diseases including androgen insensitivity syndrome and prostate cancer ( Bennett et al . , 2010 ) . Studies focusing on wild-type AR function and its role in other disease contexts can therefore shed light on SBMA pathogenesis . For instance , the main function of AR is to bind androgenic hormones , either testosterone or 5α-dihydrotestosterone ( DHT ) , in the cytoplasm , and then translocate into the nucleus to act as a DNA-binding transcription factor that regulates androgen-dependent target gene expression ( Bennett et al . , 2010 ) . SBMA pathogenesis is dependent upon the presence of circulating androgens and is therefore only observed in males , with homozygous female carriers showing only mild symptoms ( Katsuno et al . , 2012 ) . The importance of androgens to the disease has also been clearly shown in mouse models of SBMA ( Katsuno et al . , 2002; Chevalier-Larsen et al . , 2004 ) . Furthermore , the nuclear translocation of AR is also crucial for pathogenesis ( Takeyama et al . , 2002; Montie et al . , 2009; Nedelsky et al . , 2010 ) . It has also been suggested that an AR interdomain interaction known as the amino ( N ) -terminal–carboxy ( C ) -terminal ( N/C ) interaction is important for SBMA ( Orr et al . , 2010 ) , as are the DNA-binding ability of AR ( Nedelsky et al . , 2010 ) and its post-translational modification including acetylation ( Montie et al . , 2011 ) , methylation ( Scaramuzzino et al . , 2015 ) , and other modifications ( Katsuno et al . , 2012 ) . In addition , several cofactors and regulators of AR can influence SBMA disease pathogenesis ( McCampbell et al . , 2000; Taylor et al . , 2003; Palazzolo et al . , 2007; Suzuki et al . , 2009; Nedelsky et al . , 2010; Montie et al . , 2011 ) . Despite extensive studies , however , a precise molecular explanation for SBMA pathology has remained elusive . Given the importance of androgens to SBMA pathogenesis , many approaches to SBMA therapeutics have focused upon depleting androgen levels in patients ( Banno et al . , 2009; Katsuno et al . , 2010b; Fernández-Rhodes et al . , 2011; Yamamoto et al . , 2013 ) . Unfortunately , these strategies have not yielded significant results in clinical trials; hence , new approaches are necessary . It has been shown that within prostate cancer cells , wild-type AR physically interacts with Nemo-like kinase ( NLK ) and that NLK is able to regulate the activity and transcription of AR in this context ( Emami et al . , 2009 ) . Interestingly , studies show that NLK interacts either directly or indirectly with a number of neurodegenerative disease-related proteins ( Lim et al . , 2006; Ju et al . , 2013 ) , suggesting that it may play an important role in the pathogenesis of neurodegenerative proteinopathies . Indeed , we have found that loss of one copy of Nlk ( resulting in a 50% reduction in protein expression ) is beneficial in mouse models of the polyQ disease SCA1 ( Ju et al . , 2013 ) . NLK is an evolutionarily conserved mitogen-activated protein kinase -like serine/threonine kinase primarily studied in lower model organisms , where it has been linked to a number of signaling pathways ( Ishitani et al . , 1999; Ohkawara et al . , 2004; Ishitani et al . , 2010; Ishitani and Ishitani , 2013 ) . In this study , we tested the hypothesis that NLK may play a role in SBMA pathogenesis . We present evidence that NLK influences the aggregation and toxicity of polyQ-expanded AR across multiple model systems , using cell culture , Drosophila , and mouse . Loss of one copy of Nlk was able to partially rescue disease phenotypes in both Drosophila and mouse models of SBMA . Furthermore , this 50% reduction in NLK protein expression dramatically extended the lifespan of SBMA mice . Finally , we investigated the molecular mechanisms by which NLK mediates these effects on SBMA and suggest a model in which NLK interacts with and phosphorylates AR , inhibiting its intramolecular N/C interaction and thereby promoting gene transcription via the AR activation function 2 ( AF-2 ) domain . This effect on AR activity could then modulate SBMA-related aberrant AR-dependent gene transcription . In addition , reduced NLK expression can rescue the toxic effects of an N-terminal fragment of AR , suggesting that NLK can regulate the mutant AR protein—even in the absence of DNA binding and AR-responsive gene transcription .
It was previously reported that NLK could interact with the wild-type AR in prostate cancer cell lines ( Emami et al . , 2009 ) . However , since SBMA is caused by polyQ-expanded AR ( La Spada et al . , 1991 ) , and polyQ expansion can alter the ability of AR to interact with its binding partners ( Hsiao et al . , 1999; Irvine et al . , 2000; Sopher et al . , 2004 ) , we tested if NLK could bind mutant AR . We co-transfected a FLAG-tagged wild-type NLK construct ( FLAG-NLK-WT ) with either wild-type or mutant HA-tagged human AR ( HA-AR25Q and HA-AR120Q , respectively ) into NSC-34 motor neuron-derived cells ( Cashman et al . , 1992 ) and performed co-immunoprecipitation ( co-IP ) assays . We found that NLK was able to co-IP both wild-type and mutant AR ( Figure 1A ) . Interestingly , polyQ expansion led AR to be co-immunoprecipitated to a greater extent given its lower expression level ( Figure 1B ) . Although future in vitro and in vivo experiments would be needed to verify this result , it was consistent in our hands . In addition , NLK was able to co-IP an N-terminal fragment of AR spanning the first 130 amino acids and containing the polyQ repeat , suggesting that NLK binds within this region ( Figure 1C ) . It is worth mentioning that this fragment expresses as a doublet , and NLK seems to interact with only one of the forms of this fragment . We suspect that the upper band represents a post-translational modification of the fragment but further experiments would be required to confirm and expand this hypothesis . 10 . 7554/eLife . 08493 . 003Figure 1 . Nemo-like kinase ( NLK ) interacts with the mutant AR and enhances its aggregation . ( A ) NLK interacts with the AR protein in NSC-34 cells treated with 10 nM DHT . IP: immunoprecipitation . IB: immunoblot . GAPDH was used as a loading control in this and all following analyses unless otherwise specified . Asterisk marks a band corresponding to the immunoglobin heavy chain . ( B ) Quantification of co-IPed AR over total AR in input . *p < 0 . 05 ( t-test ) . n = 3 trials . Error bars are standard error of the mean in this and all following graphs unless otherwise specified . ( C ) NLK interacts with the N-terminal region of AR . Both full-length ( FL ) and an N-terminal fragment ( N , arrow ) of AR were pulled down with NLK . Asterisk marks a non-specific band . ( D–G ) NLK enhances the formation of mutant AR aggregates in a kinase activity-dependent manner . NSC-34 cells were treated with DHT as indicated and subjected to immunofluorescence using anti-AR N-20 ( green ) and anti-FLAG ( red ) antibodies to detect AR aggregation and NLK co-expression , respectively . NLK-WT: wild-type NLK . NLK-KN: kinase-dead NLK . Representative images of DHT-treated cells are shown in ( D–F ) . Images of the non-DHT-treated and AR25Q-expressing cells can be found in the Figure 1—figure supplements 1–3 . Scale bar in ( D ) is 20 μm and refers to all three images . Cells were scored as containing aggregates ( orange arrows ) or not ( white arrows ) and the ratio of aggregate-positive cells out of total scored is quantified in ( G ) . n . s . = not significant , ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n ≥ 3 trials . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 00310 . 7554/eLife . 08493 . 004Figure 1—figure supplement 1 . PolyQ-expanded AR120Q does not aggregate in the absence of DHT . Representative z-stack images of NSC-34 cells that were transfected as indicated , treated with ethanol ( as a negative control for DHT ) , and subjected to immunofluorescence using anti-AR N20 ( green; A–C ) and anti-FLAG ( red; A′–C′ ) antibodies to detect AR and NLK expression , respectively . Merged images are shown in ( A′′–C′′ ) . In the absence of hormone , AR120Q shows diffuse cytoplasmic localization , while NLK localizes to both the cytoplasm and the nucleus . Scale bars in merged images are 25 μm . There is a variation in overall cell size with NSC-34 cells that is not obviously influenced by the transfection of AR or NLK . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 00410 . 7554/eLife . 08493 . 005Figure 1—figure supplement 2 . Non-pathogenic AR25Q shows diffuse cytoplasmic localization in the absence of DHT . Representative z-stack images of NSC-34 cells that were transfected as indicated , treated with ethanol ( as a negative control for DHT ) , and subjected to immunofluorescence using anti-AR N20 ( green; A–C ) and anti-FLAG ( red; A′–C′ ) antibodies to detect AR and NLK expression , respectively . Merged images are shown in ( A′′–C′′ ) . In the absence of hormone , AR25Q shows diffuse cytoplasmic localization , while NLK localizes to both the cytoplasm and the nucleus . Asterisks in ( A′ ) mark non-specific staining in the red channel . Scale bars in merged images are 25 μm . There is a variation in overall cell size with NSC-34 cells that is not obviously influenced by the transfection of AR or NLK . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 00510 . 7554/eLife . 08493 . 006Figure 1—figure supplement 3 . Non-pathogenic AR25Q undergoes nuclear translocation in response to DHT , but largely does not aggregate . Representative z-stack images of NSC-34 cells that were transfected as indicated , treated with 10 nM DHT , and subjected to immunofluorescence using anti-AR N20 ( green; A–C ) and anti-FLAG ( red; A′–C′ ) antibodies to detect AR and NLK expression , respectively . Merged images are shown in ( A′′–C′′ ) . In the presence of hormone , AR25Q shows nuclear localization , while NLK localizes to both the cytoplasm and the nucleus . Scale bars in merged images are 25 μm . There is a variation in overall cell size with NSC-34 cells that is not obviously influenced by the transfection of AR or NLK . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 00610 . 7554/eLife . 08493 . 007Figure 1—figure supplement 4 . Mutant AR forms high molecular weight aggregates in the stacking gel of SDS–PAGE gels . NSC-34 cells were transfected as indicated and treated with 10 nM DHT . High molecular weight AR aggregates can be detected as a smear in the stacking gel . This aggregation is only seen upon mutant AR expression and is increased with co-expression of NLK . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 007 We next wondered whether NLK could modulate SBMA disease phenotypes . PolyQ expansion results in the aggregation of the host protein , and inclusion formation is a pathological hallmark of polyQ and other neurodegenerative diseases ( Orr , 2001; Todd and Lim , 2013 ) . We therefore asked if NLK could influence the ability of the polyQ-expanded AR to aggregate . Mutant AR forms large polyQ- and DHT-dependent aggregates that can be readily visualized in our cell model via immunofluorescence ( Figure 1D and Figure 1—figure supplements 1–3 ) . Co-expression of wild-type NLK ( NLK-WT ) significantly increased the number of cells containing visible aggregates in DHT-treated , mutant AR ( AR120Q ) -expressing NSC-34 cells ( Figure 1E , G ) , but did not cause a significant increase in aggregation in the absence of the AR ligand ( Figure 1G and Figure 1—figure supplement 1 ) . This increase was polyQ-dependent , as NLK co-expression resulted in only minimal aggregation in cells expressing a non-pathogenic AR25Q protein ( Figure 1G and Figure 1—figure supplements 2 , 3 ) . Furthermore , this increase in aggregation was not detected when we used NLK-KN ( Figure 1F , G ) , which harbors a lysine to methionine substitution at residue 155 and is defective for kinase activity ( Ishitani et al . , 1999 ) . Importantly , co-expression of NLK does not alter the subcellular localization of non-aggregated AR or inhibit its nuclear translocation , although cells with robust aggregation often showed a slight reduction in nuclear staining , suggesting that much of the mutant protein was sequestered into aggregates in these cells . In addition , we did not recognize any obvious changes in subcellular localization between NLK-WT and NLK-KN , which could both be detected in the cytoplasm and nucleus . We also noticed that aggregated mutant AR protein could be detected biochemically in the stacking gel when we ran DHT-treated NSC-34 cell extracts on SDS-PAGE gels . Co-expression of NLK-WT increases this aggregation ( Figure 1—figure supplement 4 ) . Taken together , these data suggest that NLK is able to affect polyQ-AR-specific defects within this cell culture system in a kinase activity-dependent manner . To test whether NLK can also increase mutant AR aggregation in an in vivo motor neuron setting , we cultured mouse primary motor neurons from spinal cord and transfected with GFP-tagged polyQ-expanded mutant AR and either a control plasmid or FLAG-tagged NLK-WT in the presence or absence of DHT . Cells were then blindly scored for the presence or absence of aggregation . We found that NLK is able to robustly increase mutant AR aggregation in DHT-treated neurons , while it only modestly increased aggregation in the absence of hormone ( Figure 2 ) . 10 . 7554/eLife . 08493 . 008Figure 2 . NLK increases mutant AR aggregation in primary motor neurons . ( A–H ) Primary motor neurons were transfected with GFP-tagged AR120Q , FLAG-tagged NLK-WT , or a pcDNA3 . 1 empty vector control and treated with 10 μM DHT . Aggregation was analyzed by immunofluorescence at 9 days in vitro ( DIV ) . An antibody to choline acetyltransferase ( ChAT ) was used to confirm motor neuron identity and is shown in red . GFP-AR120Q is shown in green and NLK co-expression ( as detected by an NLK antibody ) is in blue . All images were collected using identical confocal settings . In the absence of DHT , AR localizes to the cytoplasm ( E , G ) , while DHT induces its nuclear translocation ( F ) and its aggregation , which is enhanced by NLK ( H ) . Arrows mark aggregates , which can be detected in both the nucleus and cytoplasm . Scale bars are 10 μm . ( I ) The number of neurons containing aggregates out of total scored was quantified and averaged over different regions of the plate . At least 140 neurons were scored per condition . ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 008 Having established that NLK can modulate the aggregation of the mutant AR in our cell culture system , we went on to determine the effect of modulating NLK activity and expression on SBMA in model organisms . We began by utilizing Drosophila . When a full-length AR transgene is expressed in the Drosophila eye via the Gal4/UAS system ( Brand and Perrimon , 1993 ) , it produces a polyQ- , DHT-dependent retinal degeneration phenotype characterized by the presence of fused ommatidia and abnormal interommatidial bristles along the posterior margin of the eye ( Figure 3A , B and Figure 3—figure supplement 1 ) . This phenotype is similar to what has been reported for other full-length mutant AR Drosophila models of SBMA ( Takeyama et al . , 2002; Pandey et al . , 2007; Nedelsky et al . , 2010 ) . We crossed SBMA flies to flies that were heterozygous for a loss-of-function mutation in the fly homolog of Nlk , nemo ( nmo ) . To ensure that this was not due to a non-specific background effect , we utilized two independent nmo loss-of-function alleles ( adk1 and adk2 ) ( Verheyen et al . , 2001 ) . Both alleles were able to partially , but consistently , suppress the mutant AR-mediated rough eye phenotypes ( Figure 3C , D ) , although the nmoadk2 line showed a more profound rescue than nmoadk1 . Next , we assessed whether this phenotype correlated with a change in mutant AR aggregation . To do this , we compared protein extracts from Drosophila heads of each genotype by immunoblot . Aggregation of the mutant AR protein can be detected as a smear in the stacking gel and was increased in flies raised in the presence of DHT ( Figure 3E , F ) . Loss of one copy of nmo tended to reduce this aggregation , particularly when assessed with the nmoadk2 allele ( Figure 3E , F ) . Although this reduction in aggregation failed to reach significance by ANOVA , the difference seen between the two alleles correlates with the more profound partial rescue of the mutant AR-dependent retinal degeneration seen with nmoadk2 compared to nmoadk1 in this fly model of SBMA . 10 . 7554/eLife . 08493 . 009Figure 3 . NLK genetically interacts with the mutant AR in Drosophila . Loss of one nmo allele suppresses mutant AR-mediated SBMA phenotypes in Drosophila . ( A–D ) Light microscopy of adult Drosophila eyes is shown . In ( B ) , arrows mark a DHT-dependent retinal degeneration phenotype along the posterior margin . Flies were raised at 30°C and genotypes are as follows: ( A ) GMR-Gal4/+; UAS-EGFP/+ , ( B ) GMR-Gal4 , UAS-AR61Q/+ , ( C ) GMR-Gal4 , UAS-AR61Q/+; nmoadk1/+ , ( D ) GMR-Gal4 , UAS-AR61Q/+; nmoadk2/+ . For all panels , experiments were repeated multiple times and representative images are shown . ( E ) Western blots from three different trials show the aggregation of the mutant AR as a smear in the stacking gel at high exposure . Lower exposure reveals the AR61Q monomer at the expected size of around 110 kDa . Asterisk marks a non-specific band present in all lanes . ( F ) High molecular weight ( HMW ) or aggregated AR was quantified as compared to the tubulin loading control and averaged over trials . *p < 0 . 05 ( ANOVA with Tukey's post-hoc analysis ) . n ≥ 3 trials . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 00910 . 7554/eLife . 08493 . 010Figure 3—figure supplement 1 . Expression of a full-length AR protein in the Drosophila eye results in polyQ- and DHT-dependent retinal degeneration phenotypes . ( A , B ) Expression of a full-length AR transgene with 14Q ( wild-type , AR14Q ) in the fly eye does not produce a recognizable phenotype on the exterior fly eye , regardless of DHT treatment . ( C , D ) Expression of a full-length AR transgene with 61Q ( mutant , AR61Q ) results in a ‘rough’ eye phenotype along the posterior margin of the eye ( bracket ) in the presence of DHT . Flies were raised at 25°C . Phenotypes were consistent over multiple trials . Genotypes are as follows: ( A ) GMR-Gal4/UAS-AR14Q [without DHT] ( B ) GMR-Gal4/UAS-AR14Q [with DHT] , ( C ) GMR-Gal4/UAS-AR61Q [without DHT] , ( D ) GMR-Gal4/UAS-AR61Q [with DHT] . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 010 We next tested whether increased expression of NLK could enhance the mutant AR phenotypes in this Drosophila SBMA model . To do this , we crossed SBMA flies with flies expressing either the human NLK or an EGFP control ( Figure 4 ) . Co-expression of NLK-WT enhanced the retinal degeneration phenotype ( Figure 4B ) and , more strikingly , dramatically increased the mutant AR aggregation detected by immunoblot ( Figure 4D , lane 4 vs lane 6 ) . Once again , this phenotype was DHT dependent ( Figure 4D , lane 5 vs lane 6 ) . Importantly , we also found that expression of kinase-dead NLK-KN did not enhance the retinal degeneration phenotype ( Figure 4C ) or mutant AR aggregation ( Figure 4D , lane 4 vs lane 7 ) , a finding consistent with our cell culture data ( Figure 1 ) . Taken together , these studies strongly suggest that NLK exacerbates the toxicity of the polyQ-expanded mutant AR via a mechanism that depends upon its kinase activity . 10 . 7554/eLife . 08493 . 011Figure 4 . NLK modulates mutant AR phenotypes in Drosophila in a kinase activity-dependent manner . ( A–C ) Light microscopy of adult Drosophila eyes is shown . Flies were raised at 30°C and genotypes are as follows: ( A ) GMR-Gal4 , UAS-AR61Q/UAS-EGFP , ( B ) GMR-Gal4 , UAS-AR61Q/UAS-NLK-WT , ( C ) GMR-Gal4 , UAS-AR61Q/UAS-NLK-KN . ( D ) Mutant protein aggregation is shown by immunoblot with indicated genotypes . Aggregated mutant AR protein can be detected as a smear in the stacking gel at higher exposures , while the AR61Q monomer expresses at around 110 kDa and can be seen at lower exposures . Asterisk marks a non-specific band present in all lanes . For all panels , experiments were repeated multiple times , and representative images are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 011 Our cell culture and Drosophila data strongly suggest that reducing NLK expression or activity will be beneficial in SBMA , but we wished to confirm this at the mammalian level . We therefore decided to make use of our previously produced Nlk mutant mice ( Ju et al . , 2013 ) . Mice heterozygous for either of two gene trap alleles ( both simply referred to as Nlkgt/+ here ) show a 50% reduction in NLK expression , while mice homozygous for the gene trap alleles show an approximately 90% reduction in protein expression ( Ju et al . , 2013 ) . Importantly , this decrease can be detected in both the spinal cord and skeletal muscle ( Figure 5 ) , the two tissues primarily affected in SBMA . We also obtained mice that express a BAC transgene containing a 121Q AR and its endogenous regulatory elements ( BAC fxAR121 ) . These mice recapitulate key SBMA disease phenotypes , including motor neuron pathology , muscle atrophy , and early lethality . These phenotypes are only seen in male mice , as is consistent with the hormone specificity of this disease ( Cortes et al . , 2014 ) . As homozygous expression of the Nlk gene trap alleles is lethal , we carried out our analysis in the heterozygous background . Nlkgt heterozygous mice were crossed to BAC fxAR121 mice and their F1 male progeny were analyzed to determine if loss of one copy of Nlk could rescue the SBMA-related phenotypes seen in the BAC fxAR121 mice . For our analysis , we began by looking at motor neuron pathology . BAC fxAR121 mice , like other SBMA mouse models ( Chevalier-Larsen et al . , 2004; Yu et al . , 2006 ) , fail to show overt motor neuronal loss . There is , however , a pathogenic decrease in the area and perimeter of the spinal motor neuron soma in this model ( Cortes et al . , 2014 ) . We analyzed L4–L5 anterior horn motor neurons and found that a reduction in NLK expression resulted in significantly larger motor neuron cell bodies than those seen in SBMA littermates ( Figure 6 ) , suggesting an improvement in pathology . We next focused on muscle pathology , since muscle cramping and atrophy are prominent symptoms in SBMA patients ( Rhodes et al . , 2009; Katsuno et al . , 2012 ) , and this SBMA mouse model shows an obvious muscle atrophy phenotype ( Cortes et al . , 2014; Lieberman et al . , 2014 ) . Compared to wild-type and Nlkgt/+ mice , BAC fxAR121+/− mice showed a reduction in the Feret's diameter and cross-sectional area of muscle fibers , as well as more angulated fibers and increased connective tissue , all of which is suggestive of atrophy ( Figure 7A–E and Figure 7—figure supplement 1 ) . Although muscle atrophy phenotypes were still apparent in BAC fxAR121+/−; Nlkgt/+ mice , the average fiber size was significantly increased compared to their BAC fxAR121+/− littermates ( Figure 7C–E ) . This increase was apparent at 20 weeks ( mid-late symptomatic stage ) and 30 weeks ( late symptomatic stage ) of age , but was not seen at disease onset at 10 weeks of age and was no longer significant at very late disease stages at 40 weeks of age ( Figure 7E and Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 08493 . 012Figure 5 . Nlkgt mice show reduced NLK expression in the spinal cord and skeletal muscle . Whole spinal cord ( A ) and quadriceps ( B ) extracts from indicated genotypes were immunoblotted with a NLK antibody . Mice heterozygous for Nlkgt show a 50% reduction in protein expression , while mice homozygous for the allele show an approximately 90% reduction . GAPDH was used as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 01210 . 7554/eLife . 08493 . 013Figure 6 . Loss of one copy of Nlk improves the pathogenic change in motor neuronal soma size in SBMA mice . ( A–D ) Spinal cord cross-sections from the L4–L5 region were stained with cresyl violet ( nissl stain ) to visualize the spinal motor neuron cell bodies . Representative images from the anterior horn region of 40-week-old mice are shown . Scale bars are 50 μm . ( E , F ) The average motor neuron area ( E ) and perimeter ( F ) were measured and averaged over genotype . n = 2 , 4 , 4 , 3 per genotype , respectively . Over 100 neurons were scored per animal . *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 ( ANOVA with Tukey's post-hoc analysis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 01310 . 7554/eLife . 08493 . 014Figure 7 . Loss of one copy of Nlk significantly rescues SBMA phenotypes in mice . ( A–D ) Mouse quadriceps sections of indicated genotypes were stained for hematoxylin and eosin and representative 30-week-old images are shown . Scale bars are 50 μm . ( E ) Quantification of the average minimum Feret's diameter of muscle fibers at ages indicated . *p < 0 . 05 , **p < 0 . 005 ( t-test ) . For 10 weeks , n = 3 , 3 , 4 , and 3 per genotype , respectively . For 20 weeks , n = 4 , 3 , 5 , and 5 . For 30 weeks , n = 7 , 5 , 5 , and 8 . For 40 weeks , n = 2 , 5 , 4 , and 3 . More than 500 fibers were scored per animal . See also Figure 7—figure supplement 1 . ( F–I ) Reduced NLK expression improves the defective NADH transferase activity pattern seen in BAC fxAR121+/− mouse muscle . Six littermate sets were compared and representative images at 30 weeks of age are shown . Scale bars are 200 μm . See also Figure 7—figure supplement 2 . ( J ) Kaplan–Meier survival analysis shows a significant extension in the lifespan of BAC fxAR121+/− mice with a 50% reduction of NLK . p = 0 . 00107 ( log rank test ) . n = 27 , 27 , 51 , and 37 per genotype , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 01410 . 7554/eLife . 08493 . 015Figure 7—figure supplement 1 . Loss of one copy of Nlk increases muscle fiber size in BAC fxAR121+/− mouse quadriceps . Mouse quadriceps were stained for hematoxylin and eosin and the average cross-sectional area ( A ) and Feret's diameter ( B ) of muscle fibers were quantified at ages indicated . The BAC fxAR121+/−; Nlkgt/+ mice ( green ) have slightly larger fibers than their BAC fxAR121+/− littermate controls ( blue ) . **p < 0 . 005 ( t-test ) . For 10 weeks , n = 2 , 3 , 3 , and 2 per genotype , respectively . For 20 weeks , n = 4 , 3 , 5 , and 5 . For 30 weeks , n = 7 , 5 , 6 , and 8 . For 40 weeks , n = 2 , 5 , 4 , and 3 . More than 500 fibers were scored per animal . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 01510 . 7554/eLife . 08493 . 016Figure 7—figure supplement 2 . A 50% reduction in NLK expression reduces aberrant NADH transferase staining in 30-week-old SBMA mice . Reduced NLK expression partially rescues the defective NADH transferase activity pattern seen in BAC fxAR121+/− mouse muscle . Representative images at 30 weeks are shown in Figure 7F–I . The average mean gray value of all images is quantified here with higher values corresponding to lower intensity staining . For 10 weeks , n = 3 , 4 , 4 , and 3 per genotype , respectively . For 20 weeks , n = 4 , 3 , 5 , and 5 . For 30 weeks , n = 7 , 5 , 6 , and 8 . *p < 0 . 05 ( t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 016 We also stained muscles for NADH transferase activity ( Figure 7F–I ) , as defects in the patterning of this stain are seen in SBMA mouse models and are indicative of pathology ( Sopher et al . , 2004; Monks et al . , 2007; Palazzolo et al . , 2009 ) . As previously reported ( Cortes et al . , 2014 ) , there was a general increase in staining in the muscle of BAC fxAR121+/− mice ( Figure 7H ) , as opposed to the normal ‘checkerboard’ pattern seen in wild-type and Nlkgt/+ mouse muscle ( Figure 7F , G ) . Consistent with the increase in fiber size , BAC fxAR121+/−; Nlkgt/+ mice also showed a partial but consistent rescue in this phenotype at 30 weeks of age compared to their littermate controls ( Figure 7I ) . We quantified this change in staining intensity by measuring the mean gray value of the images ( Figure 7—figure supplement 2 ) . BAC fxAR121 mice show an early lethality phenotype that can be completely rescued by removing the mutant AR only from the skeletal muscle ( Cortes et al . , 2014; Lieberman et al . , 2014 ) . This early lethality can be recapitulated in our C57/129 hybrid genetic background ( Figure 7J; median survival of 219 days ) , although the mice live slightly longer than on the pure C57BL/6J background . Strikingly , in addition to rescuing the muscle atrophy and motor neuron phenotypes , loss of one copy of Nlk extended the lifespan of the BAC fxAR121+/− mice ( Figure 7J; increased to a median survival of 299 days ) . This effect is dramatic considering that these mice have only a 50% reduction in NLK protein expression . Loss of one copy of Nlk alone did not significantly alter lifespan ( Figure 7J , orange vs black lines , p = 0 . 854 , log rank test ) . As NLK influences the aggregation of the polyQ-expanded AR in cell culture and Drosophila ( Figures 1–4 ) , we tested if there was any change in the aggregation of mutant AR in the BAC fxAR121 mice when NLK expression was decreased . While the mutant AR shows primarily diffuse staining in the spinal motor neuron nuclei ( data not shown ) , we were able to detect aggregates in the skeletal muscle of BAC fxAR121 mice via multiple assays . First , aggregation could be detected via immunofluorescence with AR antibodies , resulting in punctate , nuclear staining that was absent from wild-type or Nlkgt/+ muscle ( Figure 8A–D ) . Loss of one copy of Nlk significantly reduced the number of nuclei containing aggregates by 20 weeks of age ( Figure 8E ) . Next , we analyzed aggregation biochemically . When muscle protein extracts were subjected to a filter trap assay , insoluble , aggregated AR was detected specifically in BAC fxAR121+/− and BAC fxAR121+/−; Nlkgt/+ samples , and not in wild-type or Nlkgt/+ samples ( Figure 8F ) . Quantification revealed that the amount of aggregated AR was significantly decreased with loss of one copy of Nlk by 20 weeks of age , although there was no longer a difference in this phenotype at very late stages of disease ( i . e . , 40 weeks ) ( Figure 8G ) . At late stages of the disease , the mutant AR could also be detected as a high molecular weight smear in the stacking gel of SDS-PAGE gels , and , once again , this was decreased with loss of one copy of Nlk ( Figure 8H , I ) . Therefore , as was seen in cell culture , primary motor neurons , and flies , NLK promotes the aggregation of mutant AR , and this aggregation positively correlates with an exacerbation of SBMA phenotypes . Conversely , loss of one copy of Nlk reduces aggregation across models , and we have found that this 50% reduction in NLK protein is sufficient to significantly improve SBMA-related phenotypes , including lifespan , in BAC fxAR121 SBMA mice . 10 . 7554/eLife . 08493 . 017Figure 8 . Loss of one copy of Nlk decreases mutant AR aggregation in mice . ( A–D ) Nuclear AR aggregates ( arrows ) can be detected in quadriceps of mice expressing the BAC fxAR121 transgene ( C , D ) , but not in controls ( A , B ) . Representative 30-week-old samples are shown . Scale bars are 50 μm . Nuclei are marked with TOTO-3 in blue . ( E ) Quantification of the ratio of nuclei containing aggregates out of total nuclei counted; 300 to 500 fibers per mouse . n . s . = not significant , *p < 0 . 05 ( t-test ) . For 10 weeks , n = 3 each . For 20 weeks , n = 5 each . For 30 weeks , n = 5 and 8 , respectively . For 40 weeks , n = 4 and 3 , respectively . ( F ) A representative filter trap assay blot from 20-week-old quadriceps samples . ( G ) The amount of insoluble ( Insol . ) AR out of total ( Insol . + Soluble ) was quantified . n . s . = not significant , *p < 0 . 05 , **p < 0 . 005 ( t-test ) . For 10 weeks , n = 3 each . For 20 weeks , n = 5 each . For 30 weeks , n = 4 each . For 40 weeks , n =4 and 3 , respectively . ( H ) A representative blot shows mutant AR retained in the stacking gel of SDS-PAGE gels as high molecular weight aggregates ( arrow ) . 30-week-old quadriceps samples are shown . An antibody to the polyQ region ( 1C2 ) was used . ( I ) Quantification of AR in the stacking gel normalized to loading control . *p < 0 . 05 ( t-test ) . n = 3 for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 017 Having established that NLK promotes SBMA phenotypes , we next wondered what was the molecular mechanism underlying this effect . Since NLK binds AR ( Figure 1A–C ) and is a kinase , we first tested whether NLK could phosphorylate AR . We noted that co-expression of AR with NLK-WT induced an electrophoretic mobility shift in the AR protein ( Figure 9A , lane 2 , blue arrow ) that was not seen with co-expression of NLK-KN ( Figure 9A , lane 3 ) . This mobility shift was reversed when cell extracts were incubated with lambda phosphatase ( Figure 9A , lane 5 , red arrow ) , suggesting that this shift represents an NLK-induced AR phosphorylation . NLK targets proline-directed serines and threonines . There are thus seven potential NLK target sites within the full-length AR protein . We were able to obtain phospho-specific antibodies for two of these sites , serine ( S ) 81 and S308 . NLK significantly increased AR phosphorylation at both of these sites in a kinase activity-dependent manner ( Figure 9B–D ) . Evidence suggests that NLK can target AR in both the presence and absence of hormone ( data not shown ) , but as the effect of NLK on the non-ligand-bound AR is unlikely to be disease relevant , we have focused on its ligand-dependent activity . Taken together , NLK was able to interact with and regulate the phosphorylation of both the wild-type ( data not shown ) and polyQ-expanded AR at two sites , although NLK can likely target other sites in AR as well . 10 . 7554/eLife . 08493 . 018Figure 9 . NLK influences the phosphorylation status of AR . ( A ) NLK can induce the phosphorylation of AR in a cell culture system . AR25Q is shown here , but the same effect is seen with polyQ-expanded AR . ( B–D ) NLK can phosphorylate the mutant AR at S81 and S308 . ( C ) Quantification of phospho-AR-S81 expression over total AR expression ( as detected by AR-N20 antibody ) . ( D ) Quantification of phospho-AR-S308 expression over total AR expression . *p < 0 . 05 ( t-test ) . n ≥ 4 trials . ( E , F ) NLK can affect mutant AR phosphorylation in SBMA mouse muscle in vivo . ( E ) Representative image of 30-week-old mouse quadriceps samples immunoblotted with phospho-AR-S81 antibody and an antibody to detect total AR . Only mutant AR protein is shown here , but a lower wild-type AR band can also be detected in all 4 genotypes . ( F ) Quantification of phospho-AR-S81 expression over total AR expression . *p < 0 . 05 ( t-test ) . n = 7 and 9 for BAC fxAR121+/− and BAC fxAR121+/−; Nlkgt/+ , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 018 We next asked whether NLK could influence the phosphorylation of AR in vivo . We used the same phospho-specific antibodies to assess the phosphorylation of the mutant AR protein in the skeletal muscle of BAC fxAR121 mice . Unfortunately , the phospho-AR-S308 antibody could not detect the mutant AR protein in these mice ( data not shown ) , and so we could not assess if NLK influences phosphorylation at this site in vivo via this approach . However , the phospho-AR-S81 antibody could detect the mutant AR protein , and we found that male mice lacking one copy of Nlk showed a reduction in the level of AR-S81 phosphorylation ( Figure 9E , F ) . This suggests that NLK regulates the phosphorylation of AR in vivo . We next tested if the NLK-mediated change in AR-S81 phosphorylation contributed to the SBMA phenotype . As NLK increases mutant AR aggregation across multiple model systems ( cultured cells , primary motor neurons , Drosophila , and mouse ) , and this positively correlates with its effects on SBMA phenotypes in vivo , we reasoned that our cell culture system could be reliably used as an initial read-out for NLK-mediated effects on mutant AR . In order to test the specific contribution of AR-S81 phosphorylation to SBMA-related phenotypes , we introduced a phospho-resistant mutation into the polyQ-expanded AR construct at S81 ( S81A; serine to alanine substitution ) . We found that the AR-S81A mutant tended to show slightly less aggregation than AR-S81 ( Figure 10A , representative images in Figure 10—figure supplement 1 ) , although this decrease was not significant by ANOVA . Interestingly , the S81A mutation significantly compromised the NLK effect on mutant AR aggregation ( Figure 10A and Figure 10—figure supplement 1 ) . This suggests that phosphorylation at S81 can contribute to the NLK-mediated effects on AR aggregation at least in our cell culture system . 10 . 7554/eLife . 08493 . 019Figure 10 . NLK regulates the aggregation and toxicity of mutant AR by influencing the phosphorylation of AR at residues including S81 . ( A ) AR-S81 phosphorylation could contribute to the NLK effect on mutant AR aggregation . NSC-34 cells were transfected with indicated constructs and treated with 10 nM DHT . Quantification of the ratio of cells containing AR aggregates out of total counted is shown . ***p < 0 . 001 ( ANOVA with Tukey's post-hoc analysis ) . n ≥ 3 trials . See also Figure 10—figure supplement 1 . ( B ) NLK induces the phosphorylation of a 130 amino acid AR N-terminal fragment at S81 in NSC-34 cells . ( C–F ) Reduced expression of NLK suppresses the toxicity induced by a mutant AR fragment in a Drosophila model of SBMA . Two independent mutant alleles ( adk1 and adk2 ) of nmo showed the same results . Flies were raised at 22°C and genotypes are as follows: ( C ) GMR-Gal4/+; UAS-EGFP/+ , ( D ) GMR-Gal4/+; UAS-trAR112Q/+ , ( E ) GMR-Gal4/+; UAS-trAR112Q/nmoadk1 , ( F ) GMR-Gal4/+; UAS-trAR112Q/nmoadk2 . More than 50 adult flies per genotype were observed at day 2 after eclosion , and five independent experiments were performed . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 01910 . 7554/eLife . 08493 . 020Figure 10—figure supplement 1 . S81 phosphorylation contributes to NLK-mediated effects on AR aggregation . Representative images of DHT-treated NSC-34 cells expressing HA-tagged AR120Q or AR120Q-S81A in the absence ( A , B ) and presence ( C , D ) of wild-type NLK ( NLK-WT ) co-expression . Cells were subjected to immunofluorescence using anti-AR N-20 ( green; A-D ) and anti-FLAG ( red; A′-D′ ) antibodies to detect AR aggregation and NLK co-expression , respectively . Merged images are shown in ( A′′-D′′ ) . Scale bars are 50 μm . Cells were scored as containing aggregates ( orange arrows ) or not ( white arrows ) . The ratio of aggregate-containing cells out of total scored is quantified in Figure 10A . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 020 To further investigate the contribution of NLK-mediated AR-S81 phosphorylation on mutant AR toxicity , we decided to make use of a previously published N-terminal fragment model of SBMA . Expression of a polyQ-expanded 130 amino acid N-terminal fragment of AR ( trAR112Q ) in the Drosophila eye results in a robust retinal degeneration and depigmentation phenotype ( Chan et al . , 2002 ) . This fragment is able to interact with NLK ( Figure 1C ) and contains only two putative NLK targets sites , S81 and S94 . We found that mutating S94 to alanine did not affect the NLK-mediated increase in full-length mutant AR aggregation in NSC-34 cells ( data not shown ) . We therefore predicted that if loss of one Nlk allele could rescue the toxicity of this polyQ-expanded AR N-terminal fragment , the mechanism would likely depend upon the interaction of NLK with AR and phosphorylation at S81 . We first confirmed that NLK could still induce phosphorylation at S81 in this fragment by co-expressing the proteins in NSC-34 cells ( Figure 10B ) . We next crossed trAR112Q flies with nmo mutant flies and assessed the eye phenotypes of the resulting progeny . We found that loss of one copy of nmo reversed the depigmentation phenotype induced by the trAR112Q fragment ( Figure 10C–F ) . This result supports the idea that NLK may regulate the aggregation and toxicity of polyQ-expanded AR via N-terminal binding and AR-S81 phosphorylation in SBMA . SBMA is caused by polyQ expansion in the full-length AR protein , but the exact molecular mechanisms underlying the disease are unclear . On one hand , it has been reported that mutant AR can be processed by proteases and the polyQ-containing AR fragments are toxic and aggregation-prone ( Merry et al . , 1998; Ellerby et al . , 1999; Chan et al . , 2002 ) . Mutant AR inclusions in patient tissue can only be detected by N-terminal AR antibodies and not by antibodies to the AR C-terminus ( Li et al . , 1998 ) . It has therefore been speculated that aggregates are comprised of mostly N-terminal AR fragments , and there is some evidence from mouse models to support this theory ( Li et al . , 2007 ) . These fragments lack the AR DNA-binding domain , suggesting that the polyQ-dependent toxicity seen , for example , in the trAR112Q Drosophila model must occur in the absence of specific DNA binding and AR-mediated gene transcription . Our data show that NLK can modulate mutant AR toxicity in this fragment model ( Figure 10C–F ) , suggesting that it can play a role in transcription-independent pathological pathways in SBMA , such as protein misfolding and aggregation . Of course , these fragment models show ligand-independent toxicity and therefore cannot recapitulate the specific features of SBMA . Furthermore , the ability of AR to bind DNA is known to be important for toxicity in a full-length mutant AR Drosophila model of SBMA , suggesting that SBMA may also arise via a mechanism that involves aberrant gene transcription ( Nedelsky et al . , 2010 ) . Consistent with this idea , changes in gene expression have been detected in SBMA mouse models and this is believed to contribute to pathology ( Sopher et al . , 2004; Ranganathan et al . , 2009; Katsuno et al . , 2010a; Mo et al . , 2010; Minamiyama et al . , 2012 ) . We therefore wondered whether NLK was also able to affect the function of the full-length AR protein , as this may contribute to the molecular mechanism by which NLK affects SBMA in vivo . We started by testing if NLK could affect the ability of the mutant AR to activate gene transcription by making use of an AR-responsive luciferase reporter . Both wild-type and mutant AR ( Figure 11A and Figure 11—figure supplement 1 ) are able to activate the expression of this reporter when expressed in DHT-treated NSC-34 cells , although , as expected ( Mhatre et al . , 1993; Thomas et al . , 2006 ) , AR120Q showed less activity than AR25Q . When NLK was co-expressed with AR , it led to a robust increase in AR-mediated gene transcription in a hormone- and kinase activity-dependent manner ( Figure 11A and Figure 11—figure supplement 1 ) . This effect was also seen with wild-type AR ( Figure 11—figure supplement 1B ) , suggesting that NLK may normally act as an AR cofactor or regulator . 10 . 7554/eLife . 08493 . 021Figure 11 . NLK promotes AR-mediated gene transcription by inhibiting the N/C interdomain interaction and promoting AF-2 cofactor binding . ( A ) NLK increases AR-dependent gene transcription in a kinase activity-dependent manner in NSC-34 cells . n . s . = not significant , **p < 0 . 01 ( ANOVA with Tukey's post-hoc analysis ) . n = 3 trials . ( B ) NLK inhibits the AR N/C interaction as measured by a mammalian two-hybrid assay in NSC-34 cells . ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n ≥ 4 trials . See also Figure 11—figure supplement 2 . ( C ) NLK can activate AR-dependent gene transcription in the absence of the N/C interaction in NSC-34 cells . **p < 0 . 01 , ***p < 0 . 001 ( ANOVA with Tukey's post-hoc analysis ) . n ≥ 3 trials . ( D ) NLK and p300 synergistically increase AR-mediated gene transcription in NSC-34 cells , suggesting NLK may promote AR-cofactor binding and function . *p < 0 . 05 ( ANOVA with Tukey's post-hoc analysis ) . n = 5 trials . ( E ) NLK increases AR-mediated gene transcription via the AR AF-2 domain in NSC-34 cells . n . s . = not significant , *p < 0 . 05 , ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n = 4 trials . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 02110 . 7554/eLife . 08493 . 022Figure 11—figure supplement 1 . NLK does not induce AR transactivation in the absence of hormone . ( A ) NSC-34 cells were transfected as indicated and treated with DHT or vehicle only , and AR transactivation activity was measured by a dual-luciferase assay using an AR-responsive reporter . In the absence of hormone , NLK does not induce the transactivation activity of AR . n . s . = not significant , ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n = 4 trials . ( B ) NLK increases the activity of the wild-type AR25Q in NSC-34 cells in a kinase activity-dependent manner . ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n = 3 trials . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 02210 . 7554/eLife . 08493 . 023Figure 11—figure supplement 2 . NLK dose-dependently inhibits the AR N/C interaction . The AR N/C interaction was measured by a mammalian two-hybrid assay in DHT-treated NSC-34 cells . Interaction between the two AR fusion constructs ( i . e . , N/C interaction ) was measured via the activity of a Gal4-dependent luciferase reporter normalized over a renilla luciferase control . ( A ) NLK inhibits the wild-type AR N/C interaction . This only partially depends upon the kinase activity of NLK . ***p < 0 . 001 , ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n ≥ 3 trials . ( B ) NLK dose dependently inhibits the AR N/C interaction . VP16-AR N-terminal fragments of indicated glutamine ( Q ) length were co-transfected with Gal4-AR-C and increasing amounts of FLAG-NLK-WT and cells were treated with 10 nM DHT . Even low levels of NLK are able to significantly repress the N/C interaction of polyQ-expanded AR . Asterisks in ( B ) refer to the comparison of indicated sample with the minus NLK control . *p < 0 . 05 , **p < 0 . 005 , ***p < 0 . 0005 ( t-test ) . n ≥ 3 trials . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 02310 . 7554/eLife . 08493 . 024Figure 11—figure supplement 3 . NLK can increase mutant AR aggregation and phosphorylation independent of an N/C interaction . ( A–C ) NLK increases the aggregation rate of the N/C interaction-defective mutant AR ( HA-AR120Q-L26A/F27A ) . NSC-34 cells were transfected as indicated and treated with 10 nM DHT . AR aggregation was detected via immunofluorescence using anti-AR N20 ( green; A , B ) . NLK co-expression was detected via immunofluorescence using anti-FLAG ( red; A′ , B′ ) . Merged images are shown in ( A′′ , B′′ ) , where an orange arrow indicates a cell with aggregates and white arrow indicates one without . Asterisks mark non-specific staining in the red channel in ( A′ ) . Scale bar in ( A′′ ) is 50 μm and applies to all panels . The ratio of cells containing aggregates out of total counted is quantified in ( C ) . ***p < 0 . 001 , ****p < 0 . 0001 ( ANOVA with Tukey's post-hoc analysis ) . n = 3 trials . ( D ) NLK increases AR phosphorylation at S81 in the absence of an N/C interaction . Data were consistent over multiple trials , and a representative immunoblot is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 024 We next wondered how exactly NLK was able to influence AR-mediated gene transcription . While most nuclear hormone receptors regulate gene transcription primarily via the interaction of their ligand binding-induced AF-2 domain with cofactors that contain a LxxLL motif , AR is unique in that it contains an LxxLL-like site in its N-terminus ( 23FQNLF27 ) that interacts with its own AF-2 domain with a greater affinity than other motifs ( He et al . , 2001 ) . This intramolecular interaction is known as the N/C interaction and it causes AR to regulate gene transcription primarily through its AF-1 domain in lieu of the AF-2 domain ( He et al . , 2001 , 2004 ) . Loss of this interaction leads to a decrease in AR-mediated gene transcription at some , but not all AR-dependent genes ( He et al . , 2002; Callewaert et al . , 2003 ) . Interestingly , it has been previously reported that the N/C interaction is upstream of mutant AR aggregation and toxicity , as well as its phosphorylation at both S81 and S308 ( Orr et al . , 2010 ) . Therefore , we predicted that NLK might be acting to promote this intramolecular interaction and thereby increase AR-mediated gene transcription , AR phosphorylation , and SBMA phenotypes . We tested this idea by performing a mammalian two-hybrid assay in which a VP16 activation domain-fused AR N-terminus ( VP16-AR120Q-N ) is co-transfected with a Gal4 DNA binding domain-fused AR C-terminus ( Gal4-AR-C ) . When these AR N- and C-terminal fragments interact , they bring together the Gal4-DBD and the VP16 activation domain , leading to an increase in the expression of a co-transfected Gal4-dependent luciferase reporter ( Figure 11B ) . When we carried out this assay in the presence of NLK , we found that , surprisingly , NLK inhibits the N/C interaction ( Figure 11B and Figure 11—figure supplement 2A ) . This inhibition was NLK dose dependent ( Figure 11—figure supplement 2B ) . The N/C interaction was also inhibited by NLK-KN , but to a lesser extent ( Figure 11B and Figure 11—figure supplement 2A ) . Once again , these effects were seen with both wild-type and mutant AR , although NLK inhibited the N/C interaction more robustly in the presence of the polyQ expansion ( Figure 11—figure supplement 2B ) . This suggests that NLK is able to prevent the N/C interaction via a mechanism that is only partially dependent on its kinase activity . In order to confirm that NLK was still able to induce its effects on the mutant AR in the absence of this AR N/C interaction , we introduced a mutation in the N-terminal 23FQNLF27 motif of AR that prevents it from binding the AF-2 domain in the C-terminus of the protein ( HA-AR120Q-L26A/F27A ) ( He et al . , 2001 ) . As previously reported ( He et al . , 2002; Callewaert et al . , 2003; Orr et al . , 2010 ) , this construct tended to be compromised in its ability to aggregate ( Figure 11—figure supplement 3A , C ) and was significantly impaired in its ability to induce AR-mediated gene transcription ( Figure 11C ) . It also showed a reduction in phosphorylation at AR-S81 ( Figure 11—figure supplement 3D ) , as was reported in a separate SBMA cell model ( Orr et al . , 2010 ) . Nonetheless , co-expression of NLK-WT still increased the aggregation rate of this mutant AR ( Figure 11—figure supplement 3A–C ) . NLK increased AR-mediated gene transcription when co-expressed with the AR N/C mutant ( Figure 11C ) . NLK was also able to increase the phosphorylation of this construct at S81 ( Figure 11—figure supplement 3D ) , as well as at S308 , although to a lesser extent ( data not shown ) . Taken together , these data indicate that NLK can influence the activity and toxicity of the mutant AR via a mechanism that is independent , but perhaps parallel to , the AR N/C interaction . One important remaining question is how NLK can increase AR-meditated gene transcription while inhibiting its N/C interaction . AR regulates target gene transcription by interacting with several cofactors at both its AF-1 and AF-2 domains ( Bennett et al . , 2010 ) , and these interactions can be altered by polyQ expansion . For example , the coactivator CREB-binding protein ( CBP ) binds the polyQ-expanded AR more robustly than its wild-type counterpart in a mouse model of SBMA ( Sopher et al . , 2004 ) and can be sequestered into mutant AR aggregates ( McCampbell et al . , 2000 ) , suggesting that the interaction of this protein with mutant AR may be important for disease pathogenesis in vivo . In addition , AR is acetylated by CBP/p300 ( Fu et al . , 2000 ) , and this acetylation is also important for SBMA pathogenesis ( Montie et al . , 2011 ) . Therefore , we wondered whether NLK regulates AR transcriptional activity by altering cofactor interactions , and chose to look specifically at p300 . As expected , we found that co-expression of p300 with NLK showed a higher level of AR-dependent transcriptional activity than with either cofactor alone ( Figure 11D ) . This suggests that NLK may enhance coactivator recruitment to the polyQ-expanded AR and thereby increase AR-mediated gene transcription . Based on our mammalian two-hybrid data ( Figure 11B ) , we speculated that the binding of NLK to the N-terminus of AR ( Figure 1C ) sterically blocks the ability of the AR N-terminus to bind the C-terminal AF-2 domain . As the N/C interaction can inhibit cofactor binding at the AR AF-2 domain ( He et al . , 2001 ) , we reasoned that NLK may be acting to relieve this inhibition and thereby promote gene transcription via the AR AF-2 domain . To test this , we introduced two different point mutations into the AR AF-2 domain to differentially inhibit cofactor binding . The AR AF-2 domain is flanked by two charged clamp residues that mediate its interaction with cofactors containing LxxLL or FxxLF motifs . K720A is a partial AF-2 mutation that neutralizes the charge of one of the clamps , preventing LxxLL motif binding and reducing FxxLF motif binding by 50% ( Dubbink et al . , 2004; Nedelsky et al . , 2010 ) . E897K is a complete AF-2 mutation that reverses the charge at the other clamp , abolishing both LxxLL and FxxLF motif binding ( Dubbink et al . , 2004 ) . We carried out the AR-responsive luciferase assay with both mutants and found that the E897K mutation alone tended to slightly decrease AR-mediated gene transcription compared to that seen with a wild-type AR , while the K720A mutation did not affect AR activity in NSC-34 cells ( Figure 11E ) . This is consistent with what was reported in COS-1 cells ( Nedelsky et al . , 2010 ) . When we co-expressed a wild-type NLK with these AR mutants , we were still able to detect an increase in AR-mediated gene transcription with the K720A mutation . In contrast , NLK-mediated enhancement in AR activity was dramatically compromised by the E897K mutation ( Figure 11E ) . These data suggest that the NLK-induced increase in AR transcriptional activity is dependent on a functional AR AF-2 domain .
SBMA is a devastating neuromuscular disease without any cure or effective therapy to date . In this study , we explored whether and how NLK could modulate the pathogenesis of SBMA . By utilizing a variety of model systems , we clearly show that NLK is a key regulatory factor capable of modulating AR activity and SBMA pathology . Using a combination of cell culture , Drosophila , and mouse models , we show that reduced expression of NLK suppresses , while increased expression exacerbates , mutant AR-associated SBMA pathology , including protein aggregation , cellular toxicity and degeneration , and animal lethality phenotypes . It is particularly intriguing that the effects of NLK on the mutant polyQ-expanded AR and SBMA are consistent across different model systems , as this suggests that the role of NLK in SBMA pathogenesis is fundamental . Furthermore , all of these effects are clearly dependent on the kinase activity of NLK . Our work therefore strongly suggests that a reduction in NLK expression or enzymatic activity could be beneficial for SBMA patients . Of particular importance is our finding that a 50% reduction in NLK expression partially rescues the phenotypes of an SBMA mouse model ( Figures 6–8 ) . This improvement in pathology was seen at 20 weeks of age in these mice and was more robust at the later time point of 30 weeks . By very late time points ( i . e . , 40 weeks ) , however , a reduction in NLK expression resulted in an improvement of only some of SBMA phenotypes assayed ( Figures 6 , 8E ) , and did not show a robust effect in other assays ( Figures 7 , 8G ) . This suggests that a reduction in NLK expression may act to delay disease progression in this model , but is not sufficient to completely prevent the onset of the full SBMA phenotype . It should be noted that the majority of BAC fxAR121 mice die before reaching this final time point , however , and so we cannot rule out the possibility that the small cohort of mice analyzed at 40 weeks represent an ‘escaper’ subset of SBMA mice that are slightly healthier than the average BAC fxAR121 mouse . The reasons for the variation in the SBMA phenotype in these mice are not known , but may be interesting to investigate in the future . It is also worth mentioning that we analyzed the mice on a C57/129 F1 hybrid genetic background . Although they can be considered to be on a pure background for the purpose of this study , we cannot rule out the possibility of mouse background effects . Future studies on a different pure genetic background and/or using NLK inhibitors would be useful in corroborating our findings . In this study , we uncovered some molecular mechanisms that we predict underlie the role of NLK in SBMA at the cellular level . First , NLK interacts with mutant AR at the N-terminal region of the protein , and , interestingly , polyQ expansion results in a more robust interaction between NLK and mutant AR in comparison to wild-type AR ( Figure 1A–C ) . Second , consistent with our data that NLK modulates SBMA features in a kinase activity-dependent manner , NLK promotes the phosphorylation of AR , either directly or indirectly , at multiple sites , including S81 and S308 ( Figure 9 ) . NLK-induced changes in AR-S81 phosphorylation can be detected in vivo in mice , and AR-S81 phosphorylation likely contributes to the effect of NLK on SBMA pathology in cell culture and Drosophila models ( Figure 10 ) . Interestingly , however , the S81A mutation decreased , but did not completely abolish , the NLK-mediated effects on mutant AR aggregation ( Figure 10A ) . This indicates that , while AR-S81 is likely an important NLK phosphorylation site in the N-terminal region of AR , there may be other NLK target sites outside of this region that also contribute to NLK-dependent AR toxicity in SBMA . Finally , NLK can affect the transcriptional activity of the mutant AR protein , and , once again , this is dependent on its kinase activity ( Figure 11 ) . Unlike aggregation , which is dependent on the presence of a polyQ expansion , this effect is seen with both wild-type and mutant AR . This suggests that NLK normally acts as an AR cofactor or regulator . PolyQ expansion , while resulting in the aggregation of the protein , also affects the activity of the AR monomer , whose altered function in target gene transcription may exert pathology in SBMA ( Nedelsky et al . , 2010 ) . The ability of NLK to promote the activity of the mutant AR could therefore exacerbate this polyQ-induced protein dysfunction . We suspect that both the modulation of mutant AR aggregation and the misregulation of its native functions ultimately contribute to SBMA pathology , and our data suggest that NLK influences both of these pathomechanisms ( Figure 12 ) . 10 . 7554/eLife . 08493 . 025Figure 12 . A potential model for the role of NLK in SBMA pathogenesis . NLK can induce the phosphorylation of the polyQ-expanded AR , which influences its aggregation and contributes to its toxicity in SBMA models . NLK can also regulate the ability of the mutant AR to act as a transcription factor , which would enhance any aberrant AR-mediated gene transcription that contributes to SBMA pathology . A combination of these toxic mechanisms and others could ultimately result in the degeneration and pathology characteristic of SBMA . These events occur downstream of AR ligand binding and nuclear translocation . In addition , NLK may inhibit the AR N/C interaction to promote AR AF-2 cofactor binding . DOI: http://dx . doi . org/10 . 7554/eLife . 08493 . 025 The binding of NLK to AR , and likely its subsequent phosphorylation , strongly inhibit the AR N/C interaction , and yet paradoxically increase AR-mediated gene transcription ( Figure 11A–C ) . This led us to investigate whether NLK could regulate AR activity via the AF-2 domain , and indeed , we found that complete inhibition of cofactor binding at the AR AF-2 domain strongly compromised the ability of NLK to increase AR transcriptional activity ( Figure 11E ) . The NLK effect on AR activity was not completely abolished by this mutation , however , suggesting that NLK may promote AF-1-dependent transcription , as well . Furthermore , the E897K mutation compromised the effect of NLK on AR activity more so than the K720A mutation ( Figure 11E ) . As these mutations both completely abolish LxxLL motif binding , this suggests that NLK may preferentially allow for FxxLF motif-containing cofactor binding at the AR AF-2 domain , and this possibility warrants further investigation . Once again , as the effect of NLK on AR transactivation and on the AR N/C interaction are seen with both wild-type and mutant AR , this role for NLK in the regulation of AR AF-2 cofactor interactions is likely a normal function of NLK in AR signaling . Given that NLK promotes SBMA pathology , this may suggest that the AR AF-2 domain is important for SBMA pathogenesis . Interestingly , a separate study found that the retinal degeneration phenotypes in a full-length mutant AR Drosophila model were also dependent upon the AF-2 domain of AR . This study also found that the E897K mutation at the AF-2 domain led to a more robust rescue of mutant AR phenotypes than K720A , again demonstrating that the ability of the AR AF-2 domain to bind FxxLF-containing cofactors may be important for SBMA pathogenesis ( Nedelsky et al . , 2010 ) . We also find it intriguing that the effect of NLK on AR molecular functions is very similar to that of the primate-specific Melanoma Antigen Gene Protein 11 ( MAGE-11 ) . MAGE-11 has been reported to bind AR specifically at the 23FQNLF27 motif in the N-terminal region of the AR protein to prevent the N/C interaction and allow for cofactor binding at the AF-2 domain ( Bai et al . , 2005 ) . MAGE-11 is also known to directly bridge interactions between AR and various cofactors , including TIF2 and p300 , resulting in a synergistic activation of AR-dependent gene transcription ( Askew et al . , 2009 , 2010 ) . Together with our data , this suggests that inhibition of the N/C interaction by specific AR cofactors represents a unique and intriguing approach to regulating AF-domain dominance in AR target gene transcription . It will be interesting to investigate whether there is any cross-talk between NLK and MAGE-11 in AR-mediated gene activation and , perhaps , even in SBMA disease pathogenesis . We suspect that the NLK-mediated increase in AR transactivation results from an increase in cofactor binding at the AR AF-2 domain , thereby supporting a model in which AF-2-mediated interactions are important for SBMA pathogenesis . And yet , it is also clear that inhibiting the N/C interaction via point mutations in the AR 23FQNLF27 motif reduces mutant AR aggregation and toxicity ( Figure 11—figure supplement 3A–C and Orr et al . , 2010 ) , features that NLK clearly promotes . One explanation for this seemingly conflicting data is that the binding of NLK to AR and the subsequent phosphorylation of the AR protein , perhaps at S81 , elicit an effect on the AR protein that is similar to the effect of the N/C interaction . In this model , NLK binding and the N/C interaction are parallel means of triggering a similar downstream pathogenic response . It should be stressed , however , that binding and phosphorylation by NLK does not preclude the need for AR ligand binding , as aggregation ( Figure 1G and Figure 1—figure supplement 1 ) , gene transcription ( Figure 11—figure supplement 1A ) , and the formation of the AR AF-2 domain ( Wärnmark et al . , 2003 ) all depend upon the presence of androgens , and NLK has no effect on these features without ligand . Furthermore , the exact details of this downstream pathway are still not completely clear . For instance , toxicity could arise from aberrant AR-mediated gene transcription ( via a combination of AF-1 and AF-2 dependent mechanisms ) , the sequestration of various cofactors into aggregates , the inability of certain cells to handle the accumulation of toxic AR conformers , or via some other as-yet-unknown pathogenic factors . Or , perhaps more likely , SBMA may arise from a combination of the above ( Figure 12 ) . Given that NLK interacts with and phosphorylates the mutant AR ( Figures 1 , 9 ) , we suspect that it is acting cell autonomously to regulate AR activity based on the mechanism we propose . As our mouse studies were carried out using a constitutive knockdown of NLK , however , we cannot at this time determine where NLK exerts its effects on SBMA pathology . In other words , it could be regulating mutant AR activity in the spinal motor neurons , the skeletal muscle , or both . Future investigations using targeted NLK inhibitors or tissue-specific knockdown using both this and other SBMA mouse models could address these questions . It is also interesting to note that , although our data show that NLK can influence the aggregation of the mutant AR across multiple models , and that NLK has a robust effect on AR transactivation activity in cells , we saw only a partial rescue of SBMA muscle and motor neuron pathology with a reduction in NLK expression in mice . Why we did not see a more robust improvement of these phenotypes is an intriguing question . A simple explanation may be that the remaining 50% of NLK expression , AR phosphorylation , and mutant protein aggregation is enough to allow for the toxic effects of the mutant AR that directly result in decreased muscle fiber and motor neuron size . A more complete knockout of NLK would thus be needed to prevent degeneration . Mice heterozygous for the Nlk gene trap allele are largely normal , suggesting either that this lower expression of NLK is enough to adequately carry out wild-type functions of NLK in the adult mouse , or that some other factor or pathway compensates for the decrease in active NLK . It is possible that such a compensatory factor or pathway may also contribute to mutant AR-induced pathology when NLK expression is reduced . Data presented here and in a previous publication ( Ju et al . , 2013 ) suggest that NLK is able to regulate the pathogenesis of two separate polyQ diseases: SBMA and SCA1 . In both cases , evidence suggests that NLK binds to and phosphorylates the mutant protein and thereby regulates its aggregation and activity . Yet , why NLK interacts with multiple polyQ proteins is an open question that warrants future investigation . We also noted that co-expression of NLK seemed to increase AR protein levels in NSC-34 cells and in the AR61Q Drosophila model , as well as separately influence its propensity to aggregate . This suggests that NLK may play a role in the stabilization of AR , specifically at the protein level , as both of these systems express AR under the control of exogenous promoters . In the BAC fxAR121 mice , however , the AR121Q protein levels between mice with full NLK expression and those with a 50% reduction in NLK are not significantly different across the population assayed ( data not shown ) . As all the mice assayed did show a rescue in the degenerative phenotype , however , we concluded that another mechanism must be playing a role in these mice and therefore investigated the possibility of a direct interaction between NLK and AR . That direct mechanism is the focus of the current study . Nonetheless , we noted that a subset of about 30–40% of the mice did show a reduction in mutant AR protein levels with a reduction in NLK expression . We therefore speculate that NLK may also play a role in protein clearance pathways , and that this , in turn , may contribute to the ability of NLK to regulate mutant protein aggregation and toxicity in varying disease cases . We ultimately suspect that both direct and indirect regulation of mutant protein expression/aggregation and activity underlies the role of NLK in disease . Lastly , although there is still much to be understood about the precise molecular mechanisms underlying SBMA and the role of NLK therein , our data clearly show that NLK normally promotes the disease condition and that reduction of NLK expression or activity is sufficient to partially rescue SBMA pathogenicity . We are confident in this conclusion because we utilized a multi-system approach to address the question . NLK is therefore a novel and interesting putative therapeutic target . We should note that complete loss of NLK function may cause severe problems , however , since NLK plays a role in multiple signaling pathways ( Ishitani et al . , 1999; Ohkawara et al . , 2004; Ishitani et al . , 2010; Ishitani and Ishitani , 2013 ) . Nonetheless , Nlkgt/+ heterozygous mice are generally healthy and our study provides convincing evidence that a 50% reduction in NLK protects against SBMA pathogenesis in vivo ( Figures 6–8 ) . Thus , this study suggests that putative treatments that target NLK may not need to completely inactivate the protein to generate a therapeutic effect . It will be very interesting to determine if pharmacologically inhibiting NLK can also rescue SBMA features at the mammalian level .
The following mutant and transgenic flies were used in this study: GMR-Gal4 ( Bloomington Stock Center ) , UAS-EGFP ( Bloomington Stock Center ) , UAS-AR14Q ( current study ) , UAS-AR61Q ( current study ) , UAS-trAR112Q ( Chan et al . , 2002 ) , UAS-NLK-WT ( Ju et al . , 2013 ) , UAS-NLK-KN ( Ju et al . , 2013 ) , nmoadk1 ( Verheyen et al . , 2001 ) , nmoadk2 ( Verheyen et al . , 2001 ) . In order to generate UAS-AR14Q and UAS-AR61Q transgenic fly lines , full-length human AR cDNAs with 14Q or 61Q were subcloned into the pUAST vector and then injected into fly embryos ( via Best Gene , Inc . Chino Hills , CA ) . After crossing with the GMR-Gal4 driver line , two independent UAS-AR lines of each Q length that showed roughly equal levels of transgene expression were used for the analysis . For the genetic interaction analyses , appropriate fly lines were intercrossed and their progeny were raised at 22°C , 25°C , or 30°C on fly food containing or lacking 100 nM DHT . All experiments were carried out multiple times . The Yale University Institutional Animal Care and Use Committee approved all research and animal care procedures . Mice were maintained on a 12/12-hr light/dark cycle with standard mouse chow and water ad libitum . Two independent Nlk gene trap ( NlkRRJ297/+ or NlkXN619/+ , or simply Nlkgt/+ ) mouse lines were maintained on the pure 129S6/SvEv background ( Ju et al . , 2013 ) . BAC fxAR121 SBMA transgenic mice were maintained on the pure C57BL/6J background ( Cortes et al . , 2014 ) . To perform the genetic interaction study , BAC fxAR121+/− heterozygote mice were bred to Nlkgt/+ heterozygote mice . The F1 male progeny ( C57/129 hybrid background ) were used in subsequent analyses . Mice were monitored for their general health and the date of death was recorded . Occasionally , mice were euthanized for humane reasons at the very end stage of disease progression , and the date of euthanasia was used as the death date in the analysis . Survival curves were generated using Kaplan–Meier statistical analysis and the log rank test was used to compare individual curves . The assay was capped at 2 years of age . Mouse quadriceps were harvested and snap frozen in liquid nitrogen-chilled isopentane . Samples were sectioned on a cryostat at 12 μm and collected on superfrost slides . Sections were then either stained with hematoxylin ( 3 min ) and eosin ( 1 min ) or incubated with 0 . 4 mg/ml NADH ( Roche ) and 0 . 8 mg/ml 4-nitro blue tetrazolium chloride ( NBT; Roche ) for 15 min , 37°C . Sections were then dehydrated with ascending ethanol solutions and incubated in xylenes . Coverslips were mounted with Permount . Slides were imaged on a compound light microscope using an Olympus camera and CellSens software . Fiber area and Feret's diameter of cross-sectional muscle fibers and the mean gray value of NADH transferase activity staining images were analyzed using ImageJ software ( National Institutes of Health ) . The NADH transferase activity images were obtained on the same day using identical camera settings . Mouse vertebral columns were dissected whole from freshly sacrificed mice and post-fixed in 4% paraformaldehyde overnight , 4°C . Samples were kept at 4°C through subsequent steps until freezing . After fixing , samples were incubated in 0 . 5 M EDTA in PBS overnight . The following day , the 0 . 5 M EDTA was replaced with fresh solution three times , rocking , with the last incubation lasting overnight . The next day , samples were moved to 10% sucrose , then 20% sucrose , and finally left in 30% sucrose overnight . Spinal cord and bone were frozen in Optimal Cutting Temperature ( OCT ) medium and later sectioned on a cryostat at 18 μm and collected on Superfrost Plus slides . After sectioning , the L4–L5 region was identified based on location and morphology as compared to a mouse spinal cord atlas . Alternating sections were stained with Cresyl violet ( 4 min ) and dehydrated in ascending ethanol solutions . Slides were incubated in xylenes , and coverslips were mounted with Permount . The entire L4–L5 region was imaged on a compound light microscope using an Olympus camera and CellSens software , and then random images periodically spaced throughout this region were used for the measurement of neuronal soma size using ImageJ . Over 100 neurons were scored per animal . Quadriceps extracts were generated as for immunoblot and prepared as 400 μl ( 1 μg/μl ) samples . Samples were then divided into 2 equal halves and ran separately though the filter trap assay using a BioRad BioDot SF apparatus according to the manufacturer's instructions , with the exception that a 0 . 22-μm cellulose acetate ( CA ) membrane ( Whatman ) was placed atop the 0 . 45-μm nitrocellulose ( NC ) membrane . The CA membrane collects insoluble AR , while the NC membrane detects soluble protein . For one sample half , both the CA and NC membranes were blocked and immunoblotted with anti-AR H280 antibody ( 1:500 , Santa Cruz ) . For the other sample half , the NC membrane was immunoblotted for the loading control using either rabbit anti-Actin ( 1:10 , 000; Sigma ) or mouse anti-Tubulin ( 1:30 , 000; Developmental Studies Hybridoma Bank ) . The amount of AR collected by each membrane was quantified using ImageJ . To generate HA-tagged AR constructs , the full-length human AR cDNAs were PCR-amplified from GFP-AR25Q or GFP-AR120Q plasmids and inserted into an HA vector using the XhoI and NotI sites . The 130 amino acid N-terminal fragment was also subcloned into HA and GFP vectors via restriction digest and Gateway cloning , respectively . All AR point mutations used in this study were introduced via site-directed mutagenesis using the Stratagene Quikchange Kit . The FLAG-tagged Nlk constructs were kindly provided by Dr Kunihiro Matsumoto ( Nagoya University , Nagoya , Japan ) and Dr Tohru Ishitani ( Kyushu University , Fukuoka , Japan ) . ARE-luciferase plasmids were kindly provided by Dr Nancy L Weigel ( Baylor College of Medicine , Houston , Texas , USA ) and Dr Zafar Nawaz ( University of Miami , Miami , Florida , USA ) . The mammalian two-hybrid constructs were kindly provided by Dr Diane E Merry ( Thomas Jefferson University , Philadelphia , Pennsylvania , USA ) . A mammalian cell culture system was used for co-immunoprecipitation and biochemical analyses , immunofluorescence , and luciferase reporter assays . Standard cell culture and plasmid transfection were conducted as described ( Ju et al . , 2013; Kim et al . , 2013 ) . Briefly , NSC-34 or HeLa cells were maintained in Dulbecco's Modified Eagle Medium ( DMEM ) supplemented with 10% fetal bovine serum ( FBS; Gibco ) . Cells were plated the day before transfection in 6- or 24-well plates . The following day , cells were transfected with indicated cDNA plasmids using lipofectamine 2000 ( Invitrogen ) according to the manufacturer's instructions , treated with 10 nM DHT ( Wako; dissolved in ethanol ) using DMEM supplemented with charcoal:dextran stripped FBS ( Gemini Bio-Products ) , and cultured until analyzed . To generate cell culture extracts , NSC-34 or HeLa cells were transfected and treated with DHT as described above . 24 hr after DHT treatment , cells were lysed in 300 μl NP40 lysis buffer ( 0 . 5% NP40 , 20 mM Tris-HCl pH 8 . 0 , 150 mM NaCl , 1 mM EDTA + Protease Inhibitor Cocktail [Roche] ) . Extracts were cleared by centrifugation for 10 min at 4°C , and soluble extract was either boiled with sample buffer to generate ‘input’ samples or incubated overnight with either anti-FLAG M2 Affinity Gel ( Sigma ) or glutathione-sepharose 4B beads ( GE Healthcare ) as indicated . IP samples were washed three times with lysis buffer and submitted to western blot analysis . For non-co-IP blots , cells were transfected and treated with DHT and cell extracts were generated as described using triple cell lysis buffer ( 0 . 5% NP40 , 0 . 5% Triton X-100 , 0 . 1% SDS , 20 mM Tris-HCl pH 8 . 0 , 180 mM NaCl , 1 mM EDTA + Protease Inhibitor Cocktail [Roche] ) . Extracts were boiled with sample buffer and ran on 8 or 12% SDS-PAGE gels . Gels were transferred to NC membranes , blocked and incubated with primary antibodies overnight in non-fat milk at 4°C . Membranes were then washed and probed with horseradish peroxidase-conjugated secondary antibodies ( GE Healthcare ) and exposed to film . Mouse tissue samples were harvested and lysed in 1 ml RIPA buffer ( 1% NP40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 50 mM Tris-HCl pH 8 . 0 , 150 mM NaCl , + Protease Inhibitor Cocktail [Roche] ) by dounce homogenization and cleared by centrifugation for 10 min at 4°C . Total protein concentration was measured using a BCA assay and equivalent concentrations of protein were ran on SDS-PAGE gels and blotted . For phosho-AR-S81 blots of mouse tissue , SuperSignal Western Femto ( Thermo Scientific ) was used in order to detect the signal . Adult Drosophila heads were collected and ground in 50 μl RIPA buffer and incubated on ice for 15 min . Samples were then spun for 10 min at 13 , 000 rpm . Supernatant was boiled with sample buffer for 5 min , ran on 8% SDS-PAGE gels , and blotted . Antibodies used include: mouse anti-HA ( 1:10 , 000; Sigma ) , mouse anti-FLAG ( 1:10 , 000; Sigma ) , mouse anti-GAPDH ( 1:20 , 000; Sigma ) , mouse anti-Tubulin ( 1:30 , 000; Developmental Studies Hybridoma Bank ) , rabbit anti-GST ( 1:10 , 000; Sigma ) , rabbit anti-AR N20 ( 1:500; Santa Cruz ) , rabbit anti-AR H280 ( 1:500; Santa Cruz ) , rabbit anti-AR phospho-S81 ( 1:500; Millipore ) , rabbit anti-AR phospho-S308 ( 1:500; Santa Cruz ) , mouse anti-polyQ 1C2 ( 1:1000; Millipore ) , and rabbit anti-NLK ( 1:5000; Abcam ) . Quantification of immunoblots was carried out using ImageJ using loading controls ran on the same SDS-PAGE gel as the samples . Multiple trials were averaged . Cells were plated unto coverslips and transfected and DHT-treated . 24 hr after DHT treatment , cells were fixed in 4% paraformaldehyde , permeabilized , blocked , and incubated with primary antibodies ( 1:1000 , mouse anti-FLAG and rabbit anti-AR N20 , as indicated ) . They were then washed , incubated with Alexa Fluor conjugated secondary antibodies ( 1:500 , Life Sciences ) , and mounted onto slides in Vectashield . Immunofluorescence was imaged using a Zeiss spinning disc confocal microscope using Volocity software . All images are composite z-stacks encompassing the entire cell . Cells were scored as containing aggregates or not based on the presence of punctate vs solely nuclear staining . The ratio of aggregate-containing cells out of total cells was recorded and averaged over at minimum three trials . Mouse quadriceps were harvested and snap frozen in liquid nitrogen-chilled isopentane . Samples were sectioned on a cryostat at 12 μm and collected on superfrost slides . Slides were blocked , incubated with rabbit anti-AR H280 ( 1:200; Santa Cruz ) , washed and incubated with Alexa Fluor 488 secondary antibody ( 1:500 , Invitrogen ) and TOTO-3 ( 1:1 , 000 , Invitrogen ) . Coverslips were mounted with Vectashield . Slides were imaged on a Zeiss LSM710 confocal microscope and images are z-stack composites encompassing the entire section . Aggregation rate was determined using ImageJ software: ‘Particle Analysis’ was used to determine the number of TOTO-3-stained nuclei and the ‘Find Maxima’ tool was used to locate all AR aggregates . The ratio of nuclei containing aggregates out of total counted was recorded for several sections and averaged over individual mice by genotype . Primary motor neurons were prepared from embryonic day 13 ( E13 ) mouse embryos as described previously with slight modification ( Gingras et al . , 2007; Montie et al . , 2009 ) . Briefly , spinal cords were dissected in ice-cold L15 medium ( Gibco ) , dissociated in 0 . 05% trypsin , and plated on poly-D-lysine- and laminin-coated plates . After 7 days , cells were transfected with GFP-AR120Q and/or FLAG-Nlk-WT plasmids by the Calcium-phosphate method . On the next day , 10 μM DHT was added to the medium . At DIV9 , cells were fixed and subjected to immunofluorescence . Primary antibodies used were mouse anti-FLAG antibody ( 1:1000 , Sigma ) , rabbit anti-GFP antibody ( 1:1000 , Abcam ) , and goat anti-ChAT antibody ( 1:100 , Calbiochem ) . Appropriate Alexa secondary antibodies ( Invitrogen ) were used to visualize the proteins . The number of aggregate-containing cells per total GFP-positive cells was counted manually . NSC-34 cells were transfected with an ARE-luciferase reporter , a pRL-TK renilla luciferase reporter and any other indicated constructs using lipofectamine 2000 and treated with DHT . 24 hr after DHT treatment , cells were lysed and subjected to a dual-luciferase assay using a Promega kit according to the manufacturer's instructions . Luciferase activity was measured using a Promega GloMax 20/20 luminometer and associated software . The ratio of the luciferase activity values was recorded for each sample and normalized to control samples in each case . Each experimental trial was performed in triplicate , and ratios were averaged over multiple trials . Protein expression was confirmed by immunoblot . Mammalian two-hybrid assays were carried out as described for the other dual-luciferase assays , except a Gal4-luciferase reporter was used in place of the ARE-luciferase construct and cells were transfected with the VP16- and Gal4-DBD-fused protein constructs as indicated . Unless otherwise noted , statistical significance between two sample sets was determined by the Student's t-test using a two-tailed distribution and assuming unequal variance . Statistical significance between multiple sample sets was determined by one-way ANOVA using Tukey's post-hoc HSD test to compare individual group differences . Statistics were calculated using Microsoft Excel and GraphPad Prism software . | Spinal and bulbar muscular atrophy ( SBMA ) is an inherited disease that eventually leads to degeneration in motor neurons and weakness in muscles . It is caused by a specific genetic mutation in the gene that encodes the androgen receptor protein , which leads to the production of a mutant protein that is larger than normal . Similar mutations in other genes can lead to the development of other so-called ‘polyglutamine’ diseases such as Huntington's disease and spinocerebellar ataxia . However , the precise details of how these mutations lead to disease symptoms are not known , and there are currently no effective ways of treating these conditions . Previous research has shown that an enzyme called Nemo-like kinase ( or NLK for short ) regulates the normal androgen receptor in cancer cells . NLK has kinase activity , that is , it adds phosphate molecules to other proteins to regulate their activity . Todd et al . used human cells , fruit flies , and mice as model systems to investigate whether NLK is involved in the development of SBMA . The experiments show that NLK promotes the development of features associated with SBMA in all three models . The kinase activity of NLK is required for these features to develop . Todd et al . also found that NLK can bind to and add phosphate molecules to the mutant version of the androgen receptor protein . This causes the mutant androgen receptor proteins to accumulate and increases the ability of the mutant proteins to activate particular genes . Todd et al . 's findings suggest that NLK promotes the development of SBMA by interacting with the mutant androgen receptor . Previous studies have shown that NLK is able to modulate the development of spinocerebellar ataxia type 1 , which suggests that NLK may also play an important role in other polyglutamine diseases . The next challenge will be to fully understand the role of NLK in these diseases , which may aid future efforts to develop new treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Nemo-like kinase is a novel regulator of spinal and bulbar muscular atrophy |
Diverse features of sensory stimuli are selectively processed in distinct brain areas . The relative recruitment of inhibitory and excitatory neurons within an area controls the gain of neurons for appropriate stimulus coding . We examined how such a balance of inhibition and excitation is differentially recruited across multiple levels of a cortical hierarchy by mapping the locations and strengths of synaptic inputs to pyramidal and parvalbumin ( PV ) -expressing neurons in feedforward and feedback pathways interconnecting primary ( V1 ) and two higher visual areas . While interareal excitation was stronger in PV than in pyramidal neurons in all layer 2/3 pathways , we observed a gradual scaling down of the inhibition/excitation ratio from the most feedforward to the most feedback pathway . Our results indicate that interareal gain control depends on the hierarchical position of the source and the target , the direction of information flow through the network , and the laminar location of target neurons .
Visual perception and visually guided actions result from the coordinated neuronal communication between multiple , functionally diverse areas of visual cortex . Within visual cortex , interareal communication is achieved through the axons of pyramidal ( Pyr ) cells carrying feedforward ( FF ) information from lower to higher areas and feedback ( FB ) signals through ‘top-down’ connections descending across the hierarchy of visual areas ( Felleman and Van Essen , 1991; Coogan and Burkhalter , 1993 ) . How neurons within such a highly interconnected network and increasing densities of inputs at higher levels of the cortical hierarchy ( Wang et al . , 2012; Elston , 2003 ) maintain stimulus-specificity without saturating their spike output has been studied by modelling the effects of inhibitory synaptic inputs and by recording the balance of excitation and inhibition in local networks of sensory cortex ( Shadlen and Newsome , 1998; Pouille et al . , 2009 ) . However , the rules by which the inhibition/excitation ( I/E ) balance changes along processing pathways from early to deep stages of the brain and back are incompletely understood . In the rodent visual system , interareal FF and FB pathways communicate through excitatory synapses contacting Pyr and GABAergic neurons ( Johnson and Burkhalter , 1996 ) . In the target area , both cell types are reciprocally connected by a fine-scale circuit embedded within the global network ( Yoshimura and Callaway , 2005; Jiang et al . , 2015; Pfeffer et al . , 2013 ) . Although interareal FF and FB connections terminate on multiple types of GABAergic neurons , most of them synapse onto PV-expressing fast-spiking interneurons ( Gonchar and Burkhalter , 1999; Gonchar and Burkhalter , 2003; Hangya et al . , 2014 ) , which provide feedforward inhibition ( FFI ) to local Pyr cells ( Dong et al . , 2004; Yang et al . , 2013 ) . FFI is a common functional motif throughout the brain , capable of regulating the I/E balance and thereby influencing the gain , the integration window , and the temporal precision of inputs ( Shadlen and Newsome , 1998; Atallah et al . , 2012; Gabernet et al . , 2005; Cardin et al . , 2010 ) . Similar to the thalamocortical and local circuits in mouse barrel cortex and in V1 ( Atallah et al . , 2012; Gabernet et al . , 2005 ) , FFI is also involved in interareal communication across the visual cortical hierarchy ( Dong et al . , 2004 ) . In fact , our studies in mouse visual cortex have shown that FF input to Pyr cells is more strongly counterbalanced by inhibition than FB input , suggesting pathway-specific differences in the gain and dynamic range of recurrent excitation involved in cortical computations ( Yang et al . , 2013; Atallah et al . , 2012; Okun and Lampl , 2008 ) . Here , we demonstrate that higher relative inhibition in FF than in FB pathways is part of a more general rule of cortico-cortical communication . We studied the strengths of FF and FB inputs interconnecting mouse V1 with Pyr and PV neurons in the extrastriate area PM ( posteromedial ) situated high in the hierarchy , and compared the I/E balance with input to and from the hierarchically intermediate area LM ( lateromedial ) . Using whole-cell patch clamp recordings and laser-scanning photostimulation of Channelrhodopsin-2 ( ChR2 ) -expressing FF and FB connections in acute cortical slices , we show that the relative strength of excitatory input to Pyr and PV cells is pathway-specific and depends on the position of the source and target areas within the hierarchy . Interareal inputs to PV interneurons in upper layers , but not in lower layers , are stronger than to Pyr cells , and the asymmetry of I/E balance was greater for V1 to PM than for LM to PM connections , suggesting weaker FFI by inputs from hierarchically higher areas . In support of the notion that pathways from sources deeper in the brain , which deliver input that vary over a narrower range than input from the outside world , would require lower levels of inhibitory control , we found that FFI is weaker in FB connections and weakest in FB input to V1 , at the bottom of the hierarchy . Our findings therefore suggest that in FF and FB pathways targeting neurons in layer 2/3 ( L2/3 ) , excitation is more strongly counterbalanced by inhibition and that the imbalance is gradually rectified according to hierarchical distance from the most feedforward to the most feedback .
To study interareal inhibition across different levels of the cortical architecture , we first asked whether the visual areas V1 , LM and PM lie at distinct levels of an interconnected hierarchical network . To do this we traced the outputs from V1 , LM and PM with the anterograde tracer biotinylated dextran amine ( BDA ) and studied the laminar patterns of axon terminals in each of the visual cortical target areas: V1 , LM , POR ( postrhinal ) , AL ( anterolateral ) , P ( posterior ) , LI ( laterointermediate ) , PM , AM ( anteromedial ) , RL ( rostrolateral ) and A ( anterior ) ( Figure 1 , Figure 1—figure supplement 1–3 ) . Projections were assigned to areas by their locations relative to retrogradely bisbenzimide-labeled callosal landmarks ( Wang and Burkhalter , 2007 ) , which we imaged in situ before sectioning the brain , and by their relative positions to each other ( Figure 1a ) . Sections were numbered from the posterior pole of cortex so that the callosal landmarks seen in the coronal plane could be matched to specific locations of the in situ pattern . BDA labeled fibers were then superimposed onto the callosal pattern observed in the same section , and projections were assigned to specific areas according to the map by Wang and Burkhalter ( Wang and Burkhalter , 2007 ) ( Figure 1a ) . Optical density maps of projections showed striking laminar differences ( Figure 1b ) . Although most projections involved L1-6 , inputs from V1 consistently showed dense terminations in L2-4 of each of the higher areas with much sparser projections in L1 . In contrast , projections from both LM and PM strongly targeted L1 of V1 while weakly targeting L2-4 ( Figure 1—figure supplement 2 and 3 ) . The selective targeting of L1 by FB projections is consistent with observations in other species ( Felleman and Van Essen , 1991; Coogan and Burkhalter , 1993; Rockland and Virga , 1989; Henry et al . , 1991 ) . To analyze these patterns quantitatively , we computed the density ratio ( DR ) of terminations in L2-4 to that in L1 of axons from V1 , LM and PM to each of the other nine areas and plotted DRs in a 3 × 9 matrix ( Figure 1c ) . We reasoned that FF projections from lower areas would , on average , have a higher DR than FB projections from higher areas . The matrix showed that the average DRs in all targets of V1 were >2 . 52 ± 0 . 31 , whereas the DRs for projections to V1 were <0 . 72 ± 0 . 08 ( Figure 1c ) . Pairwise comparisons of average DRs of projections from each of V1 , LM , and PM to the other areas showed significant ( p<0 . 001 , Mann-Whitney U test ) differences , demonstrating that V1 , LM and PM are at distinct hierarchical levels , with LM at the intermediate level between V1 and PM ( Figure 1d , e ) . 10 . 7554/eLife . 19332 . 003Figure 1 . Hierarchy between V1 , LM , and PM . ( a ) Rostrocaudal series of coronal slices through the left hemisphere showing anterogradely labelled axonal projections ( yellow/orange ) after V1 was injected with BDA . Retrogradely labelled callosally projecting neurons ( light cyan ) , upon injection of bisbenzimide in the right hemisphere , act as landmarks for identification of areas ( Wang and Burkhalter , 2007 ) . Numbers denote sections corresponding to the positions shown in inset . See Figure 1—figure supplement 1 for higher magnification of areas within dotted squares . Projection to LM adjacent to LI in section 33 is indicated . Arrowhead indicates a region in V1 near the injected site . In situ image of retrograde bisbenzimide-labelled callosally projecting neurons in the left hemisphere . Injection site in V1 ( asterisk ) and positions of coronal slices shown above are indicated . Scale bars , 1 mm . ( b ) Optical density of axonal projections in the target areas of the indicated pathways , normalized to peak density . Contours connect regions with similar optical densities . Arrowheads denote the edge of the slice and edge artifacts due to interpolation of optical density with dark background . ( c ) Color-coded heat map of L2-4:L1 density ratio ( DR ) for each of 25 distinct cortico-cortical connections . Blocks in grey indicate projections that were too weak to analyze . V1 exhibits the highest DRs , and PM the lowest , indicating the relative hierarchical positions of the areas . ( d ) The mean DR for all target areas is highest for V1 , intermediate for LM , and lowest for PM; ***p<0 . 001 , Mann-Whitney U-test . ( e ) Our schematic interpretation of the hierarchy of V1 , LM , and PM in visual processing . Feedforward pathways are denoted in green , feedback in red . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 00310 . 7554/eLife . 19332 . 004Figure 1—figure supplement 1 . Darkfield images of the termination patterns of BDA-labelled axonal projections from V1 to LM , LI , P , POR , AL , PM , RL , AM and A . Images are taken from the boxed regions shown in 1A . All projections are FF , which target L2-4 more strongly than L1 . Scale bar , 0 . 5 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 00410 . 7554/eLife . 19332 . 005Figure 1—figure supplement 2 . Coronal sections showing anterogradely labelled axons ( yellow/orange ) from LM to V1 , LI , P , POR , AL , PM , RL , AM and A , upon BDA injection into LM . Calossally projecting neurons ( light cyan ) are labelled retrogradely after injection of bisbenzimide into the opposite hemisphere . Boxed regions show projections to each of the nine target areas . Asterisk in coronal section 28 denotes injection site in LM . Scale bar , 1 mm . In situ image of left hemisphere shows retrogradely bizbenzimide-labelled neurons marking callosal landmarks . Injection site in LM ( asterisk ) and numbers of coronal slices shown above are indicated . Scale bar , 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 00510 . 7554/eLife . 19332 . 006Figure 1—figure supplement 3 . Axonal projections ( yellow/orange ) from PM to V1 , LM , LI , P , POR , AL , RL , AM and A , upon BDA injection into PM . Calossally projecting neurons ( light blue , labeled with bisbenzimide ) provide landmarks for areal identification ( Wang and Burkhalter , 2007 ) . Boxed regions show projections to target areas . Asterisk in section 48 denotes a region in PM adjacent to the injection site . Inset: Left hemisphere in situ shows injection site in PM ( asterisk ) and positions of coronal slices shown above . Scale bars , 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 006 Cortico-cortical inhibition between areas involves both , the initial excitation of interneurons by long-range axonal projections of Pyr cells , and the disynaptic inhibition of Pyr cells by interneurons . As a first step in the analysis of the recruitment of interareal inhibition , we first confirmed the role of PV interneurons in inhibiting neighboring Pyr cells . These experiments were performed in area PM , one of the targets of V1 , in acute slices from mice in which PV cells expressed tdTomato ( tdT ) ( Figure 2—figure supplement 1a ) . The axonal projections from V1 to PM were labeled by anterograde tracing with adeno associated virus ( AAV ) expressing a ChR2-Venus fusion protein ( Petreanu et al . , 2009 ) ( Figure 2—figure supplement 1a–f ) . PM is the posterior projection zone medial to the densely type 2 muscarinic acetylcholine receptor ( M2 ) -expressing area V1 ( Wang et al . , 2011; Ji et al . , 2015 ) ( Figure 2—figure supplement 1d , e ) . Similar to other cortical areas , PM contained tdT-PV cell bodies in L2-6 , with axons and dendrites reaching into L1 ( Figure 2—figure supplement 1f , Figure 2—figure supplement 2a ) . We performed paired recordings to examine whether increasing the excitation of PV cells results in stronger inhibition of neighboring synaptically connected Pyr cells ( Figure 2—figure supplement 2c ) . To do this , we evoked action potentials by injecting current steps ( 100 , 200 , 300 , and 400 pA; 50 ms; Figure 2—figure supplement 2c–f ) into PV cells and recorded inhibitory postsynaptic currents ( IPSCs ) in connected Pyr cells . Similar to recordings in other cortical areas ( Pfeffer et al . , 2013; Packer and Yuste , 2011 ) we found a high connection probability . Recordings from PV and Pyr neurons within ~100 µm of each other resulted in 11/13 ( 84 . 6% ) and 7/15 ( 46 . 6% ) synaptically connected pairs in L2/3 and L5 , respectively ( Figure 2—figure supplement 2g ) . In both layers , increasing the firing of PV cells resulted in larger IPSCs in Pyr cells ( n = 11 pairs in L2/3 , 7 pairs in L5; Figure 2—figure supplement 2d–f ) . The increase in inhibition was due to both , the increased probability of PV cells to reach spike threshold , as well as increased spiking . These findings reveal a local subnetwork that is likely tapped by interareal connections for FFI of Pyr cells in target areas . 10 . 7554/eLife . 19332 . 007Figure 2 . Subcellular ChR2-assisted mapping of V1→PM connections to L2/3 PV and Pyr cells . ( a ) Coronal slices showing injection ( left ) and target ( right ) sites two weeks after the injection of AAV2/1 . CAG . ChR2-Venus . WPRE . SV40 into V1 . Scale bar , 500 µm . Select target areas indicated in right panel . SC , superior colliculus . ( b ) Schematic of laser-scanning photostimulation of ChR2-expressing axon terminals during whole-cell recording of a biocytin-filled neuron . TTX and 4-AP are added to the bath solution , and the blue laser is delivered successively one spot at a time in a grid pattern separated by 75 µm . ( c ) EPSCssCRACM in a PV ( left ) and a neighboring Pyr ( right ) cell upon photostimulation . Grey shapes denote the location of the cell body of the recorded neuron . ( d ) Heat map of mean EPSCs within 75 µs after photostimulation for the EPSCs in 3c . Reconstructions of respective biocytin-filled neurons are superimposed on heat map . ( e ) Average heat map of 14 neighboring PV-Pyr cell pairs in L2/3 receiving V1→PM input , normalized to largest pixel value between a pair . PV cells receive substantially stronger input . ( f ) Scatter plot denoting the relative input strengths to 14 PV-Pyr cell pairs . Each data point represents a pair with the respective EPSCs in the PV ( vertical axis ) and the Pyr ( horizontal axis ) cell . The total EPSC in PV cells is significantly larger than that in neighboring Pyr cells ( p<0 . 001 , Wilcoxon signed-rank test ) . Solid black line: mean slope , blue line: mean slope after normalizing currents to mean cell conductance . ( g ) The mean time to peak of EPSCs after photostimulation is larger in Pyr cells than in PV cells ( *p<0 . 05 , paired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 00710 . 7554/eLife . 19332 . 008Figure 2—figure supplement 1 . V1→PM pathway in a PV-tdT mouse . ( a−e ) Image of coronal section two weeks after the injection of AAV2/1 . CAG . ChR2-Venus . WPRE . SV40 into V1 . Slice includes areas PM , AL , and V1 . PV cells in red ( a , c ) , ChR2-Venus-expressing axons in green ( b , c ) . Merged image in ( c ) . Dotted lines demarcate AL/V1 and V1/PM boundaries indicated by the sharp decline of M2 expression between V1 and surrounding areas ( d ) . V1 is characterized by a thick band of M2 expression ( purple ) in L4 , showing that the axonal terminations ( green ) lie outside V1 ( d , e ) . ( f ) Higher magnification view of PM from ( c ) with layers indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 00810 . 7554/eLife . 19332 . 009Figure 2—figure supplement 2 . Paired recordings of excitation-dependent , PV cell-mediated inhibition of Pyr cells . ( a ) Coronal section through V1 and PM of a PV-Cre × Ai9 mouse in which PV cells express tdT ( red ) . Scale bar , 200 µm . Inset: Higher magnification of the boxed region shows a high density of tdT-expressing dendrites in L1 even though PV cell bodies are not found in this layer . ( b ) Left , PV cell identified by tdT expression targeted for whole-cell recordings . Scale bar , 20 µm . Middle , same cell as in left panel , imaged under DIC-IR optics . PV cell shows a non-adapting , fast-spiking firing pattern ( inset , red trace ) upon current injection . Right , A Pyr cell under infrared optics exhibits an adapting spiking physiology ( blue trace ) upon current injection . ( c ) Schematic of paired recordings of a PV ( red ) and a Pyr ( blue ) cell . Successively increasing current steps ( 100 , 200 , 300 and 400 pA ) were injected into the PV cell under current clamp , and inhibitory currents ( IPSCs ) were recorded in the Pyr cell held at 0 mV under voltage clamp . ( d ) Example trace in a L2/3 PV-Pyr connected pair in PM . Increasing current injections into the PV cell results in stronger inhibitory drive to the Pyr cell . ( e ) Example traces of a connected PV-Pyr cell pair in L5 of PM . ( f ) Pooled data from connected PV-Pyr pairs in L2/3 ( left ) and L5 ( right ) show that increasing the excitation of PV cells results in stronger inhibition of synaptically connected Pyr cells ( p<0 . 001 for both sets of data; ANOVA ) . Mean IPSCs measured over 75 ms after start of current step . ( g ) Probability of a PV cell connected to a neighboring Pyr cell ( <100 µm ) in L2/3 and L5 of PM . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 009 Because the level of PV cell excitation determines the feedforward inhibitory drive to synaptically connected Pyr cells , we examined the strength of excitatory inputs to neighboring PV and Pyr cells by different pathways . We performed subcellular ChR2-assisted circuit mapping ( sCRACM ) ( Yang et al . , 2013; Petreanu et al . , 2009; Mao et al . , 2011 ) in acute slices of visual cortex to measure the input strength and the laminar location of interareal connections to PV and Pyr cells in different pathways . To study connections in the FFV1→PM pathway , we expressed ChR2-Venus in axons projecting from V1 to PM , and recorded excitatory postsynaptic currents ( EPSCs ) from PV and Pyr cells centered at the peak of the PM projection ( Figure 2a ) . Photostimulation of ChR2-expressing axon terminals was achieved by a 473 nm laser delivered one spot at a time in a grid pattern separated by 75 µm ( Figure 2b ) . Recordings were performed in the presence of 1 µm TTX ( tetrodotoxin ) and 50 µm 4-AP ( 4-aminopyridine ) in the bath to block polysynaptic currents and repolarization of axon terminals , respectively . Resulting EPSCs were measured by whole cell patch clamp recordings from PV and Pyr neurons voltage-clamped at −70 mV ( Figure 2c ) . We compared EPSCs between PV and Pyr neurons whose cell bodies were in the same layer of the same slice , within ~100 µm of each other . In the L2/3 FFV1→PM pathway , EPSCs recorded from PV cells were larger than those from Pyr cells ( Figure 2c–f ) . On average , the largest EPSCs were evoked from synaptic inputs to proximal dendrites at the bottom of L2/3 whereas inputs to distal dendrites were weaker ( Figure 2d , e ) . The mean total current in PV cells was 12 . 85 ± 4 . 48-fold stronger than that in neighboring Pyr cells ( p<0 . 001 , n = 14 pairs ) . To illustrate the relative excitation of PV and Pyr cells , we plotted the total EPSC in each PV cell against the total EPSC in its Pyr neighbor , and measured the mean slope for all such pairs in this pathway ( Figure 2f ) . We also computed the mean slope after normalizing the EPSC to the mean cell conductance to control for cell size . Similar to observations in thalamocortical and local circuits , the time to peak of EPSCs was significantly shorter in PV than in Pyr cells ( Figure 2g; n = 14 pairs , p<0 . 05 , paired t-test ) , consistent with the notion that PV interneurons can be recruited more rapidly than Pyr neurons in diverse brain areas ( Hull et al . , 2009; Povysheva et al . , 2006 ) . We next asked whether connections to PV and Pyr cells in the FBPM→V1 pathway showed a different I/E balance . Recordings in V1 showed that similar to PM , EPSCs in PV cells were larger and faster than in Pyr cells ( Figure 3a–f ) . In contrast to the FFV1→PM pathway , however , the excitation of PV cells in the FBPM→V1 pathway was only 1 . 93 ± 0 . 44-fold stronger than Pyr cell excitation . Thus , the excitation of PV cells , relative to that of neighboring Pyr cells , was weaker in the FBPM→V1 than in the FFV1→PM pathway ( Figure 3g; n = 14 pairs for FFV1→PM , n = 21 pairs for FBPM→V1; p<0 . 001 ) , similar to previous observations in the L2/3 FFV1→LM and FBLM→V1 pathways ( Yang et al . , 2013 ) . The larger EPSCs in PV cells could be a result of either a higher density of excitatory input or due to a larger area over which individual PV cells are contacted by interareal projections , or both . We therefore measured the mean EPSC per pixel and the total area over which each cell type exhibited measurable EPSCs ( Figure 3—figure supplement 1 ) . In the FFV1→PM pathway , PV cells exhibited larger EPSCs per pixel than Pyr cells ( Figure 3—figure supplement 1a ) as well as received input over a larger area ( Figure 3—figure supplement 1b ) . In contrast , in the FBPM→V1 pathway , the mean EPSC per pixel between the two cell types were not significantly different ( Figure 3—figure supplement 1c ) , indicating that the larger total EPSCs in PV cells were the result of PV cells receiving excitatory input over a larger area ( Figure 3—figure supplement 1d ) . 10 . 7554/eLife . 19332 . 010Figure 3 . Lower I/E balance in PM→V1 pathway . ( a ) Coronal slices showing AAV2/1 . CAG . ChR2-Venus ( green ) injection in PM ( top ) and axonal labelling in target areas ( bottom ) of a PV-tdT ( red ) mouse . Scale bar , 1 mm . ( b ) EPSCssCRACM in a pair of neighboring PV ( left ) and Pyr cells ( right ) in V1 . ( c ) Heat map of the currents in 4b superimposed with biocytin-filled neurons ( white ) . Note significant input into L1 of both cell types . ( d ) PV cells , on average , exhibit larger EPSCssCRACM than neighboring Pyr cells in the PM→V1 pathway ( p<0 . 02 , Wilcoxon signed-rank test ) . Solid black line: mean slope of data points; blue line: mean slope after normalization to cell conductance . ( e ) Normalized , mean heat map of all L2/3 pairs in the FBPM→V1 pathway . ( f ) EPSCs are faster in L2/3 PV than in neighboring Pyr cells upon stimulation of FBPM→V1 axon terminals ( *p<0 . 001 , paired t-test ) . ( g ) The interareal excitation of PV cells , normalized to that of neighboring Pyr cells , is on average larger in the FFV1→PM than in the FBPM→V1 pathway ( ***p<0 . 001 , Mann-Whitney U-test ) . ( h ) Total currents in each row of the 8 × 16 grid for FFV1→PM and FBPM→V1 pathways plotted against relative position of each of the 16 rows . EPSCs normalized to total EPSC in each cell-type . pia , pia mater; wm , white matter . ( i ) Interareal input to L1 is stronger in the FBPM→V1 than in the FFV1→PM pathway in both cell types ( **p<0 . 01 , ***p<0 . 001 , Mann-Whitney U-test ) . L1 input was calculated as the mean of the total input to each row of the 8 × 16 grid that resided in L1 , presented as the percentage of the total input to the neuron . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 01010 . 7554/eLife . 19332 . 011Figure 3—figure supplement 1 . Analyses of inputs to L2/3 PV and Pyr neurons in the reciprocal pathways between V1 and PM . ( a ) The EPSC per 75 × 75 µm pixel , calculated as pA/µm ( 2 ) , is larger in PV than in Pyr cells in FFV1→PM . ( b ) Individual PV cells receive interareal FFV1→PM input over a larger area than Pyr cells ( ***p<0 . 001 ) . ( c ) The EPSC per pixel in PV and Pyr cells is not significantly different in the FBPM→V1 pathway . ( d ) PV cells receive input over a larger area than Pyr cells in the FBPM→V1 pathway ( *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 011 The laminar organization of interareal input to individual neurons was significantly different for the two pathways . Unlike FFV1→PM projections , FB axons from PM provided strong inputs to L1 of V1 . We quantified L1 input by measuring the total pixel values in each row of the photostimulation grid pattern and plotting EPSCs against distance from the pial surface ( Figure 3h , i ) . The values are percentages in each row of the total EPSC in the respective cell type . Consistent with the distribution of projections ( Figure 1b ) the proportion of inputs to L1 , relative to total EPSCs , was larger in the FBPM→ V1 pathway than in the FFV1→PM pathway ( Figure 3i; p<0 . 01 for both PV and Pyr cells ) . It must be noted , however , that due to dendritic filtering of signals , EPSCs at distal dendrites are attenuated more than those near the soma . Thus , the proportion of L1 inputs to the total current may be an underestimate . Do connections originating from higher areas follow the same normalization rules as those from V1 ? We addressed this question with sCRACM experiments in L2/3 of FFLM→PM and FBPM→LM pathways ( Figure 4 ) . Similar to FF and FB pathways between V1 and PM , EPSCs were larger in PV cells than in Pyr cells in both pathways ( Figure 4a–f ) . In the FFLM→PM pathway , the mean total current in PV cells was 3 . 62 ± 0 . 75-fold larger than in neighboring Pyr cells ( n = 15 pairs , p<0 . 02; Figure 4c , g ) . Inputs to both cell types were maximal at proximal dendrites in L3 and 4 , but weak in L1 and L2 ( Figure 4a , b , i; Figure 4—figure supplement 1 ) . In the FBPM→LM pathway , the mean total EPSC to PV cells was 3 . 77 ± 0 . 81-fold the total EPSC in neighboring Pyr cells , with substantial input into L1 ( n = 18 pairs , p<0 . 01; Figure 4d–f , h , i; Figure 4—figure supplement 1 ) . Similar to connections between V1 and PM , EPSCs were faster in PV than in Pyr cells in both FFLM→PM and FBPM→LM pathways ( Figure 4—figure supplement 2d ) . Hence , the faster activation of PV cells appears to be a general rule for FFI provided by long-range connections ( Hull et al . , 2009; Povysheva et al . , 2006 ) . 10 . 7554/eLife . 19332 . 012Figure 4 . Interareal recruitment of L2/3 PV cells depends on the pathway and the hierarchical distance between areas . ( a ) EPSCssCRACM in a L2/3 PV ( left ) and Pyr ( right ) cell in the FFLM→PM pathway . ( b ) sCRACM map of EPSCs in 5a with reconstructed neuron positions . ( c ) Scatter plot of all PV-Pyr cell pairs in the L2/3 FFLM→PM pathway . PV cells exhibit larger currents than Pyr cells . ( d–f ) Similar data as in ( a–c ) but for the L2/3 FBPM→LM pathway . Note stronger L1 input in this pathway . ( g ) PV cell excitation , normalized to that of a neighboring Pyr cell , is stronger in the FFV1→PM than in the hierarchically shorter FFLM→PM pathway ( *p<0 . 05 , Mann-Whiteney U-test ) . ( h ) Normalized PV cell excitation is stronger in the FBPM→LM than in the hierarchically longer FBPM→V1 pathway ( *p<0 . 05 , Mann-Whiteney U-test ) . ( i ) Normalized plot of the total current in each row of the 8 × 16 grid , plotted against row position . EPSCs normalized to total current in each cell-type . ( j ) The total EPSCssCRACM in a PV cell , normalized to the total EPSCssCRACM in a neighboring Pyr cell , depends on the directionality of the pathway and hierarchical distance between areas . Red boxes represent data describing connections between V1 and LM from Yang et al . ( 2013 ) . ***p<0 . 001 , Kruskal-Wallis test . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 01210 . 7554/eLife . 19332 . 013Figure 4—figure supplement 1 . Stronger input to L1 in the PM-to-LM than in the LM-to-PM pathway . ( a ) Normalized heat maps of average EPSCs in L2/3 PV cells in the FFLM→PM and FBPM→LM pathways . EPSC values normalized to peak pixel in each panel . ( b ) Interareal input to L1 is stronger for PV cells in the FBPM→LM pathway than in FFLM→PM . L1 input calculated as average input to each row of the 8 × 16 grid that resided in L1 , normalized to total EPSC recorded from the cell . ( c ) Normalized heat maps of average EPSCs in L2/3 Pyr cells in FFLM→PM and FBPM→LM . EPSC values normalized to peak value within panel . ( d ) Interareal input to L1 is stronger for Pyr cells in the FBPM→LM than in the FFLM→PM pathway . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 01310 . 7554/eLife . 19332 . 014Figure 4—figure supplement 2 . Comparisons of strengths , extent , and rise times of inputs in L2/3 pathways . ( a ) Total EPSC per row plotted against row position for L2/3 PV and Pyr cells in the FFV1→PM , FFLM→PM , FBPM→LM and FBPM→V1 pathways . Strongest EPSCs in both cell types are observed in the LM→PM pathway , with weak EPSCs to Pyr cells in the FFV1→PM pathway . ( b ) FFLM→PM inputs result in the strongest current density ( measured as EPSC/pixel , converted to pA/µm [Coogan and Burkhalter , 1993] ) among all examined L2/3 pathways for both PV and Pyr cells ( #p<0 . 05 against each of the other three pathways ) ( c ) The total area over which individual PV cells receive input is not significantly different for PV cells . Pyr cells in the FFV1→PM pathway receive input over the smallest area ( #p<0 . 05 against each of the other pathways ) . ( d ) The mean time to peak for interareal EPSCs is smaller in PV than in Pyr cells in both FFLM→PM and FBPM→LM pathways . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 014 While PV cells received stronger excitatory inputs than Pyr cells in all four L2/3 pathways described here , the difference in the relative excitation of PV and Pyr was bigger in the FFV1→PM originating at the bottom and terminating at the top of the hierarchy than in the FFLM→PM pathway originating from the higher area , LM ( Figure 4g ) . In contrast , in the FBPM→V1 pathway , the difference was smaller for connections originating at the top and terminating at the bottom of the hierarchy than for terminations at an intermediate level in LM ( Figure 4h ) . These relationships are evident in a significant ( p<0 . 001 , Kruskal-Wallis test ) decrease of the EPSCPV/EPSCPyr ratios , when pathways are ordered by hierarchical distance from the most feedforward to the most feedback ( Figures 1e , 4j ) . The plot suggests that inhibitory counterbalance to long-range excitation is gradually adjusted depending on the hierarchical location of the source and target areas . Although the total input to PV and Pyr cells differed across pathways , the pathway-specific normalization was independent of the absolute strength of the excitatory input so that the EPSCPV/EPSCPyr ratios , and not the absolute values of EPSCs in PV and Pyr cells , show a hierarchy-dependent variation ( Figure 4—figure supplement 2a–c ) . We next asked if interareal inputs to L5 neurons follow a similar physiological connectivity rule as those to L2/3 . Unlike in L2/3 , the EPSCs recorded in L5 PV and Pyr cells upon stimulation of FFV1→PM and FFLM→PM axons were not significantly different ( Figure 5a–f ) . The relative excitation of L5 PV cells , expressed by the EPSCPV/EPSCPyr ratio , was smaller than that observed in L2/3 for both FF pathways ( Figure 5g , h ) . While we did not observe EPSCs in L1 for L5 Pyr cells , likely due to attenuation of signals by dendritic filtering , we detected significant input to L2-4 . In particular , L5 Pyr cells in FFLM→PM exhibited large EPSCs at apical dendrites in L2-4 , hundreds of microns distal to the cell body ( Figure 5d , e , i , j ) . The proportion of such L2-4 inputs to the total EPSC was higher in FFLM→PM than in FFV1→PM for L5 Pyr cells but not for PV cells , whose input distributions were similar in both pathways ( Figure 5i , j ) . Thus depending on the source of long-range synaptic input , L5 Pyr cells in PM receive FF input at different locations of their dendritic arbor . 10 . 7554/eLife . 19332 . 015Figure 5 . FF input to L5 neurons . ( a ) FFV1→PM EPSCssCRACM in a pair of neighboring L5 PV ( left ) and Pyr ( right ) cells . ( b ) Heat map of EPSCs in 6a superimposed with respective biocytin-filled L5 neurons . ( c ) Scatter plot , as previously described , of EPSCssCRACM in PV and Pyr cell pairs in L5 FFV1→PM . The total current in PV and Pyr cells were not significantly different . ( d–f ) Similar data as in Figure 6a—c but for L5 FFLM→PM . ( g , h ) PV cell excitation , normalized to the excitation of a neighboring Pyr cell , is stronger in L2/3 than in L5 for both V1→PM ( g ) and LM→PM pathways ( h ) . ( i ) Total EPSC in each row of the 8 × 16 grid normalized to total current recorded , plotted against row position ( 16 rows ) . Note that L5 PV cells do not show significant differences in the laminar distribution of EPSCs in the two pathways , but L5 Pyr cells receive more input in the upper layers from LM than from V1 . ( j ) Interareal input in L2-4 for L5 PV cells ( left ) in PM is not significantly different for the two pathways . L5 Pyr cells ( right ) receive more L2-4 input in the FFV1→PM than in the FFLM→PM pathway . L2-4 input calculated as the average EPSC in each row that resided in L2-4 , shown as the percentage of the total EPSC in the cell ( ***p<0 . 001 , Mann-Whitney U-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 015 Finally , we examined the two FB pathways projecting from PM to L5 neurons in V1 and LM respectively . Activation of either FBPM→V1 or the FBPM→LM projecting axons resulted in EPSCs of similar magnitudes in neighboring PV and Pyr cells ( Figure 6a–f ) , with the strongest inputs primarily recorded at proximal dendrites in L5 for both cell types ( Figure 6g ) . These results suggest that the stronger activation of PV cells observed in L2/3 is absent in L5 . Consistent with this observation , we found no significant difference between the EPSCPV/EPSCPyr ratios for L5 cell pairs for the different pathways ( Figure 6h ) , suggesting equal potency of FFI among these pathways regardless of whether they are FF or FB . Similar to L2/3 , however , the EPSCs in PV cells showed faster rise times than those in Pyr cells in all L5 pathways ( Figure 6i ) . 10 . 7554/eLife . 19332 . 016Figure 6 . FB input to L5 neurons . ( a ) FBPM→LM EPSCssCRACMin a pair of neighboring L5 PV ( left ) and Pyr ( right ) cells . ( b ) Heat map of EPSCs from 7a superimposed with the respective biocytin-filled L5 neurons . ( c ) Scatter plot of all L5 PV-Pyr neuron pairs receiving input from FBPM→LM . Total EPSC in the two cell types are not significantly different . ( d–f ) Similar data as 7a-c but for the FBPM→V1 pathway . ( g ) Total EPSC in each row of the stimulation grid plotted against row position . The grids of the two different pathways are aligned to pial surface . ( h ) EPSCs in PV cells normalized to EPSCs in neighboring Pyr cells ( EPSCPV/EPSCPyr ) for all L5 pathways arranged from most FF to most FB . Unlike in L2/3 , the EPSCPV/EPSCPyr ratios in L5 are not significantly different in different pathways ( p>0 . 2 , Kruskal-Wallis test ) . Red boxes describe data from Yang et al . ( 2013 ) . ( i ) Interareal EPSCs are faster in PV than in Pyr cells in all L5 pathways ( *p<0 . 05 , ***p<0 . 001 , paired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19332 . 016
We have mapped input strengths to inhibitory PV and excitatory Pyr cells in diverse pathways interconnecting three visual cortical areas with distinct spatiotemporal sensitivities and specialized functions ( Marshel et al . , 2011; Andermann et al . , 2011; Roth et al . , 2012; Glickfeld et al . , 2014 ) . The results in L2/3 support the notion that in FF and FB pathways , excitation is more strongly counterbalanced by inhibition and that the imbalance is gradually rectified according to hierarchical distance from the most FF to the most FB ( Figure 1e , 4j ) . The results further suggest that the hierarchical distance rule of normalization is independent of the absolute magnitude of EPSCs across the hierarchy ( Figure 4—figure supplement 2a–c ) . Our findings argue that excitation ascending across multiple hierarchical levels is gradually adjusted to keep the dynamic range of L2/3 Pyr cell firing constant and compensate for the increased density of synaptic input to Pyr cells in higher cortical areas ( Elston , 2003 ) . Strong activation of PV neurons may narrow the window for effective excitation and result in high frequency gamma-band synchronization of activity found in FF signaling ( Gabernet et al . , 2005; Cardin et al . , 2009; Bastos et al . , 2015 ) . In contrast , in FB pathways excitation is weakly counterbalanced by inhibition , which may broaden the window for synaptic integration and result in slower synchronization frequencies found in FB communications ( Bastos et al . , 2015 ) . Thus , variation in I/E balance , through the differential recruitment of PV and Pyr neurons in different cortical pathways , is a key feature of distributed hierarchical processing . Reciprocal connections between areas are a highly conserved feature of the mammalian cortex . However , the exact pattern of termination of FF and FB axonal projections in the target area appears to vary between species , particularly in the termination patterns of FF pathways in layers 2 , 3 and 4 ( Felleman and Van Essen , 1991; Price and Zumbroich , 1989; Coogan and Burkhalter , 1990 ) . Despite these differences , a consistent observation among different species is a tendency for FF projections to avoid L1 and the selective targeting of L1 by FB pathways ( Coogan and Burkhalter , 1993; Rockland and Virga , 1989; Henry et al . , 1991; Cauller , 1995 ) . We therefore used the average DR of axonal terminations in L2-4 to those in L1 to classify pathways on a sliding scale as being FF or FB . In this reference frame , V1 , LM and PM constitute a clear hierarchy , which broadly matches that of rat visual cortex ( Coogan and Burkhalter , 1993 ) and is consistent with the increasing size of receptive fields ( Wang and Burkhalter , 2007 ) . The hierarchical ordering of V1 , LM and PM based on average DRs is consistent with the ordering based on the difference of DRs between reciprocally connected pairs . This is notable because differences in the laminar patterns of reciprocal projections between two areas have traditionally been used to arrange areas in a hierarchy ( Felleman and Van Essen , 1991; Coogan and Burkhalter , 1993 ) . While our method of averaging DRs provides a hierarchy based on how individual visual areas project to every other area within the network , it is conceivable that such a hierarchical arrangement may not be consistent with defining pathways between every reciprocally connected pair of areas as being FF or FB by comparing the DRs of projections to each other . The absolute value of the difference between DRs of reciprocally connected areal pairs therefore remains an open issue for defining hierarchical distance and designating connections as FF , FB , or lateral ( Felleman and Van Essen , 1991; Coogan and Burkhalter , 1993 ) . Cortical Pyr cells typically receive thousands of synaptic contacts , raising the question of how these neurons successfully generate graded spike outputs , without saturating their spike output , in response to varying levels of excitatory input ( Shadlen and Newsome , 1998 ) . This problem is compounded by the need for deeper parts of the brain , which are further separated from the outside world than lower areas , to respond robustly and appropriately to sensory input varying in intensity over several orders of magnitude . Pertinently , Pyr cells in higher areas have been shown to have a higher density of dendritic spines than those in lower areas in both primates ( Elston , 2003 ) and rodents ( Elston et al . , 2006 ) , indicating that Pyr neurons in higher areas must integrate a larger number of excitatory inputs . To maintain a wide dynamic range over which Pyr cells can signal , inhibitory neurons have been proposed to be critical ( Shadlen and Newsome , 1998; Pouille et al . , 2009 ) . In particular , PV neurons normalize cortical activity by inhibiting Pyr cells by a level that is proportional to the latter’s excitation , thus controlling their gain ( Atallah et al . , 2012; Wilson et al . , 2012; Xue et al . , 2014 ) . Because they are strongly targeted by interareal inputs ( Gonchar and Burkhalter , 1999 ) , PV cells are also ideally suited to mediate long-range FFI between areas . Such an interareal inhibitory circuit would make Pyr cells coincidence-detectors ( Gabernet et al . , 2005; Pouille and Scanziani , 2001 ) , leading to a reduction of noise levels and the preservation of temporal precision in the target area ( Bruno , 2011; Zhu et al . , 2015 ) . Coincidence-detection has also been proposed to help achieve a wide dynamic range by allowing only a fraction of excitatory inputs to summate and evoke a spike response ( Shadlen and Newsome , 1998 ) . Our observation that L2/3 PV cells are recruited most strongly by pathways transmitting signals from V1 to higher cortical areas imply that signals sent to deeper parts of the brain from more peripheral areas are more potently controlled by inhibition than pathways originating in higher areas . Such a high level of inhibition may be crucial in order for Pyr cells to efficiently integrate excitatory input from a large number of areas . On the other hand , lower I/E levels in FB pathways would broaden the 'window of opportunity' for spikes to be integrated and trigger an output in the postsynaptic cell ( Isaacson and Scanziani , 2011 ) , suggesting that FB signals originating in association cortex require less gain control than FF signals . Rather , by broadly modulating the excitability of neurons in lower areas ( such as by targeting the primary dendrites of Pyr cells in L1/2 ) , FB pathways are well-placed to prime Pyr cells to selectively respond to FF input in a context-dependent manner ( Larkum , 2013 ) . Although synaptic inputs to L5 Pyr cells are also denser in higher areas ( Elston and Rosa , 2000 ) , we found that excitation of these neurons in FF and FB pathways is similar and appears to be less strongly counterbalanced by inhibition . This provides a putative mechanism for the previously observed sparse coding in L2/3 Pyr cells and dense excitation in intrinsically burst-spiking L5 Pyr cells , allowing for distinct computational strategies within individual neurons depending on their postsynaptic targets ( Harris and Mrsic-Flogel , 2013; Hefti and Smith , 2000; Schubert et al . , 2007 ) . The laminar difference may indicate that , similar to thalamocortical input ( Constantinople and Bruno , 2013 ) , interareal inputs to L5 are driving Pyr cells . This may enable interareal communication through cortico-thalamo-cortical loops ( Guillery and Sherman , 2002 ) as well as with subcortical motor targets , thereby linking perception and action ( Kim et al . , 2015 ) . While PV neurons are a critical component of cortical gain control , it must be noted that they are only one of a number of inhibitory sources ( Jiang et al . , 2015; Pfeffer et al . , 2013; Gonchar et al . , 2007 ) . For instance , neocortical GABAergic interneurons that express vasoactive intestinal polypeptide ( VIP ) are thought to be an important target of long-range and neuromodulatory inputs ( Fu et al . , 2014; Reimer et al . , 2014 ) , and in turn , primarily inhibit other interneurons leading to disinhibition of cortical Pyr cells ( Pfeffer et al . , 2013; Kepecs and Fishell , 2014; Pi et al . , 2013; Karnani et al . , 2016 ) . Somatostatin ( SOM ) -expressing interneurons , which include Martinotti cells , make extensive inhibitory contacts with local Pyr cells , and can consequently mediate disynaptic inhibition between neighboring Pyr cells ( Fino and Yuste , 2011; Silberberg and Markram , 2007 ) . SOM neurons have also been shown to provide inhibitory inputs to other interneurons , including PV cells , suggesting a role in the disinhibition of Pyr cells as well ( Pfeffer et al . , 2013 ) . A perhaps surprising source of inhibition and disinhibition is glutamate . By activating pre- and postsynaptic metabotropic receptors in various neocortical circuits , glutamate release can induce suppression of GABA release and inhibition of L4 neurons , respectively ( Liu et al . , 2014; Lee and Sherman , 2009 ) . Thus , multiple , partially overlapping ( Gonchar et al . , 2007 ) sources of inhibition may be differentially recruited depending on context , providing a multilayered control of cortical function ( Kepecs and Fishell , 2014; Pakan et al . , 2016 ) .
For analyzing projection patterns between cortical areas , we used 6–8 week-old C57BL/6J male and female mice . In addition , we crossed Pvalb-Cre mice ( RRID:IMSR_JAX:008069 ) with Ai9 reporter mice ( C57BL/6 background , The Jackson laboratory , Bar Harbor , ME; RRID:IMSR_JAX:007905 ) , which harbored a floxed STOP cassette that prevents transcription of the fluorescent protein tdTomato ( tdT ) . The crossing resulted in offspring in which PV neurons express tdT . All electrophysiology experiments were performed in male and female PV-tdT mice . Mice were anesthetized by intraperitoneal injection of a ketamine/xylazine ( 86 mg·kg−1/13 mg·kg−1 , IP ) mixture and secured in a headholder . Analgesia was achieved by buprenorphine ( 5 mg·kg−1 , SC ) . Callosal connections were labeled by 30–40 pressure injections ( 20 nl each ) of the retrograde tracer bisbenzimide ( BB , 5% in H2O , Sigma ) into the right occipital cortex . Interareal projections were labeled by iontophoretic injections ( 3 µA , 7 s on/off duty cycle for 7 min ) of the anterograde tracer biotinylated dextran amine ( BDA; 10 , 000 molcular weight , 5% in H2O; Invitrogen ) using a coordinate system whose origin was the intersection between the midline and a perpendicular line drawn from the anterior border of the transverse sinus at the posterior pole of the occipital cortex . The coordinates of the injected areas were ( anterior/lateral in mm ) : V1 ( 1 . 1/2 . 6 ) ; LM ( 1 . 4/4 . 1 ) ; PM ( 1 . 9/1 . 6 ) . Mice were randomly assigned for injections of a particular area . Three days after the tracer injections , the mice were overdosed with ketamine/xylazine , perfused through the heart with heparinized phosphate buffer ( PB; 0 . 1 M , pH 7 . 4 ) followed by 4% paraformaldehyde in PB ( PFA ) . Brains were postfixed with 4% PFA and equilibrated in 30% sucrose . To enable areal identification of injection and projection sites , BB labeled callosal landmarks in the left hemisphere were imaged in situ under a fluorescence stereomicroscope ( Leica MZ16F ) , equipped with UV optics . The imaged hemispheres were then cut on a freezing microtome at 40 µm in the coronal plane . Sections were collected and numbered as a complete series across the full caudo-rostral extent of the hemisphere . Sections were wet mounted onto glass slides and imaged under UV illumination under a fluorescence microscope equipped with a CCD camera . The sections were then removed from the slides and BDA labeled axonal projections were visualized with avidin and biotinylated HRP ( Vectastain ABC Elite ) in the presence of H2O2 and diaminobenzidine ( DAB ) ( Wang et al . , 2012 ) . Sections were mounted onto glass slides , coverslipped in DPX and imaged under a microscope equipped with dark field optics . 16 to 23-day-old mice were anesthetized with a mixture of ketamine/xylazine ( 86 mg·kg−1/13 mg·kg−1 , IP ) . Held in a stereotaxic apparatus , intracerebral injections of viral vector ( AAV2/1 . CAG . ChR2-Venus . WPRE . SV40 ( Addgene20071 ) ; Vector Core , University of Pennsylvania ) ( Petreanu et al . , 2009 ) were made with glass pipettes ( tip diameter 25 μm ) connected to a Nanoject II Injector ( Drummond ) . Injections were performed stereotaxically into V1 , LM or PM , 0 . 3 and 0 . 5 mm below the pial surface , to ensure infection of neurons throughout the thickness of cortex . The total volume of the viral vector at each depth was 46 nl . Successful injections resulted in the simultaneous expression of Channelrhodopsin-2 ( ChR2 ) and the fluorescent protein Venus in terminals of outgoing axons . Mice were randomly selected for the study of a particular pathway . 30 to 45 day-old mice , 14–21 days after viral injection , were anesthetized with a mixture of ketamine/xylazine ( 86 mg·kg−1/13 mg·kg−1 , IP ) , and transcardially perfused with 10 ml of ice-cold oxygenated 95% O2/5% CO2 dissection solution ( sucrose-ACSF ) containing ( in mM ) : 228 sucrose , 2 . 5 KCl , 1 . 25 NaH2PO4 , 25 NaHCO3 , 0 . 5 CaCl2 , 7 . 0 MgCl2 , and 10 D-glucose . Mice were decapitated , the brain removed from the skull , and mounted on the specimen plate of Leica Vibratome ( Leica VT1200 ) with a cyanoacrylate adhesive ( Krazy Glue ) . Visual cortex was cut coronally at 350 µm in ice-cold sucrose-ACSF . Slices were transferred to a holding chamber filled with ACSF containing ( in mM ) : 125 NaCl2 , 2 . 5 KCl , 1 . 25 NaH2PO4 , 25 NaHCO3 , 2 . 0 CaCl2 , 1 . 0 MgCl2 , and 25 D-glucose . Slices were incubated in ACSF for 30 min at 34°C and maintained at room temperature until recordings . Acute slices were superfused with recirculating oxygenated ACSF at room temperature in a submersion chamber mounted on the fixed stage of an upright microscope ( Nikon Eclipse FN1 ) . For subcellular , optogenetic mapping experiments , 1 µM TTX and 100 µM 4-AP were added to the bath in order to block action potentials ( and therefore polysynaptic excitation ) and fast repolarizing potassium currents . Whole-cell patch clamp recordings were performed with borosilicate pipettes ( 4–6 MΩ resistance ) . The pipette solution contained ( in mM ) 128 potassium gluconate , 4 MgCl2 , 10 HEPES , 1 EGTA , 4 Na2ATP , 0 . 4 Na2GTP , 10 sodium phosphocreatine , 3 sodium L-ascorbate , and 3 mg/ml biocytin . The pH was adjusted to 7 . 2–7 . 3 , and osmolarity to 290 mOsm . Fluorescence of ChR2/Venus-expressing fibers and tdT-expressing PV neurons was imaged with a CCD camera ( Retiga 2000DC , QImaging ) . Pyr and PV neurons lying within maximal levels of ChR2/Venus-expressing axonal projections were selected for recordings . PV neurons were identified by tdT expression . For sCRACM experiments ( see below ) , neurons were voltage clamped at −70 mV . All voltage-clamp and current-clamp experiments were performed using the Ephus software ( Suter et al . , 2010 ) ( Vidrio Technologies ) , an Axopatch 700B amplifier ( Molecular devices ) , and a data acquisition ( DAQ ) device ( NI USB-6259 , National Instruments Corp . , Austin , TX ) . The photostimulation of ChR2-expressing fibers was achieved by a blue laser ( 473 nm; CrystaLaser ) delivered in an 8 × 16 grid in which stimulation points were spaced 75 µm apart , one spot at a time , 400 ms between laser delivery at each spot . The grid was aligned such that the longer axis was perpendicular to the pial surface and stimulated spots in all six layers . The position of the laser beam was controlled by galvanometer scanners ( Cambridge Scanning ) , and the duration of stimulation ( 1 ms ) was controlled by a shutter ( LS6 , Uniblitz ) . The laser beam ( ~20 µm at half maximal intensity ) passed through a Pockels cell ( ConOptics ) and an air objective ( 4x PlanApo ) . Because the expression level of ChR2-Venus in interareal axons varied across slices and animals , the laser power was adjusted in every slice so that the largest EPSCsCRACM ( EPSC recorded under sCRACM conditions ) in a neuron did not increase upon increasing laser intensity . Importantly , the laser power was constant for all recordings made in the same slice , in order to compare EPSCssCRACM between neighboring neurons . The laser power measured at the image plane was 0 . 7–1 mW/cm ( 2 ) . Photostimulation was repeated three to five times for each neuron . The shutter timing and the position of galvanometer mirrors was controlled by Ephus ( Suter et al . , 2010 ) . After the recordings , slices were fixed in 4% PFA , cryoprotected in 30% sucrose and re-sectioned on a freezing microtome at 40 µm . The sections were then incubated with an antibody against the type 2 muscarinic acetylcholine receptor ( M2; 1:500 in PB; MAB367 , Millipore; RRID:AB_94952 ) and stained with Alexa Fluor 647-labeled IgG ( 1:500 in 10% NGS; A21247; Invitrogen ) . M2-expression was imaged under a microscope equipped with IR fluorescence optics . The intense M2-expression in V1 was used as a landmark for assigning Venus labeled axonal projections to LM and PM ( Wang et al . , 2011 ) . M2 stained sections containing biocytin-filled neurons were treated with 1% H2O2 , and incubated in avidin and biotinylated horseradish peroxidase ( Vectastain ABC Elite ) in the presence of DAB . The soma and dendritic arbor of biocytin-filled neurons were reconstructed under a 60x oil objective using Neurolucida ( MBF Bioscience; RRID:SCR_001775 ) . | The visual cortex is the part of the brain responsible for the conscious sense of vision . It is made up of multiple connected areas , and each area has a different expertise for analyzing images . The areas exchange information about the outside world via connections between cells called neurons . Communication between the areas works like a hierarchy with deeper , more connected areas in the brain extracting more complex information from a visual scene . Communication in the cortex requires repeated stimulation or “excitation” of pathways of neurons; this risks damage or loss of sensitivity . But all of the communication in the hierarchy is excitatory , meaning that a signal from one area activates other areas in the visual cortex . So , how does the brain avoid becoming over-stimulated ? The answer is that connections between the areas of the visual cortex also contact inhibitory neurons that suppress brain activity . However , it is not clear how the level of inhibition in different areas of the visual cortex is fine-tuned to avoid over-stimulation while maintaining accurate perception of vision . D’Souza et al . now report how three distinct areas of the mouse visual cortex communicate to process visual signals . The approach involved making particular pathways of neurons sensitive to light , such that they could be activated separately with a laser . Next , D’Souza et al . measured the activity of both inhibitory and excitatory neurons that link the different brain areas . The experiments showed that the inhibitory neurons are more strongly activated in the areas of the brain that are further up the hierarchy . This indicates that our ability to make sense of more complex features of visual signals requires higher levels of inhibitory control . The next step is to examine how the brain activates and controls inhibitory neurons , and how this depends on the situation an animal is in and the task it is performing . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Recruitment of inhibition and excitation across mouse visual cortex depends on the hierarchy of interconnecting areas |
Serostudies are needed to answer generalizable questions on disease risk . However , recruitment is usually biased by age or location . We present a nationally-representative study for dengue from 70 communities in Bangladesh . We collected data on risk factors , trapped mosquitoes and tested serum for IgG . Out of 5866 individuals , 24% had evidence of historic infection , ranging from 3% in the north to >80% in Dhaka . Being male ( aOR:1 . 8 , [95%CI:1 . 5–2 . 0] ) and recent travel ( aOR:1 . 3 , [1 . 1–1 . 8] ) were linked to seropositivity . We estimate that 40 million [34 . 3–47 . 2] people have been infected nationally , with 2 . 4 million ( [1 . 3–4 . 5] ) annual infections . Had we visited only 20 communities , seropositivity estimates would have ranged from 13% to 37% , highlighting the lack of representativeness generated by small numbers of communities . Our findings have implications for both the design of serosurveys and tackling dengue in Bangladesh .
There has been growing recognition of the utility of nationally representative serum banks to monitor the burden from infectious diseases in a population ( Metcalf et al . , 2016; Wilson et al . , 2012; Osborne et al . , 1997; Jardine et al . , 2010; De Melker and Conyn-van Spaendonck , 1998; van der Klis et al . , 2009 ) . By tracking the levels of pathogen-specific antibodies in populations , these banks are a powerful tool for public health agencies to understand a wide range of factors that can assist in the fight against diseases , including pathogen circulation patterns , vaccination levels , and the existence of spatial pockets of susceptibility ( De Melker and Conyn-van Spaendonck , 1998; Gidding et al . , 2005; Osborne et al . , 2000 ) . To date , the accumulation and use of nationally representative banked sera have focused almost exclusively on vaccine preventable childhood infections . However , serum banks also have the potential to be invaluable in efforts to understand the burden from arboviruses and optimize efforts for control , including the targeted deployment of vaccines ( Imai and Ferguson , 2018 ) . In these diseases , high levels of subclinical infection and frequent clinical misclassification mean that even in locations with good disease surveillance , the underlying risk of infection is poorly understood ( Halstead , 2007 ) . Studies that collect serum for the detection of pathogen-specific antibodies typically rely on either convenience samples ( e . g . , blood donations ) or focus on single communities , often in major cities or where the perceived risk of the pathogen is believed to be greatest ( Petersen et al . , 2012; Salje et al . , 2016a; Rodríguez-Barraquer et al . , 2014 ) . While they are often a good starting place for understanding historic pathogen circulation in the sampled communities and age-groups , the ability to generalize the results to the wider population is rarely known . By contrast , if communities are randomly chosen from across a country , we can estimate national incidence rates , the underlying spatial heterogeneity in burden , the spatial dependence across neighbouring communities and identify risk factors for infection . Such study designs therefore provide mechanistic insights into pathogen spread and facilitate the development of data-informed control policies . Dengue virus is a flavivirus , transmitted by Aedes mosquitoes , that is found across tropical and subtropical regions and causes a range of disease manifestations , ranging from asymptomatic infection to death ( Petersen et al . , 2012 ) . Transmission of arboviruses , such as dengue , appears to be driven by the interplay of individual- ( e . g . , sex , age , travel ) , household- ( e . g . , water supply , use of mosquito control ) and community-level ( e . g . , urban/rural , mosquito abundance ) factors ( Salje et al . , 2016b; Rodríguez-Barraquer et al . , 2015 ) . In order to make data-informed decisions on how best to control spread , we need to understand the relative importance of these different factors by collecting detailed data across these scales . A recent literature search found only one nationally representative dengue seroprevalence study , from Singapore , but there was only a subset of age-groups considered ( Imai et al . , 2015 ) . Outside of city states such as Singapore , Bangladesh is the most densely populated country in the world with 146 million people living in an area under 150 , 000 km2 . The dengue burden in Bangladesh is unclear . Sporadic cases were reported in the 1960 s and a major outbreak occurred in 2000 ( Rahman et al . , 2002; Sharmin et al . , 2015; Yunus et al . , 2001 ) , with clinical cases reported annually since then ( Government of the People’s Republic of Bangladesh , Ministry of health and family Welfare , 2017 ) . However , our knowledge of dengue epidemiology in the country is largely restricted to Dhaka , where a seroprevalence of 80% has been observed ( Dhar-Chowdhury et al . , 2017 ) , with the burden elsewhere unknown ( Government of the People’s Republic of Bangladesh , Ministry of health and family Welfare , 2017 ) . Here , we present the results of a study where we use sequential annual visits in randomly selected communities across Bangladesh to determine the burden of dengue and identify key risk factors for infection .
We randomly selected 70 communities from the 97 , 162 communities in the national census , where the probability of selection was proportional to the size of the community population . In rural locations ( around three-quarters of the country ) , these census-communities consist of villages , whereas in urban places , these communities are city wards . Study teams visited each of the selected communities at least twice , once during the period 08/2014-12/2014 ( Y1 ) and once during the period 10/2015-01/2016 ( Y2 ) to conduct interviews , collect serum and trap mosquitoes . A further visit was conducted in 06/2015-07/2015 in a subset of communities for additional mosquito collection only . For each visit , the study team spent at least 5 days in the community . In an attempt to select households randomly , the study staff identified the house where the most recent wedding had taken place and identified the closest neighbour . They then counted six households in a random direction to identify the first household for the study . To select each additional household for the study , they used the previous household as a starting point and counted six households in a random direction . Different households were selected in each visit . For selected households , the household head was informed of the study and invited to participate . If the household head was away during the first visit , the study team returned at a later time . If the household head agreed to participate , all household residents over the age of 6 months were also invited to participate . Residents were offered a test to determine their blood group as a benefit of participating . If some members of the household agreed and some refused , all consenting members were included in the study . Where some household members were not present at the time of the visit , study staff organised a time to come back . Data collection for a community was considered to be complete when at least 40 serum samples from at least 10 households had been collected . There were three elements to data collection: ( A ) questionnaires ( B ) serum collection and ( C ) mosquito collection . Each participant was led through a questionnaire . Where individuals were too young to answer , older individuals from the household answered for them . We asked a series of questions on demographics ( age , sex ) , whether they had ever been diagnosed with dengue and whether they had travelled outside of their community in the prior 7 days , 30 days or 6 months . In addition , the head of the household was asked to complete a separate questionnaire , which included questions about their education level , total household income , household utilities ( e . g . , access to electricity and clean water ) , whether they had used any form of mosquito control in the last week and whether they owned land away from the household . A phlebotomist collected 5 ml of venous blood from all individuals who gave consent . Individuals who were sick at the time were ineligible . These samples were centrifuged in the field and the serum extracted into separate vials before being shipped to icddr , b ( previously known as the international centre for diarrhoeal disease research , Bangladesh ) laboratories in Dhaka in nitrogen dry shippers . The samples were tested for antibodies against IgG dengue virus , which indicates historic infection , using PanBio indirect IgG ELISAs ( Alere Inc , Massachusetts , USA ) . During the first visit in 2014 , BG Sentinel traps ( Biogents AG , Germany ) were placed in eight randomly sampled households in each of the of the 70 participating communities . The traps were placed in the main living area of the households and after 24 hr , they were collected and all mosquitoes sent to icddr , b laboratories in Dhaka where an entomologist identified the species of each captured mosquito . To help ensure that communities where no Aedes mosquitoes were found during the initial mosquito trapping truly had no Aedes , mosquito trapping was repeated in these communities from June 2015 to July 2015 during which eight households were randomly selected to have sentinel traps placed in their homes for 24 hr and the traps and mosquitos were processed in the same way as they had been initially . We divided the covariates into individual-level ( age , sex , travel patterns ) , household-level ( household income , electricity in household , access to water in household , mosquito control ) and community-level ( Ae . albopictus/Ae . aegypti in community , log population size ) categories . For each covariate , we initially performed simple logistic regression to explore associations with dengue serostatus using a hierarchical model with random intercepts for household and community . We accounted for spatial correlation structure using a Matern covariance function using a stochastic partial differential equation and fitted the models using integrated nested Laplace approximations ( INLA ) in a Bayesian framework ( Lindgren et al . , 2011 ) . All covariates were then included in a multivariable analysis . As the probability of being seropositive is strongly linked to the past circulation of dengue , we also performed a sensitivity model in which we recalculated the regressions using only individuals > 20 years as seropositivity was not found to differ by age among older adults . Finally , we assessed the importance of using spatial correlation structure by calculating the coefficients in a separate regression that did not include a spatial covariance matrix . To explore the variability in dengue risk across Bangladesh , we initially placed a 5 km x 5 km grid over the country and estimated the population size in each of those grid cells using data available from worldpop . org ( Tatem , 2017 ) . We then fit a multivariable model using the data from our sampled locations using log ( population size ) , age category ( <10 y , 11-20y , 21-30y , 31-40y , 41y-50 , 51-60y , >60 y ) and sex as covariates , which represent variables that are either available for all the grid cells ( population size ) or where we can use the overall proportion of the population that is within each category ( age and sex ) from the national census . As above , we fit the model in a Bayesian framework with a Matern spatial correlation structure using integrated nested Laplace approximations . We used the fitted model to predict in the unsampled grid cells by drawing 1000 samples from the posterior for each grid cell and calculated the mean as well as 2 . 5% and 97 . 5% quantiles to quantify uncertainty . The estimated number of seropositive individuals in a cell was calculated by multiplying the estimated proportion seropositive in a cell by the population within that cell . The total number of seropositive individuals in the country was calculated as the sum of seropositive individuals across all the grid cells . As a sensitivity analysis , we also predicted the spatial distribution of dengue seropositivity in the country using a model with the Matern spatial covariance matrix only ( i . e . , without any covariates ) . We assessed the ability of different model formulations to accurately predict the level of seropositivity in unsampled locations . We considered four different models: ( i ) crude proportion seropositive , ( ii ) multivariable logistic model with sex , age-group and population size as covariates and Matern spatial covariance ( the baseline model ) , ( iii ) multivariable logistic model with sex , age-group and population size as covariates but with no spatial dependence , and ( iv ) spatial dependence model using Matern spatial covariance with no covariates . For 100 iterations and for each model in turn , we repeatedly randomly selected a subset of communities to train the model ( varied between 2 , 20 , 40 , 60 and 69 communities ) and predicted the seroprevalence in the remaining communities not used to fit the model . Separately , we considered the impact of having sampled fewer people per community . We reran each of the four models over repeated iterations using 50 randomly selected communities and with between 2 and 80 individuals sampled per community to train the models . We then estimated the seroprevalence in the 20 remaining communities . We used the probability of being seropositive as a function of age to estimate the proportion of the susceptible population that get infected each year using catalytic models , an approach which has been used frequently to reconstruct the past circulation of pathogens ( Salje et al . , 2016a; Rodríguez-Barraquer et al . , 2014; Imai et al . , 2015; Ferguson et al . , 1999 ) . We assumed a constant force of infection due to all four serotypes , λ and that there were no differences in risk by age . The proportion seropositive of age a , is given by z ( a ) =1-exp ( -λ x ( min ( a , NYears ) ) , where NYears is the number of years prior to 2014 that dengue has circulated in Bangladesh . We fixed NYears at 20 to reflect the approximate period when dengue first appeared in the country . We conducted a sensitivity analysis where this was varied between 15 and 25 years . We estimated λ using maximum likelihood where the contribution to the likelihood from seronegative individuals coming from community i is exp ( -λ x ( min ( a , NYears ) ) x 1/wt ( commi ) , where the weights , wt ( commi ) , represent the proportion of that community that was sampled ( number of people in community i/population in community i ) . This approach was used to ensure that all individuals contributed equally to the likelihood . The contribution to the likelihood from seropositive individuals is ( 1-exp ( -λ x ( min ( a , NYears ) ) ) x 1/wt ( commi ) . We calculated the force of infection for the entire sampled population as well as separate estimates by sex and for the locations from the three largest cities ( Dhaka , Chittagong and Khulna ) only versus the rest of the country . To estimate the number of people infected each year we used the estimated population by age for each year for the period 1995–2014 ( Kinsella and He , 2009 ) . We assumed that in 1995 , the entire population was susceptible . The proportion of the population that have monotypic immunity is calculated as w ( a , y ) =4 x exp ( −3 x λs x a* ) x ( 1-exp ( -λs x a* ) ) where λs is the serotype specific force of infection and is calculated as λ/4 and a* is the number of years an individual has been alive since the introduction and is calculated as min ( a , y-1995 ) . Similarly , the proportion of the population that has previously been infected with two serotypes ( w2 ( a , y ) ) is 6 x exp ( −2 x λs x a* ) x ( 1-exp ( -λs x a* ) ) 2 and the number previously infected with three serotypes w3 ( a , y ) =4 x exp ( - λs x a* ) x ( 1-exp ( -λs x a* ) ) 3 . Using these proportions we can calculate the number of primary , secondary , tertiary and quaternary dengue infections . Where N ( a , y ) x 4 x λs*exp ( −4 x λs x a* ) is the number of primary infections , N ( a , y ) x 3 x λs x w ( a , y ) , the number of secondary infections , N ( a , y ) x 2 x λs x w2 ( a , y ) the number of tertiary infections and N ( a , y ) x λs x w3 ( a , y ) the number of quaternary infections . N ( a , y ) is the size of the population of age group a in year y . We present the estimated total number of infections across primary , secondary , tertiary and quaternary infections . This study was approved by the icddr , b ethical review board ( protocol number PR-14058 ) . The U . S . Centers for Disease Control and Prevention relied on icddr , b’s ethical review board approval . All adult participants provided written , informed consent after receiving detailed explanation of the study and procedures . Parents/guardians of all child participants were asked to provide written , informed consent on their behalf .
In total , 5866 individuals fully participated ( completed questionnaire and had blood taken ) in our study across 70 communities , 2911 during August–December 2014 and 2955 during October 2015 – January 2016 ( Table 1 ) . We obtained serum from 76% of household members of participating households ( Figure 1—figure supplement 1 ) . Per community there were an average of 95 participants ( range 81–116 ) from 20 households ( range 20–23 ) . The age and sex distributions in the study largely matched those obtained by the 2011 census although we had some under-representation in those <10 years ( Figure 1—figure supplement 2 ) . The PanBio assay appeared to discriminate well between those with and without past dengue infection ( Figure 1—figure supplement 3 ) . We found that overall , 24% of individuals had evidence of a past infection . We observed substantial heterogeneity by age and sex ( Figure 1B ) , with 27% of males seropositive compared to 21% of females ( p-value<0 . 001 ) . Individuals > 20 y had 30% seropositivity compared to 14% in those under 20y . There was close correlation between the proportion seropositive in a community between 2014 and 2015 ( Pearson correlation of 0 . 92 ) ( Figure 1C ) . Overall , there was no observed difference in seropositivity across the two years of the study ( p=0 . 66 ) . While most of the study population ( 91% ) aged >10 y had heard of dengue , only 38 individuals ( 0 . 6% ) reported having had dengue , of whom only 16 had evidence of past infection . While all communities had at least one seropositive individual , there was substantial spatial heterogeneity across the country with the proportion seropositive ranging from 3% in rural Maulvibazar in Sylhet Division to 88% in urban Chittagong . Communities in the north of the country appeared largely unaffected . Communities in the northern division of Rangpur had a mean seropositivity of 9% compared to 45% for communities in Khulna division in the southeast . Even within Dhaka district ( which includes the capital and has the highest population density ) , where we visited three urban ( ‘Thana’ ) communities , there was substantial heterogeneity , with seropositivity ranging from 36 to 85% . The two urban communities we visited in the city of Chittagong had seropositivities of 84 and 88% . We found that several individual-level variables were associated with seropositivity ( Table 2 ) . In particular , males were much more likely to be seropositive ( odds ratio [OR]: 1 . 6 [95%CI: 1 . 4–1 . 9]; adjusted odds ratio [aOR]: 1 . 7 [1 . 5–2 . 0] ) , although this difference was concentrated in communities where overall seropositivity was <20% ( Figure 2—figure supplement 1 ) . Travel also appeared important , with those who had travelled in the prior 7 days having twice the odds of being seropositive compared to those that had not travelled in the prior 6 months ( OR: 1 . 9 [1 . 5–2 . 4]; aOR: 1 . 4 [1 . 1–1 . 8] ) . Household-level covariates did not appear to be important in determining risk of seropositivity , including having household electricity , household access to clean water , land ownership or household income . The use of mosquito control in the household was also not associated with seropositivity ( OR: 0 . 9 [0 . 8–1 . 1]; aOR: 0 . 9 [0 . 7–1 . 1] ) . At the community-level , we found some evidence that individuals living in locations where we had found Ae . aegypti were more likely to be seropositive , although this effect was less in the multivariable model ( OR 1 . 8 [1 . 2–2 . 8] , aOR 1 . 4 [0 . 9–2 . 2] ) . Having Ae . albopictus in the community was not linked to individual serostatus ( OR 1 . 1 [0 . 7–1 . 6] , aOR 1 . 0 [0 . 7–1 . 6] ) . Overall Ae . aegypti was found in 23 ( 33% ) and Ae . albopictus in 29 ( 41% ) of communities with a slight negative correlation between the two ( Pearson correlation of −0 . 2 , p-value 0 . 07 ) . The median seropositivity in communities with Ae . aegypti was 33% compared to 13% in the other communities . The median seropositivity in communities with Ae . albopictus was 15% compared to 18% in communities where it was not found . Individuals living in urban communities were more likely to be seropositive than those living in rural communities with each unit increase in log population size associated with a 1 . 3 times increased probability of being seropositive ( 95% CI: 1 . 2–1 . 5 ) . The intraclass correlation coefficients showed that the Matern spatial covariance matrix explained 15% of the variance , the community-level random effects explained 6% and the household random intercept explained 12% of the variance in individual level responses . In a model without the spatial covariance matrix , the community-level random intercept explained 23% of the variance with the household-level effect unchanged . Including spatial covariance was associated with a small improvement in model fit , justifying its inclusion ( Deviance Information Criterion [DIC] difference of 4 ) . While most of the coefficient estimates were largely consistent in models that did and did not include the spatial covariance structure , the impact of Ae . aegypti changed significantly , increasing to aOR 2 . 4 ( 95% CI: 1 . 3–4 . 5 ) when spatial correlation was not incorporated ( Figure 2—figure supplement 2 ) . Not including random intercepts by household and community resulted in falsely narrow confidence intervals and some changes in coefficient estimates and a substantial drop in model fit ( DIC difference of 801 ) . Coefficients of models where the data was restricted to adults only were largely unchanged . We used a spatial prediction model that incorporates the population size and sex distribution and spatial correlation structure to estimate the level of seropositivity throughout the country ( Figure 2A ) . We found that the proportion of people seropositive in communities was spatially correlated up to 108 km ( as measured from the Matern covariance function ) , consistent with that observed in a variogram of the seropositivity between communities ( Figure 2B ) . Our model performed well at estimating the observed levels of seropositivity in participating communities in leave-one-out cross-validation with a Pearson correlation of 0 . 8 between the observed and fitted values and a mean absolute error of 8% ( Figure 2C ) . These maps further suggest dengue is currently concentrated in the three largest cities of Dhaka , Chittagong and Khulna . This estimated distribution of dengue risk in the country was very similar if we used the spatial dependence information only , without age and sex covariates ( Figure 2—figure supplement 3 ) . Overall , we estimate that approximately 25% ( 95% CI: 21–29% ) of the population had been infected with dengue at some point during their lives , equivalent to 40 . 3 million individuals ( 95% CI: 34 . 3–47 . 2 ) . This estimate is consistent with that obtained using the crude proportion seropositive among our samples ( 24% or 39 . 0 million individuals ) . Using a catalytic model to estimate the proportion seropositive by age , we estimated that 1 . 6% ( 95% CI: 1 . 5–1 . 7% ) of the susceptible population gets infected each year across the four serotypes , equivalent to an average of 2 . 4 million annual infections ( 95% CI: 2 . 2–2 . 5 million ) ( Figure 2—figure supplement 4 ) . However , estimates were much higher for the three major urban hubs of Dhaka , Chittagong and Khulna compared to the rest of the country . Within these hubs , 6 . 4% ( 95% CI: 5 . 4–7 . 6% ) of the population gets infected annually with no differences by sex , whereas this drops to 1 . 0% ( 95% CI: 0 . 9–1 . 2% ) for females outside these areas and 1 . 6% ( 95% 1 . 4%–1 . 8% ) for males ( Figure 2D–E ) . We assessed the sensitivity of our results to the number of participating communities and the model framework used . Over repeated iterations , we used a subset of our communities to estimate the overall proportion seropositive and to train a suite of models that were then used to estimate seropositivity in the remaining communities . We found that if we had only visited twenty communities , the seropositivity among the samples would have ranged from 13% to 37% , depending on the communities visited . Spatially explicit models which incorporated data on sex , age and population size , did not result in substantial improvements with the range of seropositivity estimates similarly wide ( Figure 3A ) . By contrast , had we visited 60 communities , the range would have been much smaller ( 22–28% ) , with similar results in the spatial models . The accuracy of our predictions in unsampled locations improved substantially with increasing numbers of communities visited ( Figure 3B ) . In spatial models with no covariates , the mean absolute error in the predictions per community fell from 13 . 6% when 20 communities were sampled to 10 . 5% when 69 communities were sampled , with a corresponding rise in the correlation between the observed and predicted seroprevalence from 0 . 39 to 0 . 81 ( Figure 3C ) . Incorporating information on age , sex and population size resulted in a small improvement in performance when between 20 and 60 communities were sampled ( e . g . , when 20 communities were sampled , the mean correlation was 0 . 39 when no covariates were used and 0 . 49 when covariates were incorporated ) . Multivariable models with age , sex and population size as covariates but with no spatial correlation performed poorly , with no improvements with increasing numbers of communities visited . We found that sampling fewer people per community had little effect on our estimates , with the performance of nationwide and community-level seropositivity similar if 20 people were sampled per location compared to 80 ( Figure 3D–F ) .
We have presented the results of a large , nationally-representative , serostudy that provides a comprehensive description of dengue infection in Bangladesh . Our results demonstrate that , to date , dengue risk is very heterogeneous across the country . It also shows that the vast majority of the country has never been infected . The framework presented here can act as a strategy for future efforts to estimate nationally-representative infection risks in a population . Our results suggest that since dengue re-emerged in the late 1990 s , the virus has only established a pattern of sustained endemic transmission in a few urban settings , and not throughout the country , as is the case in nearby Myanmar , Thailand and Cambodia ( Rahman et al . , 2002; Government of the People’s Republic of Bangladesh , Ministry of health and family Welfare , 2017; van Panhuis et al . , 2015 ) . Part of the reason for this may be the limited presence of the principal vector , Ae . aegypti , which was found in only one third of the communities that participated in this study . By contrast , more ( especially rural ) communities had the secondary vector , Ae . albopictus , but its presence was not associated with infection . Our finding of a negative correlation between Ae . aegypti and Ae . albopictus presence is consistent with the species occupying different environmental niches or competition between the two species , as has previously been suggested ( Braks et al . , 2004 ) . All communities had at least one seropositive individual , suggesting that external viral introductions may be common and that there may be factors preventing large-scale outbreaks , including from the limited abundance of the Ae . aegypti vector . It is unclear how stable these vector populations are . Characterizing the drivers of Ae . aegypti spread and maintenance , especially in the context of changing land use and climate , appears key to understanding future risk of spread . We were able to use our study to identify risk factors for infection . At the individual level , we found that males had 1 . 6 times the odds of having been infected as females , although this difference was concentrated in communities where overall seropositivity was low , suggesting that comparing infection proportions between males and females can be a good marker of local dengue endemicity . In addition , those that travelled more ( as indicated by having left the community recently ) were more likely to have been infected . These findings suggest that in the current scenario in which infection risk is heterogeneously distributed around the country , individuals from communities with low or no transmission are likely to get infected when they travel to higher transmission areas . If the vector presence expands throughout the country , these individuals could act as sources of outbreaks within these communities . Our findings are in marked contrast to what has been observed with chikungunya in Bangladesh , where women in a community that had a widespread chikungunya epidemic were found to be at significantly increased risk of being infected compared to males , with the increased risk of infection linked to greater time spent in and around the home ( Salje et al . , 2016b ) . These findings suggest that it may be difficult to generalize inferences across arboviruses due to differences in vector species and the frequency of introductions and risk of onwards spread . While we incorporated spatial correlation into our risk factor regression analyses , in practice this only resulted in a relatively small improvement in model fit compared to hierarchical models with random intercepts at the community and household levels . The biggest impact of incorporating the spatial correlation was to move the coefficient estimate for Ae . aegypti presence towards the null . This suggests that the covariance structure is a better predictor than the basic mosquito absence/presence data as the covariance is driven by , and absorbs , the true underlying environmental drivers including mosquito distributions . Overall , we estimated that around a quarter of the population , around 40 million individuals , have been infected by dengue with an average of 2 . 4 million annual infections . This figure is much less than the 16 . 7 million previously estimated through a modelling exercise ( Bhatt et al . , 2013 ) , highlighting the need for representative data to help support these models and the importance of considering immunity in the estimation of annual case numbers . We did not detect a significant difference in seropositivity between the two study years , though we were not sufficiently powered to detect small changes in seropositivity across the population . This points to a clear trade-off between resampling the same individuals across the two study years , which would have facilitated quantification of the incidence between the two years and our approach , which allowed us to maximize our sample size . While there was substantial spatial heterogeneity in the risk of being seropositive across the country , ultimately our crude estimates of the proportion seropositive in the country ( 24% ) was very close to modelled estimates that incorporated the age , sex and population distribution in the country ( 25% ) . This provides strong support for the sampling frame we used to capture population-level estimates of population exposure . While spatial prediction models did not help improve overall estimates of national burden , they did allow us to build maps of how infection risk is distributed throughout the country . Incorporating covariates ( e . g . , sex , age , population distribution ) did not result in substantive improvements in predictive accuracy compared to models that included a spatial covariance term only . This highlights the importance of spatial dependence in obtaining accurate estimates in unsampled locations . The use of environmental and climate covariates were not considered here and may improve estimates , especially where insufficient ( or no ) sampled locations exist to make use of spatial correlation structure . Our approach is particularly relevant to settings like Bangladesh where there was little prior understanding of the distribution in disease burden in the country . Alternative strategies may exist in other settings where there is already some existing knowledge of where risk is concentrated . Although , even in these settings , unmeasured spatial differences in healthcare seeking or in surveillance system infrastructure may mean that it is preferable to randomly sample communities without specifically focusing on specific areas of populations in the country . Our study provides key insight for the national vaccine policy for dengue . For the only currently licensed vaccine , Dengvaxia , the WHO has recommended that countries perform nationally representative serosurveys to inform vaccine rollout as the vaccine only provides protection in people with existing antibodies ( World Health Organization , 2016 ) . The vaccination of seronegative individuals has been linked to increased risk of subsequent severe disease ( Salje et al . , 2018; Hadinegoro et al . , 2015 ) . For Bangladesh , our findings suggest that any vaccine rollout should be concentrated to the urban areas of Dhaka , Chittagong and Khulna . However , even in these communities , the proportion seropositive at age 9 years is far below the threshold of 80% where vaccine rollout is potentially feasible without pre-vaccination screening ( World Health Organization , 2018 ) . Therefore , any rollout will require the screening of individuals for presence of antibodies before vaccination to avoid placing large numbers of individuals at risk for more severe disease manifestations . The approach presented here could be used as a strategy for other countries interested in obtaining national estimates of disease risk . The optimal number of communities to visit will depend on the size and distribution of the population , the underlying level and heterogeneity of infection in the population , the required level of precision and the available budget . Therefore , it is difficult to make general recommendations about the number of communities which should be sampled in other settings . However , our finding that spatial correlation exists within 100 km suggests that communities should be sampled at a density to ensure that there is at least one sampled community within 100 km of all residents and preferably more . Sampling as few as 20 individuals per community still provides robust nationwide estimates , however , in practice , there are fewer budget and time constraints to sampling additional individuals within a community than visiting additional communities . Cross-reactivity of antibodies is a problem for all seroprevalence studies , especially with flaviviruses . This prevents us from quantifying the relative importance of the different dengue serotypes . In addition , some seropositive individuals may have been infected with Japanese encephalitis rather than dengue . However , Japanese encephalitis is typically only found in rural communities where it circulates at low levels ( estimated at 2 . 7 cases/100 , 000 ) ( Paul et al . , 2011 ) . As we estimated only low levels of dengue seropositivity in rural communities , the number of false positives from Japanese encephalitis cross-reactivity is likely to be small . Individuals who participated in the study may not be representative of all members in the community . In particular , individuals who travel frequently may have been away . We attempted to minimize this risk by organizing times to meet with household members who were not present in the initial visit . Around 90% of households that we approached agreed to take part in the study . We used BG sentinel traps that have been shown to be well suited to trapping Aedes mosquitoes ( Maciel-de-Freitas et al . , 2006; Obenauer et al . , 2010 ) . In addition , we revisited all communities where we did not find Aedes mosquitoes in the initial visit for additional mosquito trapping . However , given the heterogeneous nature of mosquito distributions within communities , we may have nevertheless failed to find Aedes mosquitoes in communities where they breed . This would mean that Aedes may be more widespread than we found . We used the time period since individuals last left the community as a marker of travel . While this is likely to broadly capture trends in mobility , it remains a crude marker and more detailed measures of movement ( from e . g . , movement diaries , global positioning system monitors ) would help provide a more detailed understanding of how people move . To randomly select households in a community , we would ideally have used a sampling frame of all households in the community . However , in this setting there were no detailed community maps and enumerating all households in the communities would have added an additional day in the field per community . Therefore , we used a quasi-random approach that identified a starting point for household sampling based on the area of the community where the most recent wedding took place; given that >95% of adults marry in Bangladesh , it is unlikely that this approach could bias the sample we obtained . However , in cultural contexts where marriage rates may vary by community or location within a community , this method of choosing a random starting point could produce a biased sample . We found that simply asking people about whether or not they had been infected with dengue was not informative of past infection . This highlights that studies that only use questionnaires can only provide limited burden information for pathogens such as dengue . For example , Demographic Health Surveys , which collect detailed health questionnaires from randomly selected individuals across many countries could benefit significantly from adding a serological component to their studies . In Bangladesh , where dengue is still emerging , surveillance for vectors could be a way to monitor risk of future outbreaks and continued efforts to understand drivers of transmission could point to interventions to reduce its geographic spread . | Dengue is a mosquito-borne virus that infects millions of people each year . Often the countries most affected by the virus , such as Bangladesh , do not have the resources needed to tackle the disease . For resources sent to these countries to have the greatest impact , it is important to know which areas are most affected , and which subsets of the population are most at risk . A way to gather this information is to test for dengue virus antibodies a protein produced by the immune system in response to the infection in the blood of individuals . However , previous efforts to use these tests to understand dengue risk in communities have generally only been done in single locations , typically a major city , and the findings of these tests are unlikely to be applicable to the wider population . Now , Salje et al . have visited 70 different communities from all around Bangladesh and used these tests on blood samples collected from over 5 , 000 individuals from a range of age-groups . From these measurements it was estimated that an average 2 . 4 million people are infected with dengue each year in Bangladesh , with major cities , such as Dhaka , experiencing more concentrated levels . The exposure to dengue outside major cities was much lower , and men , who tend to travel more , were found to be at greater risk of infection . Salje et al . also showed that using a small number of communities to estimate national levels of infection led to misleading results . This highlights the danger of using information collected from a limited number of places to represent the effects of a disease on the wider population . Public health agencies in Bangladesh will be able to use this information to tackle dengue more effectively , focusing on the areas and the populations most affected by the disease . In addition , the design and analytical approaches used in this study could be applied to other countries , and to different diseases . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2019 | Nationally-representative serostudy of dengue in Bangladesh allows generalizable disease burden estimates |
Increasing evidence highlights the important roles of microRNAs in mediating p53’s tumor suppression functions . Here , we report miR-139-5p as another new p53 microRNA target . p53 induced the transcription of miR-139-5p , which in turn suppressed the protein levels of phosphodiesterase 4D ( PDE4D ) , an oncogenic protein involved in multiple tumor promoting processes . Knockdown of p53 reversed these effects . Also , overexpression of miR-139-5p decreased PDE4D levels and increased cellular cAMP levels , leading to BIM-mediated cell growth arrest . Furthermore , our analysis of human colorectal tumor specimens revealed significant inverse correlation between the expression of miR-139-5p and that of PDE4D . Finally , overexpression of miR-139-5p suppressed the growth of xenograft tumors , accompanied by decrease in PDE4D and increase in BIM . These results demonstrate that p53 inactivates oncogenic PDE4D by inducing the expression of miR-139-5p .
microRNAs ( miRNAs ) represent a class of cellular short non-coding RNAs responsible for modulating the expression of their target genes at the post-transcriptional level . Abnormal regulation of miRNAs is associated with human cancer development and staging ( Lu et al . , 2005 ) . Over the past decade , increasing attention has been drawn to the role of miRNAs in p53 signaling network , and a number of miRNAs have been identified as p53 target genes ( Liao et al . , 2014a ) . These miRNAs are involved in multiple biological processes , including cell cycle arrest , apoptosis , glycolysis and so on . Also , they often connect p53 with other signaling pathways ( Christoffersen et al . , 2010; Sachdeva et al . , 2009; Liang et al . , 2013 ) . Although miRNAs have been appreciated as important mediators of p53’s tumor suppression functions , a lot still remain unexplored to better understand the fine-tuning of p53 signaling and crosstalk with other pathways by these RNAs . In this study , we identified miR-139-5p as a novel p53 target gene and demonstrated a new pathway connecting p53 and miR-139-5p with an oncogenic protein PDE4D as a new target of this miRNA and its downstream cAMP signaling .
In order to identify potential p53 target microRNAs , we used colon cancer cell lines , HCT116 p53+/+ and HCT116 p53-/- , the latter of which were genetically engineered to lose the expression of wild type ( WT ) p53 ( Bunz et al . , 1998 ) . Both cells were treated with 4 μM Inauhzin-C ( INZ-C ) , which is a p53 activating compound discovered in our lab ( Zhang et al . , 2012 ) . After confirmation of p53 and its targets induction through Western blot ( Figure 1A ) and quantitative real-time PCR ( qRT-PCR ) ( Figure 1B ) , the total RNA was extracted and sent to ArrayStar for miRNA-sequencing analysis . The results revealed that in addition to known p53 target microRNAs , such as miR-34a , miR-1246 and miR-143 ( Liao et al . , 2014a; Suzuki et al . , 2009 ) , miR-139 was also significantly induced in HCT116 p53+/+ , but not HCT116 p53-/- , cells , suggesting miR-139 as a potential p53 target ( Figure 1C ) . We independently confirmed this observation by detecting miR-139-5p expression after treating HCT116 p53+/+/HCT116 p53-/- and H460 ( WT p53 ) /H1299 ( p53 null ) cells with DMSO or 4 μM INZ-C . miR-139-5p was significantly induced only in p53 positive , but not in p53 null , cells ( Figure 1D ) . In contrast , p53 knockdown decreased miR-139-5p expression by more than 50% in H460 and U2OS cells ( Figure 1E ) . These data indicate that miR-139-5p is a possible p53 target gene . 10 . 7554/eLife . 15978 . 003Figure 1 . Identification of miR-139 as a novel p53-responsive gene . Validation of induction of p53 and its targets gene . HCT116 p53+/+ and HCT116 p53-/- cells were treated with DMSO or 4uM INZ-C for 18 hr in duplicate and the protein samples and RNA samples were prepared for Western Blot ( A ) and qRT-PCR ( B ) , respectively . ( C ) . microRNA sequencing data analysis shows significantly induced microRNAs uniquely or commonly in HCT116 p53-/- and HCT116 p53+/+ cells . miR-139 was highlighted in red among other known p53 targets . ( D ) . Induction of miR-139-5p by p53 . p53 positive ( H460 , HCT116 p53+/+ ) and p53 negative ( H1299 , HCT116 p53-/- ) cells were treated as in ( A ) and ( B ) , and qRT-PCR analysis followed . ( E ) . Decrease of miR-139-5p by knocking down p53 . H460 and U2OS cells were transfected with scramble siRNA ( SC ) or siRNA specific to p53 ( p53 ) and subjected to Western blot and qRT-PCR analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 003 After carefully analyzing the genomic sequence of the miR-139 gene using p53MH algorism ( Hoh et al . , 2002 ) , we found a highly conserved putative p53 responsive element ( RE ) located at the 5’ flanking region ( Figure 2A ) . To test if endogenous p53 binds to this p53RE sequence , we conducted a chromatin-associated immunoprecipitation ( ChIP ) assay after treating H460 or HCT116 p53+/+ cells with 0 . 5 μM Doxorubicin for p53 induction . In both cell lines , the binding of p53 with the p53RE was dramatically increased upon Doxorubicin treatment as compared to non-treatment control , indicated by p53RE sequence pulled down with p53 specific antibody DO-1 , but not non-specific immunoglobulin G ( Figure 2B ) . Also , we assessed luciferase expression driven by either a wild type or a mutant-p53RE-motif-containing miR-139 sequence ( Figure 2C ) in H1299 cells . GFP-p53 markedly induced luciferase activity in a dose-dependent manner when wild type miR-139 p53RE , but not the mutant p53RE , was used ( Figure 2D ) . These results clearly show that p53 binds to the miR-139 promoter region and thus regulates the transcription of miR-139 . 10 . 7554/eLife . 15978 . 004Figure 2 . miR-139 is a direct target of p53 . ( A ) Diagram shows putative p53 responsive element ( p53RE ) located upstream of miR-139 gene . ( B ) Increased binding of p53 and the endogenous p53RE-containing miR-139 promoter in response to Doxorubicin . H460 or HCT116 p53+/+ cells were treated with 0 . 5μM Doxorubicin for 18 hr to stimulate the endogenous p53 before chromatin-associated immunoprecipitation assays were conducted with DO-1 p53 antibodies and the p53RE specific primers listed in Supplementary file 1 online . Western blot analysis was performed to confirm induction of p53 . ( C ) Schematic of the pGL3 luciferase reporter constructs used . The plasmids either contain wild-type or mutant p53RE sequences of the miR-139 promoter , as underlined . ( D ) . Enhancement of miR-139-promoter-driven luciferase activity by p53 . H1299 cells were co-transfected with pGL3-139 or pGL3-139mut and increasing amount of GFP-p53 and collected 48 hr after transfection for assessment of luciferase activity , which was normalized against β-gal expression . Western blot was also performed to confirm the expression of GFP-p53 . Error bars represent standard deviation ( n = 3 ) . β-gal , β-galactosidase; Ctrl , control; Dox , Doxorubicin; IgG , immunoglobulin G; mut , mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 004 By using the online microRNA target prediction tool ( Maragkakis et al . , 2009 ) , we searched for potential RNA targets of miR-139-5p . After screening several candidates , PDE4D turned out to be an ideal target as its 3’-untranslated region ( 3’-UTR ) contains miR-139-5p targeting sequence ( Figure 3A ) . Overexpression of miR-139-5p mimic markedly reduced the expression of PDE4D in H460 and A549 cells ( Figure 3B ) . The type-I insulin-like growth factor receptor ( IGF-IR ) , a previously reported miR-139-5p target ( Shen et al . , 2012 ) , was also decreased by miR-139-5p mimic , indicating the activity of miR-139-5p used here ( Figure 3—figure supplement 1 ) . PDE4D belongs to the family of phosphodiesterases and is a cyclic AMP ( cAMP ) specific phosphodiesterase with several splice variants ( Omori and Kotera , 2007 ) . It is an oncogenic protein that regulates cancer cell proliferation , angiogenesis and apoptosis ( Lin et al . , 2013; Ogawa et al . , 2002; Pullamsetti et al . , 2013; Rahrmann et al . , 2009 ) . To further test whether the downregulation of PDE4D by miR-139-5p is through direct regulation on its mRNA , we constructed the WT or mutant predicted miR-139-5p target sites into the pMIR-Report system that contains the luciferase reporter gene subjected to regulation mimicking the microRNA target ( Figure 3C ) . When co-transfecting H1299 cells with miR-139-5p , only the pMIR-PDE4D-WT , but not the pMIR-PDE4D-mutant , displayed suppressed expression ( Figure 3D ) , suggesting that miR-139-5p regulates PDE4D expression by directly binding to the target sequence at 3’-UTR of PDE4D mRNA . 10 . 7554/eLife . 15978 . 005Figure 3 . p53 modulates PDE4D expression via miR-139-5p . ( A ) Bioinformatic analysis shows the miR-139-5p-targeted 3’-untranslated region ( 3’UTR ) sequence of PDE4D mRNA . ( B ) Overexpression of miR-139-5p decreases the level of PDE4D protein in cells . H460 and A549 cells were collected 48 hr after transfection with miR-139-5p mimic for Western blot analysis . ( C ) Schematic of the pMIR–PDE4D luciferase reporter constructs used , which contain either a wild-type or a mutant miR-139-5p target site derived from the PDE4D mRNA . ( D ) Overexpression of miR-139-5p specifically inhibits luciferase activity from the plasmid harboring a wild-type , but not a mutant , miR-139-5p targeted sequence . H1299 cells were co-transfected with the indicated plasmids and collected 48 hr after transfection for luciferase assay . Luciferase activity was measured and normalized against β-gal expression . Even amount of oligos was achieved by adding Negative control oligos accordingly . ( E ) p53 modulates PDE4D expression . H460 and A549 cells were collected for Western blot analysis 18 hr after treatment with Doxorubicin ( left panel ) or 48 after transfection with si-SC or si-p53 ( right panel ) . ( F and G ) p53 modulation of PDE4D expression is through miR-139-5p . H460 cells were co-transfected with indicated plasmids and 24 hr later were treated with Doxorubicin for 18 hr followed by measurement of luciferase activity ( F ) or Western blot analysis ( G ) . 139-5p m , miR-139-5p mimic; 139-5p Inh , miR-139-5p inhibitor . Cells were treated with solvent for Dox or transfected with the same concentration of negative control oligos as miR-139-5p mimic or inhibitor . Error bars represent standard deviation ( n = 3 ) . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 00510 . 7554/eLife . 15978 . 006Figure 3—figure supplement 1 . HepG2 cells were collected 48 hr after transfection with miR-139-5p mimic at 0 , 20 or 40 nM for Western blot analysis on IGF-IR expression . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 00610 . 7554/eLife . 15978 . 007Figure 3—figure supplement 2 . PC-3 cells were collected 48 hr after transfection with miR-139-5p mimic at 40 nM for Western blot analysis on PDE4D expression . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 00710 . 7554/eLife . 15978 . 008Figure 3—figure supplement 3 . HCT116 p53+/+ and HCT116 p53-/- cells were treated with 10 μM Nutlin-3 ( Nut ) or 5 nM Actinomycin D ( ActD ) for 18 hr , and then collected for Western blot analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 008 In line with the above results , activation of p53 by doxorubicin in H460 and A549 cells significantly reduced the protein levels of PDE4D ( Figure 3E , left panel ) . Notably , the PDE4D variants with smaller molecular weight have been reported to possess stronger oncogenic activities due to the lack of functional inhibitory domains as compared to the longer forms ( Lin et al . , 2013 ) . Nevertheless , miR-139-5p transfection and doxorubicin treatment led to similar expression inhibition of both the long and the short variants of PDE4D , indicating that this newly identified pathway has broader inhibition on PDE4D . In contrast , knocking down p53 elevated PDE4D expression in these two cell lines ( Figure 3E , right panel ) . We then assessed the effect of miR-139-5p on PDE4D in PC-3 , a p53 null cell line , and found similar reduction of PDE4D by miR-139-5p mimic ( Figure 3—figure supplement 2 ) . To further validate the regulatory role of p53 on PDE4D , we also treated HCT116 p53+/+ and HCT116 p53-/- cells with 10 μM Nutlin-3 ( Nut ) , which disrupts MDM2-p53 interaction and therefore activates p53 ( Vassilev et al . , 2004 ) , and 5 nM Actinomycin D ( ActD ) , which causes ribosomal stress-mediated p53 activation ( Iapalucci-Espinoza and Franze-Fernández , 1979 ) , and found PDE4D was reduced by both of these two drugs only in p53 positive , but not p53 negative , cells ( Figure 3—figure supplement 3 ) . These results suggest that this suppression of PDE4D is p53 dependent in response to various stresses . Furthermore , doxorubicin inhibited pMIR-PDE4D activity , but introduction of miR-139-5p inhibitor reversed this inhibition ( Figure 3F ) . Consistently , the suppression of PDE4D protein by doxorubicin was also compromised in the presence of miR-139-5p inhibitor ( Figure 3G ) . Collectively , these findings demonstrate that activation of p53 can induce the expression of miR-139-5p that in turn suppresses the expression of oncogenic protein PDE4D . Since PDE4D is a cAMP specific phosphodiesterase , ectopic expression of miR-139-5p in A549 cells led to significant increase of cellular cAMP levels ( Figure 4A ) . Also , consistent with our results shown in Figure 3F and Figure 3G , doxorubicin treatment increased cellular cAMP levels in A549 cells , which were alleviated by the miR-139-5p inhibitor ( Figure 4—figure supplement 1 ) . In addition , Nutlin-3 , a more specific p53-activating reagent , showed similar effect on cAMP level , which was also reversed in the presence of miR-139-5p inhibitor ( Figure 4B ) . In contrast , in H1299 cells , which are p53-null and express non-detectable PDE4D , neither doxorubicin nor miR-139-5p affected cellular cAMP level ( Figure 4—figure supplement 2 ) . Notably , the rescue effect of miR-139-5p on A549 cell growth inhibition by Nutlin-3 was correlated with the change of cAMP levels ( Figure 4—figure supplement 3 ) . These data indicate that p53 activation could modulate cAMP levels through miR-139-5p . 10 . 7554/eLife . 15978 . 009Figure 4 . miR-139-5p induces cAMP/BIM mediated cell growth arrest . ( A ) . Overexpression of miR-139-5p increases cellular cAMP levels . A549 cells were transfected with miR-139-5p mimic control or miR-139-5p mimic and were subjected to cAMP measurement 48 hr after transfection . ( B ) . Nutlin-3 increases cellular cAMP levels via miR-139-5p . A549 cells were transfected with miR-139-5p inhibitor control or miR-139-5p inhibitor for 48 hr followed by 10 μM Nutlin-3 treatment for 18 hr and subjected to cAMP measurement . ( C ) . Overexpression of miR-139-5p induces BIM expression . H460 and A549 cells were transfected with miR-139-5p mimic and Western blot performed 48 hr after transfection . ( D ) . Knockdown of BIM rescues miR-139-5p induced cell growth arrest . A549 cells were transfected with miR-139-5p mimic or siRNA against BIM ( si-BIM ) or both and subjected to MTT assay 48 hr after transfection . Cells were treated with solvent for Nutlin-3 or transfected with the same concentration of negative control oligos as miR-139-5p mimic or BIM siRNA . Error bars represent standard deviation ( n = 3 ) . *p<0 . 05 as compared to negative control oligos . Δ , p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 00910 . 7554/eLife . 15978 . 010Figure 4—figure supplement 1 . Doxorubicin increases cellular cAMP levels via miR-139-5p . A549 cells were transfected with miR-139-5p inhibitor control or miR-139-5p inhibitor for 48 hr followed by 0 . 5 μM Doxorubicin treatment for 18 hr and subjected to cAMP measurement . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 01010 . 7554/eLife . 15978 . 011Figure 4—figure supplement 2 . H1299 cells were treated with vehicle or 0 . 5 μM Doxorubicin for 18 hr followed by cAMP measurement , or transfected with miR-139-5p mimic control or miR-139-5p mimic and subjected to cAMP measurement 48 hr after transfection . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 01110 . 7554/eLife . 15978 . 012Figure 4—figure supplement 3 . miR-139-5p inhibitor rescues Nutlin-3 inhibition of cell growth . A549 cells were transfected with miR-139-5p inhibitor control or miR-139-5p inhibitor for 48 hr followed by 10 μM Nutlin-3 treatment for 24 hr and subjected to MTT analysis . *p<0 . 05 as compared to control siRNA . Δ , p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 01210 . 7554/eLife . 15978 . 013Figure 4—figure supplement 4 . HCT116 p53+/+ and HCT116 p53-/- cells were treated with vehicle , or 10 μM Nutlin-3 for 18 hr , and subjected to Western blot analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 01310 . 7554/eLife . 15978 . 014Figure 4—figure supplement 5 . BIM knockdown alleviates Nutlin-3 inhibition of cell growth . A549 cells were transfected with control siRNA or BIM siRNA and 48 hr after transfection , the cells were treated with 10 μM Nutlin-3 for 24 hr , followed by MTT assay and Western blot analysis . Error bars represent standard deviation ( n = 3 ) . *p<0 . 05 as compared to control siRNA . Δ , p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 01410 . 7554/eLife . 15978 . 015Figure 4—figure supplement 6 . A549 cells were transfected with miRNA inhibitor control or miR-139-5p inhibitor , and 48 hr after transfection , the cells were treated with vehicle , 2 μM INZ-C , 5 nM ActD or 10 μM Nutlin-3 for 48 hr , followed by MTT assay to determine cell viability . *p<0 . 05 as compared to the respective miRNA inhibitor control . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 015 Depletion of PDE4D was previously reported to induce BIM-mediated apoptosis through activation of the cAMP pathway ( Lin et al . , 2013; Zambon et al . , 2011 ) . We found that introduction of miR-139-5p into H460 and A549 cells dramatically increased BIM protein expression ( Figure 4C ) . As expected , treatment with 10 μM Nutlin-3 also induced BIM expression in HCT116 p53+/+ , but not HCT116 p53-/- , cells ( Figure 4—figure supplement 4 ) , supporting that BIM induction is p53-dependent . Consequently , A549 cell growth was significantly inhibited by miR-139-5p or Nutlin-3 , both of which were attenuated by knocking down BIM using siRNA ( Figure 4D , Figure 4—figure supplement 5 ) . Inversely , introduction of miR-139-5p inhibitor into A549 cells significantly alleviated cell growth inhibition by several p53 activating drugs , including INZ-C , Actinomycin D and Nutlin-3 ( Figure 4—figure supplements 3 and 6 ) . These results demonstrate that in response to p53 activation , increased miR-139-5p induces BIM-mediated cell growth arrest via the PDE4D/cAMP pathway . In order to determine the clinical relevance of miR-139-5p regulation of PDE4D , we obtained 50 paired human colon tumor specimens and their adjacent normal tissues and conducted qRT-PCR analysis on these specimens . miR-139-5p expression was significantly lower while PDE4D was higher in these tumor specimens than that in their normal tissues ( Figure 5A ) . Pearson’s correlation analysis of the expression results from the tumor specimens and normal tissues revealed that miR-139-5p is inversely correlated with PDE4D expression ( Figure 5A , bottom panel ) . These results provide clinical evidence supporting miR-139-5p as a negative regulator of PDE4D , consistent with our above results . 10 . 7554/eLife . 15978 . 016Figure 5 . miR-139-5p is negatively correlated with PDE4D expression in human colorectal tumor samples , and represses the growth of SW480 xenograft tumors . ( A ) miR-139-5p is negatively correlated with PDE4D expression in colon tumor samples . miR-139-5p ( top panel ) and PDE4D ( middle panel ) RNA expression was determined in 50 tumors ( T ) and paired adjacent normal tissues ( ANT ) . The Pearson’s correlation of miR-139-5p and PDE4D RNA expression was analyzed combining tumor samples and normal tissues ( bottom panel ) . ( B ) The sizes of SW480 xenograft tumors stably overexpressing pEZX-control ( Vector ) or pEZX-miR-139-5p ( miR-139-5p ) were measured every three days starting at seven days after inoculation . Mean tumor sizes were presented . Error bar , SD; *p<0 . 05 . ( C ) Expression of miR-139-5p of xenograft tumors was determined by qRT-PCR . ( D ) Representative images of the xenograft tumors stained with hematoxylin and eosin ( H&E ) , or immunohistochemistry analyzed with PDE4D or Ki67 antibody . ( E ) Xenograft tumors were subjected to Western blot analysis of BIM expression . ( F ) Proposed model of the p53/miR-139-5p/PDE4D pathway . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 01610 . 7554/eLife . 15978 . 017Figure 5—figure supplement 1 . Quantification of PDE4D and Ki67 IHC analysis . The IHC staining intensity was categorized to low , medium and high as determined by ImageJ software . Five random fields were chosen from each slide to obtain average intensity of each category , and all six mice from each group were included in the analysis . *p<0 . 05 as compared to control . DOI: http://dx . doi . org/10 . 7554/eLife . 15978 . 017 To validate this clinical correlation , we established a xenograft model using SW480 cells stably expressing either scramble oligos ( Control ) or miR-139-5p . As expected , miR-139-5p expressing tumors grew significantly slowly as compared to control tumors starting at day 16 after inoculation ( Figure 5B ) . The difference of miR-139-5p expression in these two groups of tumors was comparable to that observed in human specimens ( Figure 5C vs Figure 5A ) . Immunohistochemistry analysis revealed that in the miR-139-5p overexpressing tumors , PDE4D expression was markedly repressed , while tumor cell proliferation was significantly inhibited as reflected by Ki67 staining ( Figure 5D and Figure 5—figure supplement 1 ) . Consistent with our in vitro observation , BIM expression was also elevated in miR-139-5p tumors ( Figure 5E ) . These findings suggest that the tumor suppressor role of miR-139-5p is likely attributed to its regulation of the PDE4D/BIM pathway . In summary , this study for the first time demonstrates that p53 can induce the expression of miR-139-5p ( Figure 1 and Figure 2 ) , which in turn suppresses the expression of an oncogenic protein PDE4D ( Figure 3 ) and leads to cAMP/BIM-mediated cell growth arrest ( Figure 4 ) . Significantly , miR-139-5p expression is negatively correlated with PDE4D in human colorectal tumor and normal tissues , and overexpression of miR-139-5p is associated with slower tumor growth in the xenograft model , which is accompanied with PDE4D suppression , BIM induction and cell proliferation inhibition ( Figure 5 ) . As a potential tumor suppressor , miR-139 was previously shown to be downregulated in human hepatocellular carcinoma and colorectal cancer with characterized targets including Rho-kinase 2 , IGF-IR and RAP1B ( Guo et al . , 2012; Shen et al . , 2012; Wong et al . , 2011 ) . Here , we identified PDE4D , an oncogenic protein that is upregulated in various human cancers ( Lin et al . , 2013 ) , as another novel target of this miRNA . Inhibition or depletion of PDE4D significantly induces apoptosis and inhibits proliferation of cancer cells ( Lin et al . , 2013; Ogawa et al . , 2002; Rahrmann et al . , 2009 ) . Notably , the oncogenic property of PDE4D involves the cAMP/BIM pathway ( Lin et al . , 2013; Zambon et al . , 2011 ) . cAMP is an important secondary messenger mediating diverse cellular processes with protein kinase A as its main effector ( Taskén and Aandahl , 2004 ) . In particular , lower cAMP levels favor cancer cell survival and proliferation , which can be reversed by inhibition of PDE4D , the cAMP specific phopsphodiesterase ( Goldhoff et al . , 2008; Lin et al . , 2013; Murata et al . , 2000; Ogawa et al . , 2002 ) . The tumor suppressor role of p53 has been extensively documented over the last two decades , and is highly attributable to its regulation of target genes involved in cell cycle arrest , apoptosis and senescence ( Bieging et al . , 2014; Levine , 2011 ) . More recent discoveries revealed that p53 is also a critical mediator of metabolic pathways that are important for tumor survival ( Bieging et al . , 2014; Jiang et al . , 2015; Wang and Gu , 2014 ) . Based on our findings , we propose a p53-miR-139-5p-PDE4D-cAMP-BIM pathway as a novel pathway that can mediate p53’s tumor suppression function to modulate cellular cAMP levels by inhibiting PDE4D expression via miR-139-5p , and deregulation of this pathway would be highly associated with tumorigenesis ( Figure 5F ) .
Human HCT116 p53+/+ ( RRID: CVCL_0291 ) and HCT116 p53-/- ( RRID: CVCL_S744 ) cells were generous gifts from Dr . Bert Vogelstein at the John Hopkins Medical Institutes . A549 ( RRID: CVCL_0023 ) , HepG2 ( RRID: CVCL_0027 ) , U2OS ( RRID: CVCL_0042 ) , H460 ( RRID: CVCL_0459 ) and H1299 ( RRID: CVCL_0060 ) cells were purchased from American Type Culture Collection ( ATCC ) . All cells were cultured in Dulbecco’s modified Eagle’s medium ( DMEM ) and PC3 ( RRID: CVCL_0035 ) cells ( also from ATCC ) in RPMI 1640 medium supplemented with 10% fetal bovine serum , 50 U/ml penicillin and 0 . 1 mg/ml streptomycin at 37°C in 5% CO2 . STR profiling was performed to ensure cell identity . No mycoplasma contamination was found . The pGL3-miR-139 luciferase reporter plasmid was constructed from the miR-139 promoter with primers as listed in Supplementary file 1 . The fragment was inserted into pGL3 at the MluI and XhoI sites . The pGL3-miR-139-mut was generated by site-directed mutagenesis with primers listed in Supplementary file 1 using pGL3-miR-139 as template . The pMIR-PDE4D and pMIR-PDE4D mutant plasmids were constructed by inserting the miR-139-5p-targeted PDE4D mRNA-coding sequence or its mutant into the pMIR vector ( Ambion ) at the SpeI and HindIII sites . The pSIF-H1-miR-139-5p was constructed by inserting annealed oligos as listed in Supplementary file 1 into the pSIF-H1 vector at BamHI and EcoRI sites , as per manufacturer’s instruction . The anti-p21 ( Thermo Fisher Scientific , Waltham , MA , RRID: AB_10986834 ) , anti-p53 ( DO-1 , Santa Cruz , Dallas , TX , RRID: AB_628082 ) , anti-PDE4D ( Aviva Systems Biology , San Diego , CA , RRID: AB_10879817 ) , and anti-BIM ( Cell Signaling , Danvers , MA , RRID: AB_10692515 ) antibodies used here were commercially purchased . The anti-MDM2 ( 2A10 and 4B11 ) antibodies were described previously ( Jin et al . , 2002 ) . Inauhzin-C ( INZ-C ) was reported previously ( Zhang et al . , 2012 ) . Doxorubicin ( Dox ) was purchased from Sigma-Aldrich . qRT-PCR for mature microRNAs was carried out by using the methods as described previously ( Tang et al . , 2006 ) . qRT-PCR for other genes were conducted using the protocol as described before ( Sun et al . , 2008 ) . Relative gene expression was calculated using the ΔΔCT method . All reactions were carried out in triplicate . As described previously ( Jin et al . , 2008 ) , briefly , cells were transfected with plasmids as indicated in each figure by using TurboFect ( Thermo Scientific , Waltham , MA ) and following the company’s manuals . Cells were harvested and lysed in lysis buffer 48 hr post transfection . The total protein concentrations were determined using a BioRad protein assay kit and equal amounts of total proteins ( 50 μg , otherwise indicated specifically ) were then subjected to SDS-PAGE , followed by WB with antibodies as indicated in each figure . Hsa-mir-139-5p mimic and Negative control were purchased from Gene Pharma ( Shanghai , China ) . Anti-miR miRNA Inhibitor and Anti-miR miRNA Inhibitors—Negative Control were purchased from Ambion . p53 siRNA was purchased from Santa Cruz . Two BIM siRNA ( siBIM-1 , ID: 19 , 474 and siBIM-2 , ID: 195012 ) were purchased from Ambion . Transfection of miRNA inhibitors was performed using the same method as that for normal siRNA as described previously ( Sun et al . , 2008 ) . Cells were transfected with pCMV-β-galactoside and indicated plasmids ( total plasmid DNA 1 μg/well ) as indicated in figures . Luciferase activity was determined and normalized by a factor of β-gal activity in the same assay as described previously ( Jin et al . , 2006 ) . ChIP analysis was performed as described previously using p53 ( DO-1 ) antibodies for endogenous p53 ( Liao and Lu , 2013; Liao et al . , 2014b ) . Immunoprecipitated DNA fragments were analyzed by quantitative real-time PCR ( qRT-PCR ) amplification using primers for miR-139 gene . The primers are listed in Supplementary file 1 online . cAMP-GloTM Assay ( Promega , Madison , WI ) was performed to measure the cellular cAMP concentration as per manufacturer’s instructions . Fifty paired colorectal tumors and adjacent non-tumor tissues were collected and histopathologically diagnosed at the Departments of Gastrointestinal Surgery and Pathology , the First Affiliated Hospital , Sun Yat-sen University . Patient consent and Institutional Research Ethics Committee approval were obtained prior to the use of these clinical materials for research purposes . SW480 cells ( 1 × 107 ) stably expressing pEZX-scramble control sequence ( Vector ) or pEZX-miR-139-5p were inoculated subcutaneously into the right flank of female BALB/c nude mouse ( four weeks old , n = 6 per group ) . The tumor volume was measured every three days and calculated as 0 . 524 × length × width2 ( Gleave et al . , 1992 ) . At the conclusion of the experiments , tumors were removed and fixed in 10% formalin for paraffin embedding and histological analysis , or flash-frozen in liquid nitrogen for Western blot and qRT-PCR analysis . All experimental procedures were approved by the Medical Ethical Committee of the First Affiliated Hospital , Sun Yat-sen University ( Guangzhou , China ) . H & E staining and immunohistochemistry were described previously ( Cao et al . , 2013; Zhang et al . , 2013 ) . Quantitative analysis of IHC staining was achieved by categorizing the staining intensity to low , medium and high as determined by ImageJ software ( NIH ) . The Student’s two-tailed t test was used to compare the mean differences between treatment and control groups , unless otherwise indicated . Data are presented as Mean ± SD ( standard deviation ) . p<0 . 05 was determined as statistically significant . | The human body is kept mostly free from tumors by the actions of so-called tumor suppressor genes . One such gene encodes a protein called p53 , which prevents tumors from growing by regulating the activity of many other genes that either inhibit cell growth or cause cells to die . For example , p53 regulates genes that encode short molecules called microRNAs , which in turn suppress the activity of other target genes . Although a number of microRNAs have been reported as p53-regulated genes , there are still more to find . Discovering these genes would in turn help researchers to better understand exactly how p53 acts to suppress the growth of tumors , and to treat cancers caused by mutations in this tumor suppressor gene . Cao , Wang et al . now discover a new microRNA – called miR-139-5p – as one that is activated by p53 in human cells . Colon tumors produce much lower levels of this microRNA than normal tissues , while the cancer cells with a higher level of miR-139-5p grow slower than do the cancer cells with less miR-139-5p . Further experiments showed that this is because miR-139-5p can suppress the production of a protein called PDE4D , which is often highly expressed in human cancers . The suppression of PDE4D by this microRNA results in an increase in the levels of a protein that can cause cancer cells to die . Cao , Wang et al . suggest that miR-139-5p and PDE4D form part of a signaling pathway that plays an important role in suppressing the growth of colon cancer cells . Since microRNAs often have more than one target , future studies could explore if miR-139-5p regulates the production of other cancer-related proteins as well . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2016 | Inactivation of oncogenic cAMP-specific phosphodiesterase 4D by miR-139-5p in response to p53 activation |
The interrelationship between endogenous microbiota , the immune system , and tissue regeneration is an area of intense research due to its potential therapeutic applications . We investigated this relationship in Schmidtea mediterranea , a model organism capable of regenerating any and all of its adult tissues . Microbiome characterization revealed a high Bacteroidetes to Proteobacteria ratio in healthy animals . Perturbations eliciting an expansion of Proteobacteria coincided with ectopic lesions and tissue degeneration . The culture of these bacteria yielded a strain of Pseudomonas capable of inducing progressive tissue degeneration . RNAi screening uncovered a TAK1 innate immune signaling module underlying compromised tissue homeostasis and regeneration during infection . TAK1/MKK/p38 signaling mediated opposing regulation of apoptosis during infection versus normal tissue regeneration . Given the complex role of inflammation in either hindering or supporting reparative wound healing and regeneration , this invertebrate model provides a basis for dissecting the duality of evolutionarily conserved inflammatory signaling in complex , multi-organ adult tissue regeneration .
Host and resident microorganism interactions are not only integral to the maintenance of normal physiological functions , but also to the development of pathology . Microbial dysbiosis , for example , underlies persistent inflammatory disorders , chronic non-healing wounds , and scar formation ( Dowd et al . , 2008; Price et al . , 2009; Carvalho et al . , 2012; Scales and Huffnagle , 2013; Brothers et al . , 2015; Shin et al . , 2015 ) . Long-term management of these conditions constitutes a substantial and sharply rising burden on our healthcare system ( Sen et al . , 2009 ) . Significant evidence spanning a diverse array of organisms and tissues indicates that the immune response can play a central role in either promoting or hindering wound healing and tissue regeneration ( Eming et al . , 2009; Karin and Clevers , 2016 ) . The precise determinants facilitating these diametrically opposed outcomes is an area of intense investigation . As such the development of diverse and robust model systems capable of elucidating basic molecular mechanisms governing the interplay of the immune response and tissue regeneration should provide new insights for the design of novel therapies to combat this rising medical issue ( Sen et al . , 2009 ) . Planaria are a classic model system for studying adult wound healing and tissue regeneration ( Reddien et al . , 2004 ) . These free living members of the phylum Platyhelminthes contain a persistent pool of adult pluripotent stem cells , termed neoblasts , capable of regenerating all of the tissues and cell types of the organism ( Wagner et al . , 2011 ) . Ablation of this population of mitotic cells via irradiation eliminates regenerative capabilities , resulting in regression of anterior structures and eventual tissue lysis ( Reddien et al . , 2005 ) . Development of RNAi methodologies has enabled the interrogation of genes involved regeneration ( Sanchez Alvarado and Newmark , 1999; Reddien et al . , 2005 ) . These studies have informed our understanding of how wound responses , the recognition of lost tissues , dynamic establishment of positional identity , and activation of appropriate stem cell and differentiation programs all serve to accomplish large-scale complex tissue regeneration ( Wenemoser and Reddien , 2010; Petersen and Reddien , 2011; Wenemoser et al . , 2012; Scimone et al . , 2014 ) . While wounding is the stimulus for regeneration across all organisms studied to date , its incidence also presents an additional challenge . Disruption of barrier epithelia and exposure of mucosal surfaces poses an increased risk for bacterial invasion of internal tissues . Yet the role to which , if any , changes in endogenous microbiota or the planarian immune response have in regeneration is entirely unknown . The exemplary regenerative capabilities and conservation of multiple immune signaling genes in planarians make this organism an attractive model for understanding how robust tissue regeneration capabilities can be balanced with an effective immune response ( Peiris et al . , 2014 ) . Previous studies in planaria have uncovered components of the immune system conserved in humans , but absent from the well-studied innate immunity models of ecdysozoa ( e . g . , flies and nematodes ) ( Abnave et al . , 2014 ) . This indicates that planaria can serve as a complement to previously established invertebrate models to inform our understanding of the human immune response . While their potent regenerative capabilities and robust capacity for pathogen clearance render them quite resilient , planaria are not invincible . Planaria reared using traditional static culture methods can exhibit features of declining health including decreased appetite , loss of motility , dorsal tissue lesions , tissue degeneration , and lysis . Many of these symptoms can be temporarily alleviated and managed with antibiotics , suggesting that endogenous bacteria can play an antagonistic role in normal regenerative and tissue homeostatic functions . To date much of the planarian immune system and its potential integration with tissue regenerative processes are still a mystery . Importantly , the composition of the planarian microbiome is still unresolved and no direct link between microbiota , the immune system , and tissue regeneration has been established . In this study , we investigated the relationship between endogenous microbiota , host response , and regeneration . We performed deep sequencing of bacterial 16 s rDNA under various perturbations to elucidate the composition and dynamics of the planarian microbiome . Importantly , we found that transitions in culture conditions or tissue amputation elicited a robust Proteobacteria expansion that coincided with an increase in the susceptibility of intact worms to develop ectopic tissue lesions . Isolation and identification of bacteria that expanded during this time yielded a strain of Pseudomonas capable of inducing progressive tissue degeneration that resulted in either total lysis or resolved to permit regeneration of lost structures . The development of novel low septic culture methods facilitated a candidate RNAi screen examining the mediators of this pathological progression in planaria . We successfully identified a conserved TAK1/MKK/p38 signaling module that both mediated infection-induced tissue degeneration and specifically impeded regeneration during infection . In situ phospho-signaling analyses permitted us to identify tissue and cell types that activate p38 signaling in response to infection , revealing a role in the initiation of localized apoptosis preceding tissue degeneration . This apoptotic and tissue degeneration response coincided with an immune response mediated by jun D , an AP-1 family homolog downstream of TAK1 signaling . Finally , while TAK1/MKK/p38 signaling facilitates apoptosis during infection , it plays a contrasting role in the inhibition of apoptosis during normal regeneration . Altogether , our study uncovered potential mechanisms by which immunity and regeneration intersect to mediate distinct tissue regeneration outcomes , and introduces planarians as the first animal model linking the expansion of endogenous Proteobacteria to inhibition of complex tissue regeneration via the activation of distinct TAK1/MKK/p38 signaling .
To increase the scale and efficiency planarian husbandry , we developed a novel recirculation culture system for co-cultivation of massive numbers of worms . In brief , this system enables the constant recirculation and UV sanitization of planarian water to mitigate pathogenic levels of bacteria ( Figure 1A ) . The result is a permissive context for rearing large biomass levels of healthy , fissioning , lesion-free planarians . Adoption of this culture system offered a unique opportunity to study the heretofore unknown composition and dynamics of the planarian microbiome . We hypothesized that bacteria tightly associated with the physiology of healthy worms would be preserved within this recirculation culture while additional , potentially harmful species may accumulate upon exit to traditional static culture conditions . Previously , declining worm health within the static culture was remedied by administration of antibiotic . We assayed parallel cohorts of worms transitioned from recirculation to static culture with or without the antibiotic gentamycin for characterization of the composition and dynamics of the planarian microbiome ( Figure 1A ) . 10 . 7554/eLife . 16793 . 003Figure 1 . The planaria microbiome dynamically responds to changes in culture conditions and regeneration . ( A ) Diagram of transition of planaria from recirculation to static culture . ( B ) Percentage of the phyla Bacteroidetes and Proteobacteria following exit from recirculation culture ( * = t-test p<0 . 05 ) . ( C ) Heatmap of the percentage of the top 10 bacterial genera across all the samples of CIW4 strain planaria following release from recirculation culture in the presence or absence of the antibiotic gentamycin ( n = pool of ~30 worms with 2–3 generated 16 s rDNA libraries , independently sequenced twice ) . ( D ) Heatmap of the percentage of the top 18 bacterial genera of amputated head , trunk , and tail fragments during regeneration in comparison to intact worms ( n = pool of 2–16 intact worms or 30–60 fragments for 16 s rDNA library generation ) . ( E ) Heatmap of the percentage of the top 5 bacterial genera across the samples of individual wild-type planaria . The heatmap of the genera Vogesella , Chryseobacterium , and Pseudomonas across wild worm samples are included for reference ( n = 1–4 16 s rDNA libraries generated per worm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 00310 . 7554/eLife . 16793 . 004Figure 1—source data 1 . 16 s rDNA sequencing results . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 00410 . 7554/eLife . 16793 . 005Figure 1—figure supplement 1 . Analysis of changes in bacterial levels and composition following exit from recirculation culture . ( A ) Composition of bacterial phyla during the transition from recirculation to static culture in the absence or presence of the antibiotic gentamycin ( n = pool of ~30 worms with 2–3 generated 16 s rDNA libraries , independently sequenced twice ) . Percentage of the prominent ( B ) Bacteroidetes genus Candidatus Amoebophilus and Proteobacteria genera ( C ) Vogesella and ( D ) Pseudomonas during the aforementioned changes in culture conditions . ( E ) Composition of bacterial phyla during regeneration of head , trunk , or tail fragments in comparison to intact worms ( n = a pool of 2–16 intact worms or 30–60 fragments for 16 s rDNA library generation ) . Percentage of the genera ( F ) Candidatus Amoebophilus , ( G ) Vogesella , and ( H ) Pseudomonas during regeneration . ( I ) Composition of bacterial phyla of individual wild type sexual planaria ( n = 1–4 16 s rDNA libraries generated per worm ) . ( J ) Percentage of the genera Candidatus Amoebophilus , Vogesella , and Pseudomonas amongst wild worm samples . Proportional Venn diagram comparison of the bacterial genera of ( K ) individual wild worms relative to one another or ( L ) compared in aggregate to the genera of the CIW4 lab strain . Overlaps less than 1% not pictured . ( M ) Bacterial CFU quantification of prominent bacterial strains from S2F2 sexual lab strain of planaria 3 days after exit from the fill and drain system ( n = 3 pooled worms ) . ( N ) qPCR of relative 16 s rDNA levels during the recirculation culture to static culture transition in the presence or absence of antibiotic ( n = 3 technical replicates of pooled samples of > 30 worms ) . ( O ) Total bacterial CFUs per planaria following exit from recirculation culture in the presence and absence of the antibiotic gentamycin ( n = 4 samples of 4 pooled worm homogenates , experiment independently repeated > 4 times ) . ( P ) Total bacterial CFUs per planaria or regenerating fragment after amputation ( n = 3–8 pooled worm or fragment homogenates per timepoint ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 005 We surveyed the endogenous bacteria of the asexual CIW4 lab strain of S . mediterranea via deep sequencing of the 16 s rDNA variable region ( Figure 1—source data 1 ) . As a control , bacteria present in the beef liver food source of lab-raised planarians was also analyzed but yielded bacterial levels comparable to negative controls and was not considered further . Initial analysis of the planarian microbiome revealed as many as 350 distinct species ( Table 1 ) , making it more akin to zebrafish than flies in terms of overall complexity ( Lee and Brey , 2013 ) . The most prevalent bacterial phyla within planaria cultured in the recirculation system was Bacteroidetes ( Figure 1—figure supplement 1A ) . This phylum represents a diverse array of key symbionts that support homeostatic functions in mammals ( Rakoff-Nahoum et al . , 2004; Mazmanian et al . , 2005 , 2008; Alegado et al . , 2012 ) . Bacteroidetes comprised 78% of the bacterial phyla of planaria from the recirculation system ( static culture day 0 ) , but this proportion declined to 57% after 3 days of static culture ( Figure 1B , Figure 1—figure supplement 1A ) . Administration of gentamycin exacerbated this effect , reducing Bacteroidetes composition to 17% . In the wake of this Bacteroidetes decrease , the proportion of bacteria of the phylum Proteobacteria steadily increased from 9% to 40–44% upon exit from the recirculation system ( Figure 1B , Figure 1—figure supplement 1A ) . This phylum comprises a wide variety of pathogens and increases in its abundance are a hallmark of microbial dysbiosis and pathological inflammation ( Carvalho et al . , 2012; Clark et al . , 2015; Shin et al . , 2015 ) . Thus , transitions in culture conditions are coincident with shifts in bacterial composition with potential implications for planarian health . 10 . 7554/eLife . 16793 . 006Table 1 . Distribution of Species reads in CIW4 S . mediterranea . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 006No . of species ( Average ) No . of species ( Standard deviation ) Sample≥1 read≥10 reads≥100 reads≥1 read≥10 reads≥100 readsRecirc D0268 . 785 . 337 . 536 . 313 . 995 . 3D1 -Gent245 . 879 . 634 . 217 . 45 . 10 . 8D1 +Gent281 . 889 . 441 . 210 . 87 . 82 . 7D3 -Gent36814458 . 857 . 22811 . 6D3 +Gent354 . 6126 . 457 . 226 . 514 . 855 . 8 We next analyzed compositional changes amongst the most abundant genera of the planarian microbiome following removal from the recirculation system ( Figure 1C ) . The most abundant bacterial genera present in recirculation-cultured worms largely belonged to the phylum Bacteroidetes ( Candidatus Amoebophilus , Segetibacter , and Chitinophaga ) and declined in relative abundance in static culture ( Figure 1C , Figure 1—figure supplement 1B ) . The genera Oxalobacter , Rhodoferax , and Vogesella , belonging to the phylum Proteobacteria , exhibited proportional increases ranging from 6 to 200 fold in worms transferred to static culture ( Figure 1C , Figure 1—figure supplement 1C ) . Two genera of the phylum Bacteroidetes , Chryseobacterium and Pedobacter , also increased dramatically after transfer despite the net decrease in the phylum overall . Members of the genus Chryseobacterium are potential pathogens in immune-compromised individuals and newborns undergoing prolonged antibiotic treatment ( Bloch et al . , 1997; Calderón et al . , 2011 ) , suggesting that dramatic shifts in abundance may constitute an opportunistic infection . All of the aforementioned genera were largely gentamycin sensitive . In contrast , proportional increases of Pseudomonas in static culture were exacerbated by gentamycin treatment ( Figure 1C , Figure 1—figure supplement 1D ) , consistent with its adept nature to adapt to the selective pressure of antibiotics ( Hoffman et al . , 2005; Price et al . , 2009 ) . Additionally , accumulation of Delftia , Arthrospira , Elizabethkingia , and Enterobacter was largely gentamycin dependent ( Figure 1C ) . Collectively , these data suggest that enrichment of a particular cohort of bacterial species may be more conducive to optimal planarian health and destabilization of this microbiota permits amplification of potentially harmful species . Planaria exhibit a robust capacity to replace and integrate missing tissues through a regenerative process combining morphogenesis with the remodeling of pre-existing tissues . The extent to which this process impacts the resident microbiome is heretofore unknown . We transferred worms from recirculating into static culture conditions and amputated the animals above and below the pharynx 3 days later . To mitigate bacterial bloom after exit from recirculation culture , worms were starved for 2 weeks prior to transfer and cultured at low density with frequent water changes . Amputated head , trunk , and tail fragments were separated and 16 s sequences were analyzed at 0dpa , 3dpa , 7dpa , and 14dpa . Un-amputated worms were analyzed in parallel as a control ( Figure 1—source data 1 ) . Analysis of planarian bacterial phyla during regeneration revealed dramatic shifts in composition that mirrored those seen during the transition from recirculation to static culture ( Figure 1—figure supplement 1A , E ) . Prolonged starvation and additional water changes successfully mitigated the Proteobacteria bloom , with Bacteroidetes comprising 75 to 81% of the bacterial phyla of intact worms and freshly amputated worm fragments on day 0 . Three days after amputation , worms exhibited a reciprocal decrease of Bacteroidetes and amplification of Proteobacteria phyla ( Figure 1—figure supplement 1E ) . In contrast , un-amputated worms cultured in parallel maintained a relative abundance of Bacteroidetes until the final 2-week time point . This eventual shift is likely the result of prolonged static culture conditions . Next , we analyzed proportional changes in the top genera of intact and regenerating worms . The genera Candidatus Amoebophilus and Segetibacter , both members of the phylum Bacteroidetes , decreased in abundance 3 days after amputation ( Figure 1D , Figure 1—figure supplement 1F ) . In parallel , the genera Acidovorax , Chryseobacterium , Pseudomonas , Vogesella , and Oxalobacter exhibited transient or sustained increases in regenerating worm fragments 3 to 7 days after amputation ( Figure 1D , Figure 1—figure supplement 1G , H ) . Thus , the expansion of Proteobacteria observed in both regeneration and static culture time courses can be traced to common genera amplified under both conditions . To determine how the bacterial composition of the asexual CIW4 lab strain compared to that of wild planaria , we performed 16 s sequencing on individual sexual S . mediterranea samples collected from multiple sites in Sardinia . Negative controls were performed in tandem during sample and library preparation to exclude contamination from resident lab strains . We succeeded in amplifying and sequencing 16 s libraries from 5 out of 8 worms covering 2 collection sites ( Figure 1—source data 1 ) . Phyla composition varied between individual animals . Proteobacteria was either the most abundant ( 3 out of 5 worms ) or the second most abundant ( 2 out of 5 worms ) phyla across all worm samples ( Figure 1—figure supplement 1I ) . Bacteroidetes or Firmicutes were the dominant phyla in the two cases in which Proteobacteria was not the most abundant . These differences in phyla composition were also reflected in the most abundant bacterial genera across individuals . The genera in the highest proportion was Rhodoferax , Lactococcus , Plesiomonas , Giesbergeria , and Candidatus amoebophilus in worms 1 , 2 , 3 , 4 , and 5 , respectively ( Figure 1E ) . This level of variation in endogenous bacterial composition amongst individuals is consistent with previous observations in other animals ( Caporaso et al . , 2011; Human Microbiome Project Consortium , 2012; Wong et al . , 2013 ) . Furthermore , the phylum Proteobacteria was proportionally higher in wild collected versus lab-raised worms consistent with observations flies ( Wong et al . , 2013 ) . Interestingly , we found that the genera that increased during transitions in culture conditions and regeneration , ( Chryseobacterium , Vogesella , and Pseudomonas ) , were detectable in wild planaria samples but present at relatively low levels ( Figure 1E , Figure 1—figure supplement 1J ) . Despite this heterogeneity in bacterial composition , more bacterial genera were common across all four worms collected from site#1 in Sardinia than were unique to any individual ( Figure 1—figure supplement 1K ) . Furthermore , greater than 90% of the bacterial genera detected among all of our CIW4 lab strain planaria samples were also detected among our wild worm samples ( Figure 1—figure supplement 1L ) . With respect to wild worms , this level of overlap was ~46% , suggesting that a significant proportion of bacterial genera is distinct in our wild sexual samples . To determine possible differences between asexual versus sexual biotypes , we sampled prominent culturable bacteria from lab raised sexual planaria . We utilized sexual S2F2 genome strain planaria from a fill and drain culture system that were transitioned to traditional static culture . We detected abundant Vogesella , Pseudomonas , Aminobacter , and Hafnia bacterial colonies from sexual worms at day 3 of static culture ( Figure 1—figure supplement 1M , Table 2 ) . This limited sampling demonstrates that asexual and sexual lab cultured planaria have both similar and distinct bacterial composition . Therefore , with respect to wild sexual bacteria , it is unclear to what extent the observed differences with asexual lab planaria are attributable to ( 1 ) intrinsic microbiome differences between the asexual and sexual biotypes , ( 2 ) region-specific bacterial genera , and/or ( 3 ) shifts in genera during lab culture . Importantly , our analyses demonstrate that the vast majority of genera from CIW4 lab strain planaria are common to those of wild type sexual worms . 10 . 7554/eLife . 16793 . 007Table 2 . 16 s rDNA homology of emergent bacterial strains . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 007Planarian sourceColony morphologyTop 10 sequence hitsMax scoreTotal scoreQuery coverE valueIdentAccessionSexual strain S2F2Large WhiteHafnia paralvei strain ATCC 29927 16 sribosomal RNA gene , partial sequence1991199198%096%NR_116898 . 1Obesumbacterium proteus strain 42 16 s ribosomal RNA gene , partial sequence1932193298%096%NR_025334 . 1Hafnia alvei strain JCM 1666 16 s ribosomal RNA gene , partial sequence1927192798%096%NR_112985 . 1Obesumbacterium proteus strain NCIMB 8771 16 s ribosomal RNA gene , partial sequence1927192798%095%NR_116603 . 1Hafnia alvei strain ATCC 13337 16 s ribosomal RNA gene , complete sequence1917191798%095%NR_044729 . 2Ewingella americana strain CIP 81 . 94 16 s ribosomal RNA gene , complete sequence1908190893%097%NR_104925 . 1Rouxiella chamberiensis 16 s ribosomal RNA , partial sequence1881188193%097%NR_135871 . 1Hafnia psychrotolerans strain DJC1-1 16 s ribosomal RNA , partial sequence1875187593%097%NR_134741 . 1Serratia liquefaciens strain ATCC 27592 16 s ribosomal RNA gene , complete sequence1871187198%095%NR_121703 . 1Serratia grimesii strain DSM 30063 16 s ribosomal RNA gene , partial sequence1871187198%095%NR_025340 . 1Medium yellowish beigePseudomonas peli strain R-20805 16 s ribosomal RNA gene , partial sequence2002200298%097%NR_042451 . 1Pseudomonas guineae strain M8 16 s ribosomal RNA gene , partial sequence1953195398%096%NR_042607 . 1Pseudomonas anguilliseptica strain S 1 16 s ribosomal RNA gene , partial sequence1908190898%096%NR_029319 . 1Pseudomonas cuatrocienegasensis strain 1N 16 s ribosomal RNA gene , partial sequence1897189798%095%NR_044569 . 1Pseudomonas pseudoalcaligenes strain Stanier 63 16 s ribosomal RNA gene , partial sequence1868186898%095%NR_037000 . 1Pseudomonas pseudoalcaligenes strain NBRC 14167 16 s ribosomal RNA gene , partial sequence1866186698%095%NR_113653 . 1Pseudomonas indoloxydans strain IPL-1 16 s ribosomal RNA gene , partial sequence1866186698%095%NR_115922 . 1Pseudomonas alcaligenes strain ATCC 14909 16 s ribosomal RNA gene , partial sequence1864186498%095%NR_114472 . 1Pseudomonas alcaligenes strain NBRC 14159 16 s ribosomal RNA gene , partial sequence1864186498%095%NR_113646 . 1Pseudomonas composti strain C2 16 s ribosomal RNA gene , partial sequence1864186498%095%NR_116992 . 1Small Beige with circleVogesella mureinivorans strain 389 16 s ribosomal RNA gene , partial sequence2039203999%097%NR_104556 . 1Vogesella perlucida strain DS-28 16 s ribosomal RNA gene , partial sequence2039203999%097%NR_044326 . 1Vogesella amnigena strain Npb-02 16 s ribosomal RNA , partial sequence1967196799%096%NR_137334 . 1Vogesella oryzae strain L3B39 16 s ribosomal RNA gene , partial sequence1934193499%095%NR_135212 . 1Vogesella lacus strain GR13 16 s ribosomal RNA gene , partial sequence1895189599%095%NR_116268 . 1Vogesella fluminis strain Npb-07 16 s ribosomal RNA gene , partial sequence1890189099%095%NR_109463 . 1Vogesella indigofera strain ATCC 19706 16 s ribosomal RNA gene , complete sequence1869186999%094%NR_040800 . 1Vogesella alkaliphila strain JC141 16 s ribosomal RNA gene , partial sequence1862186299%094%NR_108891 . 1Gulbenkiania mobilis strain E4FC31 16 s ribosomal RNA gene , complete sequence1805180597%094%NR_042548 . 1Gulbenkiania indica strain HT27 16 s ribosomal RNA gene , partial sequence1762176297%093%NR_115769 . 1Sexual strain S2F2Tiny WhiteCarbophilus carboxidus strain Z-1171 16 s ribosomal RNA gene , complete sequence1951195198%096%NR_104931 . 1Aminobacter aminovorans strain DSM 7048 16 s ribosomal RNA gene , partial sequence1951195198%096%NR_025301 . 1Aminobacter lissarensis strain CC495 16 s ribosomal RNA gene , complete sequence1945194598%096%NR_041724 . 1Aminobacter niigataensis strain DSM 7050 16 s ribosomal RNA gene , partial sequence1940194098%096%NR_025302 . 1Aminobacter aganoensis strain TH-3 16 s ribosomal RNA gene , partial sequence1940194098%096%NR_028876 . 1Aminobacter anthyllidis strain STM4645 16 s ribosomal RNA gene , partial sequence1934193498%096%NR_108530 . 1Aminobacter ciceronei strain IMB-1 16 s ribosomal RNA gene , complete sequence1929192998%096%NR_041700 . 1Mesorhizobium australicum strain WSM2073 16 s ribosomal RNA gene , complete sequence1882188294%096%NR_102452 . 1Mesorhizobium qingshengii strain CCBAU 33460 16 s ribosomal RNA gene , partial sequence1882188294%096%NR_109565 . 1Mesorhizobium shangrilense strain CCBAU 65327 16 s ribosomal RNA gene , partial sequence1882188294%096%NR_116163 . 1Asexual strain CIW4Small Beige with circleVogesella perlucida strain DS-28 16 s ribosomal RNA gene , partial sequence2167216794%097%NR_044326 . 1Vogesella mureinivorans strain 389 16 s ribosomal RNA gene , partial sequence2156215694%097%NR_104556 . 1Vogesella lacus strain GR13 16 s ribosomal RNA gene , partial sequence2019201995%095%NR_116268 . 1Vogesella oryzae strain L3B39 16 s ribosomal RNA gene , partial sequence2013201395%095%NR_135212 . 1Vogesella fluminis strain Npb-07 16 s ribosomal RNA gene , partial sequence2012201294%095%NR_109463 . 1Vogesella indigofera strain ATCC 19706 16 s ribosomal RNA gene , complete sequence1989198994%094%NR_040800 . 1Vogesella alkaliphila strain JC141 16 s ribosomal RNA gene , partial sequence1980198094%094%NR_108891 . 1Gulbenkiania mobilis strain E4FC31 16 s ribosomal RNA gene , complete sequence1884188494%093%NR_042548 . 1Pseudogulbenkiania gefcensis strain yH16 16 s ribosomal RNA gene , partial sequence1857185795%092%NR_118145 . 1Aquaphilus dolomiae strain LMB64 16 s ribosomal RNA gene , partial sequence1855185594%093%NR_118538 . 1Medium YellowChryseobacterium lactis strain KC1864 16 s ribosomal RNA gene , partial sequence2017201798%096%NR_126256 . 1Chryseobacterium viscerum strain 687B-08 16 s ribosomal RNA gene , partial sequence2012201298%095%NR_117206 . 1Chryseobacterium tructae strain 1084-08 16 s ribosomal RNA gene , partial sequence1995199598%095%NR_108531 . 1Chryseobacterium oncorhynchi strain 701B-08 16 s ribosomal RNA gene , partial sequence1995199598%095%NR_108481 . 1Chryseobacterium ureilyticum strain F-Fue-04IIIaaaa 16 s ribosomal RNA gene , partial sequence1980198098%095%NR_042503 . 1Chryseobacterium indologenes strain NBRC 14944 16 s ribosomal RNA gene , partial sequence1969196998%095%NR_112975 . 1Chryseobacterium gleum strain NBRC 15054 16 s ribosomal RNA gene , partial sequence1967196798%095%NR_113722 . 1Chryseobacterium gleum strain CCUG 14555 16 s ribosomal RNA gene , partial sequence1967196798%095%NR_042506 . 1Chryseobacterium indologenes strain LMG 8337 16 s ribosomal RNA gene , partial sequence1964196498%095%NR_042507 . 1Chryseobacterium artocarpi strain UTM-3 16 s ribosomal RNA , partial sequence1960196098%095%NR_134001 . 1Asexual strain CIW4Large BeigePseudomonas fluorescens Pf0-1 strain Pf0-1 16 s ribosomal RNA , complete sequence2242224294%099%NR_102835 . 1Pseudomonas koreensis strain Ps 9-14 16 s ribosomal RNA gene , partial sequence2220222094%098%NR_025228 . 1Pseudomonas reinekei strain MT1 16 s ribosomal RNA gene , partial sequence2217221794%098%NR_042541 . 1Pseudomonas moraviensis strain 1B4 16 s ribosomal RNA gene , partial sequence2215221594%098%NR_043314 . 1Pseudomonas vancouverensis strain DhA-51 16 s ribosomal RNA gene , partial sequence2215221594%098%NR_041953 . 1Pseudomonas helmanticensis strain OHA11 16 s ribosomal RNA gene , partial sequence2193219394%098%NR_126220 . 1Pseudomonas baetica strain a390 16 s ribosomal RNA gene , partial sequence2193219394%098%NR_116899 . 1Pseudomonas jessenii strain CIP 105274 16 s ribosomal RNA gene , partial sequence2185218594%098%NR_024918 . 1Pseudomonas umsongensis strain Ps 3-10 16 s ribosomal RNA gene , partial sequence2170217094%098%NR_025227 . 1Pseudomonas mucidolens strain NBRC 103159 16 s ribosomal RNA gene , partial sequence2167216794%098%NR_114225 . 1 We quantitated endogenous planarian bacterial levels by either 16 s rDNA qPCR or plating tissue homogenates onto LB plates . Removal of animals from the recirculation culture system elicited a 17-fold expansion in 16 s rDNA over three days that was largely mitigated by gentamycin treatment ( Figure 1—figure supplement 1N ) . Additionally , both removal from the recirculation culture system and amputation elicited a 2- to 3-log fold increase in score-able bacterial species per worm within 3 days ( Figure 1—figure supplement 1O , P ) . Representative colony 16 srDNA amplification and sequencing revealed that the observed bacterial bloom was dominated by three bacterial species belonging to the following genera: Vogesella , Chryseobacterium , and Pseudomonas ( Figure 2A , Table 2 ) . The expansion of these bacteria during culture transition was highly reproducible in multiple withdrawals from two independent recirculation culture systems ( data not shown ) . These findings are consistent with our 16 s rDNA sequencing data showing proportional increases of these genera within worms upon exit from recirculation culture ( Figure 1C , Figure 1—figure supplement 1C , D ) . Thus , shifts in microbial composition are coincident with a robust expansion in both absolute bacterial levels and particular bacterial species . 10 . 7554/eLife . 16793 . 008Figure 2 . Bacterial infection compromises planarian tissue homeostasis and regeneration . ( A ) Bacterial CFU quantification of prevalent bacterial strains following exit from the recirculation system ( n = 4 homogenates from 4 worms per time point , experiment independently repeated > 4 times ) ; Vogesella ( red ) , Chryseobacterium ( orange ) , and Pseudomonas ( green ) . ( B ) Depiction and enumeration of pathological stages following bacterial infection . Stage 2 . 5 refers to worms that reach stage 3 but ultimately regenerated anterior structures . ( C ) Comparison of the effects of infection of 1e8 CFU/ml Vogesella , Chryseobacterium , or Pseudomonas ( green ) on pathological stage progression over time ( n = 15–25 , experiment independently repeated > 3 times ) . ( D ) Anti-Pseudomonas antibody staining ( green ) and DAPI nuclear counterstain ( blue ) of surface and gut epithelia following infection ( n = 2 , experiment independently repeated > 2 times ) . ( E ) Histological sections stained with Alcian Blue/ PAS following Pseudomonas infection ( n = 2 ) . ( F ) Representative images depicting the effects of increasing concentrations of Pseudomonas on regenerating head , trunk , and tail fragments . Worms were amputated 1 day following infection and images were taken seven days after amputation ( n = 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 00810 . 7554/eLife . 16793 . 009Figure 2—figure supplement 1 . Effects of bacterial infection on worm tissue homeostasis . ( A ) Images of worms 30 days post recirculation culture administered food , antibiotic , or neither ( experiment independently repeated > 3 times ) . Bacterial CFU/worm following infection with Vogesella either by ( B ) direct administration to planaria water or ( C ) by feeding bacteria mixed with beef liver paste ( CFU determined from a pool of four worms per time point ) . ( D ) Percentage of worms exhibiting demarcated phenotypes following infection ( n = 15–25 , experiment independently repeated > 3 times ) . Phenotypes are color coded as follows: blue = normal , teal = posterior lesion , dark green = anterior lesion , light green = head regression , orange = partial lysis , red = full lysis , light green+squares = head regeneration following regression . ( E ) Anti-Pseudomonas antibody staining of histological sections of planaria immediately following the consumption of beef liver paste alone or mixed with the specified bacteria ( n = 2 ) . ( F ) Labeling of neoblasts via piwi WISH during Pseudomonas infection and pathological progression ( n = 5–10 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 009 Following the transition from recirculation to static culture , poorly managed worms are highly susceptible to the development of dorsal lesions , tissue degeneration , and lysis . This susceptibility can be exacerbated by increased feedings or temporarily mitigated by antibiotics ( Figure 2—figure supplement 1A ) . We hypothesized the etiology of this tissue degeneration lie in the observed bacterial bloom upon exit from recirculation system ( Figures 1C , 2A , Figure 1—figure supplement 1C , D ) . To determine whether this bacteria was sufficient to induce tissue homeostatic defects we introduced emergent Vogesella , Chryseobacterium , or Pseudomonas bacterial strains to planaria water immediately upon transition to static culture . Worms were infected at a concentration of 1E8 CFU/ml and washed every 3–4 days for administration of fresh planaria water containing bacteria . This method of infection lead to a dose dependent increase in endogenous bacterial levels over time ( Figure 2—figure supplement 1B ) similar to those observed in the worms following removal from the recirculation system ( Figure 2A ) . In comparison feeding of bacteria mixed with beef liver paste elicited an initial spike and steady decline in bacterial levels as previously reported ( Abnave et al . , 2014 ) , but the endogenous bacterial levels of control fed worms also increased to match that of bacteria fed worms ( Figure 2—figure supplement 1C ) . This expansion of endogenous bacteria complicates analysis and obfuscates effects of tested bacterial strains on host biology . We therefore utilized supplementation of bacteria directly in planaria water to determine effects on tissue homeostasis . In response to infection , worms exhibited pathological defects largely resembling planaria of declining health in static culture ( Figure 2—figure supplement 1A ) . We categorized the observed pathologies into the following discrete categories: normal , posterior lesions , anterior lesions , head regression , partial lysis , and full lysis . Infected worms exhibited a consistent development of tissue defects over time varying in the rate of progression by bacterial strain ( Figure 2—figure supplement 1D ) . Generally phenotypically normal worms initially developed either posterior or anterior dorsal lesions , anterior lesions increased in severity to result in full head regression , and worms exhibiting head regression ultimately initiated full tissue lysis . The notable exception was the incidence of anterior blastema formation and regeneration as opposed to lysis in a subset of Pseudomonas infected worms with regressed heads . These data demonstrate that both the rate of pathological progression and eventual outcome of the infection in planaria varies by bacterial strain . We were able to infer a chronological hierarchy and assigned specific stages to the observed progressive tissue degeneration: Normal = 0 , Posterior Lesion = 1 , Anterior Lesion = 2 , Head regression = 3 , Partial Lysis = 4 , Full Lysis = 5 ( Figure 2B ) . Worms that regenerated lost anterior structures were assigned stage 2 . 5 , reflecting tissue restoration following stage 3 head regression ( Figure 2B ) . Using these stages we plotted tissue degeneration over time , highlighting differential effects of tested strains on planarian tissue homeostasis ( Figure 2C ) . We chose Pseudomonas for subsequent analyses based on the following criteria: it offered the opportunity to study loss and regeneration of tissues during infection , it is the most well studied of the tested strains , and members of this genus are relevant human pathogens ( Driscoll et al . , 2007 ) . We used a validated anti-Pseudomonas antibody and histological sections in order to determine tissues likely affected by infection ( Figure 2—figure supplement 1E ) . Initially , Pseudomonas mainly localized to the barrier epithelia of the worm in the epidermis and intestine ( Figure 2D ) . By 3 days some Pseudomonas appear to have crossed epithelial barriers and localize to the mesenchyme . In order to study the effects on the mucosal epithelia , we performed Alcian Blue/ Periodic Acid-Schiff staining of histological sections following infection ( Figure 2E ) . Infected worms exhibited a gradual apical compression of the morphology of outer epithelial cells with a slight distortion in the organization of the basement membrane . In the intestine , Pseudomonas infection resulted in a pronounced lumenal constriction with changes in epithelial organization and morphology over time . Thus , our data indicate that barrier epithelia represent the initial point of interaction with Pseudomonas and that sustained exposure to this bacterium results in observable cellular and tissue morphology changes . Given that Pseudomonas infection progressively compromised tissue homeostasis ( Figure 2C , E ) , we next determined whether neoblasts and regeneration potential were similarly affected . We analyzed neoblast levels and distribution using WISH analysis . While infected worms exhibited abundant numbers of piwi+ neoblasts both prior to and following tissue degeneration , the frequency of neoblasts within the interior of the animal relative to the periphery appeared to progressively decrease during infection ( Figure 2—figure supplement 1F ) . To determine how bacterial challenge altered regeneration potential , we infected worms with increasing dosages of Pseudomonas and amputated above and below the pharynx one day after exposure to bacteria . Phenotypic effects on resulting head , trunk , and tail fragments were documented one week later . Pseudomonas infection inhibited tissue regeneration of amputated fragments in a dose dependent manner ( Figure 2F ) . Interestingly , we observed that segments of the worm along the A/P axis had different tolerances for bacterial infection . Head fragments were the most susceptible , tail fragments exhibited intermediate sensitivity , and trunk fragments were the most resistant . These data suggest that differences in tissue composition along the A/P axis and/or the extent of tissue generation versus remodeling required for re-establishment of the body plan confer resistance or susceptibility to infection . Given its effects on tissue regeneration , we chose the dosage of 2e8 CFU/ml Pseudomonas for subsequent experiments . To identify the host transcriptional changes underlying shifts in planarian microbial composition , we conducted RNAseq analyses on worms transferred from recirculation to static culture in the presence or absence of the antibiotic gentamycin ( Figure 3A , Figure 3—source data 1 ) . This transition results in an increase in absolute bacterial levels characterized by a relative decrease in the phyla Bacteroidetes , a proportional increase in the phyla Proteobacteria , and the emergence of newly dominant bacterial genera and species ( Figures 1B , C , 2A , Figure 1—figure supplement 1N , O ) . Clustering of the 1 , 218 genes significantly altered relative to static culture at day 0 revealed transcriptional patterns that were either dependent ( clusters 1 , 2 , 4 , 5 , 7 , 8 , and 9 ) or independent ( clusters 3 and 6 ) of antibiotic treatment ( Figure 3B , C , Figure 3—source data 1 ) . Gene clusters sensitive to antibiotic treatment could be divided based on up-regulation ( clusters 1 , 4 , 8 , 9 ) or down-regulation ( clusters 2 , 5 , 7 ) , and then further subdivided into phases of early ( clusters 4 and 7 ) , mid ( cluster 8 ) , late ( clusters 1 and 5 ) , or sustained ( clusters 2 and 9 ) expression changes . Of particular interest was cluster 1 , consisting of genes upregulated gradually after exit from the recirculation system in the absence of antibiotic ( Figure 3D ) . This cluster had clear potential enrichment in genes that correlated with increased bacterial levels over time . Interestingly , this gene cohort contained two transcripts with homology to peptidoglycan recognition proteins ( pgrps ) , upstream receptors of the IMD pathway in D . melanogaster . Along with Toll signaling , the IMD pathway represents a conserved branch of the innate immune system with analogous architecture to the inflammatory signaling pathways in vertebrates ( Panayidou and Apidianakis , 2013 ) . RT/qPCR and WISH analysis confirmed that Pseudomonas infection significantly induced one of these homologous receptors , that we designated pgrp-4 ( Figure 3E–G ) . These data suggest that conserved components of the innate immune and inflammatory signaling pathways may dynamically respond to shifts in the microbial composition in planaria . This data as well as the established role of this pathway in immunity and apoptosis lead us to take a candidate based investigation of the role of inflammatory signaling pathways in altered tissue homeostasis and regeneration during infection . 10 . 7554/eLife . 16793 . 010Figure 3 . Elucidating bacterial contribution to the transcriptional changes underlying the transition of worms from recirculation to static culture conditions . ( A ) Diagram depicting comparison of RNAseq samples of planaria following exit from a recirculation culture system in the absence or presence of the antibiotic gentamycin ( n = 4 replicates each containing 4 planaria ) . ( B ) Hierarchical clustering of RNAseq results displaying significant clusters ( 1 through 9 , y axis ) . ( C ) Visualization of gene expression patterns of significant clusters . ( D ) Differentially expressed annotated genes from cluster 1 ( Day 0 vs Day 4 adj . p-value <0 . 05 ) . Peptidoglycan recognition protein genes are highlighted in bold . RT and QPCR of ( E ) pgrp-1 and ( F ) pgrp-4 following Pseudomonas infection ( n = 3 biological replicates , 3 technical replicates , * = t-test p<0 . 05 ) . ( G ) WISH of pgrp-4 following Pseudomonas infection ( n = 5–7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 01010 . 7554/eLife . 16793 . 011Figure 3—source data 1 . RNAseq analysis of worms during recirculation to static culture transition . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 011 The identification of genes modulated by a bacterial infection as well as the appearance of a visibly score-able progression of tissue degeneration presented an opportunity to elucidate genes mediating this process via RNAi-mediated genetic interference . Given their low levels of Proteobacteria and in particular the emergent pathogenic strains we identified ( Figures 1B , C , 2A ) , planaria from the recirculation culture system presented the greatest opportunity for studying the maximal differential effect caused by Pseudomonas infection in terms of phenotypic and underlying molecular responses . However , in order to use these animals and conditions for an RNAi screen to identify immune response mediators , we needed to devise a way to reduce the bacterial bloom observed after removal from recirculation culture without using antibiotics ( Figure 4—figure supplement 1A ) . As we observed in our experiments , the antibiotic gentamycin substantially altered endogenous microbiota composition , diminishing the phylum Bacteroidetes while enriching for Proteobacteria , including Pseudomonas itself ( Figure 1B , C ) . Its usage would preclude analysis of phenotypes that are dependent on dynamic shifts in endogenous microbiota . Additionally , a shared recirculation culture system precludes the delivery of multiple , distinct dsRNAs to separate experimental cohorts of planaria . To overcome these limitations , we developed a unidirectional flow system in which worms were maintained in multiple individual containers and steadily flushed with fresh planaria water ( Figure 4—figure supplement 1A–C ) . Culture of planaria under a steady flow rate prevented a 4-log increase in culturable bacteria levels observed in static conditions following exit from the recirculation culture system ( Figure 4—figure supplement 1D ) . Additionally , introduction of static worms to flow vessels resulted in a nearly 2 log reduction in culturable bacterial levels over 5 days ( Figure 4—figure supplement 1E ) , and worms administered 3 feedings of control dsRNA mixed with beef liver paste maintained levels comparable to those observed in recirculation culture ( Figure 4—figure supplement 1F ) . We conclude from these findings that the unidirectional flow system provides the necessary conditions to systematically carry out an RNAi screen of infection-modulated genes and to effectively analyze their resulting phenotypes ( Figure 4—figure supplement 1A ) . We carried out a candidate gene RNAi screen for mediators of planarian tissue degeneration in response to infection by focusing on conserved members of the mammalian inflammatory signaling cascade and the analogous , infection-responsive IMD pathway in Drosophila melanogaster . We identified 32 homologous genes using reciprocal BLAST of either Homo sapiens or D . melanogaster reference genes and categorized them with respect to known functions in mediating ( activators ) or reducing ( inhibitors ) signal transduction ( Figure 5—source data 1 ) ( Kopp et al . , 1999; Mogensen , 2009; Dai et al . , 2012; Xie , 2013; Fernando et al . , 2014; Herrington and Nibbs , 2016 ) . Additional genes with roles in regulating apoptosis , bcl2-1 and bcl2-2 , and one TIR domain containing gene , tehao , were also included in our analysis ( Luo et al . , 2001; Pellettieri et al . , 2010 ) . The RNAi screen was performed in low septic conditions outside of the recirculation culture as follows ( Figure 4 ) . Worms in individual flow vessels were administered 3 dsRNA feedings . Four days after RNAi treatment , worms were infected with Pseudomonas and ocularly scored for pathological progression using established criteria ( Figure 2B ) every 1–3 days over a period of 30 days . Scoring resulted in a series of 5 to 6 integers representing the state of each worm in the dish for each RNAi condition for each day . By taking the average score of the worms on each day and converting that average to a color-based scale from blue to red to denote pathological severity , we were able to reduce the complexity of the data to facilitate its visualization . For each RNAi condition , color-coded average scores for each day were stacked along the y-axis in ascending order from top to bottom forming a column . Individual RNAi columns were sorted in ascending order from left to right along the x-axis by their average column scores . The result was a heatmap that ranked the pathological progression of the worms following RNAi treatment in order of increasing overall severity over time . Simply stated , genes in which RNAi treatment reduces pathological progression ( presumptive activators ) reside left of the control , while genes in which RNAi treatment enhances pathological progression ( presumptive inhibitors ) reside right of the control . 10 . 7554/eLife . 16793 . 012Figure 4 . Diagram depicting workflow and data analysis of an RNAi screen to identify genes modulating pathological progression . Planaria are transferred directly from the recirculation system to a flow culture system permitting maintenance of a reduced septic state during dsRNA feedings . Following 3 RNAi feedings , planaria are removed for infection with Pseudomonas . Observations of pathological stages are recorded every 1 to 3 days , and planaria are replenished with fresh water containing Pseudomonas every 3 to 4 days . Following enumeration of pathological stages , data for each day is reduced to an average pathological score and converted to a heat color code . Days for each RNAi condition are aligned in ascending order along the y-axis of a column . The average score of each column is calculated and used to sort the effects of RNAi conditions in ascending order along the x-axis . The result is a heat map visualization ranking the effects of RNAi treatments on the pathological progression in response to bacterial infection in planarians over time . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 01210 . 7554/eLife . 16793 . 013Figure 4—figure supplement 1 . A novel flow culture method for planarian RNAi in low septic conditions . ( A ) Diagram depicting comparison of worm health following execution of the RNAi protocol in static versus flow culture . Worms removed from recirculation culture and placed in static culture experience a robust amplification of bacterial levels and are susceptible to lesion formation . RNAi of components that normally inhibit lesion formation or bacterial infection would exacerbate this effect . In comparison , maintenance of worms in a flow culture system can maintain the healthy , low septic state independent of RNAi manipulation . Diagram of ( B ) individual planarian flow culture vessels and ( C ) an integrated unidirectional flow system ( planaria water is administered to all cups but depicted only in the left column and the upper row to simplify visualization ) . ( D ) Comparison of bacterial CFU/worm during culture in static versus flow RNAi conditions following exit from recirculation culture . ( E ) Effects of flow culture on bacterial CFU of worms with pre-existing heightened bacterial levels . ( F ) Bacterial CFU/worm prior to and after 3 RNAi feedings in the flow culture system ( experiment independently repeated > 3 times ) . Bacterial CFUs were determined from samples of 4 pooled worms per timepoint . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 013 A total of 20/35 tested candidate genes resulted in a significant change ( 2-WAY ANOVA p<0 . 05 ) in pathological progression following infection relative to unc controls , while 2/35 genes , bcl2-1 and bcl3-1 aka NFκB-p105 , exhibited total lysis prior to infection and were not considered further ( Figure 5A ) ( Pellettieri et al . , 2010; Forsthoefel et al . , 2012 ) . Among these , 15/20 significantly reduced pathological progression while 5/20 significantly enhanced it; we termed these genes activators and inhibitors , respectively . Interestingly , the homologs of 13/15 activators ( mkk6-1 , mkk4 , p38-1 , pgrp-2 , tab1-1 , jun D , hep , tak1 , jnk , pgrp-3 , pgrp-1 , xiap , pgrp-1 ) have been previously shown to be positive regulators of inflammatory or innate immune signaling ( Figure 5B , Figure 5—source data 2 ) . Conversely , the homologs for 4/5 inhibitors ( pp6 , ppm1b , cyld-1 , ppm1a ) have been previously shown to inhibit inflammatory and innate immune signaling ( Figure 5B , Figure 5—source data 2 ) . In the case of the inhibitor traf2-1 , the homolog traf2 functions as mediator of inflammatory signaling , but it appears that the function of this gene in planaria is more akin to traf3 , which functions as an inhibitor ( Xie , 2013 ) . Overall the data suggest that conserved inflammatory signaling may be largely responsible for mediating tissue degeneration after bacterial infection . 10 . 7554/eLife . 16793 . 014Figure 5 . Heatmap depicting results of an RNAi screen for mediators of pathological progression following bacterial infection in planaria . ( A ) Heatmap depicting average pathological scores for each RNAi treatment following 2e8 CFU/ml Pseudomonas infection ( n = 5–12 worms per condition ) . Days versus average pathological score over time are aligned along the y- , and x- axes , respectively . Unc control sample is indicated by a dashed box . RNAi-targeted genes that significantly reduce or enhance pathological progression are highlighted in green and red , respectively ( 2-way ANOVA p<0 . 05 ) . ( B ) Focus on RNAi of genes that result in a significant reduction ( activators ) or enhancement ( inhibitors ) or pathological progression . ( C–F ) Heatmaps depicting the percentage of worms exhibiting pathological stage ( C ) 0 , ( D ) 3 , ( E ) 5 , and ( F ) 2 . 5 for each RNAi treatment following Pseudomonas infection . Ordering and significance based on average pathological score over time are maintained for reference . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 01410 . 7554/eLife . 16793 . 015Figure 5—source data 1 . Homologous innate immune and inflammatory genes from RNAi screen . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 01510 . 7554/eLife . 16793 . 016Figure 5—source data 2 . Pathological scores of worms in RNAi screen for mediators of infection induced tissue degeneration . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 016 Our data set also permitted us to track the progression of worms through individual pathological stages in response to infection . We mapped the percentage of worms that remained lesion free ( stage 0 ) , exhibited head regression ( stage 3 ) , fully lysed ( stage 5 ) , or regenerated their heads ( stage 2 . 5 ) onto our average pathological score heatmap ( Figure 5C–F ) . Control worms remained lesion free for up to 12 days after infection ( Figure 5C ) . RNAi of activators could extend this period up to 20 days . Conversely , RNAi of inhibitors reduced this lesion free period to 6 days . With respect to head regression ( stage 3 ) , RNAi of activators largely delayed the initiation of and the extended the duration of this stage while RNAi of inhibitors expedited and contracted it ( Figure 5D ) . Following head regression , control worms began undergoing lysis by day 12 , and 100% of worms lysed 27 days following infection ( Figure 5E ) . RNAi of activators delayed and reduced the incidence of lysis and 100% of mkk6-1 and mkk4 RNAi worms survived the entire 30-day duration of our assay . RNAi of inhibitors induced lysis as early as day 4 with no worms surviving after 9 to 23 days . While the vast majority of worms eventually lysed with the administered dosage of 2e8 CFU/ml Pseudomonas , a rare fraction of control worms regenerated their heads by day 23 ( Figure 5F ) . While RNAi of many activators increased the incidence , frequency , and duration of head regeneration and subsequent survival during infection , we observed no incidence of head regeneration following RNAi of the inhibitors pp6 , ppm1b , cyld-1 or traf2-1 . Amongst activators , mkk4 and mkk6 RNAi worms had the highest incidence of head regeneration while p38-1 RNAi had the lowest despite having a comparable frequency and timing of the initiation of head regression ( Figure 5D ) . In summation , our RNAi screen has identified genes that play a role in mediating tissue degeneration in response to infection and reveal distinct effects on the timing and duration of individual stages of pathological progression . Our candidate RNAi screen identified cohorts of genes that serve as activators or inhibitors of the tissue degeneration response to Pseudomonas infection . These opposing phenotypes provided an opportunity to order these components with respect to their hierarchy in mediating anterior tissue degeneration . We chose 6 activator genes ( mkk6-1 , mkk4 , p38-1 , jun D , tak1 , and jnk ) and 3 inhibitor genes ( ppm1b , pp6 , cyld-1 ) analogous to core inflammatory and innate immune signaling in H . sapiens ( Figure 6A ) ( Xue et al . , 2007; Mogensen , 2009; Dai et al . , 2012; Panayidou and Apidianakis , 2013 ) . To order these components we performed combinatorial RNAi experiments and assayed the resulting effects on anterior tissue degeneration . The dsRNA concentration of the inhibitors was reduced by 50% and balanced with a 50% increase in the concentration of activators . This allowed us to determine whether simultaneous knock down of activator genes was sufficient to overcome the sensitized tissue degeneration phenotype from partial knockdown of inhibitors . Worms were analyzed at a tissue degeneration 'tipping point' at which point the following occurred: infected control worms developed anterior lesions ( stage 2 , Mild ) , RNAi of single activators resulted in posterior or no lesions ( stage 0–1 , None ) , and RNAi of single inhibitors produced head regression and/or lysis ( stage 3–5 , Severe ) ( Figure 6B ) . The resulting phenotypic stages of worms given each of 18 combinations of inhibitor and activator dsRNA were analyzed ( Figure 6C , D , Figure 6—source data 1 ) . Median pathological scores were used to order components in an effort to ascertain phenotypic outcomes while mitigating changes due to variation in phenotypic penetrance . 10 . 7554/eLife . 16793 . 017Figure 6 . Epistatic analysis of activators and inhibitors of planarian pathological progression . ( A ) Diagram depicting relevant components of the TAK1 pathway in Homo sapiens . ( B ) Diagram depicting combinatorial RNAi experiment and assay of tissue degeneration outcome . ( C ) Representative images and median pathological stage of worms following RNAi treatment with each combination of 6 activators and 3 inhibitors 12 days post Pseudomonas infection ( n = 6–24 ) . ( D ) Table of phenotypic outcomes following combinatorial RNAi treatment and Pseudomonas infection . ( E ) Diagram depicting phenotypic hierarchy of the mediators of pathological progression in Schmidtea mediterranea ( bold = order is consistent with Homo sapiens TAK1 Pathway ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 01710 . 7554/eLife . 16793 . 018Figure 6—source data 1 . Pathological scores of worms during combinatorial RNAi analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 018 After 12 days of infection the median scores of worms given control dsRNA alone , activator dsRNA mixed with control dsRNA , or inhibitor dsRNA mixed with control dsRNA were 2 , 0 , and 3 or 4 , respectively ( Figure 6C , D ) . Introduction of activator dsRNA in combination with inhibitor dsRNA yielded clear and largely binary shifts in pathologic bacterial responses . Addition of either mkk6-1 , p38-1 , tak1 , and jnk dsRNA rescued head regression and lysis observed in worms following RNAi of pp6 alone . In contrast , knockdown of pp6 in combination with mkk4 phenocopied pp6 dsRNA treatment while knockdown with jun D yielded a partial rescue . RNAi of ppm1b in combination with mkk4 , tak1 , and jnk phenocopied the bacterial-induced head regression observed in worms with ppm1b RNAi alone . In contrast , RNAi of ppm1b in combination with mkk6-1 , p38-1 , and jun D rescued head regression and phenocopied activator dsRNA treatment alone . Finally , head regression induced by cyld-1 knockdown alone could be largely rescued by simultaneous knockdown of mkk6-1 or p38-1 but not mkk4 or tak1 . Curiously , combinatorial knockdown of cyld-1 with jun D or jnk actually accelerated tissue degeneration suggesting additional roles for these components in other aspects of tissue homeostasis . The results of our analyses allowed us to order these components with respect to their roles in mediating anterior tissue degeneration in response to infection: mkk4 , pp6 , tak1 , jnk , ppm1b , cyld-1 , jun D , mkk6-1 , and p38-1 . Remarkably , the Phenotype Hierarchy of these components is largely consistent with the TAK1 signaling pathway in humans ( Figure 6A , E ) . Notably , there are some key distinctions . Given that we did not observe any effects of mkk4 dsRNA when combined with any inhibitor dsRNA tested , our analysis placed mkk4 either upstream of or lateral to these other components with respect to tissue degeneration . Additionally , our analysis placed cyld-1 further downstream in the tissue degeneration process relative to its role in signaling in H . sapiens . The homologous deubiquitinating enzyme CYLD regulates ubiquitin mediated signaling and proteolysis . It is possible that its position within this Phenotype Hierarchy reflects a role in deubiquitination of proteins targeted for proteosomal degradation that are downstream of jnk rather than or in addition to targeting the homolog of the ubiquitin dependent kinase TAK1 ( Reiley et al . , 2007; Xue et al . , 2007 ) . While the phenotype hierarchy derived from our combinatorial RNAi analysis is internally consistent and largely resembles the TAK1 signaling pathway in H . sapiens , one must take caution in extrapolating that this relationship represents the TAK1 signaling pathway in S . mediterranea . Traditionally , studies in more genetically amenable model organisms have utilized both epistatic analysis with null alleles to order genes and co-immunoprecipitation to demonstrate protein-protein interactions in the elucidation of signaling pathways . Thus , it remains to be determined to what extent the Phenotype Hierarchy uncovered here reflects the true S . mediterranea TAK1 signaling cascade . Following the identification of conserved TAK1/MKK/p38 signaling components in the mediation of tissue degeneration planaria in response to infection , we explored the roles of these components in amputation-induced regeneration . As previously described , infection with Pseudomonas compromises phenotypic regeneration of lost tissues and remodeling of existing tissues to various extents and in a dose dependent manner ( Figure 2F ) . We examined the effects of RNAi knockdown of mkk6-1 , mkk4 , p38-1 , jun D , tak-1 , jnk , and pp6 on regeneration in the presence or absence of infection ( Figure 7 ) . The concentration of dsRNA for mkk6 , mkk4 , p38-1 , jun D , tak-1 , jnk was doubled to increase the phenotypic penetrance while the concentration of dsRNA of pp6 was halved to ensure the absence of any tissue degeneration during the initial period of infection prior to amputation . RNAi treated worms were infected with Pseudomonas for 2 days and then amputated above and below the pharynx to generate head , trunk , and tail fragments . Resulting phenotypic tissue regeneration was recorded 2 weeks later in comparison to uninfected worms for control and RNAi treated worms ( Figure 7A ) . For quantitative characterization of RNAi effects , we categorized the resulting observed phenotypes and reported the proportion of regenerating fragments displaying the following: normal regeneration , abnormal regeneration , lesions , tissue regression , and lysis ( Figure 7B ) . 10 . 7554/eLife . 16793 . 019Figure 7 . Effects of planarian pathological progression mediators on regeneration in the presence and absence of infection . ( A ) Representative images of regenerating head , trunk , and tail fragments following specified RNAi treatment in the presence or absence of Pseudomonas infection . Worms were amputated above and below the pharynx 2 days post infection and imaged 14 days post amputation ( n = 5 , non-lysed fragments shown when present ) . ( B ) Quantitation of phenotypes of regenerating head , trunk , and tail fragments following specified RNAi treatment in the presence or absence of Pseudomonas infection ( dark blue = normal regeneration , light blue = abnormal regeneration , green = lesions , orange = tissue regression , red = lysis ) . Animals were scored at the same time when representative images were taken at 16dpi 14dpa . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 019 The vast majority of head , trunk , and tail fragments regenerated normally following RNAi of target genes in the absence of infection ( Figure 7A , B ) . Notably 20% of trunk and tail fragments lysed after pp6 RNAi and 40–60% of trunk fragments exhibited abnormal regeneration ( forked posterior balstemas ) following mkk4 and jnk RNAi . Knockdown of other positive regulators within this signaling module ( mkk6-1 , p38-1 , jun D , and tak1 ) did not phenocopy these regeneration defects so they are likely the result of an additional role for jnk . This is consistent with previous studies highlighting defects in planarian regeneration following jnk RNAi ( Almuedo-Castillo et al . , 2014; Tejada-Romero et al . , 2015 ) . The relatively minor defects we observed may reflect different extents of jnk knockdown as the result of direct injection versus feeding of dsRNA . In contrast to the relative lack regeneration phenotypes observed in the absence of infection , RNAi effects were much more pronounced in regenerating fragments from infected worms . Overall , RNAi of the activators largely prevented infection-associated defects in regeneration while RNAi of the inhibitor pp6 exacerbated them ( Figure 7A , B ) . Following infection , 100% of regenerating head fragments in control worms lysed while RNAi of mkk6-1 , p38-1 , and jnk not only prevented tissue lysis but re-established normal regeneration for 80% of the fragments . Similarly , RNAi of mkk4 , tak1 , and jun D reduced infection-associated defects in head fragment regeneration observed in controls . RNAi of p38-1 or tak-1 prevented infection-associated regenerative defects in tail fragments while mkk6-1 or jun D RNAi reduced them . Regenerating trunk fragments were relatively resistant to the effects of the infection and RNAi of activators could largely reverse the minor defects observed in controls . Consistent with its role as an inhibitor of TAK1/MKK/p38 signaling , RNAi of pp6 exacerbated infection-associated defects resulting in lysis of 100% , 60% , and 100% of regenerating head , trunk , and tail fragments , respectively ( Figure 7A , B ) . These results demonstrate that conserved TAK1/MKK/p38 signaling components also play a pivotal role in determining regenerative outcomes induced by amputation during infection in a manner consistent with the positional sensitivity of these regenerating fragments along the A/P axis . WISH analysis of members of the TAK1/MKK/p38 signaling pathway revealed that these components were expressed in largely overlapping patterns of tissues along the planarian body axis . The most prominent expression pattern was evident in the planarian gut which was shared in both ordered activators ( tak1 , jnk , jun D , mkk6-1 , and p38-1 ) and inhibitors ( pp6 , ppm1b , and cyld-1 ) ( Figure 8A , Figure 8—figure supplement 1A , B ) . Weaker staining was also visible in the mesenchymal space between the gut and the epidermis with a slight anterior enrichment ( Figure 8A , Figure 8—figure supplement 1C ) . While mRNA analysis is useful in identifying structures competent to respond to stimulation , it fails to resolve those cells or tissues dynamically responsive to infection . Fortunately , the phosphorylation-dependent signal transduction of the TAK1/MKK/p38 pathway presented an opportunity to analyze dynamic activation via phospho-antibody based detection of the downstream target , p38 . The human homolog of the kinase p38 is activated via phosphorylation at Thr180 and Tyr182 , a region highly conserved in planaria ( Figure 8—figure supplement 2A ) ( Han et al . , 1994 ) . We tested one polyclonal and one monoclonal phospho-p38 ( Thr180/Tyr182 ) antibody for reactivity against the planarian phospho-epitope . UV treatment is a common inducer of P-p38 and western blot analysis revealed both tested antibodies reacted to an induced band at 45kDa corresponding to the size of planarian p38 ( Figure 8—figure supplement 2B ) . In situ antibody staining revealed that both polyclonal and monoclonal antibodies recognized discrete cells above the pharynx ( Figure 8B , Figure 8—figure supplement 2C ) . Initially tested bleaching treatments that were conducive to WISH protocols yielded high background fluorescence in the epidermis precluding analysis of reactivity in this peripheral tissue at this point ( Figure 8—figure supplement 2C ) . Treatment with UV irradiation induced a broad antibody staining pattern throughout the gut , mirroring the localized expression of the p38-1 transcript ( Figure 8A , Figure 8—figure supplement 1A ) . We utilized the monoclonal antibody for subsequent analyses since it produced stronger staining . 10 . 7554/eLife . 16793 . 020Figure 8 . Signaling dynamics in gut tissue during infection and regeneration visualized with a phospho-p38 antibody . ( A ) Colorimetric WISH of genes comprising the TAK/MKK/p38 signaling module ( n > 3 worms ) . Combinatorial fluorescent whole mount ISH and immuno-labeling of gut marker mat1 and phospho-p38 ( P-p38 ) ( B ) prior to stimulation , ( C ) following infection with 2e8 CFU/ml Pseudomonas , or ( D ) after amputation ( n = 2–11 worms ) . ( E ) Higher magnification images of P-p38+ cells co-labeled with gut markers mat1 , porc , hnf4 , and nkx-2 . 2 following amputation ( 35-65 mpa ) . Quantification of P-p38 ( F ) Fraction Volume and ( G ) Average Intensity following infection or amputation ( * = t-test p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 02010 . 7554/eLife . 16793 . 021Figure 8—figure supplement 1 . Co-expression of TAK1/MKK/p38 pathway components . Fluorescent double WISH of mkk6-1/p38-1 , pp6/cyld-1 , p38-1/cyld-1 , and pp6/mkk6-1 in ( A ) whole worms and ( B ) posterior gut tissue ( n = 6–8 worms ) . ( C ) Fluorescent double WISH of pp6/cyld-1 and p38-1/cyld-1 in the anterior mesenchyme ( n = 6–8 worms ) . Identical settings used for posterior gut and anterior mesenchyme for comparison of cells surrounding the gut . Negative controls treated identically but lacking DIG and DNP labeled probes were used to determine non-specific background staining . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 02110 . 7554/eLife . 16793 . 022Figure 8—figure supplement 2 . Validation of an antibody that recognizes planarian P-p38 . ( A ) Conservation of the amino acids surrounding the Thr180/Tyr182 p38 phosphorylation in Homo sapiens and Schmidtea mediterranea . ( B ) Western blot of planarian protein lysate that is untreated or treated with 30 min of UV irradiation and stained with P-p38 antibody #1 ( polyclonal ) or P-p38 antibody #2 ( monoclonal ) . ( C ) Whole mount P-p38 antibody of staining of planaria that are untreated or treated with 30 min of UV irradiation ( n > 4 worms ) . ( D ) Whole mount P-p38 antibody staining of planaria following unc or p38-1 RNAi and subjected to the following treatments: untreated , 30 min post amputation , 30 min post UV irradiation ( n > 3 worms ) . Confocal images of ( E ) infected posterior gut or ( F ) amputated fragments following WISH of the gut marker mat1 and anti-P-38 staining . ( H ) Whole mount ISH of porc and P-p38 staining of gut tissue injured via poking or lateral cut ( n > 2 worms ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 02210 . 7554/eLife . 16793 . 023Figure 8—figure supplement 3 . High resolution analysis of P-p38 signaling in gut tissue following amputation . Whole mount ISH of porc and P-p38 staining of anterior gut tissue 5 min post amputation . Z-series of images . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 023 We tested antibody specificity by determining if RNAi against planarian p38-1 could eliminate the observed signal ( Figure 8—figure supplement 2D ) . Worms fed beef liver pasted mixed with in vitro synthesized dsRNA had a relative lack of discrete cell staining near the pharyngeal intestinal border , while worms fed E . coli producing dsRNA had robust staining in the anterior gut . We found that in addition to UV irradiation , amputation of worms produced an induction of signal in discrete cells throughout the gut within 30 min . Under all observed instances of signal induction ( E . coli feeding , amputation , and UV irradiation ) we found no detectable staining after p38-1 RNAi ( Figure 8—figure supplement 2D ) . Thus , we conclude that this antibody specifically recognizes P-p38 in planaria . Using the characterized P-p38 antibody , we analyzed phospho-signaling in the planarian gut during infection . Prior to Pseudomonas infection , we observed both P-p38+ cells that co-labeled with gut marker mat1 as well as P-p38+mat1- cells adjacent to the gut ( Figure 8B , C ) . P-p38+ cells were relatively constant for the first 3 days of infection ( Figure 8C ) . Six days after infection , large nuclear dense mounds consisting of P-p38+ cells could be observed in the anterior lumen of the gut with more intense lumenal staining also seen in the posterior and regions ( Figure 8C , Figure 8—figure supplement 2E ) . Eight days after infection the gut appeared more distorted and P-p38 staining had consolidated to distinct cells throughout the gut ( Figure 8C , Figure 8—figure supplement 2E ) . We compared the kinetics of P-p38 activation in response to injury versus bacterial infection . We observed a robust P-p38 staining throughout the interior gut of the head , trunk , and tail fragments within 5 min of amputation overlapping with lumenal nuclear dense mounds ( Figure 8D , Figure 8—figure supplement 2F , Figure 8—figure supplement 3 ) . Thirty five to sixty minutes after amputation , signaling coalesced into discrete P-p38+ cells throughout the gut of the resulting fragment ( Figure 8D ) . Double labeling analysis revealed that these P-p38+ cells co-expressed the gut markers mat1 ( low ) , porc , hnf4 , and nkx-2 . 2 ( Gurley et al . , 2008; Wagner et al . , 2011; Tu et al . , 2015 ) ( Figure 8E ) . Gut injuries resulting from either lateral incision or needle poke both elicited a similar but more localized P-p38 activation within 30 min ( Note: epidermal staining in lateral wound could not be distinguished from background ) ( Figure 8—figure supplement 2G ) . The amputation induced P-p38 staining pattern persisted for 24 hr and by one week regenerating worms largely re-established the discrete P-p38 staining pattern observed in intact worms ( Figure 8D ) . Quantitative image analysis confirmed that while infection resulted in an increase in P-p38 fraction volume over time , amputation induced a sharp increase in both P-p38 fraction volume and average intensity followed by a gradual decrease ( Figure 8F , G ) . At present , it is difficult to distinguish the extent to which P-p38 signaling in the gut is directly activated as a result infection versus indirectly activated by damage resulting from infection . Conversely , it is similarly difficult to evaluate the extent to which tissue injury that results in the invasion of bacterial populations normally excluded from internal tissue structures elicits activation of P-p38 . In either case , our data indicate that the downstream kinase of the TAK1/MKK/p38 signaling module is dynamically modulated in response to infection and/or injury within the planarian gut Phospho-signaling analyses revealed that p38 was activated within the planarian gut in response to infection or injuries . Curiously , RNAi of the inhibitors pp6 and cyld-1 did not elicit intense P-p38 staining in the gut of intact worms prior to stimulation or during the first 3 days of infection even though the appearance of lesions and head regression immediately followed ( data not shown ) . As previously indicated , initial P-p38 staining experiments utilizing bleaching methods conducive to WISH elicited background staining in peripheral tissues ( Figure 8—figure supplement 2C ) . Use of an alternative bleaching solution eliminated background , revealing a previously unresolvable anterior/dorsal enriched mesenchymal/epidermal P-p38 staining pattern that was induced by Pseudomonas infection ( Figure 9A , Figure 9—figure supplement 1A , B ) . P-p38+ cells were relatively absent prior to infection but could be visualized in anterior regions within 24 hr of infection and further increased in frequency over the next two days . A lower frequency of P-p38+ cells was observed in the posterior region with little to no signal in the mid body region . This signal represents the earliest observed P-p38 activation in response to infection in a pattern that overlaps with and precedes the occurrence of anterior lesions and eventual head regression ( Figure 2B ) . Furthermore , its localization correlates with the differential sensitivity of regenerating tissue fragments to Pseudomonas infection ( Figure 2F ) . We analyzed levels of apoptosis in response to Pseudomonas infection and observed a similar anteriorly biased induction of apoptosis that coincided with P-p38 activation ( Figure 9B ) . 10 . 7554/eLife . 16793 . 024Figure 9 . TAK1/MKK/p38 signaling mediates contrasting regulation of apoptosis in infection versus regeneration . ( A ) P-p38 staining or ( B ) TUNEL of Unc RNAi worms following Pseudomonas infection . Focus on the effects of pp6 and cyld-1 RNAi on ( C ) P-p38 signaling and ( D ) TUNEL in the anterior regions of planaria during Pseudomonas infection . Quantification of RNAi effects on ( E ) P-p38+ and ( F ) TUNEL+ cells in the anterior during infection ( n = 5–9 ) . Confocal images of TUNEL in amputated tail fragments following RNAi of TAK1/MKK/p38 signaling components . RNAi effects assayed at ( G ) 72 hr or ( H ) 48 hr post-amputation ( two independent experiments ) . ( I , J ) Quantification of RNAi effects on corresponding TUNEL experiment ( n = 1–7 ) ( * = t-test p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 02410 . 7554/eLife . 16793 . 025Figure 9—figure supplement 1 . Effects of TAK1/MKK/p38 signaling on proliferation and apoptosis . ( A ) Imaging and ( B ) quantification of the effects of p38-1 RNAi on P-p38 staining of cells in the anterior region prior to or following infection ( n = 7–9 ) . ( C ) Confocal images and ( D ) quantification of H3P labeling in worms prior to and following Pseudomonas infection of cyld-1 RNAi worms in comparison to unc control ( n = 2–5 ) . Magnified images of cells in which TUNEL+ cells either ( E ) overlap with or are in ( F ) proximity of P-p38+ cells . Quantification of the effects of TAK1/MKK/p38 signaling on proliferation within tail fragments following amputation ( n = 1–13 ) ( * = t-test p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 025 We determined the role of TAK1/MKK/p38 signaling in the observed infection induced apoptosis with a focus on the anterior region of the worm . We analyzed the results of ectopic induction of this pathway via RNAi of the negative regulators pp6 and cyld-1 . RNAi of either of these genes resulted in elevated numbers of anteriorly biased P-p38+ and TUNEL+ cells both prior to and following Pseudomonas infection ( Figure 9C–F ) . Additionally , cyld-1 RNAi resulted in pre-emptive increase in cell proliferation that mirrored the levels observed in control infected worms ( Figure 9—figure supplement 1C , D ) . These data support a role for TAK1/MKK/p38 signaling in the control of both a general proliferative response and an apoptotic response during infection . The frequency of P-p38+ cells was consistently lower than that of TUNEL+ cells with only 1–8% overlap depending on RNAi treatment or days post infection ( Figure 9E , F , Figure 9—figure supplement 1E , Table 3 ) . Interestingly , we observed a higher frequency of TUNEL+ cells that were proximal to P-p38+ cells ( ranging from 12–80% ) , suggesting a possible extrinsic , rather than an intrinsic role for the mediation of apoptosis ( Figure 9—figure supplement 1F , Table 3 ) . 10 . 7554/eLife . 16793 . 026Table 3 . Quantitation of overlapping and proximal P-p38 and TUNEL signal . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 026RNAi conditionDays post infection% of P-p38 signal overlapping with TUNEL% of P-p38 signal proximal to TUNELunc01 . 2712 . 66cyld-105 . 7845 . 66pp604 . 2967 . 86unc11 . 1639 . 31cyld-114 . 5262 . 05pp618 . 8779 . 31unc21 . 9034 . 29cyld-123 . 4860 . 65pp625 . 7357 . 96unc32 . 5544 . 19cyld-133 . 9854 . 23pp638 . 2957 . 46 We next determined whether TAK1/MKK/p38 signaling effects on apoptosis and proliferation were general or specific to infection . Previous research has demonstrated that tissue amputation elicits a robust induction of both proliferation and apoptosis ( Pellettieri et al . , 2010; Wenemoser and Reddien , 2010 ) . We analyzed the effects of RNAi of both activators and inhibitors of the TAK1/MKK/p38 pathway on amputated tail fragments in the absence of infection . We observed no effects of RNAi of any TAK1/MKK/p38 signaling component on the proliferative burst following amputation ( Figure 9—figure supplement 1G ) . To our surprise , amputation induced apoptosis was increased by RNAi of tak1 or p38-1 , and decreased by RNAi of cyld-1 ( Figure 9G–J ) . This stands in direct contrast to the observed effects of ectopic activation of this pathway in enhancing infection-induced apoptosis ( Figure 9D , F ) . Interestingly , tak1 and p38-1 rnai worms all exhibited phenotypically normal regeneration in the absence of infection ( Figure 7 ) , suggesting that the observed enhancement in apoptosis is relatively innocuous during normal regeneration . Overall , our data indicate that TAK1/MKK/p38 signaling has dual contrasting roles in the control of apoptosis during either infection or regeneration . We evaluated the potential role of TAK1 signaling pathway homologs in the planarian immune response . Planaria were infected and maintained in planaria water containing 2e8 CFU/ml Pseudomonas without reinfection to assay clearance of the initial bacterial dosage . Subsequent changes in bacterial levels and tissue degeneration of control and activator/inhibitor RNAi worms were monitored over time . Pseudomonas levels in control RNAi worms increased to ~1e5 CFU/worm by 1 to 3 days and 2e5 CFU/worm by 6 days ( Figure 10A–D ) . During this time worms exhibited no phenotypic signs of tissue degeneration ( Figure 10E ) . By 12 days post infection , nearly all control worms exhibited head regression or anterior lesions , coincident with a ~80% decrease in Pseudomonas levels per worm ( Figure 10A–E ) . RNAi of the activator jun D significantly increased bacterial load over time while RNAi of p38-1 had no overall significant effects on Pseudomonas levels ( Figure 10A , B ) . Importantly , jun D RNAi worms still exhibited a reduction in bacterial levels from day 6 to day 12 , suggesting that jun D independent immune responses also mediate bacterial clearance ( Abnave et al . , 2014 ) . The roles of the inhibitors pp6 and cyld-1 in the immune response were less clear as RNAi of either of these components resulted in a non-significant overall increase in Pseudomonas over time ( Figure 10C , D ) . We hypothesize that increased basal apoptosis and induced tissue degeneration in these RNAi conditions compromises barrier epithelia , leading to a larger influx of bacteria and complicating analysis of the immune response ( Figures 9 , 10E ) . 10 . 7554/eLife . 16793 . 027Figure 10 . Effects of TAK1 components on the immune response . Bacterial CFU/worm following infection with a single dose of 2e8 CFU/ml Pseudomonas and RNAi of ( A ) jun D , ( B ) p38-1 , ( C ) pp6 , or ( D ) cyld-1 ( n = 1–4 pools of 1–4 worms each ) . ( E ) Pathological state of individual infected worms ( squares ) pooled for bacterial CFU/worm analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 16793 . 027 These data suggest that the planarian immune response is normally coordinated with but not the result of induced apoptosis and tissue degeneration . While control RNAi worms exhibit the largest decrease in Pseudomonas levels following phenotypic tissue degeneration ( Figure 10A–E ) , RNAi of p38-1 results in an identical reduction of bacterial levels in the absence of phenotypic tissue degeneration ( Figure 10B , E ) . Similarly , pp6 and cyld-1 RNAi enhanced tissue degeneration but not bacterial clearance ( Figure 10C–E ) . Thus , Pseudomonas infection induces a jun D and p38-1 dependent tissue degeneration response in conjunction with a jun D dependent , p38-1 independent immune response .
This study elucidates the heretofore unknown S . mediterranea microbiome . We observed how microbial composition dynamically shifted in response to changes in culture conditions or during regeneration ( Figure 1 , Figure 1—figure supplement 1 ) . The ratio of Bacteroidetes to Proteobacteria within the planaria from the recirculation culture system mirrors the distribution of these phyla within the human lower intestinal tract ( Human Microbiome Project Consortium , 2012 ) . This phyla composition is not observed in other invertebrate ( e . g . , C . elegans and D . melanogaster ) or even vertebrate ( e . g . , D . rerio ) model systems that are amenable to large scale genetic screens and have played a pivotal role in shaping our understanding of immunity and host-microbe interactions ( Apidianakis and Rahme , 2011; Roeselers et al . , 2011; Novoa and Figueras , 2012; Lee and Brey , 2013; Wong et al . , 2013; Ermolaeva and Schumacher , 2014; Berg et al . , 2016 ) . An enrichment in Bacteroidetes in S . mediterranea is significant , as this phylum represents a diverse array of key symbionts known to support proper tissue homeostasis , the development of the immune system , as well as the prevention of inflammatory diseases in mammals , and , intriguingly , have even been linked to the early evolution of multicellular organisms ( Rakoff-Nahoum et al . , 2004; Mazmanian et al . , 2005 , 2008; Alegado et al . , 2012 ) . Furthermore , changes in culture conditions or tissue transection results in an expansion in Proteobacteria , a phenomenon linked to a myriad of human inflammatory disorders ( Frank et al . , 2007; Shin et al . , 2015 ) . Models studying this phenomenon in subsets of genetically-altered or aged individuals have proven instrumental in the analysis of the effects of this expansion , but have been less useful in dissecting factors mediating its causation ( Carvalho et al . , 2012; Clark et al . , 2015 ) . Our ability to induce shifts in the microbiota of wild type animals , uniquely positions planaria as a model for the study of the interrelationship between endogenous bacterial dynamics , tissue homeostasis , and complex tissue regeneration with implications for human health . Our study has highlighted the impact that the microbiota can have on tissue homeostasis and regeneration in planaria . Maintenance and care of planarians using traditional static methods can yield variable bacterial levels and composition over time , potentially influencing experimental outcomes . Furthermore , prolonged use of antibiotics can lead to the gradual accumulation of resistant pathogenic strains , such as Pseudomonas . Given that many published phenotypes describing gene knockdowns mirror some of the observed effects of bacterial infection ( Reddien et al . , 2005; Forsthoefel et al . , 2012; Labbé et al . , 2012; Tu et al . , 2012 ) , it will be important to examine the extent to which alterations in endogenous bacteria or stimulation of immune signaling underlie these phenotypes . This issue is particularly relevant to the common practice of feeding worms E . coli producing dsRNA as a means of gene function interrogation . As we have demonstrated , this method of RNAi elicits robust P-p38 activation ( Figure 8—figure supplement 2E ) . To mitigate the issues of inconsistent bacterial composition and variable immune stimulation across experiments , we have developed an RNAi workflow using recirculation and flow culture systems ( Figure 4 , Figure 4—figure supplement 1 ) . Previous screens examining genes involved in planarian regeneration have utilized either static cultured or E . coli fed worms of indeterminate bacterial content . It would be of interest to employ our novel culture systems to determine whether published phenotypes describing tissue homeostatic defects similar to those observed in this study have inadvertently uncovered genes involved in the interplay between the immune system and regeneration . Our study revealed a distinct role for TAK1/MKK/p38 mediated cellular apoptosis during infection versus normal regeneration in planaria . Hyper activation of TAK1/MKK/p38 signaling via pp6 or cyld-1 RNAi elevated basal apoptotic rates and increased stem cell proliferation prior to infection ( Figure 9D , F , Figure 9—figure supplement 1C , D ) . In contrast , TAK1/MKK/p38 signaling repressed apoptosis during regeneration in the absence of infection ( Figure 9G–J ) . Importantly , tak1 , mkk6-1 , p38-1 , and jun D RNAi worms exhibited phenotypically normal regeneration in the absence of infection , suggesting that this alteration in apoptosis did not ultimately hinder overall tissue morphogenesis and remodeling ( Figure 7 ) . It is unclear at this point whether these divergent outcomes of TAK1/MKK/p38 signaling are the result of unique co-stimulation or the activation of discrete cell types induced during infection versus amputation . Nevertheless , our study provides further evidence that TAK1/MKK/p38 signaling is multifaceted and can play supportive or antagonistic roles in tissue regeneration ( Zarubin and Han , 2005; Cuenda and Rousseau , 2007; Karin and Clevers , 2016 ) . The role of TAK1/MKK/p38 in apoptosis during infection is not unique to planaria . In C . elegans , oral Pseudomonas infection initiates programmed cell death of gonadal cells in a p38-dependent manner ( Aballay and Ausubel , 2001 ) . In Drosophila , oral infection induces p38-mediated production of reactive oxygen species ( ROS ) , damaging gut epithelia , and triggering intestinal stem cell activation ( Buchon et al . , 2009; Ha et al . , 2009 ) . Interestingly , attenuation of TAK1/MKK/p38 signaling in nematodes and flies largely compromises host survival to infection , but substantially enhances host survival in planaria ( Figure 5 ) ( Kim et al . , 2002; Buchon et al . , 2009; Chen et al . , 2010 ) . One explanation is that the persistence of a population of adult pluripotent stem cells capable of regenerating all missing tissues uniquely endows planaria with a resilience to somatic tissue damage . Planaria may utilize TAK1 innate immune signaling for a coordinate mounting of antibacterial responses while actively clearing infected tissue via apoptosis prior to the initiation of regeneration . Under our sustained infection protocol , pathogen levels were constantly maintained , resulting in failure of complete bacterial clearance and continuous TAK1/MKK/p38 mediated apoptosis . Once this level of tissue turnover outpaces the ability of neoblasts to replace these cells , tissue homeostasis and regeneration capabilities are compromised . Under this context attenuation of infection-induced TAK1/MKK/p38 signaling rescued tissue degeneration and restored regenerative potential . The management and care of chronic non-healing wounds poses a substantial and sharply rising burden to our healthcare system and economy ( Sen et al . , 2009 ) . Of interest is the observation that planaria are capable of resolving tissue degenerative wounds induced by sustained Pseudomonas infection and regenerating lost anterior structures ( Figure 2C , Figure 2—figure supplement 1D ) . Antibiotic resistant Pseudomonas is increasing in prevalence in chronic wounds and poses an impediment to effective healing ( Dowd et al . , 2008; Fazli et al . , 2009; Price et al . , 2009; Goldufsky et al . , 2015 ) . Further research is needed for the development of treatments by which to effectively manage infection and tissue repair . Limited regenerative capabilities combined with the severe lethality of Pseudomonas infection has precluded the study of this phenomenon in other model systems ( Kim et al . , 2002; Vodovar et al . , 2005 ) . In contrast , results from our RNAi screen have already indicated that mkk4 and mkk6-1 may play a role in inhibiting this post-infection regeneration in a p38-1 independent manner ( Figure 5F ) . The molecular mechanism by which this is accomplished is still unclear . Furthermore , the observation that this phenomenon was unique to Pseudomonas infection relative to the other pathogenic strains tested is equally fascinating . In plants , some species of Pseudomonas are capable of attenuating host immune response via secretion of a tyrosine phosphatase that inhibits signaling ( Macho et al . , 2014 ) . It would be of great interest to compare the differential host responses of planaria to other pathogens or mutant Pseudomonas strains to determine unique gene regulatory networks supportive of the observed tissue regeneration during chronic infection . The novel tools and findings of this study not only open up new areas of inquiry for planaria as a model system , but also broaden our understanding of the relationship between endogenous microbiota , the innate immune system , and regeneration . We took advantage of the amenability of planarians to RNA-mediated genetic interference to identify mediators of Pseudomonas-induced tissue degeneration and its lytic versus regenerative outcomes . We identified a TAK1 innate immunity signaling module underlying tissue degeneration and compromising regeneration capacity during infection . Interestingly , while TAK1/MKK/p38 signaling enhanced apoptosis during infection , it repressed apoptosis during tissue regeneration in the absence of infection . Our data indicate that activation of this pathway has discrete and seemingly opposite roles in host immunity versus normal regeneration . Given the complex role of inflammation in either the hindrance or support of reparative wound healing and regeneration , S . mediterranea provides a basis for dissecting the duality of this evolutionarily conserved inflammatory signaling module in complex , multi-organ adult tissue regeneration . Furthermore , the conservation of bacterial phyla composition and dynamics we uncovered in this invertebrate make it the first non-murine model organism for the study of pathological shifts in endogenous bacteria with relevance to human disease . In summation , our work not only advances planarians as a unique and tractable in vivo model system for the molecular dissection of regeneration , but also opens the door to the identification and characterization of conditions that promote or inhibit the natural execution of regenerative processes .
Schmidtea mediterranea clonal CIW4 strain was maintained in 1X Montjuic salt as previously described for traditional static culture methods ( Newmark and Sánchez Alvarado , 2000 ) , supplemented with 50 μg/ml gentamycin where indicated . A novel recirculation culture system was the predominant source of CIW4 planaria used in this study . The planarian recirculation culture system is composed of three culture trays ( 96' L × 24' W × 12' H ) stacked vertically on top of each other over a sump . Planarian water flows from the sump pump through a chiller , canister filter , and a UV sterilizer into the top tray . Water subsequently flows down drains at the opposite ends of the source through the series of three trays . Water then re-enters the sump where it is filtered through two vertically stacked 400 μm and 200 μm sieves . Beyond the sieves , water is gravity fed and mechanically filtered through a set of filter/floss pads . Finally , water passes through Water Garden Oasis Pond Matrix and Kaldness media to remove nitrogenous waste . Water is then able to flow back through the chiller , canister filter , and UV sterilizer back into the top tray . For the sexual S2F2 strain of planaria , worms were obtained from a fill and drain system . The system is composed of 32 tanks and a sump . All tanks drain into a common drain line , flowing into the filter box resting on top of the sump . The sump contains a heavily aerated bed of Kaldness media for biofiltration . Water is next pumped through finer sets of sock filters ( 50 and 20 μm ) and then through ultraviolet irradiation prior to being returned to the culture tanks . Each of the culture tanks utilizes a recurring fill and drain action to provide more frequent water turnover for each tank . Animals are housed within cylinders with porous mesh bottoms within these tanks . For recirculation to static time course , 4 day starved worms were removed from the system and placed in containers filled with planaria water with or without the antibiotic gentamycin . For regeneration time course , 14 day starved worms were removed from recirculation culture at low density , washed after two days and transferred to petri dishes after 3 days . Twenty intact worms or 60 amputated tissue fragments were allocated per dish per time point . For intact worms , only worms that had not undergone fission were used for analysis . Worms were washed every 3–4 days after amputation . Wild worms were collected from sites in Sardinia in RNA later for sample storage . DNA isolation was performed as follows . Negative controls from the initial DNA isolation were carried out for regeneration time course and wild worms . Collected worms were crushed in 100 μl in chilled lysis buffer containing 20 mM Tris pH 8 . 0 , 100 mM NaCl , 50 mM EDTA , 2 mM spermidine , and 0 . 2 mg/ml Proteinase K , and 0 . 2% β-Me . Lysate was added to 800 μl of lysis buffer and warmed to 50C . 100 μl of 10% SDS was added and the tube was nutated at 50°C for 1 hr . 1 ml of phenol:chloroform:IAA was added to each tube and mixed on nutator for 10 min . Samples were spun at 4°C 1600 g for 5 min to get phase separation . The top layer was transferred to a new tube and 1 ml chloroform was added . Samples were spun again for phase separation and the top layer was transferred to a new tube . Samples were precipitated by adding 12% volume of 2 . 5 M NaAc and 1 volume isopropanol . After 1 hr incubation at −20°C , samples were pelleted by centrifugation and rinsed 3 times with 70% EtOH . The pellet was dried and resuspended in H2O . Samples were RNAse treated in a solution of 5 mM Tris pH 8 . 0 and 1 μg/ml RNAse A for 15 min at room temperature . Samples were precipitated with 12 . 5% NaAc , 2 . 5 volumes of EtOH and incubation at −80°C for 1 hr . Samples were pelleted by centrifugation and washed 3 times with 70% EtOH . Pellets were dried and resuspended in H2O . Sample preparation was performed as per manufacturer's instructions with some modifications . In brief , the V3 and V4 primer pair was used for template amplification using 2X KAPA HiFi Hotstart ReadyMix ( KAPA Biosystems , Wilmington , MA ) in PCR strip vials . Reactions were placed in a thermal cycler at 95°C 3 min , followed by 25 to 35 cycles of 95°C 30 s , 55°C 30 s , and 72°C 30 s with a 5 min 72°C extension . For static culture timecourse and regeneration timecourse , 35 and 32 cycles were used respectively . For wild Sardinia worm material , 25 cycles were initially used for all samples . Those that failed to amplify in addition to a control successfully amplified sample were rerun with 35 cycles . Negative controls of H2O were run in parallel from the start of library preparation for static time course sample set . Negative controls from the start of the DNA isolation were run for wild worm and regeneration sample sets . Following the initial PCR amplification , the template was purified using Agencourt AMPure XP beads and a magnetic stand . Next , indexing pcr was performed using the Nextera XT index kit ( Illumina ) . Reactions were placed in a thermal cycler using the identical PCR program with only 8 cycles . A final round of cleanup with Agencourt AMPure XP beads was performed and libraries were pooled , requantified , and sequenced as 2 × 250 cycle paired reads on the Illumina MiSeq , using MiSeq Control Software v2 . 5 . 0 . 5 . Following sequencing , the MiSeq Reporter v2 . 5 . 1 . 3 ( recirculation to static time course experiment ) or v2 . 6 . 2 . 3 ( regeneration time course and individual wild worm experiments ) Metagenomics workflow was used to de-multiplex reads for all libraries , generate FASTQ files , and perform taxonomic classification of the reads . Reads of negative controls were used to determine background bacterial amplification . After elimination of background levels , a low signal threshold was set at 10 reads per OTU . For wild worms , 16 s rDNA was successfully amplified and sequenced for 4 out of 6 worms from Sardinia collection site#1 ( wild worm#1 jj04-9 , wild worm#2 jj04-11 , wild worm#3 jj04-11 , wild worm#4 jj04-13 ) and 1 out of 1 worms from collection site#2 ( wild worm#5 jj08-1 ) . A 5th worm from collection site #1 was also amplified and sequenced ( jj04-10 ) , but since this sample did not have a negative control run in tandem during DNA isolation ( as was performed for the other wild samples ) , it was not possible to determine potential contamination of the sample with bacteria from the lab . Therefore , this sample was not used for further analysis . Graphs were generated using Prizm software . Heatmaps were made using TM4 MeV . Venn Diagrams were generated using Venneuler . Images of colorimetric WISH samples , histology samples , confocal images , and live worm images were acquired using Zeiss Lumar , Leica DM 600B , LSM510-VIS , and Leica M205 microscopes , respectively . Individual tiled whole worm or 10X zoomed head images were acquired on a Perkin Elmer Ultraview spinning disk microscope for P-p38 , TUNEL , and phospho H3 quantification . Stitching was performed using Fiji plugins combined with customized batch processing macros or wrapper plugins where necessary . Worms were segmented by DAPI labeling with custom plugins and spots were counted using the ‘Find Maxima’ function via batch processing macros . All macros and plugins are available at https://github . com/jouyun . For P-p38 average intensity and fraction whole quantification , spots were segmented and manually filtered from false worm boundary objects and used to measure spot area and compute intensities . The whole worm or amputated fragment was segmented using DAPI to compute the total volume . The specified number of worms or fragments were washed three times and then crushed in 100 μl H2O . Homogenate was serially diluted 4X by 10-fold and 25 μl was plated onto each quadrant of an LB plate . Plates were incubated at room temperature for 3–4 days and the morphology and the number of colonies was recorded . For identification of abundant strains , colony morphology was categorized by color , shape and size . The 16 s variable region from representative colonies of each category was PCR amplified with the following primers: forward primer 63f ( 5′-CAG GCC TAA CAC ATG CAA GTC-3′ ) and reverse primer 1378r ( 5′-GGG CGG WGT GTA CAA GGC-3′ ) ( Marchesi et al . , 1998 ) . The resulting template was sequenced and nucleotide BLAST was performed for putative identification . Strains isolated from LB plating of planaria homogenate were identified by 16 s sequence homology and stored in glycerol stocks . Concentration was determined by diluting bacteria to a 600 nm absorbance of 1 and plating serial dilutions on LB plates . Average colony count across replicates was used to determine an absorbance to bacterial CFU/ml conversion factor . Worms were transferred to petri dishes and allowed to rest overnight . Bacterial strains stored in glycerol stocks were used to inoculate LB for overnight culture at 30°C . Bacterial cultures were spun down , washed , and resuspended in planaria water at the specified concentrations using 600 nM absorbance and previously calculated bacterial CFU/ml conversion factors . Planaria were washed several times and bacteria resuspended in planaria water was added . Infection was refreshed every 3 to 4 days ( unless specified otherwise ) by washing worms several times , transferring to a new petri dish , and adding fresh bacteria resuspended in planaria water . Worms were ocularly scored every 1 to 3 days using a Zeiss Stemi 200-C stereo microscope using the following criteria . Worms exhibiting no phenotypic lesions were scored 0 . Worms exhibiting lesions in either the mid-body or posterior were scored 1 . Worms with lesions in the anterior but displaying at least one remaining photoreceptor were scored 2 . Worms with anterior tissue degeneration that resulted in no remaining photoreceptors were scored 3 . Worms with regressed anterior regions that displayed partially disrupted or lysed tissue were scored 4 . Worms that fully lysed were scored 5 . Worms that regressed heads and formed blastemas with both visible photoreceptors were scored 2 . 5 . Worms inadvertently lost or damaged during reinfection procedures and those that crawled out of sample dishes were not scored further . For histology preparation , animals were treated with 7 . 5% NAC ( 8 min ) , fixed in 4% PFA in PBSTx 0 . 3% ( 20 min ) , and then dehydrated by washing in 30% , 50% , and 70% methanol ( 5 min each ) . For paraffin embedding , animals were soaked in ethanol with 5% glycerol , washed in xylene ( 7 min ) and clear-rite ( 2 × 7 min ) , and soaked in paraffin ( 2 × 14 min , then 2 × 30 min ) . After serial sectioning ( 6 μm thickness ) , slides were heated to 60°C for 20 min , deparaffinized with three 3 min washes in xylene , washed 3 × 1 min in 100% ethanol , then 80% ethanol , rinsed in tap water . Subsequent staining utilized the Leica Infinity system and was performed in a Leica Autostainer For Alcian Blue/ Periodic acid-Schiff ( AB/PAS ) staining , hydrated slides were stained with 1% Alcian Blue made in 3% acetic acid for 15 min . Afterwards , slides were washed ( 2 min ) in running tap water , rinsed in DiH2O , incubated in 0 . 5% periodic acid ( 5 min ) , and rinsed in DiH2O . Slides were then stained in Schiff’s reagent ( 10 min ) and rinsed under tap water ( 5 min ) . Nuclei were stained with hematoxylin ( 1 min ) and rinsed in running tap water ( 2 min ) . Slides were placed in acid alcohol ( 1 min , Leica Infinity , Differentiator ) , rinsed in tap water ( 1 min ) , and incubated in Bluing Agent ( 1 min ) followed by tap water rinse ( 1 min ) . Finally , the samples were dehydrated through washes in 80% ethanol , 4 × 100% ethanol , cleared in xylene , and mounted with coverslip onto slides . For anti-Pseudomonas staining , antigen retrieval was performed using citrate buffer , pH 6 . 0 for 15 min at 95°C , cooled 20 min , and rinsed in DiH2O . Slides were treated 30 min in Background buster ( VWR #NB306 ) for 30 min and rinsed . Staining was performed with anti-Pseudomonas fluorescens antibody ( abcam #25182 ) 1:750 , overnight at 4°C , rinsed , stained 1 hr with donkey anti-goat 488 for 1 hr , and rinsed . Slides were mounted under coverslip with Fluoromount G with DAPI ( VWR #102092–102 ) . For RNAseq analysis , 4 biological replicates of 4 worms were collected at the specified times after the transition from recirculation to static culture in the presence or absence of 50 μg/ml gentamycin for TRIZOL isolation . mRNAseq libraries were generated from 500 ng of high quality total RNA , as assessed using the Bioanalyzer 2100 ( Agilent ) . Libraries were made according to the manufacturer’s directions for the TruSeq Stranded mRNA LT– set A and B ( Illumina , San Diego , CA; Cat . No . RS-122-2101 and RS-122-2102 ) and using NEXTflex DNA Barcodes ( BiooScientific , Austin , TX; Cat . No . 514104 ) . Resulting short fragment libraries were checked for quality and quantity using a LabChip GX ( Perkin Elmer ) and Qubit Fluorometer ( Life Technologies , Carlsbad , CA ) . Libraries were pooled , requantified and sequenced as 50 bp single reads on the Illumina HiSeq 2500 instrument using HiSeq Control Software v2 . 2 . 38 . Following sequencing , Illumina Primary Analysis version RTA v1 . 18 . 61 and Secondary Analysis version CASAVA-1 . 8 . 2 were run to demultiplex reads for all libraries and generate FASTQ files . RNAseq analysis was performed as previously described using a minimum fold change of 1 . 4 and an adjusted p-value < 0 . 05 ( Tu et al . , 2015 ) . Reverse transcriptase ( for pgrp genes ) and quantitative PCR ( for pgrp genes and 16 s rDNA ) was performed using Superscript III ( Invitrogen ) and Fast SYBR Green Master mix ( ThermoFisher ) for the following gene targets: universal 16 s ( f: 5’-GTG STG CAY GGY TGT CGT CA-3’ , r: 5’-ACG TCR TCC MCA CCT TCC TC-3’ ) , pgrp-1 ( f: 5’-CTG CCA TCC GAT AAG ATG AGT T-3’ , r: 5’-TAT CGT TTC TCG TCG GCA TTT A-3’ ) , pgrp-4 ( f: 5’-GAC TCT CGA TCC GAA AGT AGG A-3’ , r: 5’-GGG TTG TCC ATT CCC AGA AAT A-3’ ) , β-actin ( f: 5’-CCG TGC CAA TTT ATG AAG GGT AT-3’ , r: 5’-GAA GAT GAA GAG GCC GCA GTT T-3’ ) , and gapdh ( f: 5’-GAT GGG CAT GCT ATT TCG GTT TAT-3’ , r: 5’-CTT TGC TCG GTT GTT TTT GGT ATG-3’ ) . Gene expression was normalized across the samples using β-actin and gapdh levels . For RNA expression analysis , worms were stained using previously established WISH protocols ( King and Newmark , 2013 ) with additional modifications for increased reagent penetrance and washing efficiency . These modifications permitted effective in situ analysis of worm or worm fragments ranging from 1 to 8 mm in size . All washes were performed in 50 ml Falcon tubes or 2 ml tubes in accordance with worm numbers in place of 24 well plates . All incubations and washes were performed for 10 min with nutation at room temperature unless otherwise specified . Mucus was removed with 7 . 5% NAC in PBS for 8 min . Worms were fixed in 4% PFA 0 . 3% Triton X for 30 min and washed twice with 1X PBS 0 . 3% Triton-X ( PBSTx ) . Worms were then washed with 10% SDS for increased permeabilization . Reduction was performed at 37°C for 10 or 20 min with a solution of 1X PBS , 1% NP-40 , 0 . 5% SDS , 50 mM DTT . Worms were washed twice with PBSTx 0 . 3% , and then dehydrated with a 50% MeOH: 50% PBSTx 0 . 3% solution , and washed and stored in 100% MeOH O/N at −20°C . In preparation for sample bleaching , worms were rehydrated with a 50% MeOH: 50% PBStX 0 . 3% wash , washed twice with PBSTx 0 . 3% , and washed once in 1X SSC . For bleaching , worms were incubated in a 1 . 2% H2O2 , 5% Formamide , 0 . 5X SCC solution for 4 hr over a light box . Worms were washed once with 1X SSC and twice with 1X PBS 0 . 3% Tween-20 ( PBSTw 0 . 3% ) . Worms were permeabilized for 20 min with a 1X PBS 0 . 1% SDS solution containing 2–4 μg/ml Proteinase K ( in accordance with worm size ) . Worms were post-fixed in 4% PFA 0 . 3% Tw for 10 min and washed twice with PBSTw 0 . 3% . Worms were prepared for hybridization with a 10 min 50% solution wash and subsequent 2 hr 56°C incubation with a prehybridization solution containing 50% Formamide , 1X Denhardts solution , 100 ug/ml Heparin , 1% Tween-20 , 50 mM DTT , and 1 mg/mL Sigma Torula Yeast RNA in 5X SSC . DIG-labeled and/or DNP-labeled probes were denatured 5 min at 70°C , and incubated O/N at 56°C in a hybe solution identical to prehybe solution but with Deionized Formamide in place of Formamide , Calbiochem Yeast RNA 0 . 25 mg/ml in place of Sigma Torula Yeast RNA , and an addition of 5% Dextran Sulfate . The next day non-specific probe binding was washed out at 56°C . The samples were washed twice for 30 min in a Wash Hybe solution containing 50% Formamide , 0 . 5% Tween-20 , and 1% Denhardts in 5X SSC . Worms were then washed twice in 50% Wash Hybe:50% 2X SSC 0 . 1% Tw for 30 min , thrice with 2XSSC 0 . 1% Tw for 20 min , and thrice with 0 . 2X SSC 0 . 1% Tw for 20 min . The samples were returned to room temperature and washed twice with 1X MAB 0 . 1–03% Tw ( MABTw ) pH 7 . 5 . The samples were blocked for 2 hr in a solution of 1X MABTw containing 10% Horse Serum and 0 . 5% Roche Western Blocking Reagent and stained with anti-DIG and/or anti-DNP Fab fragment conjugated to alkaline phosphatase ( colorimetric development ) or peroxidase ( fluorescent development ) at a 1:1000 dilution O/N at 4°C . For the final antibody incubation prior to development and slide mounting , this solution was supplemented with 1:5000 DAPI . Non-specific binding was removed with washing 6 times with MABTw for 2–3 hr and then processed for colorimetric or fluorescent development . For colorimetric development , worms were incubated for 15 min in solutions of 0 . 1 M Tris pH 9 . 5 , 0 . 1 M NaCl , 0 . 05 M MgCl2 , 0 . 1% Tween containing 0% , 50% , and then 80% PVA . The final solution of 80% PVA was supplemented with 1:188 BCIP or 1:94 NBT and worms monitored for color development . The reaction was stopped by rinsing 1–2 times in PBS , fixed in 4% PFA 0 . 3% Tx for 45 min , and washed 3 times with PBSTx 0 . 3% . The samples were then washed with 100% EtOH for 20 min and then with 50% EtOH:50% PBS for 5 min . The samples were the rehydrated with 1X PBS washes and mounted in 80% glycerol . For fluorescent development , worms were pre-incubated for 15 min in a development solution of 0 . 1 M Boric Acid , 2 M NaCl , pH 8 . 5 supplemented at 1:2000 with tyramide . Worms were incubated another 45 min in this solution after the addition of 0 . 006% H2O2 . Samples were washed twice with PBSTw 0 . 3% and peroxidase activity was terminated with a 1 hr incubation in 200 mM NaN3 in PBSTw 0 . 3% . The samples were washed twice in PBSTw 0 . 3% and either mounted in a 4 M Urea , 0 . 1% Triton-X , 2 . 5% DABCO , 20% glycerol solution . If additional immunostaining was performed , worms were washed 6 times in PBSTw 0 . 3% for 2–3 hr , 4 times with MABTw 0 . 1–0 . 3% , and then incubated O/N with the specified primary antibody . Planaria were cultured in paper cups ( 24oz ECO products containers or 9oz Dixie cups ) lined with a pocket of 150 μm nylon mesh ( Saatitech , Italy ) . Cups were perforated with holes at the desired planaria water level for outflow to complete the flow vessel . These flow vessels were placed in an enclosed container on a perforated tray suspended above a collection basin and centralized waste drain . Water was pumped from a 55 gallon tank into a series of 44 spigots directly above the flow vessels and adjusted to a constant drip rate of 1 drop per 5 s for 9oz cups and 1 drop per 2 s for 24oz cups . The system actively pumped water for 1 . 5 hr every 4 hr achieving total dispensation of approximately 55 gallons of planarian water every 3 . 5 days . Flow was turned off during RNAi feedings . Flow vessels were washed after RNAi feeding and worms were washed and transferred to fresh flow vessels the day before RNAi feeding or the start of experiments . Candidate genes analyzed in this study were cloned from a CIW4 cDNA library into a pPR-T4P vector ( J . Rink ) as described elsewhere ( Adler et al . , 2014 ) . These served as a template for in vitro dsRNA synthesis for RNAi feedings ( Rouhana et al . , 2013 ) . RNAi food was prepared by mixing 1 volume of dsRNA at 50–200 ng/μl ( depending on the experiment ) with 1 . 5 volumes of beef liver paste . The amount of food administered was typically 1 to 2 μl of food per worm depending on animal size . Worms were allowed to feed for 6 to 10 hr with 2 rounds of agitation by gentle pipetting to facilitate additional consumption . Worms were fed every 3 days for a total of 3 RNAi feedings . Experiments were conducted at 4 days or more after the last RNAi feeding . Following enumeration of individual worm pathological scores , data for each day was reduced to an average pathological score . Days for each RNAi condition were aligned in ascending order along the y-axis of a column . The average score of each column was calculated and used to sort the effects of RNAi conditions in ascending order along the x-axis . For the frequency of individual stages , the percentage of worms exhibiting the corresponding score each day was plotted in place of the average score . Heat maps were generated using TM4 MeV . Significance was calculated using 2-WAY ANOVA , p<0 . 05 . Western blot analyses were performed as previously described ( Hatton et al . , 2011 ) . The P-p38 antibodies used in this study were acquired from Cell Signaling Technologies . Both polyclonal ( #9211 ) and monoclonal ( #4511 ) antibodies were used for the initial assessment of cross-reactivity with the planarian phospho-epitope . Subsequent validation and in situ staining analysis was conducted with the monoclonal antibody due to higher in situ staining signal . Worms were prepared for TUNEL or immunostaining identically to those prepared for WISH up to the point of sample dehydration and storage in MeOH . Worms were rehydrated and then bleached in a solution of 3% H2O2 , 0 . 075% NH4O4 in PBSTx for 5 hr over light . Worms were then washed 3X with PBSTx , permeabilized with 4 μg/ml for 10 min , post-fixed in 4% PFA PBSTx for 10 min , and washed another 3 times in PBSTx . For H3P and P-p38 staining , the samples were blocked and then stained for 48 hr at 4°C with primary anti-phospho H3 at 1:500 or anti-phospho p38 at 1:800 and then with secondary anti-rabbit HRP at 1:1000 . For TUNEL labeling , the samples were stored O/N in PBS at 4°C . TUNEL was performed as previously described ( Tu et al . , 2015 ) . In brief , the samples were incubated in equilibration buffer for 15 min and then the TdT reaction was performed for 4 hr at 37°C . Afterwards , the samples were washed 6 times with PBSTx for 20 min , blocked , and incubated O/N at 4°C with anti-DIG POD at 1:1000 . The samples were processed for fluorimetric development identically to those in the WISH protocol and either mounted or prepared for additional immunostaining . | Regeneration , the ability to replace missing or damaged tissue , has fascinated biologists for years and has inspired a new direction for the medical field . Figuring out how some animals easily accomplish this while others do not may help us to develop new therapies that enhance regeneration in humans . Previous work has indicated that the immune system , which is normally used to defend the body against bacteria , plays an important but complicated role in regeneration . By studying the relationships between bacteria , the immune system and regeneration in simple systems , it may be possible to see how their interactions either support or prevent the replacement of lost tissues . Flatworms called planaria can regenerate all of their tissues . Arnold et al . have now investigated what bacteria exist in planaria , how the planarian immune system responds to these bacteria , and how this response affects regeneration . The results reveal that the two main types of bacteria that are present in planaria are also found in humans . In fact , conditions that encourage the growth and spread of one of these types of bacteria ( called Proteobacteria , many of which can make humans ill ) damaged the worms and prevented them from regenerating . Arnold et al . then looked to see if the worms had genes that were similar to human genes that control the key immune process of inflammation , and found evidence of several such genes . Reducing the activity levels of these genes enabled worms that had been infected with Proteobacteria to regenerate again . However , these genes only seem to be responsible for regeneration when the planaria are infected with bacteria . Thus , planaria could be used as a simple model to discover how changes in resident bacteria can be detected by the immune system and affect the ability to regenerate tissues . Future studies could use planaria to identify even more genes that control regeneration during infection . Also , since the main types of bacteria in planaria are similar to those in humans , planaria may help us to learn how animals can properly balance the levels of these bacteria in order to remain healthy . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Pathogenic shifts in endogenous microbiota impede tissue regeneration via distinct activation of TAK1/MKK/p38 |
Glutamate delta ( GluD ) receptors belong to the ionotropic glutamate receptor family , yet they don’t bind glutamate and are considered orphan . Progress in defining the ion channel function of GluDs in neurons has been hindered by a lack of pharmacological tools . Here , we used a chemo-genetic approach to engineer specific and photo-reversible pharmacology in GluD2 receptor . We incorporated a cysteine mutation in the cavity located above the putative ion channel pore , for site-specific conjugation with a photoswitchable pore blocker . In the constitutively open GluD2 Lurcher mutant , current could be rapidly and reversibly decreased with light . We then transposed the cysteine mutation to the native receptor , to demonstrate with high pharmacological specificity that metabotropic glutamate receptor signaling triggers opening of GluD2 . Our results assess the functional relevance of GluD2 ion channel and introduce an optogenetic tool that will provide a novel and powerful means for probing GluD2 ionotropic contribution to neuronal physiology .
The delta glutamate receptors , GluD1 and GluD2 , belong to the ionotropic glutamate receptor ( iGluR ) family , yet they don’t bind glutamate ( Yuzaki and Aricescu , 2017 ) . They are considered as glutamate receptors solely based on their strong sequence and structure homology with AMPA , NMDA and kainate receptors ( Lomeli et al . , 1993; Araki et al . , 1993; Schmid and Hollmann , 2008; Elegheert et al . , 2016; Burada et al . , 2020a; Burada et al . , 2020b ) . GluD receptors are widely expressed throughout the brain , GluD1 predominantly in the forebrain , while GluD2 is highly enriched in cerebellar Purkinje cells ( PCs ) ( Konno et al . , 2014; Hepp et al . , 2015; Nakamoto et al . , 2020b ) . Both GluD1 and GluD2 play a role in the formation , stabilization , function and plasticity of synapses through their interaction with members of the cerebellin ( Cbln ) family ( Fossati et al . , 2019; Tao et al . , 2018; Matsuda et al . , 2010; Kakegawa et al . , 2008 ) . Cbln1 notably binds both the N-terminal domain of postsynaptic GluD2 and presynaptic neurexins , leading to a trans-synaptic bridge that promotes synaptogenesis and is essential for GluD2 signaling in vivo ( Elegheert et al . , 2016; Suzuki et al . , 2020 ) . Deletion of genes coding for GluD1 or GluD2 in mouse results in marked behavioral alterations ( Yadav et al . , 2012; Lalouette et al . , 2001; Yadav et al . , 2013; Nakamoto et al . , 2020a ) , and mutations in these genes in humans have been associated with neurodevelopmental and psychiatric diseases ( Griswold et al . , 2012; Treutlein et al . , 2009; Greenwood et al . , 2011; Cristino , 2019 ) , attesting to their functional importance in brain circuits . Despite their structural similarity with other iGluRs , and notably the presence of a ligand-binding domain ( LBD ) , GluDs stand out because they are not activated by glutamate ( Araki et al . , 1993; Lomeli et al . , 1993 ) . Nonetheless , recent studies revealed that GluD pore opening could be triggered indirectly , through the activation of Gq-coupled metabotropic receptors , and contributes to neurotransmission and neuronal excitability ( Ady et al . , 2014; Dadak et al . , 2017; Benamer et al . , 2018; Gantz et al . , 2020 ) . Indeed , the activation of GluD channels triggered by metabotropic glutamate receptors ( mGlu1/5 ) underlies the slow excitatory postsynaptic current in cerebellar PCs ( GluD2 , Ady et al . , 2014 ) and in midbrain dopaminergic neurons ( GluD1 , Benamer et al . , 2018 ) . Moreover , the demonstration that GluD1 channels carry the noradrenergic slow excitatory current in dorsal raphe neurons ( Gantz et al . , 2020 ) , suggests that the contribution of GluD channels to neuronal excitability and synaptic physiology may be widespread . The above studies relied largely on genetic tools , such as dead-pore mutants or targeted deletions , to assess the ion channel function of GluDs . Yet , due to the absence of specific pharmacological tools to block their ion permeation , the role of GluD1/2 channels in the regulation of neural activity remains largely elusive . Pore blockers for GluDs , such as pentamidine and 1-Naphthyl acetyl spermine ( NASPM ) , were previously identified using a point mutation ( A654T ) in GluD2 that confers constitutive ion flow and causes the degeneration of cerebellar PCs in Lurcher ( GluD2Lc ) mice ( Wollmuth et al . , 2000; Zuo et al . , 1997 ) . These molecules are however also pore blockers of NMDA and AMPA receptors , respectively . Other ligands such as D-serine and glycine bind to the LBD and reduce spontaneous currents in GluD2Lc , which suggests a coupling between the LBD and the channel ( Naur et al . , 2007; Hansen et al . , 2009 ) , but these molecules have broad spectrum activity . Finally , 7-chlorokynurenic acid has been identified to modulate GluD2Lc current by binding to the D-serine site but it is also a GluN1 competitive antagonist ( Kristensen et al . , 2016 ) . To fill this gap , we bestowed light-sensitivity to the GluD ion channel pore using a photoswitchable tethered ligand ( PTL ) approach ( Paoletti et al . , 2019; Mondoloni et al . , 2019 ) . Using structure-based design , we incorporated a cysteine point mutation at the surface of GluD2 , right above the hypothetical channel lumen , onto which can be anchored a photoswitchable pore blocker . Different wavelengths of light are then used to modify the geometry of the PTL , thereby presenting/removing the blocker to/from the channel , resulting in optical control of ionotropic activity . Here we demonstrate rapid and reversible , optical control of ion current through a cysteine-substituted GluD2 receptor . This novel tool , called light-controllable GluD2 ( LiGluD2 ) , allows rapid , reversible and pharmacologically-specific control of ionic current through GluD2 , and may help provide a mechanistic understanding of how this receptor contributes to brain circuit function and behaviors .
Our approach to probing the functionality of the ion channel in GluD is to install a photo-isomerizable pore blocker at the extracellular entrance to the channel lumen ( Figure 1A ) . The tethered ligand is site-specifically attached to a cysteine-substituted residue . In darkness or under green light ( 500–535 nm ) , the PTL adopts an elongated shape and reaches the lumen , resulting in ion channel blockade , while under violet light ( 380–390 nm ) , it switches to a twisted , shorter configuration , relieving blockade . Our design of the PTL was based on the chemical structure of pentamidine ( Figure 1B ) , a pore blocker that efficiently blocks current through GluD2Lc receptors ( Williams et al . , 2003 ) . The PTL , called MAGu , contains a thiol-reactive maleimide ( M ) moiety , a central photo-isomerizable azobenzene ( A ) chromophore , and a guanidinium ( Gu ) head group that resembles the amidinium groups of pentamidine ( Figure 1C ) . MAGu was selected notably because its synthesis route has been described ( referred to as PAG1c in the original article ) and because it was shown to have no adverse effect on native brain tissue ( Lin et al . , 2015 ) . In aqueous solution , MAGu could be converted to its cis form using 380 nm light , and converted back to trans either slowly in darkness ( t1/2 ~ 20 min ) or rapidly upon illumination with 525 nm light ( Figure 1—figure supplement 1A–B ) , in agreement with previous reports ( Lin et al . , 2018 ) . To find the best attachment site for MAGu on GluD , we developed a homology model of the GluD2 receptor , based on the structure of the recently crystallized GluA2 receptor ( Twomey et al . , 2017 ) ( see methods ) . Using this model , we selected a series of 15 residues , located on the peptide that links the LBD to the third transmembrane domain ( M3 ) that lines the channel lumen , for mutation to cysteine ( Figure 1D–E ) . Since no known ligand directly gates the ion channel of GluD2 , we used a Lc mutant , A654T , which displays a constitutively open channel ( Wollmuth et al . , 2000; Zuo et al . , 1997 ) , for screening the 15 single-cysteine mutations . Accordingly , we found that heterologous expression of GluD2-A654T , but not of the wild-type ( WT ) protein , in HEK cells produces large currents that reverse at membrane potential close to 0 mV and are reduced by externally-applied pentamidine ( Figure 2A ) . Subtracted Lc current showed clear rectification at positive potentials , as reported with the blockade by NASP , another GluD blocker ( Kohda et al . , 2000 ) . Therefore , the A654T Lc mutant was subsequently used as a screening platform to find the best attachment site for MAGu on GluD2 . Each of the 15 residues identified in Figure 1D were mutated individually to cysteine on the A654T background , and tested using patch-clamp electrophysiology . Cells were treated with MAGu ( 20 µM , 20 min ) and Lc currents were measured in voltage-clamp mode ( −60 mV ) under different illumination conditions to toggle MAGu between its cis and trans states . As expected , current through A654T was not affected by light , indicating that in the absence of a properly-positioned cysteine , MAGu has no effect on this Lc channel . In contrast , we found several cysteine mutants for which current was significantly larger under 380 than under 535 nm light , and one mutant ( Q669C ) for which there was a tendency for ‘reverse photoswitching’ , that is larger currents under 535 than under 380 nm light ( Figure 2B ) . We then quantified the degree of photoswitching by comparing the block in darkness ( trans state ) to the block evoked by a saturating concentration of pentamidine ( 100 μM ) . We excluded from the analysis mutants that displayed no pentamidine-decreased leak current ( i . e . mutants for which pentamidine block was significantly smaller than that observed on A654T , Figure 2—figure supplement 1A–B ) , because they were likely either not expressed or not functional . Photoswitching was significant for Q666C , Q669C , D670C , Q674C , and I677C , suggesting that MAGu covalently reacted with these cysteine mutants and that , once tethered , it could modulate current in one of its conformer ( Figure 2C ) . Importantly , photomodulation was absent in the control A654T and the other cysteine mutants , indicating that the effect of light is specific to the attachment of MAGu to the above-mentioned cysteine mutants . From a structural point of view , the photocontrollable mutants are all located at the very top of the linker , that is further away from the membrane domain compared to other tested residues ( Figure 2D ) . We then selected the best mutant GluD2-A654T-I677C for further characterization . Because pentamidine does not fully block GluD2 , even at saturating concentrations ( Williams et al . , 2003 ) , we quantified the extent of photoswitching by blocking leak current completely , using impermeant N-Methyl-D-glucamine ( NMDG ) . We found that MAGu blocked about 33% of the leak current in its trans form ( Figure 3A ) . Photoregulation was fully reversible over many cycles of 380 and 535 nm light ( Figure 3B ) , in agreement with the fact that azobenzenes photobleach minimally ( Beharry and Woolley , 2011 ) . Under our illumination conditions , light pulses of 200 ms were sufficient to fully unblock the current , while shorter illumination times could be used to finely tune the degree of blockade ( Figure 3C ) . Once in the cis configuration , MAGu relaxes back to its thermodynamically stable trans state slowly , with a half-life of about 20 min in solution ( Figure 1—figure supplement 1B ) . Accordingly , relief of blockade persisted for many seconds in darkness after a brief flash of 380 nm light ( Figure 3D ) , eliminating the need for constant illumination , an important feature for future neurophysiology experiments . From a pharmacological point of view , current blockade occurred in the trans state ( 535 nm ) and was relieved in the cis configuration ( 380 nm ) for all membrane potential tested , with very little voltage-dependence ( Figure 4A ) , which contrasts with the profound voltage-dependence of block observed with pentamidine . This suggested to us that the positive charge of MAGu may not sense the electrical field of the membrane as much as pentamidine does , and thus that the two molecules may bind to different sites . To investigate whether MAGu and pentamidine compete for the same binding site , we evaluated the dose-response relationship of pentamidine block on GluD2-A654T-I677C conjugated with MAGu , under both 380 and 535 nm light ( Figure 4B ) . We found the IC50s under both wavelengths to be virtually indistinguishable , favoring the idea that MAGu and pentamidine have distinct , non-overlapping binding sites . To get further molecular insight into trans MAGu-induced reduction of current , we performed molecular modeling experiments . After inserting the cysteine mutation , trans MAGu was docked by covalent docking , that is the reactive maleimide moiety was forced to be in contact with the cysteine while the rest of the molecule was free to move . We found that the guanidinium headgroup of trans MAGu couldn’t reach the membrane-embedded lumen of GluD2 ( Figure 4C ) , in agreement with our electrophysiology data . The effect of trans MAGu on the ion current was tested with the MOLEonline webserver ( Pravda et al . , 2018 ) , which allowed to compute the geometry of the ion channel . We found that the photoswitch has a direct steric effect on the size of the cavity above the channel , as shown by the comparison of the computed channel in presence or absence of the photoswitch ( Figure 4D ) . In addition , the charge of the photoswitch could modify the electrostatic potential in the cavity and thereby affect ion transfer . We next sought to determine whether the non-Lc , native channel could be photocontrolled after installation of MAGu on the cysteine-substituted GluD2-I677C receptor . In heterologous expression system , activation of mGlu1 using the selective agonist 3 , 5-Dihydroxyphenylglycine ( DHPG ) was reported to trigger opening of GluD2 receptors ( Ady et al . , 2014; Dadak et al . , 2017 ) . Therefore , we co-expressed the b isoform of mGlu1 , which displays low basal activity ( Prézeau et al . , 1996 ) , together with GluD2 in HEK cells . Cells were labeled with MAGu and DHPG currents were recorded while alternating between 380 and 535 nm light . We found that DHPG-induced currents were reversibly reduced by about 23% under 535 nm compared to 380 nm light for I677C , indicating that optical blockade with MAGu could be transposed to the native , non-Lc GluD2 ( Figure 5A ) . Importantly , DHPG-induced currents were identical in both wavelengths of light for the WT receptor ( Figure 5B ) , confirming that the effect of light is specific to the attachment of MAGu to I677C ( Figure 5C ) . In addition , we observed that the holding current increased when switching from darkness to 380 nm light for I677C , and decreased when switching back to 535 nm light , but remained constant in both wavelengths of light for WT ( Figure 5D ) . This suggests that a fraction GluD2 receptors are constitutively open prior to DHPG application , likely due to some basal mGlu1 activity in these cells . Altogether , these results show that the GluD2 I677C mutant labeled with MAGu ( a . k . a . LiGluD2 ) possesses a functional ion channel , which can be gated through the mGlu signaling pathway , and which can be reversibly blocked and unblocked with green and purple light , respectively . In order to evaluate the usefulness of LiGluD2 in neurons , we verified that MAGu treatment does not lead to photosensitization of native glutamate currents in PCs of WT mice . Since GluD2 is enriched at the parallel fiber-PC synapse ( Landsend et al . , 1997 ) , we used local application of DHPG ( 200 µM ) to induce inward current in both MAGu- and vehicle-treated slices . We found that the DHPG current amplitude remained unchanged under 535 nm compared to 380 nm in the two treatment conditions , resulting in a ratio of current amplitude I380/I535 measured for each cell not significantly different from 1 ( Figure 5—figure supplement 1A ) . We then recorded AMPA-mediated excitatory post-synaptic currents ( EPSCs ) in both vehicle- and MAGu-treated PCs . We found that the amplitude of electrically evoked EPSCs was stable under 535 nm compared to 380 nm , and that the ratio of EPSC amplitudes I380/I535 was not significantly different from one in both conditions ( Figure 5—figure supplement 1B ) . These control experiments demonstrate that wild-type GluD and GluA receptors , which lack a properly-positioned cysteine residue near the pore lumen , remain insensitive to light after MAGu treatment .
The PTL strategy has been successfully applied to several members of the iGluR family , including kainate ( Volgraf et al . , 2006 ) and NMDA ( Berlin et al . , 2016 ) receptors . In these former studies , the photoswitches were made with a glutamate head group , and were tethered to the LBD in proximity to the glutamate binding pocket , providing photocontrol of channel gating ( Reiner et al . , 2015 ) . Because their activation mechanism is still unknown , we adopted a different strategy for photocontrolling GluD receptors . We installed the photoswitchable ligand MAGu in proximity to the pore lumen , in hope to alter ion conduction through non-competitive antagonism . We found several cysteine mutants for which current was specifically modulated by light after attachment of MAGu , notably I677C ( a . k . a . LiGluD2 ) . In LiGluD2 , trans MAGu likely does not reach the pore lumen as originally designed . Nevertheless , it reversibly modulates current through the open GluD2 channel with high temporal and pharmacological precision . The compounds traditionally used to probe the ionic function of GluDs , such as pentamidine and NASPM ( Kohda et al . , 2000; Williams et al . , 2003 ) , are not specific of GluD and also block NMDA and AMPA receptors . As to D-serine and glycine , they partially inhibit GluD2Lc and mGlu1-gated GluD currents ( Naur et al . , 2007; Ady et al . , 2014; Benamer et al . , 2018 ) , but they are also co-agonists of NMDA receptors . Here , the pharmacological specificity of LiGluD2 is exquisite: after MAGu treatment , only the I677C mutant , and not the WT receptor , became sensitive to light . Likewise , MAGu did not photosensitize other WT glutamate receptors expressed in native brain tissue , in agreement with previous report demonstrating that MAGu has no off-target effects on WT GABA receptors , glutamate receptors and voltage-gated ion channels ( Lin et al . , 2015 ) . Indeed , even though the maleimide group of MAGu reacts in principle with any extracellular cysteine freely accessible on the cell surface , the ability of the tethered ligand ( here guanidinium ) to reach a particular site of action on any given protein in one configuration , but not the other ( e . g . trans but not cis ) , is highly improbable . In fact , the PTL approach has already demonstrated exquisite pharmacological specificity for a large variety of cysteine-substituted ion channels and receptors ( Paoletti et al . , 2019; Mondoloni et al . , 2019 ) , even in complex biological settings such as brain slices ( Berlin et al . , 2016 ) or intact neuronal circuits in vivo ( Lin et al . , 2015; Durand-de Cuttoli et al . , 2018 ) . Attachment of MAGu to GluD2 requires a single amino acid substitution , which is unlikely to disrupt the function of the receptor . In line with this , we found that the functional coupling of GluD2 with mGlu1 signaling ( Ady et al . , 2014; Dadak et al . , 2017 ) was intact in LiGluD2 . This enabled us to validate that activation of mGlu1 triggers the opening of the GluD2 channel in heterologous expression system , in support of earlier evidence that opening of the ion channel of GluD receptors can be triggered in response to metabotropic signaling mechanisms ( Ady et al . , 2014; Dadak et al . , 2017; Benamer et al . , 2018; Gantz et al . , 2020 ) . Even though light-induced blockade in LiGluD2 is partial , the rapid kinetics of block/unblock , coupled to the genetic specificity of the methodology , provide a unique opportunity to detect even small variations in GluD2 current , such as the tonic current we observed in heterologous expression system . LiGluD2 remains to be deployed in neuronal setting , yet we believe it will be a crucial tool for probing the ionotropic contribution of this orphan receptor to synaptic physiology .
Bio-grade Chemicals products was provided by Sigma-Aldrich from Merck . MAGu was synthesized as previously described ( Lin et al . , 2015 ) and provided by Enamine Ltd . , Kyiv , Ukraine ( www . enamine . net ) . MAGu was stored at −80°C as stock solutions in anhydrous DMSO . UV-visible spectra were recorded on a Nanodrop 2000 ( Thermo Scientific , 1 mm path ) with 100 μM MAGu in PBS pH 7 . 4 ( 10% final DMSO ) . The sample was illuminated for 1 min using ultra high-power LEDs ( Prizmatix ) connected to an optical fiber ( URT , 1 mm core , Thorlabs ) , followed by an immediate measurement of absorbance . Light intensity at the tip of the 1 mm fiber was 100 mW for the 390 nm LED , and 150 mW for the 520 nm LED . The single-cysteine mutations of GuD2 were generated by site-directed mutagenesis using the Quick Change II kit ( Agilent technology ) performed on pcDNA3-GluD2 ( Ady et al . , 2014 ) . All mutants were verified by sequencing . We used human Embryonic Kidney cells ( HEK tsA201 , Sigma-Aldrich # 96121229 ) . Cells were certified by Sigma-Aldrich . Mycoplasma contamination status were negative . Cells were cultured in 25 cm2 tissue culture flask ( Falcon , Vented Cap , 353109 ) with a culture medium composed of Dulbeco’s Modified Eagle Medium ( Gibco life technologies , 31966047 ) containing Glutamax and supplemented with Fetal Bovine Serum ( 10% , Gibco life technologies , 10500064 ) , Nonessential Amino-Acids ( 1% , Life Technologies , 11140–035 ) , ampicillin , streptomycin ( 50 , 000 U , Gibco , life technologies , 15140–122 ) and mycoplasma prophylactic ( 2 . 5 mg , InvivoGen ) antibiotics . HEK tsA201 cells were freshly seeded and plated out in a 6-well plate , on coverslips ( 10 mm ) treated with poly-L-lysine hydrobromide ( Sigma , P6282-5MG ) . Cells were transiently transfected using calcium-phosphate precipitation , as described in Lemoine et al . , 2016 , using 1 µg of cDNA of GluD2 cysteine mutant per well . For co-transfection experiments , we used mGlu1b/GluD2 ratio from 0 . 7 to 1 , with a maximum of 2 µg of total DNA . The plasmid pRK5-mGlu1b used in this study is a generous gift of L . Prezeau ( IGF , Montpellier ) . Electrophysiological currents were recorded on HEK tsA201 cells at room temperature ( 21–25°C ) , 24–48 hr after transfection . Prior to whole-cell patch-clamp experiments , cells were incubated for 20 min with an extracellular solution containing 20 μM MAGu , and then washed for at least 5 min with a fresh external solution . Cells were perfused with an external solution containing ( in mM ) : 140 NaCl , 2 . 8 KCl , 2 CaCl2 , 2 MgCl2 , 12 glucose , 10 HEPES and NaOH-buffered at pH 7 . 32 . The external NMDG solution contained ( in mM ) : 140 NMDG , 2 . 8 KCl , 2 CaCl2 , 2 MgCl2 , 12 glucose , 10 HEPES and was KOH-buffered at pH 7 . 32 . Cells were patched with a borosilicate pipette ( 4–5 MΩ ) containing an intracellular solution containing ( in mM ) : 140 KCl , 5 MgCl2 , 5 EGTA , 10 HEPES , and pH-adjusted to 7 . 32 with KOH . For recording metabotropic activation of GluD2 by mGlu1 , the internal solution contained ( in mM ) : 140 K-gluconate , 6 KCl , 12 . 6 NaCl , 0 . 1 CaCl2 , 5 Mg-ATP , 0 . 4 Na-GTP , 1 EGTA , 10 HEPES , and was adjusted to pH 7 . 32 with KOH . Pentamidine and NMDG solutions were applied using a fast-step perfusion system equipped with three square tubes ( SF77B , warning instruments ) , as described in Lemoine et al . , 2016 . Illumination was carried out using a high-power LED system ( pE-2 , Cooled ) mounted directly on the epifluorescence port of a vertical microscope ( SliceScope Pro 6000 , Scientifica ) . Light output at the focal plane was 5 and 11 . 7 mW/mm2 for the 380 and 535 nm LEDs , respectively . Currents were recorded with an axopatch 200B and digitized with a digidata 1440 ( Molecular devices ) . Signals were low-pass filtered ( Bessel , 2 kHz ) and collected at 10 kHz using the data acquisition software pClamp 10 . 5 ( Molecular Devices ) . Electrophysiological recordings were extracted using Clampfit ( Molecular Devices ) and analyzed with R . Animal breeding and euthanasia were performed in accordance with European Commission guidelines and French legislation ( 2010/63/UE ) and procedures were approved by the French Ministry of Research ( Agreement APAFIS#16198–2018071921137716 v3 ) . Mice at age P30-40 were anesthetized with isoflurane and decapitated . Cerebella were rapidly extracted and transferred into ice-cold ACSF supplemented with 50 mM sucrose and 1 mM kynurenic acid . The composition of ACSF in mM was as follows: 126 NaCl , 26 NaHCO3 , 2 . 5 KCl , 1 . 25 NaH2PO4 , 1 MgCl2 , 2 CaCl2 , 20 glucose . pH was adjusted to 7 . 4 by continuous gassing with carbogen . Sagittal slices ( 250 µm ) were sectioned from the vermis on a vibratome ( Leica VT 1200S ) and transferred to oxygenated ACSF . Slices were incubated for 15 min at 30°C then transferred at room temperature before recording . Unless stated otherwise , all the steps were performed at room temperature . Purkinje cells were visually identified using infrared Dodt contrast imaging with a 60 × water immersion objective . Whole-cell recordings from Purkinje cells in cerebellar lobules IV-VI ( voltage-clamped at −70 mV , liquid junction potentials not corrected ) were performed with borosilicate glass pipettes ( WPI , 2–4 MΩ ) pulled with a horizontal micropipette puller ( Sutter instruments ) . Internal pipette solutions contained ( in mM ) : 140 Cs-gluconate , 5 CsCl , 2 MgCl2 , 0 . 5 EGTA , 2 Na-ATP ( pH 7 . 3 , adjusted with CsOH ) . Whole-cell currents were recorded at 20 kHz and filtered with a Bessel low-pass filter at 4 kHz using a patch-clamp amplifier ( Multiclamp 700B , Molecular Devices ) connected to a Digidata 1440A interface board ( Molecular Devices ) . Only Purkinje cells with a series resistance <12 MΩ ( not compensated; monitored during experiments by applying 200 ms , −5 mV voltage pulses ) were used for the analyses . Slices were incubated for 20 min under continuous oxygenation with MAGu 20 µM or vehicle ( DMSO 0 . 4% ) dissolved in 750 µl ACSF in a well of 24-well plate . The slices were then washed in ACSF for at least 20 min , and transferred in the recording chamber . Photoswitching was achieved by illuminating the slice as described above alternatively at 380 nm and 535 nm for 1 s . ( RS ) -DHPG ( 200 µM ) was diluted in ACSF and locally pressure-applied using a patch pipette placed in the dendrites of the recorded Purkinje-cell . A pneumatic microinjector ( Picopump , WPI ) was used to deliver 0 . 1–0 . 2 ms air pressure pulses ( 4–10 PSI ) every minute . DHPG-mediated currents were recorded at room temperature in presence of CNQX 10 µM , D-APV 25 µM , gabazine 10 µM and CGP 55845 0 . 5 µM . DHPG was applied immediately after the light stimulation . Parallel fiber stimulation was achieved every 10 s with a glass pipette filled with ACSF and placed in the outer half part of molecular layer . A constant voltage isolation unit ( DS3 , Digitimer Ltd ) was used to deliver 10 µs rectangular pulses ( 50–200 µA ) for extracellular stimulation . Parallel fibers inputs were identified by paired-pulse facilitation at 50 ms inter stimulus interval . Amplitudes of evoked EPSCs were averaged from 6 to 12 traces . EPSCs were recorded at 30°C in presence of gabazine 10 µM . Electrical stimulations were performed immediately after the light stimulation . The model of the GluD receptor has been obtained by homology modeling using the software modeller version 9 . 19 ( Webb and Sali , 2016 ) . The template was that of the glutamate receptor GluA2 ( PDB code 5weo ) ( Twomey et al . , 2017 ) . The automodel class has been used with slow level of MD refinement and the optimization has been repeated three times for each model . 500 models were prepared and the best , as assessed by the DOPE score , was retained for further studies . The structure of the protein and ligand were converted to pdbqt files with the software open babel 2 . 4 . 1 . Covalent docking was then performed with the software smina ( Koes et al . , 2013 ) . The box of 25*25*25 angstrom was defined manually to encompass the mutated residue and extend to the axis of symmetry . Covalent docking forced the maleimide to be in direct contact with the SG atom of the cysteine with which it is shown experimentally to form a covalent bond . The geometry of the ion channel has been computed with MOLEonline webserver , with the 'pore' mode . The resulting ion channel was color-coded as a function of the diameter of the channel allowing to illustrate the reduction of the diameter from a large ( green ) to a small ( red ) diameter . Data are plotted as mean ± SEM . Total number ( n ) of cells in each group and statistics used are indicated in figure and/or figure legend . Comparisons between means were performed using parametric tests ( two-sample t-test , Normality always verified , Shapiro-Wilk test of normality ) . Homogeneity of variances was tested preliminarily and the t-tests were Welch-corrected accordingly . For comparison with theoretical values of 0 or 1 , we performed either one-sample t-tests when Normality was verified , or a non-parametric test ( one-sample Wilcoxon tests ) when Normality was not verified . #p<0 . 1 , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . Time-course of current unblock and of thermal relaxation were fitted with the following mono-exponential function: ( 1 ) y=100×1-e-kxwith k the decay constant , and ln2/k the half-life . Dose-response relationships were fitted with the following equation: ( 2 ) y=MIN+MAX-MIN1+ ( xIC50 ) nHwith MAX the maximal current , MIN the minimal current , IC50 the pentamidine concentration yielding half block , and nH the Hill number . | Neurotransmitters are chemicals released by the body that trigger activity in neurons . Receptors on the surface of neurons detect these neurotransmitters , providing a link between the inside and the outside of the cell . Glutamate is one of the major neurotransmitters and is involved in virtually all brain functions . Glutamate binds to two different types of receptors in neurons . Ionotropic receptors have pores known as ion channels , which open when glutamate binds . This is a fast-acting response that allows sodium ions to flow into the neuron , triggering an electrical signal . Metabotropic receptors , on the other hand , trigger a series of events inside the cell that lead to a response . Metabotropic receptors take more time than ionotropic receptors to elicit a response in the cell , but their effects last much longer . One type of receptor , known as the GluD family , is very similar to ionotropic glutamate receptors but does not directly respond to glutamate . Instead , the ion channel of GluD receptors opens after being activated by glutamate metabotropic receptors . GluD receptors are produced throughout the brain and play roles in synapse formation and activity , but the way they work remains unclear . An obstacle to understanding how GluD receptors work is the lack of molecules that can specifically block these receptors’ ion channel activity . Lemoine et al . have developed a tool that enables control of the ion channel in GluD receptors using light . Human cells grown in the lab were genetically modified to produce a version of GluD2 ( a member of the GluD family ) with a light-sensitive molecule attached . In darkness or under green light , the light-sensitive molecule blocks the channel and prevents ions from passing through . Under violet light , the molecule twists , and ions can flow through the channel . With this control over the GluD2 ion channel activity , Lemoine et al . were able to validate previous research showing that the activation of metabotropic glutamate receptors can trigger GluD2 to open . The next step will be to test this approach in neurons . This will help researchers to understand what role GluD ion channels play in neuron to neuron communication . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"neuroscience"
] | 2020 | Probing the ionotropic activity of glutamate GluD2 receptor in HEK cells with genetically-engineered photopharmacology |
Proteasomes are central regulators of protein homeostasis in eukaryotes . Proteasome function is vulnerable to environmental insults , cellular protein imbalance and targeted pharmaceuticals . Yet , mechanisms that cells deploy to counteract inhibition of this central regulator are little understood . To find such mechanisms , we reduced flux through the proteasome to the point of toxicity with specific inhibitors and performed genome-wide screens for mutations that allowed cells to survive . Counter to expectation , reducing expression of individual subunits of the proteasome's 19S regulatory complex increased survival . Strong 19S reduction was cytotoxic but modest reduction protected cells from inhibitors . Protection was accompanied by an increased ratio of 20S to 26S proteasomes , preservation of protein degradation capacity and reduced proteotoxic stress . While compromise of 19S function can have a fitness cost under basal conditions , it provided a powerful survival advantage when proteasome function was impaired . This means of rebalancing proteostasis is conserved from yeast to humans .
Maintaining the integrity of the proteome is vital for all cells . Protein chaperone systems and the ubiquitin-proteasome pathway are essential components of the global architecture that sustains protein homeostasis . The importance of these systems is underscored by the fact that perturbation of protein homeostasis is central to diverse human diseases ( Balch et al . , 2008; Labbadia and Morimoto , 2015 ) . In neurodegenerative diseases , cells succumb to an overload of toxic protein aggregates . Whereas in cancer , malignant cells co-opt protein quality control systems to accommodate the severe proteotoxic stresses that arise from high mutation loads , relentless biomass accumulation and protein turnover ( Mendillo et al . , 2012; Deshaies , 2014; Scherz-Shouval et al . , 2014 ) . Because the ubiquitin-proteasome system is the major mechanism regulating protein turnover , cells are highly vulnerable to insults that impair the function of this system ( Varshavsky , 2012; Petrocca et al . , 2013 ) . Proteasome function can be hampered in many ways . For example , a multitude of environmental stresses cause the toxic accumulation of misfolded and aggregated proteins ( Parsell et al . , 1994; Gidalevitz et al . , 2011 ) . Such proteins can bind the proteasome and stabilize an inactive closed conformation ( Deriziotis et al . , 2011; Ayyadevara et al . , 2015 ) . In addition , chemically diverse natural product inhibitors of the ubiquitin-proteasome pathway are elaborated by organisms as diverse as terrestrial and marine bacteria , fungi , and plants ( Kisselev et al . , 2006; Schneekloth and Crews , 2011; Kisselev et al . , 2012 ) . These proteasome inhibitors can thwart the proliferation of neighboring organisms and thereby provide a competitive growth advantage . Applying this lesson from nature toward the treatment of cancer has been an effective strategy . Malignant cells in which the rates of protein synthesis outstrip proteasome degradation capacity are exceedingly vulnerable to proteasome inhibition ( Adams et al . , 2000; Cenci et al . , 2012; Deshaies , 2014 ) . Targeting the 26S proteasome with the proteasome inhibitor bortezomib ( Velcade , PS-341 ) is particularly effective in hematopoietic tumors . Indeed , it is broadly applied as a therapy for patients with myeloma or mantle cell lymphoma ( Chen et al . , 2011; Crawford and Irvine , 2013 ) . The 26S proteasome is an elaborate multi-subunit protein complex that is present in all eukaryotes . This complex is comprised of a 20S catalytic core that orchestrates peptide bond cleavage and a 19S regulatory complex that can be attached to either or both ends of the 20S core ( Besche et al . , 2009; Matyskiela et al . , 2013; Tomko and Hochstrasser , 2013 ) . The 19S complex recognizes ubiquitin-tagged substrates , cleaves ubiquitin chains , unfolds substrates and translocates the unfolded proteins into the catalytic chamber of the 20S core ( Hochstrasser , 1996; Finley , 2009; Matyskiela et al . , 2013 ) . Despite an exquisitely detailed understanding of proteasome function and of the mechanism of action of proteasome inhibitors , we have a limited understanding of the molecular mechanisms that cells deploy to resist the cytotoxic effects of reduced flux through the proteasome ( Kale and Moore , 2012 ) . Such an understanding is of great importance in the dynamics of natural ecosystems in the face of diverse proteotoxic stresses and in the clinic , where pre-existing intrinsic resistance and acquired resistance following drug exposure have limited the effectiveness of bortezomib as a therapeutic . To gain insights into the mechanisms that allow cells to withstand reduced flux through the proteasome , we took advantage of highly specific chemical inhibitors . While proteasome function can be impaired by many factors , none can be controlled with the dosage-dependent precision of proteasome inhibitors such as bortezomib and the peptide aldehyde MG132 . These inhibit both 20S and 26S proteasomes by targeting the core proteolytic catalytic activity of the 20S subunits ( Kisselev et al . , 2006 , 2012; Goldberg , 2012 ) . We used these inhibitors at toxic levels in an unbiased , genome-wide screen . We selected for cells that were resistant to the inhibitors from a library of 100 million gene-trap insertions , using a human cell line that is haploid for all chromosomes except chromosome 8 ( Carette et al . , 2009 , 2011a ) . This approach is analogous to screens so broadly and effectively used in haploid yeast and identifies loss-of-function events that allow human cells to survive diverse toxic insults ( Guimaraes et al . , 2011; Reiling et al . , 2011; Carette et al . , 2011a , 2011b; Winter et al . , 2014 ) . The results of our screen led us to the discovery of a surprising and highly conserved strategy by which organisms can protect themselves from the toxic effects of reduced flux through the proteasome .
For our screens , we used a library of near-haploid human chronic myeloid leukemia cells ( KBM7 ) containing approximately 100 million retroviral gene-trap insertions that target over 98% of transcribed genes . To identify genes that increase resistance to proteasome inhibition , we exposed cells for 4 weeks to either MG132 or bortezomib . We then further expanded the pools of resistant cells to enable the amplification and sequencing of the insertion sites ( Figure 1A ) . 10 . 7554/eLife . 08467 . 003Figure 1 . The 19S regulatory subunits of the proteasome are the most significant mediators of resistance to proteasome inhibitor toxicity . ( A ) Schematic representation of the screen . One hundred million KBM7 cells subjected to random gene deletion using retroviral gene-trap insertions were exposed to either MG132 ( 700 nM ) or bortezomib ( 18 nM ) for 4 weeks . Surviving cells were expanded and insertions identified by sequencing . ( B ) The p-values of the recovered insertions from the MG132 screen are plotted ( log2 ) . Bubble sizes represent the number of insertions . ( C ) Compilation of the most significant gene deletions conferring resistance to MG132 with the gene name , number of inserts , and p-value . The subunits of the 19S regulatory complex are highlighted with orange . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 00310 . 7554/eLife . 08467 . 004Figure 1—figure supplement 1 . ( A ) Schematic representation of the knockout strategy to generate the mutant PSMD12 and PSMC2 ES cells . Genetraps are in antisense ( top ) , and sense ( bottom ) orientation . ( B ) Brightfield ( top panels ) and fluorescence ( lower panels ) microscopic imaging of FACS sorted PSMD12 and PSMC2 clones , stably expressing Cre IRES mCherry fusion transcripts ( 40× mag ) . ( C ) Genotyping of the ES clones by PCR before and after Cre-mediated inversion . ( D ) Relative gene expression of PSMD12 and PSMC2 in control and mutants as quantified by RT PCR ( n = 16–18 for each gene ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 004 For the MG132 resistance screen , we identified 992 independent insertion sites in the pool of surviving cells . Surprisingly , all insertions that reached a high level of statistical enrichment ( p-value < 1 e−7 ) lay in genes encoding subunits of the proteasome 19S regulatory complex ( Figure 1B , C; Supplementary file 1 ) . These included both ATPase subunits ( PSMC2 , PSMC3 , PSMC4 , PSMC5 , and PSMC6 ) as well as non-ATPase subunits ( PSMD2 , PSMD6 , PSMD7 , and PSMD12 ) . No insertions were recovered in genes encoding subunits of the 20S catalytic core ( Supplementary file 1 ) . For the bortezomib resistance screen , we recovered 538 independent insertions . The results of the 2 screens were remarkably similar with seven of the ten most highly enriched genes encoding subunits of the 19S complex ( Supplementary file 1 ) . Such strikingly similar results from 2 unbiased screens with chemically distinct proteasome inhibitors strongly indicated that altering the 19S complex can protect cells against compounds that inhibit the 20S catalytic core . From the resistant pools of cells , we then attempted to isolate stable clones that contained 19S subunit gene insertions . We were unable to do so . Next , we attempted to delete PSMD12 in a near-haploid fibroblast cell line ( HAP1 ) ( Carette et al . , 2010; Essletzbichler et al . , 2014 ) . With this targeted approach using CRISPR constructs , we were only able to recover diploid cell variants in which just one of the two PSMD12 alleles was disrupted . Finally , from a collection of haploid mouse embryonic stem cells that harbor reversible gene-trap cassettes ( Elling et al . , 2011 ) , we identified two clones with cassettes located in the first intron of the PSMC2 or PSMD12 genes . In these cells , inversion of the cassettes would generally be expected to inactivate the targeted gene . We induced Cre-mediated inversion in over 3000 cells harboring each cassette , but less than 1% of the cells survived . We confirmed that inversion had occurred in the surviving cells . However , all of the stable clones that emerged retained expression of the targeted subunits ( Figure 1—figure supplement 1 ) . These findings confirm that , as others have found in yeast and Drosophila , the function of the 19S regulatory complex is essential for sustained proliferation of mammalian cells under basal conditions . Presumably , cells carrying 19S mutations had become enriched in our initial MG132 and bortezomib screens because they provided cells with a short-term advantage for growth over several generations . Next , we asked if a simple reduction in the expression of 19S subunits could protect against the toxicity of proteasome inhibitors . We assembled a panel of shRNA-expressing lentiviruses targeting seventeen 19S subunits and three 20S subunits ( PSMB5 , PSMB7 , and PSMA3 ) . Each gene was targeted separately using four different shRNAs ( Figure 2A , Supplementary file 2 ) and we averaged the effects on viability from these four hairpins . 10 . 7554/eLife . 08467 . 005Figure 2 . Reducing expression of 19S subunits increases the levels of active 20S proteasomes and protects cancer cells from proteasome inhibition . ( A ) HepG2 cells were infected with 80 shRNAs targeting 20 different subunits of the proteasome and 10 control shRNAs . Infected cells were then exposed to 12 nM bortezomib and cell number was examined 4 days later . The plot represents the average ( ± SEM ) of four different hairpins targeting the indicated proteasome subunits and their relative cell number following bortezomib treatment . 19S subunits are depicted with orange bars , the control with a blue bar , and 20S subunits with green bars . ( B ) HepG2 cells harboring either a control shGFP or shRNAs targeting two proteasome subunits ( shPSMC5 , shPSMD2 ) displayed significant growth differences in the presence of 12 nM bortezomib . ( C ) Proteasome complex content in the shGFP- , shPSMC5- , and shPSMD2-expressing HepG2 cells was analyzed by native gel electrophoresis after 24 hr of treatment with or without 12 nM bortezomib revealing an increase in 20S proteasome levels and activity in cells knocked down for PSMC5 or PSMD2 . ( D ) The relative cell number of cells harboring a control shLacZ or each of 4 individual shRNAs targeting shPSMD2 was analyzed 4 days after addition of the indicated concentrations of bortezomib . ( E ) HepG2 cells stably expressing four different shRNAs targeting the PSMD2 subunit and a control shRNA ( lacZ ) were analyzed by Western blot for the indicated proteins 24 hr with or without bortezomib treatment . ( F ) Proteasome complex levels and activity in a control HepG2 cells and 4 cell lines with reduced PSMD2 levels with and without a 24 hr incubation with bortezomib ( 12 nM ) . Proteasome complex levels were detected by immunoblot analysis and 20S proteasome activity by measuring the hydrolysis of Suc-LLVY-AMC by substrate overlay assays . In B and C , the graph represents the average of four replicas and their SEM . The p-values were obtained by conducting a two-tailed unpaired t-test . *p < 0 . 05 **p < 0 . 01 Bortz- Bortezomib ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 00510 . 7554/eLife . 08467 . 006Figure 2—figure supplement 1 . ( A–D ) Examining the effect of proteasome subunit knockdown in different cell lines . 80 shRNAs targeting 20 different proteasome subunits and control hairpins were expressed in HepG2 ( A ) , H838 ( B ) , T47D ( C ) , and H1792 ( D ) cells by viral transduction . Each subunit was targeted by 4 different shRNAs and the cell viability was measured as the relative cell number compared to the average of non-targeting shRNAs 5 days after the initial introduction of the shRNAs . ( E , F ) HepG2 cells with shRNAs targeting PSMC5 , PSMD2 , and green fluorescent protein ( GFP ) were grown out . Their relative growth was analyzed in the absence of bortezomib ( E ) and their protein content was analyzed 24 hr after the addition of either 8 or 12 nM of bortezomib ( F ) . ( G ) The relative cell number of cells harboring a control shLacZ ( black ) or each of 5 individual shRNAs targeting shPSMC5 ( Cayenne ) was analyzed 4 days after addition of the indicated concentrations of bortezomib . ( H ) HepG2 cells stably expressing shRNAs targeting the PSMC5 subunit and a control shRNA ( lacZ ) were analyzed by Western blot for the indicated proteins 24 hr with or without bortezomib treatment . ( I–N ) The HepG2 cells with shRNAs targeting PSMC5 , PSMD2 , and GFP ( described above ) were further exposed to a short panel of stress inducers including bortezomib ( I ) , tunicamycin ( J ) , rohinitib-RHT ( K ) , Hsp90 inhibition ( L ) , withaferin A ( M ) , and cyclohexamide ( N ) at indicated concentrations and the relative cell number ( RFU ) was examined after 4 days . The graphs represent the average of at least 4 different measurements and the SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 006 With or without bortezomib , knockdown of any of the subunits of the 20S catalytic core reduced viability ( Figure 2A , Figure 2—figure supplement 1A–D ) . In contrast , the effects of knocking down several different 19S subunits had opposing effects , depending on the absence or presence of the inhibitor . In its absence , 19S subunit reduction had a fitness cost and decreased cell viability; in its presence , 19S subunit reduction provided a survival advantage and increased viability ( Figure 2A , Figure 2—figure supplement 1A–D ) . To investigate further , we sought to recover stable clones with reduced levels of 19S subunits . Long-term 19S subunit reduction impeded the growth of most cells , but we were able to propagate two lines that proliferated normally ( Figure 2—figure supplement 1E ) . These lines stably expressed shRNA targeting either PSMC5 or PSMD2 . In both cases , the lines had only a modest reduction in protein levels ( Figure 2—figure supplement 1F ) . At a concentration of bortezomib that completely inhibited the proliferation of control cells ( 12 nM ) , these cells continued proliferating ( Figure 2B , figure 2—figure supplement 1G ) . We next used native gels to assess the levels and ratios of 20S and 26S proteasome complexes following a 24-hr exposure to bortezomib ( Figure 2C ) . Bortezomib-induced cytotoxicity was not observed in control cells at this time point ( Figure 2B , star ) . In cells with reduced levels of either PSMC5 or PSMD2 , 26S proteasome complexes were reduced and 20S proteasome complexes were increased ( Figure 2C , upper panel ) . These changes persisted in the presence of bortezomib ( Figure 2C ) . Next , we assayed 20S proteasome complex activity with a fluorescently labeled substrate on native gels . Exposing control cells to 12 nM bortezomib for 24 hr led to a nearly complete inactivation of 20S proteasome function ( Figure 2C , lower panel ) . Cells with reduced levels of either PSMC5 or PSMD2 maintained a significant level of 20S proteasome activity ( Figure 2C ) . To validate our findings , we selected four additional cell lines each with a different shRNA driving modest reduction in PSMD2 protein levels . Again , all of these lines were significantly more resistant to bortezomib than the parental line ( EC50 values increased by three to sixfold , Figure 2D ) . In these cells , we also observed that 26S proteasome complexes were reduced and that both the levels and the activity of 20S proteasome complexes were sharply increased ( Figure 2E ) . Notably , in the presence of bortezomib , reducing 19S subunits enabled the relative preservation of the levels and activity of 20S proteasome complexes ( Figure 2E ) . A likely mechanism by which 19S subunit reduction might promote resistance is by induction of the cytoprotective stress responses that allow cells to cope with the increase in proteotoxic stress caused by the proteasome inhibitor . One major response to such inhibition is the activation of NRF1 , a transcriptional regulator of proteasome gene expression that increases the expression of proteasome subunits , elevates proteasome content , and promotes resistance ( Radhakrishnan et al . , 2010; Steffen et al . , 2010; Radhakrishnan et al . , 2014; Sha and Goldberg , 2014 ) . A second major response to proteasome inhibitors is activation of heat-shock factor 1 ( HSF1 ) , the master regulator of the heat-shock response , which increases levels of HSP70 and other protein chaperones . Surprisingly , in the four cell lines with modest reductions of PSMD2 , we did not detect constitutive activation of NRF1 , and correspondingly , the expression of 20S subunits was unaltered ( Figure 2F ) . Moreover , HSF1 was not activated as reflected by the stable expression of HSP70 , the protein that is most highly responsive to proteotoxic stress ( Figure 2F ) . Consistent with these findings , polyubiquitinated proteins did not accumulate in the PSMD2 knockdown cells ( Figure 2F ) suggesting that modestly reducing 19S subunit levels did not itself induce a cytoprotective stress response . Not only were stress response pathways not activated , but the response to bortezomib was blunted in cells with reduced PSMD2 levels . The accumulation of polyubiquitinated proteins was reduced relative to control cells treated with the inhibitor . The activation of NRF1 was also reduced ( Figure 2F ) . We obtained very similar results in cells with PSMC5 knockdown except in this case HSP70 levels were also reduced ( Figure 2—figure supplement 1H ) . Notably , the protective effect of 19S subunit reduction was specific to the toxicity caused by proteasome inhibitors . Cells with reduced PSMD2 levels remained fully sensitive to small molecule mediators of ER stress , HSP90 inhibition , thiol adduct formation or blockade of translation initiation or translation elongation among other stresses ( Figure 2—figure supplement 1I–N ) . Thus , modest 19S subunit reduction protected cells by selectively lowering the proteotoxic stress that is generated by proteasome inhibition . To examine in detail the transcriptional changes that characterize cells with increased resistance to proteasome inhibition , we performed whole-genome RNA-sequencing in two lines with modest reductions in PSMD2 . Sequence data confirmed that PSMD2 mRNA levels were reduced by ∼50% in both lines . We examined basal gene expression and the effects of bortezomib treatment on these cells , comparing them to cells carrying a control lacZ shRNA construct . Under basal conditions ( in the absence of bortezomib ) , cells with reduction in PSMD2 showed a strong induction of components of the ribosome ( gene set enrichment analysis false discovery rate [FDR] q-value = 4 . 0 e−22 ) ( Supplementary file 3 ) . Genes encoding the 20S subunits of the proteasome were not induced , consistent with our earlier observation that NRF1 is not activated and 20S subunit levels are unchanged ( Figure 2F ) . The most strongly downregulated gene category involved components of diverse proteotoxic stress responses ( FDR q-value = 2 . 31 e−9 ) ( Supplementary file 3 ) . These included genes for sentinel proteins that respond to heat-shock ( e . g . , HSPA1A , HSPA1B , HSPA8 , HSPB1 , and HSP90AA1 ) , oxidative stress ( e . g . , HMOX1 ) , and ER stress ( e . g . , CHOP/DDIT3 , TRIB3 , and HERPUD1 ) ( Figure 3A ) . Genes previously identified as an HSF1-regulated cancer-specific transcriptional program were also downregulated ( Mendillo et al . , 2012; Santagata et al . , 2013 ) ( FDR q-value = 0 . 042; normalized enrichment score ( NES ) 1 . 57 ) ( Figure 3—figure supplement 1A ) . Therefore , 19S subunit reduction not only fails to induce classic adaptive stress responses , but actually lowers basal levels of proteotoxic stress . 10 . 7554/eLife . 08467 . 007Figure 3 . Inhibition of bortezomib-mediated transcriptional responses in PSMD2 knockdown cells . RNA-seq gene expression profiling was conducted on HepG2 cells that harbor two different shRNAs targeting PSMD2 ( PSMD2-1 , PSMD2-2; same cells used in Figure 2F ) and on control cells ( shLacZ ) . The effects of reducing PSMD2 levels on gene expression were highly correlated in both basal conditions ( Pearson's r = 0 . 99 ) and following bortezomib treatment ( Pearson's r = 0 . 94 ) . ( A , B ) Heat-shock- ( black ) , oxidative stress- ( blue ) , and ER stress- ( green ) related gene expression were all lower in the PSMD2 knockdown cells vs control cells under both basal conditions ( A ) and upon introduction of 12 nM bortezomib for 24 hr ( B ) . ( C ) Gene set enrichment analysis of genes upregulated in control but not in PSMD2 shRNA cells following bortezomib treatment . Enrichment was calculated for the indicated gene sets and is presented as a normalized enrichment score ( NES ) . Statistically significant enrichment ( false discovery rate [FDR] q-value < 0 . 05 ) is shown in red; non-significant enrichment is shown in gray . ( D ) Expression levels of genes previously characterized as suppressors of bortezomib-induced toxicity ( Chen et al . , 2010 ) are down-regulated in the PSMD2 knockdown cells following the addition of bortezomib . ( E ) Heat map depicting fold change in mRNA levels of genes differentially expressed in cells harboring control shRNA or PSMD2 shRNAs in the presence or absence of 12 nM bortezomib . Gene ontology enrichment is shown to the right of the panel . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 00710 . 7554/eLife . 08467 . 008Figure 3—figure supplement 1 . ( A ) Gene set enrichment analysis using the set of genes that are bound by HSF1 in MCF7 cancer cells under 37° basal conditions ( Mendillo et al . , 2012 ) was performed on genes negatively regulated in PSMD2 knockdown cells ( siPSMD2 ) vs control cells ( LacZ ) . ( B , C ) Gene set enrichment analysis using the set of genes that are induced following heat shock ( B ) , or genes that when knocked down confer resistance to bortezomib ( C ) was performed on genes negatively regulated in PSMD2 knockdown cells treated with bortezomib ( siPSMD2_Velcade ) vs control cells treated with bortezomib ( LacZ_Velcade ) . Enrichment plot and statistics are displayed . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 008 As expected , in control cells , bortezomib treatment unleashed a powerful transcriptional response characterized by a sharp increase in stress-response transcripts ( FDR q-value = 7 . 07 e−13 ) ( Supplementary file 3 ) . This potent bortezomib-induced stress response was markedly attenuated in cells with reduced PSMD2 levels ( Figure 3B ) . The suppressed genes include ones involved in oxidative stress and ER stress responses , as well as genes in the HSF1-mediated heat-shock response ( FDR q-value < 0 . 01; NES 2 . 317 ) ( Figure 3B , C and Figure 3—figure supplement 1B ) . For example , induction of HSP70 family heat-shock genes HSPA1A and HSPA1B was reduced by eightfold , while HSPA6 was entirely suppressed ( Figure 3B ) . Previous work has identified 28 mediators of bortezomib toxicity , many of which are upregulated upon bortezomib exposure ( Chen et al . , 2010 ) . When we treated control cells with bortezomib , the mRNA transcripts of many of these genes increased significantly ( e . g . , ATG4A , DDX27 , GADD45A , NUP54 , ODC1 , PMAIP1/NOXA , SETX , SNIP1 , and TAX1BP1 ) ( Supplementary file 3 ) . This response was also strongly attenuated in cells with reduced PSMD2 levels ( Figure 3C , D , and Figure 3—figure supplement 1C ) . Overall , selectively compromising 19S subunit expression broadly reduces the diverse transcriptional responses that normally ensue when flux through the proteasome is reduced . To further characterize the transcriptional effects of 19S subunit reduction , we performed a cluster analysis on the genes displaying the highest differential expression in our RNA-seq experiment ( Figure 3E ) . This analysis confirmed that PSMD2 reduction strongly blunted the bortezomib-mediated induction of stress response genes ( FDR q-values 1 . 2 e−5 to 1 . 2 e−13 ) . It also revealed broad changes in genes involved in small molecule metabolism , which remain to be deciphered . One group of genes highlighted by this analysis revealed a connection between the suppression of the cell cycle and increased resistance to bortezomib . In cells with reduced levels of PSMD2 , bortezomib treatment strongly repressed genes involved in DNA replication ( FDR q-value = 1 . 4 e−32 ) and cell cycle control ( FDR q-value = 1 . 8 e−90 ) . These genes include replication factors , polymerases , cyclins , and cyclin-dependent kinases . This accentuated anti-proliferation response suggests that cells with reduced 19S subunits are primed to enter a protected , quiescent-like state when flux through the proteasome is compromised . To model the effects of transient reduction of 19S subunits , we developed a cell line in which a PSMD2-targeting shRNA is transiently expressed from a doxycycline-regulated promoter ( Figure 4 ) . The effects of transient reduction of PSMD2 mirrored the effects of stable PSMD2 reduction . Most notably , it significantly increased resistance to both bortezomib ( Figure 4B , Figure 4—figure supplement 1A ) and MG132 ( Figure 4—figure supplement 1B ) . Again , this resistance was selective , and not accompanied by increased resistance to other small molecule stressors ( Figure 4—figure supplement 1C–F ) . In the absence of bortezomib , transient 19S reduction did not activate NRF1 or HSF1 ( Figure 4C ) and reduced activation of NRF1 and HSF1 following bortezomib exposure ( Figure 4C ) . 10 . 7554/eLife . 08467 . 009Figure 4 . Transient induction of PSMD2 shRNA is sufficient to promote resistance to proteasome inhibition . ( A ) Schematic representation of the experimental design . ( B ) T47D cells harboring a doxycycline-inducible PSMD2 shRNA were grown in the presence or absence of 1 μg/ml doxycycline for 48 hr . Cells were then collected , washed , and plated in the absence of doxycycline for 24 hr prior to exposure to increasing concentrations of bortezomib . Relative cell numbers were measured 3 days later . ( C ) Protein content analysis by immunoblot for the indicated proteins on lysates from control or cells pre-treated for 48 hr with doxycycline ( Dox ) , followed by a recovery of 24 hr and then incubation with 10 nM bortezomib for an additional 24 hr . ( D ) Native gel analysis of proteasome complexes in cells pre-treated as in ( C ) . The proteasome complex levels and activity of the 20S proteasome were assessed by native gel electrophoresis . Loading controls were analyzed by immunoblot for PSMD2 and tubulin following SDS-PAGE . ( E ) Glycerol gradient fractionation ( 10–40% ) was conducted on cells pre-treated for 48 hr with doxycycline ( Dox ) , followed by a recovery of 24 hr and then incubation with 15 nM bortezomib for an additional 24 hr . Proteasome activity in the fractions collected was assessed with proteasome-Glo and proteasome content was analyzed by immunoblotting with PSMD1 and 20S subunits specific antibodies . ( F ) The rate of degradation was analyzed in cells with reduced levels of PSMD2 ( green bars ) vs control ( red bars ) in the presence or absence of 10 nM bortezomib ( treatment for 20 hr ) by monitoring the release of H3-Phe in pre-labeled cells . ( G ) Rate of total protein synthesis was determined in cells with reduced levels of PSMD2 ( green bars ) vs control ( red bars ) in the presence or absence of 10 nM bortezomib ( treatment for 20 hr ) by measuring the rate of incorporation of 3H-phenylalanine for 1 hr . The p-values were obtained by conducting a two-tailed unpaired t-test . **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 00910 . 7554/eLife . 08467 . 010Figure 4—figure supplement 1 . ( A–F ) PSMD2 shRNA was induced for 48 hr with 1 μg/ml doxycycline . Cells were then collected , washed , and plated in the absence of doxycycline 24 hr prior to exposure to increasing concentration of bortezomib ( A ) , MG132 ( B ) , cyclohexamide ( C ) , withaferin A ( D ) , tunicamycin ( E ) , and rotenone ( F ) . ( G–I ) PSMD2 KD was induced as described above and then cells were grown in the presence or absence of 15 nM bortezomib for the indicated time points ( 3 , 6 , 24 hr ) . ( G ) Immunoblot analysis of the total proteasome subunits following SDS-PAGE ( upper panels ) and the 20S proteasome complex levels after native gel electrophoresis . These results were quantified by Imagelab software and plotted ( H ) . ( I ) Lysosomal degradation rate was measured in control and PSMD2 knock down ( Dox ) cells in the presence or absence of 10 nM Bortezomib treatment for 20 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 010 We used native gels and glycerol gradient fractionation to monitor the levels and activity of proteasome complexes following transient reduction of PSMD2 ( Figure 4D , E , Figure 4—figure supplement 1G , H ) . Transient reduction of this 19S subunit increased the ratio of 20S/26S proteasomes and total levels of 20S proteasome activity ( Figure 4D , E ) without increasing total levels of 20S subunits ( Figure 4C , Figure 4—figure supplement 1G , H ) . Similar to the results we obtained in stable knockdown cells ( Figure 2E ) , transient 19S subunit reduction preserved a fraction of the 20S proteasome complexes following bortezomib treatment and a corresponding portion of their activity remained ( Figure 4D , E , Figure 4—figure supplement 1G , H ) . Next , we examined the impact of reducing PSMD2 levels on protein degradation and protein synthesis in cells labeled with tritiated-phenylalanine for 24 hr . We then separately measured the rates at which labeled proteins were degraded through the proteasome and through the lysosome . In the absence of bortezomib , transiently reducing PSMD2 lowered rates of proteolysis by the lysosome ( Figure 4—figure supplement 1I ) , but it did not affect protein degradation by the proteasome ( Figure 4F ) . This suggests that the 26S proteasome is normally present in excess . To measure rates of protein synthesis , cells were pulse labeled with tritiated-phenylalanine for 1 hr . Reducing PSMD2 levels resulted in a significant and reproducible 7% decrease in the rate of protein synthesis ( Figure 4G ) . Thus , even though PSMD2 knockdown did not reduce protein degradation capacity , it did trigger a reduction in protein translation ( Figure 4G ) , a change that may contribute to lowering basal levels of proteotoxic stress . Next , we measured protein degradation and synthesis rates after a 20-hr treatment with bortezomib . The rate of protein degradation sharply decreased in control cells ( Figure 4F ) . Rates of proteolysis by the lysosome remained unchanged ( Figure 4—figure supplement 1I ) . Transiently reducing PSMD2 strongly counteracted the inhibition of proteasome degradation ( Figure 4F ) . Reducing the levels of 19S subunits partially preserved proteasome complex levels following the inhibition of flux through the proteasome by bortezomib . Correspondingly , it partially preserved 20S activity and the cells overall ability to degrade proteins ( Figures 2 , 4D , E , Figure 4—figure supplement 1G , H ) . Following bortezomib treatment , the rate of protein synthesis also sharply decreased in control cells ( Figure 4G ) . This drop reflects the global repression of protein synthesis that normally follows strong proteotoxic stress ( Holcik and Sonenberg , 2005; Shalgi et al . , 2013 ) . Transiently reducing PSMD2 protein levels strongly counteracted the bortezomib-mediated suppression of protein synthesis ( Figure 4G ) . Thus , reducing PSMD2 levels counteracts the effects of bortezomib on both protein degradation and protein synthesis . Human cancer cell lines have a broad range of sensitivities to proteasome inhibition . We asked if this might correlate with changes in 19S subunit expression . We analyzed the Genomics of Drug Sensitivity in Cancer ( GDSC ) database , a public resource of transcriptional data and drug responsiveness collected from a spectrum of human cancer cell lines with diverse tissue origins and diverse oncogenic lesions ( Garnett et al . , 2012 ) . We ranked the 310 cell lines in the data set by their half maximal inhibitory concentration ( IC50 ) to either MG132 or to bortezomib ( highest to lowest ) . The cells comprising the top 10% were defined as the ‘resistant’ group and those in the bottom 10% were defined as the ‘sensitive’ group . From the 31 cell lines in each group , we averaged the expression levels of all of the 20S subunits ( PSMA and PSMB mRNA ) and the expression levels of all of the 19S subunits ( PSMC and PSMD mRNA ) . We found no significant difference in the average expression of 20S subunits between the two groups ( Figure 5A , B left panels ) . However , cells that were the most resistant to either MG132 or to bortezomib had significantly lower levels of 19S transcripts ( PSMC and PSMD mRNA ) than cells that were sensitive ( Figure 5A , B right panels; p-value = 0 . 003 for MG132; p-value = 0 . 0008 for bortezomib ) . This observation is striking as the expression levels of all proteasome subunits , both 20S and 19S , are regulated by similar mechanisms and are normally highly correlated ( Jansen et al . , 2002; Radhakrishnan et al . , 2010 , 2014; Sha and Goldberg , 2014 ) . 10 . 7554/eLife . 08467 . 011Figure 5 . Reduced expression of 19S subunits correlates with resistance to proteasome inhibitors . ( A , B ) Analysis of expression data from 315 cell lines in the Genomics of Drug Sensitivity in Cancer ( GDSC ) database ( Garnett et al . , 2012 ) . The levels of 20S proteasome subunit ( PSMAs and PSMBs ) gene expression ( A and B left panels ) and 19S subunit ( PSMCs and PSMDs ) gene expression ( A and B right panels ) were analyzed in the cell lines that are the 10% most sensitive or the 10% most resistant to either MG132 ( A ) or bortezomib ( B ) . ( C ) The relative expression level of each 19S complex subunit was analyzed in the bortezomib resistant and sensitive groups . Expression levels with deviation of more than twofold from the average were color-coded ( red-up , green-down ) . The p-values were obtained by conducting a two-tailed unpaired t-test . **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 01110 . 7554/eLife . 08467 . 012Figure 5—figure supplement 1 . The relative expression level of each 19S complex subunit was analyzed in the MG132 resistant and sensitive groups . Expression levels with deviation of more than twofold from the average were color-coded ( red-up , green-down ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 012 We next assessed the expression of the individual 19S regulatory complex subunits in each of the resistant and sensitive cell lines . A heat map of genes with significantly altered expression ( >twofold deviation from average ) revealed that bortezomib-sensitive cells commonly showed increased expression of many different 19S subunits ( Figure 5C , right panel-red ) . Resistant cells generally had at least a twofold reduction in expression of one or more 19S subunits ( Figure 5C , left panel-green ) . This was also true in the case of MG132 ( Figure 5—figure supplement 1 ) . Thus , alterations in 19S subunit expression commonly occur in the evolution of cancer cells . Human cancers are increasingly viewed as complex ecosystems comprised of cells harboring enormous genetic , functional and phenotypic heterogeneity ( Meacham and Morrison , 2013 ) . We asked if heterogeneity arising from 19S subunit expression can alter population dynamics and confer a fitness advantage in the face of exposure to proteasome inhibitors . To do so , we investigated the effects of transiently reducing PSMD2 expression in only a subpopulation of cells . We created two cell lines—one line that expresses red fluorescent protein ( turboRFP ) and the doxycycline-inducible PSMD2-targeting shRNA and another line that expresses green fluorescent protein ( GFP ) and a doxycycline-inducible control shRNA ( Figure 6A ) . First , we induced shRNA expression with doxycycline for 48 hr . After recovery , we mixed shPSMD2-RFP and shControl-GFP cells at different ratios ( 1:1 , 1:2 , 1:5 , or 1:10 ) , adding the cells with reduced PSMD2 as the minority subpopulation . 24 hr after plating , we treated these mixed populations of cells for 48 hr with increasing concentrations of bortezomib ( 5 , 7 . 5 , or 10 nM ) . We allowed the cells to recover in the absence of bortezomib and then we quantified the red and green cells by fluorescence-activated cell sorting ( FACS ) analysis ( Figure 6A ) and captured representative images by fluorescence microscopy ( Figure 6B ) . 10 . 7554/eLife . 08467 . 013Figure 6 . Transient 19S subunit reduction confers a competitive survival advantage in the presence proteasome inhibitors . ( A , B ) T47D cells that harbor a doxycycline-inducible control shRNA ( GFP ) or a doxycycline-inducible PSMD2 shRNA ( TurboRFP ) were incubated with doxycycline for 48 hr . Cells were collected , counted , and plated at the indicated ratios of TurboRFP-expressing PSMD2 shRNAs/GFP expression control shRNAs ( 1:1 , 1:2 , 1:5 , and 1:10 ) . 24 hr later , bortezomib was added at the specified concentrations and incubation continued for 48 hr . Cells were allowed to recover in the absence of bortezomib for another 48 hr and then visualized by microscopy ( B ) or analyzed by FACS after 6 days of recovery ( A , and pie charts in B ) . The green and red images were overlaid using ImageJ . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 013 In the absence of proteasome inhibitors , the initial plating ratios of these cells were maintained for 6 days ( 1:1 , 1:2 , 1:5 , and 1:10 ) ( Figure 6A ) . In contrast , even low concentrations of bortezomib ( 5 nM ) substantially shifted the populations of surviving cells and higher concentrations ( 7 . 5 nM and 10 nM ) elicited even more substantial shifts ( Figure 6A , B ) . In the presence of proteasome inhibitors , cells with modestly reduced levels of PSMD2 had a strong competitive advantage . Finally , because the essential role of the proteasome in maintaining protein homeostasis is conserved across all eukaryotes ( Hilt and Wolf , 1995 ) , we asked whether reducing expression of 19S subunits confers resistance to proteasome inhibitors in an evolutionary distant organism—the yeast Saccharomyces cerevisiae . As the 19S regulatory complex components are essential for viability in yeast , we utilized a library of hypomorphic ( DAmP ) alleles for essential yeast genes . In this library , the expression of individual mRNA species is reduced from two to 10-fold by replacing the mRNA's 3′ untranslated region ( Breslow et al . , 2008 ) . DAmP-strains were available for 12 genes comprising the 19S regulatory complex . These strains showed no significant growth-impairment under basal conditions ( Figure 7—figure supplement 1 ) . However , five of these twelve strains had significantly increased resistance to proteasome inhibition by MG132 ( Figure 7 ) . Most notable were Rpn5 and Rpt6 , the yeast orthologs of PSMD12 and PSMC5—the two most significantly enriched genes in our MG132 screen in human cells ( Figure 1C ) . 10 . 7554/eLife . 08467 . 014Figure 7 . Reducing the levels of 19S subunits is an evolutionarily conserved mechanism to acquire resistance to proteasome inhibition . Proteasome subunit DAmP strains and the BY4741 control strain were grown in the presence or absence of 50 μM MG132 for 48 hr . The relative change in OD induced by MG132 is plotted . Five proteasome subunit DAmP strains exhibited significantly reduced toxicity in the presence of MG132 . **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 01410 . 7554/eLife . 08467 . 015Figure 7—figure supplement 1 . Proteasome subunit DAmP strains and the BY4741 control strain were grown in YPD media and OD600 was measured after 48 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 08467 . 015
We have identified a highly conserved mechanism that enables organisms as diverse as yeast and humans—separated in evolution by over 1 billion years—to withstand inhibition of protein flux through the proteasome . Very surprisingly , when the proteasome is inhibited to toxic levels , suppressing individual components of the 19S regulatory complex increases cell survival . While strong reduction of any of these subunits is not tolerated , modest reduction is protective . In this partially protective state , 26S proteasomes decrease and the levels and activity of 20S proteasomes sharply increase . In the absence of proteasome inhibitors , the change in 20/26S proteasome complex ratios does not reduce protein degradation , polyubiquitinated substrates are not elevated , and hallmark stress responses are not activated . Moreover , at concentrations of proteasome inhibitors that normally unleash these responses , they are suppressed . A number of mechanisms could potentially mediate this protective effect . Our data support a significant role for shifts in the ratio of 20S/26S proteasome complexes . Normally , bortezomib treatment suppresses proteasome catalytic activity . We find that bortezomib treatment also leads to a sharp decrease in the levels of proteasome complexes . Remarkably , the protection against bortezomib that is conferred by reducing 19S subunits correlates with a preservation of both the activity of the proteasome and of the levels of the proteasome complexes themselves . Other mechanisms that may contribute include the induction of autophagy , the induction of alternative proteasome regulatory components , or the disruption of normal gating and drug–target interaction affinities . These possibilities require further investigation . However , our general findings on the surprising ability of 19S subunit reductions to increase resistance to proteasome inhibitors are complementary to those of D Acosta-Alvear , P Walter , JS Weissman , and M Kampmann ( Personal Communication ) who arrived at this conclusion through very different approaches . While changes in the ratio of 20S/26S proteasomes have not previously been shown to protect cells from inhibition of flux through the proteasome , it is well documented that they occur . Indeed , a broad range of genetic , metabolic , and environmental factors can elicit such changes . In human stem cells , manipulation of just a single subunit of the 19S regulatory complex modified the 20S/26S proteasome ratio ( Vilchez et al . , 2012 ) . In cancer cells , many chromosomal regions are recurrently lost , and these often harbor genes that encode 19S subunits ( Nijhawan et al . , 2012; Davoli et al . , 2013 ) . Indeed , mining data from a survey of 310 cancer cell lines , we find that those that have increased resistance to proteasome inhibitors have strongly reduced expression of at least one but often of several genes encoding 19S subunits . The metabolic and environmental factors that can elicit reversible shifts in the 20S/26S proteasome ratio are many and varied . For example , nutrient depravation in yeast ( Bajorek et al . , 2003 ) , the activation of glutamate receptor signaling in neurons ( Tai et al . , 2010 ) , mitochondrial dysfunction in yeast and mammalian cells ( Livnat-Levanon et al . , 2014 ) , and various states of increased oxidative stress ( Wang et al . , 2010; Livnat-Levanon et al . , 2014 ) can all increase the levels of 20S proteasome complexes . The ratio of proteasomes can also be regulated by cellular levels of NADH , a co-enzyme that directly binds 19S subunits and influences 26S proteasome stability ( Tsvetkov et al . , 2014 ) . Further , the ratio is shifted by post-translational modifications mediated by NAD+ ADP-ribosyltransferases ( Ullrich et al . , 1999; Cho-Park and Steller , 2013 ) . The modest reductions in 19S subunits in our experiments did not reduce the overall rates of protein degradation and did not activate proteotoxic stress responses suggesting that these cells normally have a buffer , or excess , of 26S proteasomes . Such a buffer has been recently described in neurons ( Asano et al . , 2015 ) , and if true more broadly , might generally allow cells to tolerate reductions in 19S subunit expression without altering basal rates of proteolysis by the proteasome . In our hands , 19S subunit reduction was accompanied by accumulation of active 20S proteasome complexes . These complexes are highly effective in degrading oxidized ( Reinheckel et al . , 1998; Grune et al . , 2003 ) and intrinsically disordered proteins ( Baugh et al . , 2009; Tsvetkov et al . , 2009a; Wiggins et al . , 2011; Ben-Nissan and Sharon , 2014 ) in an ubiquitin-independent manner . Our results suggest that cells with expanded 20S capacity might be even more broadly positioned to cope with toxic products that accumulate following inhibition of the proteasome . The increase in 20S proteasomes may also have other , pleiotropic effects that contribute to the protective state . First , these complexes mediate the endoproteolytic cleavage of translation initiating factors eIF4G1 , eIF4F , and eIF3a ( Baugh and Pilipenko , 2004 ) , which could directly contribute to the inhibition of protein synthesis that occurs following 19S subunit reduction . Second , 20S proteasomes preferentially degrade newly synthesized substrates ( Tsvetkov et al . , 2009b; Adler et al . , 2010 ) . In addition , 20S proteasome complexes degrade numerous intrinsically disordered proteins involved in cell cycle regulation , cell cycle control , and oncogenesis ( Asher et al . , 2006; Jariel-Encontre et al . , 2008; Ben-Nissan and Sharon , 2014 ) . The degradation of such 20S substrates could underlie the robust anti-proliferation response that follows bortezomib treatment of cells with reduced 19S subunits . Such a shift into a quiescent-like state likely triggers adaptive cytoprotection . This is reminiscent of yeast that transition into stationary phase when nutrients are exhausted . In this setting , there is also a reversible reduction in protein translation ( Fuge et al . , 1994 ) and levels of 26S proteasomes sharply decrease in favor of active 20S proteasomes , which are essential for viability during prolonged periods of nutrient depravation ( Bajorek et al . , 2003 ) . Thus , the protective mechanism that is generated upon 20S formation is likely conserved from yeast to human and may be part of the natural transitions used to established stress-resistant quiescent states . In our experiments , this mechanism for increasing resistance was revealed by the use of highly controllable chemical compounds . However , in nature , mechanisms for rebalancing the 20S/26S proteasome ratio most likely emerged to help cells contend with perturbations that cause protein misfolding . In fact , intracellular and environmental insults that generate large protein aggregates are known to impair the proteolytic function of the proteasome ( Deriziotis et al . , 2011; Ayyadevara et al . , 2015 ) . Such mechanisms may have also helped organisms withstand naturally occurring 20S proteasome inhibitors that are elaborated by microorganisms cohabiting their niches ( Schneekloth and Crews , 2011 ) . Such selective pressures could have shaped the evolution of this ancient survival mechanism , one that emerged long before the advent of the use of proteasome inhibitors as anti-cancer therapeutics . In agreement with our results indicating that yeast cells are protected from proteasome inhibitors by reducing 19S subunits , Breslow et al . found that reducing 19S subunits can rescue yeast strains that are growth inhibited by reductions in 20S subunits ( Breslow et al . , 2008 ) . Suppressing the expression of many different 19S subunits provided resistance to proteasome inhibitors and there are many potential routes for suppressing their expression ( e . g . , genetic , metabolic , epigenetic , environmental ) . This raises the intriguing possibility that large populations of cells might harbor functional heterogeneity for surviving altered flux through the proteasome . At one extreme , some cells might be highly proliferative yet highly sensitive to proteasome inhibition , while at the other extreme , some cells could be slowly proliferative yet highly resistant to proteasome inhibition . Because of their slower proliferation capacity , the latter would have generally reduced relative fitness , analogous to the small populations of drug-tolerant ‘persister’ cells that reside within tumor populations ( Sharma et al . , 2010; Glickman and Sawyers , 2012; Knoechel et al . , 2014 ) . Developing a strategy to address this state of resistance could have significant therapeutic value . | Cells have numerous methods for removing proteins that have been damaged or are no longer needed . One of these methods is carried out by a large protein complex called the proteasome . Because of its central role in maintaining protein balance , drugs that stop the proteasome functioning often kill cancer cells grown in dishes . However , these proteasome inhibitors tend not to work against most tumors in patients . Moreover , tumors that do respond to these drugs ultimately become resistant to them . Tsvetkov et al . used a genetic screen to find the mutations that allowed cancer cells to withstand exposure to proteasome inhibitors . The proteasome complex contains two types of subunits: regulatory subunits that recognize the proteins that need to be degraded; and catalytic subunits that degrade the proteins . Surprisingly , individually inactivating the genes for many different regulatory subunits provided protection against proteasome inhibitors . When the regulatory subunits were reduced , the proteasomes shifted into a state that ultimately protected the cells . This mechanism was observed to protect both yeast and human cells and may be a widespread mechanism for establishing stress-resistant states . The next challenge will be to identify the vulnerabilities of cells that have reduced regulatory subunits . Research is also needed to find out if this reduction varies from cell to cell , making some cells more able to withstand treatment . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"cell",
"biology"
] | 2015 | Compromising the 19S proteasome complex protects cells from reduced flux through the proteasome |
Huntington's disease ( HD ) represents an important model for neurodegenerative disorders and proteinopathies . It is mainly caused by cytotoxicity of the mutant huntingtin protein ( Htt ) with an expanded polyQ stretch . While Htt is ubiquitously expressed , HD is characterized by selective neurodegeneration of the striatum . Here we report a striatal-enriched orphan G protein-coupled receptor ( GPCR ) Gpr52 as a stabilizer of Htt in vitro and in vivo . Gpr52 modulates Htt via cAMP-dependent but PKA independent mechanisms . Gpr52 is located within an intron of Rabgap1l , which exhibits epistatic effects on Gpr52-mediated modulation of Htt levels by inhibiting its substrate Rab39B , which co-localizes with Htt and translocates Htt to the endoplasmic reticulum . Finally , reducing Gpr52 suppresses HD phenotypes in both patient iPS-derived neurons and in vivo Drosophila HD models . Thus , our discovery reveals modulation of Htt levels by a striatal-enriched GPCR via its GPCR function , providing insights into the selective neurodegeneration and potential treatment strategies .
Neurodegenerative disorders refer to a number of diseases caused by progressive loss of neurons , and they currently have no cure . Many similarities appear in these diseases , such as selective loss of neurons in certain brain regions and accumulation of aggregation-prone proteins ( Soto , 2003 ) . In order to study these fundamental features and find treatment strategies of these diseases , Huntington's disease ( HD ) is often used as an important model because of its clear genetics ( The Huntington's Disease Collaborative Research Group , 1993 ) , which facilitates establishment of genetic models as well as early diagnosis . The major cause of HD is the cytotoxicity of the mutant Htt protein ( mHtt ) ( Rubinsztein and Carmichael , 2003 ) , which is expressed throughout the brain and peripheral tissues , but elicits selective neurodegeneration of the corpus striatum and lesser damage to the cerebral cortex in HD patients ( Cowan and Raymond , 2006 ) . This selectivity is likely contributed , at least partially , by striatal-enriched modulators of mHtt toxicity and stability ( Subramaniam et al . , 2009; Tsvetkov et al . , 2013 ) . Consistent with this idea , the neuronal longevity correlates with mHtt turnover , which is slower in striatal than in cortical neurons ( Tsvetkov et al . , 2013 ) , suggesting expression of striatal-enriched mHtt stabilizers . Discovery of such stabilizers may help understanding the selective pathology of HD . More importantly , it provides potential therapeutic entry points for HD: while the mechanism of mHtt toxicity is unclear , lowering its level should suppress its downstream toxicity and treat the disease ( Yu et al . , 2014 ) . Meanwhile , reducing the wild-type Htt protein ( wtHtt ) at the same time seems to be well-tolerated ( Boudreau et al . , 2009; Grondin et al . , 2012; Lu and Palacino , 2013 ) . Thus , modulators of Htt levels are attractive targets for potential HD treatment .
To identify modulators of Htt levels in the striatal cells , we screened through a number of candidates in STHdhQ7/Q111 cells , a well-established and easily-transfectable striatal-derived cellular HD model expressing endogenous full length mHtt ( Trettel et al . , 2000 ) . We tested the endogenous mHtt levels following knock-down of 104 candidate modulators using pooled siRNAs . We selected these candidates based on our previous screening results in the stably-transfected Drosophila S2 cells ( Lu et al . , 2013 ) and tested the mHtt level changes by western-blots ( Figure 1—figure supplement 1 ) . This effort revealed six potential modulators of mHtt levels: Gpr52 and Eaf1 siRNAs lower mHtt , whereas Gclc , Grid2 , Ndrg3 and Hdhd3 siRNAs increase its level ( Figure 1—figure supplement 1 ) . Among them , Gpr52 ( a GPCR ) is of special interest . First , GPCRs locate on the plasma membrane and their functions are modulated by extracellular molecules , placing them among the most druggable targets: highly accessible to drugs and the functions are modulated by small molecules . Second , Gpr52 has been recently characterized as a Gαs-coupled receptor highly enriched in the striatum , especially D2 neurons ( Sawzdargo et al . , 1999; Komatsu et al . , 2014 ) , which are amongst the earliest affected in HD ( Raymond et al . , 2011 ) . The coincidence between Gpr52 expression and selective neurodegeneration suggests that Gpr52 may contribute to the selective early loss of striatal neurons in HD . To confirm Gpr52's effect on Htt , we tested an additional set of siRNAs ( Gpr52_si1∼3 ) in the STHdhQ7/Q111 cells , and observed robust reduction of both wild-type and mutant endogenous Htt levels ( Figure 1A ) . Consistently , in the HdhQ140/Q140 knock-in mice ( Menalled et al . , 2003 ) , shRNA mediated knock-down of Gpr52 lowers Htt in primary cultured striatal but not the cortical neurons ( Figure 1B ) . More importantly , by crossing the Gpr52 heterozygous knockout mice with the HdhQ140/Q140 knock-in mice , we have observed robust lowering of endogenous Htt levels in the striata but not in the cortices by heterozygous knockout of Gpr52 ( Figure 1C ) ; confirming Gpr52 mediated modulation of Htt levels in vivo . 10 . 7554/eLife . 05449 . 003Figure 1 . Gpr52 modulates Htt levels . All data plots: average and S . E . M . ; ‘*’: p < 0 . 05 , ‘**’: p < 0 . 01 , ‘***’: p < 0 . 001 by the two-tailed Mann–Whitney U test . The number on top of each bar indicates the biological replicate number . ( A ) Transfection of Gpr52 siRNAs ( Gpr52_si1∼3 ) in the mouse striatal cells ( STHdhQ7/Q111 ) lowers Htt levels , as detected by Htt antibodies MW1 , 2166 , ab1 and 2B7 . MW1 is the polyQ antibody that detects only the mHtt protein , whereas 2166 , 2B7 and ab1 detects both mHtt and wtHtt . Left panels: representative western-blots; Hdh5 is the Htt siRNA used as the positive control for Htt knock-down . Neg is the non-targeting siRNA used as the negative control . Right panel: western-blot quantification from multiple replicates . ( B ) Infection of lentiviruses expressing Gpr52 shRNAs ( Gpr52_sh1∼2 ) lowers Htt in primary striatal but not cortical neurons cultured from HdhQ140/Q140 knock-in mice . Left panels: representative western-blots . Right panel: western-blot quantification for the normalized 3B5H10 signals from multiple replicates . ( C ) Heterozygous knockout of Gpr52 lowers Htt in vivo in the striata but not cortices of HdhQ140/Q7 knock-in mice in vivo . The mice were obtained by crossing the heterozygous Gpr52 knockout mice with the HdhQ140/Q140 knock-in mice . Littermates between 40 to 69 days of age were analyzed . Left panels: representative western-blots . Right panel: western-blot quantification of the normalized MW1 signals from multiple mouse samples . Each dot represents the signal from a single mouse . ( D ) Left panels: Immunostaining of HD patient iPS-derived striatal-like neurons . Differentiated neurons from HD patient's iPS cells express molecular markers for striatal medium spiny neurons: Tuj1 , GABA and DARPP32 . Scale bar: 50 μM . Right panels: Transfection of human Gpr52 siRNAs ( hGpr52_si1∼2 ) in the HD patient iPS-derived neurons lowers Htt levels detected by both western-blots and HTRF . HTT3 is the Htt siRNA used as the positive control for Htt knock-down . Bar plot represents the normalized mHtt levels detected by HTRF using the 2B7/MW1 antibody pair . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 00310 . 7554/eLife . 05449 . 004Figure 1—figure supplement 1 . Screening for modulators of Htt levels in the striatal cells ( STHdh ) . ( A ) Representative western-blots in the STHdhQ7/Q111 cells of candidate modifiers of mHtt levels . 104 candidate modifiers were selected based on our previous screening results ( doi: 10 . 1038/nn . 3367 ) . The genes that had averaged Z score values larger than 1 . 2 and were not previously identified as validated hits in the patient fibroblasts were selected for testing in STHdh cells by pooled siRNAs ( Dharmacon , custom library ) . The MW1 detected full length mHtt is quantified and normalized to the tubulin signal using the ImageJ software . The siRNAs that change mHtt levels by more than 30% in the same direction in three biological repeats were selected as hits . The number on top of each bar indicates the biological repeat number of siRNA knock-down . ( B ) The bar plot of mHtt level changes upon knock-down of the identified modifiers . The numbers on top of each bar represent the number of biological replicates tested . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 004 To test Gpr52's effect in human neurons , we generated HD patient iPS-derived neurons mimicking the striatal medium spiny neurons ( Figure 1D ) . These cells exhibit neuronal morphology with a high percentage expressing the neuronal marker Tuj1 , as well as the medium spiny neuronal marker Darpp32 ( Ouimet et al . , 1984 ) and the neurotransmitter GABA ( Figure 1D ) . Because HD patient striatal neurons are unavailable for culture , these neurons represent their closest culture model . Consistently , detected by both western-blots and Homogenous Time-Resolved Förster Resonance Energy Transfer ( HTRF ) assays ( Weiss et al . , 2009 ) , knocking-down Gpr52 significantly reduces Htt in these cells ( Figure 1D ) . The effect is relatively specific , because levels of loading control proteins ( Figures 1 , 2 ) and another polyQ protein ataxin3 ( Figure 2A ) remain unchanged . The Htt lowering is not caused by antibody binding artifacts or Htt cleavage , as multiple antibodies detect comparable reduction ( Figures 1 , 2 ) and there is no obvious increase of possible Htt fragments of lower molecular weights that may account for the lowering of the full length protein ( Figure 2B ) . We further tested different biochemical fractions of the lysates including P1 , P2 and S2 , representing the crude nuclear fraction , the membrane/organelle fraction , and the cytosolic soluble fractions , respectively ( Kegel et al . , 2005 ) . The major Htt lowering occurs in both the P2 and S2 fractions , but not the P1 fraction ( Figure 2C ) . In addition , the Htt mRNA level is not affected by Gpr52 ( Figure 2D ) , and the Gpr52's effect is completely blocked by treatment with proteasome inhibitors ( Figure 2E ) , suggesting that the modulation is mainly mediated by proteasomal degradation . 10 . 7554/eLife . 05449 . 005Figure 2 . The Gpr52 mediated change of Htt levels is via protein degradation . ( A ) Representative western-blots of STHdhQ7/Q111 cell lysates showing no reduction of Ataxin3 levels by Gpr52 knock-down . Bar graph: quantification of Atxn3 levels , n = 4 . ( B ) Full membrane images of Htt blots showing no increase/appearance of lower molecular weight bands in the STHdh cell lysates , suggesting that the Htt reduction by Gpr52 knock-down is not due to protein cleavage modulations . ( C ) Representative western-blots for different biochemical fractions of the protein extract from STHdhQ7/Q111 cells transfected with non-targeting control siRNA ( Neg ) or the Gpr52 siRNA ( Gpr52_si2 ) . Gpr52 knock-down reduced Htt levels in both the P2 and S2 fractions , but not the P1 fraction . ( D ) In the Gpr52 siRNAs ( triangles ) transfected STHdhQ7/Q111 cells , Htt mRNA levels ( Y-axis ) and Gpr52 mRNA ( X-axis ) levels were measured by qPCR . Both the Gpr52 siRNAs and the Htt siRNAs show substantial knock-down of their targets , whereas the Gpr52 knock-down by Gpr52 siRNAs do not reduce Htt mRNA levels . No reverse-transcriptase control samples have been assayed to eliminate potential contaminations from genomic DNA . ( E ) Left: representative western-blots of STHdhQ7/Q111 cells transfected with non-targeting control siRNA ( Neg ) or the Gpr52 siRNA ( Gpr52_si2 ) with or without proteasome or autophagy inhibitors . Right: the bar plot of western-blot quantification of the Htt level change by Gpr52 siRNA transfection with each compound treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 005 Gpr52 is enriched in the striatum ( Sawzdargo et al . , 1999; Komatsu et al . , 2014 ) , and thus may contribute to the selective stabilization of Htt there ( Tsvetkov et al . , 2013 ) . Consistent with this , over-expression of Gpr52 cDNA ( Figure 3—figure supplement 1 ) or treatment with an agonist reserpine ( Komatsu et al . , 2014 ) ( Figure 3A ) leads to a dose-dependent increase of Htt ( Figure 3B–C ) , confirming Gpr52 as an Htt stabilizer . 10 . 7554/eLife . 05449 . 006Figure 3 . Gpr52 modulates Htt levels via cAMP dependent but PKA independent pathways . All experiments are performed in the mouse striatal cell line STHdhQ7/Q111 , and all data are plotted as average and S . E . M . ‘*’: p < 0 . 05 , ‘**’: p < 0 . 01 , ‘***’: p < 0 . 001 . The number on top of each bar indicates the biological replicate number . ( A ) Changes of cAMP levels measured by the cAMP-Glo assay ( Promega ) . Gpr52 siRNAs were tranfected for 3 days , whereas the compound treatment ( forskolin: 1 µM; reserpine: 10 µM ) lasts for 24 hr; statistical analyses performed by the two-tailed Mann–Whitney U test . ( B ) Htt level measured by the 2B7/2166 HTRF ( Liang et al . , 2014 ) upon treatment of different doses of the Gpr52 agonist reserpine for 48 hr , with transfection of Gpr52 siRNA ( Gpr52_si1 ) vs the non-targeting control ( Neg_si ) , n = 4 . ( C ) Htt level measured by the 2B7/2166 HTRF when transfected with hGPR52 cDNA titrated with the empty control vector at different percentages ( X-axis ) , n = 6; statistical analysis performed by the one-way ANOVA and post-hoc Dunnett's test . ( D ) Left and middle: Representative western-blots of STHdhQ7/Q111 cells transfected with the Gpr52 siRNA ( Gpr52_si2 ) and then treated with the indicated PKA modulators or cAMP analogs ( forskolin and cAMP analogs: 1 µM; H89: 50 µM ) . Calnexin has been used as a loading control . The Bars plot: Htt level changes ( % ) measured by 2B7/2166 HTRF ( Liang et al . , 2014 ) of the total lysates with the same treatments as in the western-blot samples . ( E ) Confocal microscopy experiments showing that HTT proteins are enriched in the perinuclear and co-localize with the endoplasmic reticulum ( ER ) marker calreticulin upon treatment of Rp-cAMP or 8-pCpt-2′-O-Me-cAMP ( 8-pCpt-cAMP for short ) . Upper panels: representative images showed the immunofluorescent signals of Htt ( green ) , ER marker ( red , only in the third and fourth columns ) and DAPI ( blue ) in STHdh Q7/Q111 cells treated by vehicle , 1 μM Rp-cAMP or 1 μM 8-cpt-cAMP . Scale bars , 20 μM . The two panels on the right side are magnified images from the left for visualizing the co-localization . Yellow pixels indicate co-localization . Lower left plot: the percentage of cells showing clear perinuclear pattern in each samples . The pattern was judged blindly . Lower middle and right plots: co-localization parameters including Pearson's coefficient and overlap coefficient ( mean and S . E . M . ) . Numbers in indicate the number of cells analyzed for each treatment from five or more biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 00610 . 7554/eLife . 05449 . 007Figure 3—figure supplement 1 . Over-expression of human Gpr52 cDNA in STHdh cells . Representative western-blots showing increased Htt levels and over-expression of human Gpr52 ( hGpr52 ) in STHdhQ7/Q111 transfected with the hGpr52 cDNA . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 00710 . 7554/eLife . 05449 . 008Figure 3—figure supplement 2 . Modulation of PKA activity by PKA inhibitor or cAMP analogs . PKA kinase activities are measured by the cAMP-Glo kit ( Promega ) with a modified protocol in the in vitro condition . Basically , PKA activity is determined by the phosphorylation of the kinase substrate , which is detected by the reduction of the luciferase signal after 1 hr . The average signal without any cAMP or analogs is used as the baseline ( 0% PKA activity , bar 1 ) . The average signal with 200 nM purified cAMP ( a component in the kit ) is set as the 100% PKA activity . The PKA inhibitor H89 was applied together with 200 nM cAMP and blocked ∼90% of the PKA activity . Different cAMP analogs ( 200 nM ) are tested for their ability to activate PKA . n = 6 for each sample . Rp-cAMP and 8-pCPT-2′-O-Me-cAMP do not activate PKA , whereas Sp-cAMP , 8-Br-cAMP and 6Bn-cAMP activate PKA as efficiently as cAMP . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 00810 . 7554/eLife . 05449 . 009Figure 3—figure supplement 3 . Gpr52's effect is not mediated by Rap1 or Rap2 . Left: Representative western-blots of STHdhQ7/Q111 cells transfected with the Gpr52 siRNA ( Gpr52_si2 ) versus control siRNA , and then with constitutively active or dominant negative Rap1 ( Rap1CA or Rap1DN ) or Rap2 ( Rap2CA or Rap2DN ) . The Bars plot: Htt level reduction ( % ) measured by 2B7/2166 HTRF of the total lysates of cells with same transfections as in the western-blots . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 009 As a Gαs-coupled receptor , Gpr52 increases intracellular cyclic adenosine monophosphate ( cAMP ) levels when activated ( Komatsu et al . , 2014 ) . Consistent with this , the intracellular cAMP level is reduced by knocking-down Gpr52 ( by 15 . 6 ± 2 . 5 nM , 35 . 8 ± 2 . 5 nM and 21 . 1 ± 3 . 1 nM , respectively; Figure 3A ) , suggesting some levels of constitutive activity . Uncovering the possible involvement of this cAMP reduction in lowering Htt is critical , because it determines whether we can target Gpr52's GPCR function to lower Htt . Thus , we tested the effect of the adenylyl cyclase activator forskolin , which increases intracellular cAMP levels ( by 54 . 8 ± 5 . 7 nM; Figure 3A ) . Treatment of forskolin completely inhibits the Htt reduction caused by the Gpr52 knock-down ( Figure 3D , left panel , lane 5 and 6 compared to lane 1 and 2; Bar plot ) , suggesting that the reduction of cAMP is required for the Gpr52 mediated Htt modulation . The canonical cAMP sensor is protein kinase A ( PKA ) ( Meinkoth et al . , 1993 ) . To test possible PKA-dependence , we treated the cells with Rp-cAMP , a non-hydrolysable cAMP analog that does not activate PKA ( Rothermel and Parker Botelho , 1988 ) ( Figure 3—figure supplement 2 ) . Rp-cAMP completely inhibits the Gpr52's effect on Htt levels ( Figure 3D , middle panel , lane 5 and 6 compared to lane 1 and 2; bar plot ) , indicating that the Htt reduction by knocking-down Gpr52 is mediated via a PKA-independent sensor of cAMP levels . Note that Rp-cAMP was applied at 1 µM , which is much lower than the Ki value for Rp-cAMP's inhibition of PKA ( Rothermel and Parker Botelho , 1988 ) and thus should not influence the PKA activity . Similarly , another cAMP analog 8-pCPT-2′-O-Me-cAMP that does not activate PKA ( Enserink et al . , 2002 ) ( Figure 3—figure supplement 2 ) completely abolishes the Htt reduction ( Figure 3D; Bar plot ) as well . In contrast , the PKA inhibitor H89 ( Takuma and Ichida , 1994 ) ( Figure 3—figure supplement 2 ) and a PKA-specific agonist 6-Bnz-cAMP ( Christensen et al . , 2003 ) fail to inhibit the Htt reduction by Gpr52 knock-down ( Figure 3D ) , confirming the PKA-independence . Finally , the hydrolysable cAMP analog 8-Br-cAMP has little effect on the Gpr52 mediated Htt reduction ( Figure 3D , middle panel , lane 3 and 4 compared to lane 1 and 2; bar plot ) , whereas the non-hydrolysable cAMP analogs 8-pCPT-2′-O-Me-cAMP and Rp-cAMP block the Gpr52's effect entirely ( Figure 3D ) , indicating that a sustained increase of the cAMP level is needed to abolish the Gpr52's effect . Other than PKA , the guanine exchange factor ( GEF ) Epac , Rapgef2 and potentially other GEFs have been reported as cAMP sensors that mediate downstream signals ( de Rooij et al . , 1998; Emery and Eiden , 2012; Emery et al . , 2013 ) . Thus , Gpr52 may function via a PKA-independent and cAMP-dependent signaling mechanism by activating GEFs . The reported small GTPases downstream of these GEFs are Rap1 and Rap2 , and thus we tested their potential involvement by dominant negative and constitutively active Rap proteins ( Fu et al . , 2007 ) . None of these showed obvious effects on the Htt lowering by Gpr52 knock-down ( Figure 3—figure supplement 3 ) , suggesting novel mechanisms potentially involving other GEFs and/or small GTPases , which is suggested in other models as well ( Emery and Eiden , 2012; Kuwayama et al . , 2013 ) . Given that the major function of small GTPases is trafficking , we tested potential Htt translocation events upon treatments of PKA-insensitive analogs Rp-cAMP or 8-pCPT-2′-O-Me cAMP , and found that they lead to Htt enrichment at the perinuclear regions and co-localization with the endoplasmic reticulumn ( ER ) marker ( Figure 3E ) . Translocation of Htt to the ER may prevent its proteasomal degradation due to lack of proteasomes in the ER ( Plemper et al . , 1997 ) . In the genomes of vertebrates including human , monkey , mouse , rat , dog , chicken and zebrafish , Gpr52 is a single exon gene that is located in the intron of another gene Rabgap1l , which encodes a GTPase-activating protein ( GAP ) in the same orientation . Gpr52 is located within the intron of the GAP domain-containing splice variants ( GAP transcripts ) only , but not the non-GAP transcripts ( Figure 4—figure supplement 1A ) , suggesting possible functional links . GAPs function in opposition to GEFs in regulating small GTPase activities , and thus Rabgap1l may function to balance certain GEFs activated by Gpr52 via cAMP . The ‘co-localization’ of Gpr52 and Rabgap1l in the genome might facilitate co-regulations of their expressions in certain cell types , so that they can balance each other's function in regulating Htt levels . Consistent with this hypothesis , knocking-down Rabgap1l increases Htt levels ( Figure 4—figure supplement 1B ) and blunts the Gpr52's effect ( Figure 4A , Htt lowering drops from 60 . 3 ± 7 . 1% to 6 . 3 ± 7 . 7% ) . Interestingly , both Gpr52 and Rabgap1l mRNA levels are lowered when HTT is knocked-down ( Figure 4B , dark axis ) . As a control , Gapdh mRNA levels do not change ( Figure 4B , blue axis ) . This suggests that Gpr52 and Rabgap1l could be co-regulated together , possibly facilitated by their shared genomic locus , via possible feedback regulation by Htt . In summary , Gpr52 modulates Htt via a cAMP-dependent and PKA-independent signaling pathway , and its ‘host’ gene Rabgap1l exhibits epistatic effects on this modulation . 10 . 7554/eLife . 05449 . 010Figure 4 . Rabgap1l interferes with Gpr52-mediated Htt modulation . ( A ) Left: Representative western-blots of STHdhQ7/Q111 cells transfected with the Gpr52 siRNA ( Gpr52_si2 ) vs the control siRNA , along with or without the Rabgap1l siRNA . Rabgap1l knock-down blunts the Gpr52's effect on the Htt level . Bar plot: HTRF quantification of the transfected cells as indicated , showing consistent results with the western-blots; statistical analyses by the two-tailed Mann–Whitney U-test . ( B ) mRNA levels of Gpr52 ( left Y-axis ) or Gapdh ( right Y-axis ) vs the ones of Rabgap1l ( X-axis ) upon transfection with HTT siRNAs or control ( Neg ) . Both Gpr52 and Rabgap1l mRNA levels are lowered upon HTT knock-down , and a correlation between the lowering is observed . ( C ) Left panels: Representative confocal microscopic immunofluorescent images of Htt and Rab39B in STHdhQ7/Q111 cells . Treatments with 1 μM RP-cAMP or 1 μM 8-pCPT-2′-O-Me-cAMP for 48 hr leads to increased co-localization between Htt and Rab39B . Red: Rab39B; Blue: DAPI; Green: Htt ( using antibody 2051 ) ; The lower panels are magnified images from the left for visualizing the co-localization . Yellow pixels indicate co-localization . Scale bars: 20 μm . Right panels: co-localization parameters including Pearson's coefficient and overlap coefficient ( mean and S . E . M . ) . Numbers in indicate the number of cells analyzed for each treatment from five or more biological replicates . ( D ) A model figure explaining the modulation of Htt levels by Gpr52 . Gpr52 increases cAMP when activated , which leads to activation of an unknown GEF ( Guanine Exchange Factor ) that activates the downstream Rab39B protein . Rab39B then co-localize with Htt and translocate it to the ER , where the proteasomal degradation is prohibited due to lack of proteasomes in the ER . The Gpr52 gene locates in an intron of the Rabgap1l gene , which expresses the GAP ( GTPase Activating Protein ) for Rab39B , and thus blocks the modulation . Thus , Gpr52 and Rabgap1l provide balanced regulation of Htt in striatal cells , and the shared genomic loci may facilitate their balance in modulating Htt levels via co-regulated expression in striatal cells . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 01010 . 7554/eLife . 05449 . 011Figure 4—figure supplement 1 . Rabgap1l genomic information and siRNA validation . ( A ) Genomic loci of Rabgap1l and Gpr52 . ( B ) qPCR quantification of the Rabgap1l mRNA level of STHdh cells transfected with Rabgap1l siRNAs or the non-targeting control siRNA ( Neg ) . 50–80% knock-down could be achieved by siRNA transfection in these cells . ( C ) Western-blot ( left ) and HTRF ( right ) experiments showing that Rabgap1l knock-down by siRNA increases the Htt level in the STHdh cells . Hdh5 and B01 are Htt siRNAs used as positive controls . For HTRF , the 2B7/2166 antibody pair was used . Data are plotted as mean and S . E . M , n = 16 for non-targeting siRNA control ( Neg ) samples , and n = 12 for Rabgap1l siRNA transfected samples . ‘***’: P < 0 . 001 by the two-tailed Mann–Whitney U-test . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 01110 . 7554/eLife . 05449 . 012Figure 4—figure supplement 2 . No increase of Htt co-localization with Rab22A . Representative confocal microscopic immunofluorescent images of Htt and Rab22A in STHdhQ7/Q111 cells and the quantification analysis of co-localization . Treatments with 1 μM RP-cAMP or 1 μM 8-pCPT-2′-O-Me-cAMP for 48 hr leads to no change of co-localization between Gpr52 and Rab22A . Red: Rab22A; Blue: DAPI; Green: Htt ( using antibody 2051 ) ; Scale bars: 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 01210 . 7554/eLife . 05449 . 013Figure 4—figure supplement 3 . Constitutively active Rab39B blunts Gpr52's effect . Left: Representative western-blots of Htt ( ab1 ) , consitutively active Rab39B ( anti-FLAG for Rab39B_CA ) , and the loading control ( Tubulin ) . ‘−’ indicates the empty vector or scrambled siRNA controls . Right: quantification of the normalized Htt signals , n = 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 013 The involvement of Rabgap1l gives the clue for identifying the downstream small GTPase ( s ) that mediates the modulation of Htt levels . There are three reported substrates for Rabgap1l: Rab22A , Rab39B and Rab34 ( Itoh et al . , 2006 ) , among which Rab22A and Rab39B are expressed in the brain ( http://www . biogps . org/; http://www . brain-map . org/ ) . We thus tested whether Rab22A or Rab39B is the small GTPase downstream of Gpr52 and cAMP to mediate the Htt modulation . Interestingly , Rab39B but not Rab22A shows significantly increased co-localization with Htt when the cells are treated with cAMP analogs Rp-cAMP or 8-pCPT-2′-O-Me cAMP ( Figure 4C; Figure 4—figure supplement 2 ) , suggesting that Rab39B may participate in the translocation of Htt to the ER . Consistent with this , expressing the constitutively active form of Rab39B blunts the Gpr52's effect ( averaged Htt lowering reduces from 54 . 1 ± 6 . 3% to 17 . 2 ± 10 . 6% ) , confirming Rab39B's involvement in the pathway ( Figure 4—figure supplement 3 ) . Taken together ( Figure 4D ) , the presence of Gpr52 leads to sustained elevation of intracellular cAMP levels , which activates potential GEF ( s ) and subsequently its substrate Rab39B . Rab39B co-localizes with Htt and changes the localization of Htt , protecting it from degradation . Rab39B is inactivated by its GAP Rabgap1l , which shows epistatic effects on Gpr52 mediated Htt modulation . Gpr52 and Rabgap1l locate in the same genomic loci , which may facilitate their balanced regulation of Htt levels in the striatal cells . To test whether targeting Gpr52 may benefit HD neurons , we examined several HD-related phenotypes . STHdhQ111/Q111 cells exhibit mHtt-dependent caspase 3 and/or 7 activation by stress , such as serum removal ( Miller et al . , 2012; Lu et al . , 2013 ) . This provides readout for mHtt-dependent toxicity in striatal cells . Knocking-down Gpr52 remarkably reduces serum starvation-induced caspase 3 and/or 7 activity in STHdhQ111/Q111 cells ( by 67 . 8 ± 1 . 1%; Figure 5A ) , indicating a suppression of the mHTT-induced toxicity . Treatment of forskolin substantially blunts this effect ( to 17 . 3 ± 4 . 8%; Figure 5A ) , confirming that the Gpr52's effect is mainly mediated via the cAMP-dependent Htt reduction . Consistently , in the HD patient iPS-derived striatal neurons , knocking-down Gpr52 suppresses the mHtt-dependent neuronal loss and caspase 3 activation induced by withdrawn of the brain-derived neurotrophic factor ( BDNF ) ( HD iPS Consortium , 2012; Lu and Palacino , 2013 ) ( Figure 5B–C ) . In addition , knocking-down Rabgap1l exacerbate the neuronal loss and attenuates Gpr52's rescue ( Figure 5C ) , confirming the involvement of Rabgap1l . 10 . 7554/eLife . 05449 . 014Figure 5 . Lowering Gpr52 rescues human HD neurons and the in vivo fly HD models . ( A ) Caspase-glo of STHdhQ7/Q111 ( HD ) or STHdhQ7/Q7 cells ( WT ) with indicated transfections ( Neg: the non-targeting controls siRNA; Gpr: the Gpr52 siRNA smartpool; Htt: the Htt siRNA Hdh5 ) and compound treatments; statistical analyses by the two-tailed Mann–Whitney U-test . ( B ) Caspase 3 activity of patient iPS-derived neurons ( Q47 ) measured by the fluorescent indicator dye before and after BDNF removal ( scale bar: 200 µm ) . Bar plots: quantification of caspase 3 signals corrected by the total cell number ( by DAPI ) and normalized to the HD controls ( second bar ) . Statistical analyses were performed by the two-tailed Mann–Whitney U-test . ( C ) Immunostaining of Tuj1 and DAPI showing loss of neurons of HD patient iPS-derived neurons ( Q47 ) cultured under the BDNF-deprived condition with indicated transfections , and the rescue by knocking-down Htt or Gpr52 or Rabgap1l or Rabgap1l + Gpr52 ( scale bar: 200 µm ) . Bar plots: quantification of the area in each field covered by the Tuj1 signal ( Tuj1 area ) and the nuclei counts . All data normalized to the non-targeting siRNA transfected control samples . Statistical analyses were performed by the two-tailed Mann–Whitney U-test . Sholl plots: the sholl analysis ( Sholl , 1953 ) results plotted for each sample , ‘n’ indicated the number of analyzed neurons . Statistical analysis performed by two-way ANOVA tests . ( D ) Age-dependent motor performance in normal flies expressing elav-GAL4 driver alone ( Negative control , blue dotted lines ) , HD flies expressing elav-GAL4 driven NT-Htt128Q or FL-Htt200Q alone ( black dotted lines ) , or HD flies crossed to loss of function mutation ( LOF ) or knock-down ( shRNA ) lines of Drosophila homologs of Gpr52 or Rabgap1l ( red lines ) . Lowering Gpr52 rescues the motor behavior deficits , whereas lowering Rabgap1l enhances the phenotype . n = 15 , statistical analysis performed by one way ANOVA and Dunnett's post-tests . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 014 To validate the above observations in vivo , we tested the effect of modulating Gpr52 levels in a Drosophila model of HD expressing the N-terminal human Htt fragment with 128Q ( Figure 5D , NT-Htt128Q ) ( Al-Ramahi et al . , 2006 ) . Decreased levels of Drosophila Gpr52 either by shRNA knockdown or by loss-of-function mutation ( LOF ) causes a robust suppression of the motor deficits induced by expression of this mHtt fragment in the Drosophila central nervous system ( Figure 5D ) . Meanwhile , consistent with cellular models where Rabgap1l functions in opposition to the effect of Gpr52 , knock-down of the Drosophila homolog of Rabgap1l exacerbates the HD motor deficits in the in vivo fly model ( Figure 5D ) . The rescue effect by knocking-down Gpr52 is further validated in a Drosophila model of HD expressing the full length human Htt protein with 200Q ( FL-Htt200Q ) . Expression of Gpr52 shRNA substantially improves the motor defects in those flies as well ( Video 1 ) . To test if Gpr52 knock-down rescue the HD cell survival in vivo , we tested the retina degeneration phenotype of the fly HD model and failed to observe a significant rescue ( not shown ) , probably due to a lack of expression of Gpr52 in the retina and the supporting cells . 10 . 7554/eLife . 05449 . 015Video 1 . The representative video showing that lowering Gpr52 rescues the in vivo fly full-length HD models–linked to Figure 5D , the lower panel . DOI: http://dx . doi . org/10 . 7554/eLife . 05449 . 015
We have revealed Gpr52 as a striatal-enriched modulator of Htt levels . This provides a possible mechanism of selective neurodegeneration in the striatum , although other mechanisms may contribute as well; including striatal-enriched post-translational modifiers of Htt such as Rhes ( Subramaniam et al . , 2009 ) , or non–cell-autonomous mechanisms ( Wang et al . , 2014 ) . Gpr52 is enriched in the D2 neurons ( Komatsu et al . , 2014 ) , which are amongst the earliest affected in HD ( Raymond et al . , 2011 ) , possibly reflecting their susceptibility to mHtt due to expression of Gpr52 . Further evidence needs to be obtained to reveal whether Gpr52 is one of the major causes of the regional specificity . For example , Gpr52 could be exogenously expressed in other brain regions in HD in vivo models to see if the regional specificity of neurodegeneration is shifted . Targeting Gpr52 is effective in lowering Htt and remarkably rescues HD phenotypes in vitro and in vivo ( Figure 5 ) , confirming its therapeutic potential . Further efforts need to be made to confirm its activity in mammalian and/or large animal models and to screen for its antagonists , which are currently unavailable . While the possible link between cAMP and Htt has been suggested previously ( Williams et al . , 2008; Lin et al . , 2013 ) , those studies focus on exogenously expressed or transgenic Htt aggregates or fragments , and show inconsistent results . Our data confirm that the sustained lowering of cAMP is required for GPCR-mediated lowering of endogenous soluble full-length Htt through PKA-independent mechanisms . It is intriguing why other striatal GPCRs , such as dopamine receptors , were not identified as Htt modulators by our screenings . There are several major possibilities besides a random missing of these targets . First , in the knock-down screen that we performed , the GPCRs need to have a certain level of constitutive basal activity in order to be identified as hits . Second , a sustained change of cAMP is required to modulate Htt levels ( Figure 3D ) , and this may not be achieved by knocking-down many of the GPCRs in absence of agonists . Gpr52 may have a relatively high endogenous activity in the striatal cells , enabling it to provide a significant contribution to the endogenous cAMP levels . Alternatively , the endogenous Gpr52 ligand might be present at a certain level under endogenous conditions , and ligand is predicted to be fatty acids ( Kakarala and Jamil , 2014 ) . Finally , cAMP may only be a required part of the mechanism , while other signaling pathway or the receptor itself may contribute as well . Gpr52 is located in the intron region of Rabgap1l gene in the mammalian genome , and the two genes are functionally linked in regulating Htt levels . This functional link is validated not only by the epistatic effects of Rabgap1l on Gpr52 mediated modulation ( Figure 4A ) , but also confirmed by the involvement of Rab39B ( Figure 4D ) , the substrate of Rabgap1l ( Itoh et al . , 2006 ) . To our knowledge , this is so far the only pair of intronic and host genes showing functional links . The shared genomic locus of Gpr52 and Rabgapl1 may facilitate their balance in modulating Htt levels via co-regulated expression in certain cell types . For example , they may share the same enhancer regions or transcription factors in striatal cells . Alternatively , Gpr52 mRNA might be cleaved and processed from the intron of Rabgap1l pre-RNA , possibly in a similar way as the biogenesis of intronic lncRNAs that have been recently identified ( Yin et al . , 2012; Zhang et al . , 2013 ) . These possibilities could be potentially tested by reconstitution of the relevant genome regions with reporters in a different system , and are beyond the scope of this study . In summary , we have identified Gpr52 as a striatal-enriched GPCR that stabilizes Htt in vitro and in vivo . It may contribute to the selective neurodegeneration and serve as a potential target for HD drug discovery . Our finding also implies that the selective vulnerability of striatal neurons may be contributed by striatal-enriched modulator of Htt levels . Identifying these modulators opens new therapeutic avenues for HD .
The Rap1-V12 ( constitutively active Rap1 ) , Rap1-N17 ( dominant negative Rap1 ) , Rap2-V12 ( constitutively active Rap2 ) and Rap1-N17 ( dominant negative Rap2 ) are kind gifts from Dr Daniel Pak . Their sequences and expression were further verified . All Rap mutants are in the mammalian expression vector pGW1 with the addition of an N-terminal HA ( influenza hemagglutinin ) epitope tag as previously described ( Fu et al . , 2007 ) . The human GPR52 cDNA was synthesized in vitro and cloned into pDEST Gateway Vectors for transfections ( Life Technologies , Grand Island , NY ) . The Rab39B cDNA was amplified from human brain first-strand cDNA library ( Clonetech , Mountain View , CA cat no . S1199 ) and cloned into Hind III/Xho I sites of pcDNA3 . 0 ( Life Technologies ) with a FLAG-tag fused to C-teminus . The constitutively active forms ( Q68L ) were generated by point mutagenesis and verified by sequencing Rab39B . The mouse striatal cells ( STHdh ) were obtained from Coriell Cell Repositories ( Camden , NJ ) . STHdh cells were cultured in DMEM ( Life Technologies , cat . no . 11965 ) with 10% ( vol/vol ) FBS ( Life Technologies , cat . no . 10082–147 ) . For the generation of Huntington's disease iPSC lines , patient fibroblasts obtained from Coriell Cell Repositories ( Q70 ) or from an HD patient ( Q47 ) and his healthy sibling ( WT , Q19 ) from one Mongolian family were transduced with the retroviral STEMCCA polycistronic reprogramming system ( Millipore , Billarica , MA ) . The iPS lines were confirmed positive for Tra-1-81 , Tra-1-60 , SSEA-4 and Nanog by immunofluorescence and flow-cytometry and all four vector-encoded transgenes were found to be silenced . The study was approved by the ethic community of IBS at Fudan University ( No . 28 ) . Verbal and written consent was obtained from all patients . iPSCs were maintained on a feeder layer of irradiated mouse embryonic fibroblasts ( MEFs ) . The neuronal differentiation was performed as previously described ( Ma et al . , 2012 ) . In briefly , iPSCs were differentiated to Pax6-expressing primitive neuroepithelia ( NE ) for 10–12 days in a neural induction medium . Sonic hedgehog ( SHH , 200 ng/ml ) was added at days 10–25 to induce ventral progenitors . For neuronal differentiation , neural progenitor clusters were dissociated and placed onto poly-ornithine/laminin-coated coverslips at day 26 in Neurobasal medium , with a set of trophic factors , including brain derived neurotrophic factor ( BDNF , 20 ng/ml , Peprotech , Rocky Hill , NJ , cat . no . 450–02 ) , glial-derived neurotrophic factor ( GDNF , 10 ng/ml , Peprotech , cat . no . 450–10 ) , insulin-like growth factor 1 ( IGF1 , 10 ng/ml , Peprotech , cat . no . 100–11 ) and Vitamin C ( Sigma , St . Louis , MO cat . no . D-0260 , 200 ng/ml ) . Primary mice cortex and striatum neurons were obtained from P0 pups , and tissues were digested with papain ( Worthington Biochemical Corporation , Lakewood , NJ , cat . no . LS003119 , 12 units/ml ) and DNaseI ( Roche Life Science , Indianapolis , IN , cat . no . 10104159001 , 1 mg/ml ) for 30 min with occasional mixing; digestion was stopped with 10% serum . Tissues were then triturated and plated in the Neurobasal @ medium ( Life Technologies , cat . no . 21103–049 ) supplemented with 2% B27 ( Life Technologies , cat . no . 17504044 ) , 0 . 5× Pen/Strep ( Life Technologies , cat . no . 15070–063 ) and 2 mM Glutamax ( Life Technologies , cat . no . 35050–061 ) for striatal neurons ( validated by Darpp32 staining ) , or 1% N2 ( Life Technologies , cat . no . 17502–048 ) , 2% B27 , 0 . 5× Pen/Strep ( Life Technologies , cat . no . 15140 ) and 2 mM Glutamax ( Life Technologies , cat . no . 35050 ) for cortical neurons . All the cells were maintained at 37°C incubator with 5% CO2 , except STHdh cells , which were maintained at 33°C with 5% CO2 . All the cells were tested for mycoplasma contamination , but they have not been authenticated by STR profiling . The generation and characterization of the Hdh140Q knock-in mice have been previously described ( Menalled et al . , 2003 ) . The Gpr52 knock-out was generated by Cyagen Biosciences Inc . ( Guangzhou , China ) using TALEN technology . A pair of TALEN constructs for Gpr52 knockout were cloned into a mammalian expression vector pCMV-TALEN and capped , polyA-tailed mRNA for injection were produced using the Ambion mMessage mMachine kit . The knockout mice were produced by microinjecting TALEN mRNAs into fertilized eggs from C57BL/6 strain . The knockout allele has been sequence validated to have eight missing base pairs ( GAATGTGT , 57–64 of the ORF ) , causing a frameshift and an early stop ( sequencing primers , forward: 5′-agccaaagctgcaaactccct-3′; reverse: 5′-gaaccaagcaggtaactccaacg-3′ ) . The mRNA transcribed from targeted allele with frameshift undergoes nonsense-mediated decay . The mice were back-crossed to the wild-type background for five generations before crossing to the HD mice ( with the same genetic background ) or performing other experiments . The mouse experiments were carried out following the general guidelines published by the Association for Assessment and Accreditation of Laboratory Animal Care . The Animal Care and Use Committee of the School of Medicine at Fudan University approved the protocol used in animal experiments ( Approval #20140904 ) . For protein extraction from the mouse brain , the brains were collected and the striata and cortices were acutely dissected . The siRNAs were reversely transfected into the STHdh cells with Lipofectamine 2000 ( Life Technologies , cat . no . 11668 ) and into the iPSC-derived neurons with Lipofectamine RNAiMAX ( Life Technologies , cat . no . 13778 ) according to the manufacturer's protocol . The cDNA were transfected with Lipofectamine 3000 ( Life Technologies , cat . no . L3000 ) according to the manufacturer's protocol . Cells were collected 3 days after siRNA transfection or 2 days after cDNA transfection for western-blot , HTRF or immunofluorescence . For experiments with both cDNA and siRNA transfections , the cDNA transfection was performed 1 day after the siRNA transfection with culture medium replacement , and the cells were collect 2 days after the cDNA transfection . siRNA target sequences: mouse Gpr52_si1: GGATCGATATCTTGCAATA , Gpr52_si2: GGATAACTAGCGTGTTTTA , Gpr52_si3: GTGGATCGATATCTTGCAATA; human hGpr52_si1: TGTCGCTTGAGAATTTGCATTATTT , hGpr52_si2: GACAATCCAACTCTGTCCTTCTTAA; mouse Htt siRNA: ACCGTGTGAATCATTGTCTAA ( Hdh5 ) , CTCATTGTGAATCACATTCAA ( B01 ) , CTGGTTGGTATTCTTCTAGAA ( C01 ) ; human Htt siRNA ( HTT3 ) : CAGGTTTATGAACTGACGTTA; mouse Rabgap1l siRNAs: mouse Rabgap1l siRNA1: GCAGUGAAGUGGAGGCUUUTT , Rabgap1l siRNA2: GCUAUGAUGGGAGAGCUUA; human Rabgap1l siRNA: GCUAUGAUGGGAGAGCUUA ( target both human and mouse ) . For compound treatment , the cells were plated at the same density as the siRNA transfection , and the compounds were diluted in OPTI-MEM to 10× concentrations and added 1–2 days later . The cells were then collected 1–2 days later for further analysis . Compound ordering information is as follows: forskolin ( Sigma , cat . no . F6886 ) , Rp-cAMP ( Sigma , cat . no . A165 ) , 8-pCPT-2O′-Me-cAMP ( Sigma , cat . no . C8988 ) , 6-Bnz-cAMP ( Sigma , cat . no . B4560 ) , 8-Br-cAMP ( Sigma , cat . no . B5386 ) , H-89 ( Sigma , cat . no . B1427 ) , Reserpine ( Selleck , Houston , TX , cat . no . S1601 ) , Bafilomycin A ( Sigma , cat . no . B1793 ) , Epoxomicin ( Cayman Chemical , Ann Arbor , MI , cat . no . BU4061T ) , MG132 ( Sigma , cat . no . M7449 ) , chloroquine ( Sigma , cat . no . C6628 ) , Ammonium chloride ( Sigma , cat . no . A9434 ) . The assays utilize the cAMP-Glo Assay kit purchased from Promega ( Fitchburg , WI , cat . no . V1501 ) . For cAMP measurement in cells , chemicals were added 24 hr before measurement . On the day of measurement , change medium containing compound into PBS with phosphodiesterase inhibitors , and incubate 2 hr and then the procedures were performed following manufacture's instruction . Purified cAMP provided by the kit was diluted into different concentrations to plot the standard curve . The slope of the standard curve was utilized to determine the cAMP concentration change per unit change of the signal . For PKA activity measurement , compounds diluted in Opti-MEM at indicated concentrations were directly added into 384 well plates , and then the experiments were performed following manufacture's instruction . For standard western-blots , the cell pellets were collected and lysed on ice for 30 min in PBS + 1% ( vol/vol ) Triton X-100 + 1× Complete Protease Inhibitor ( Roche Diagnostics , Indianapolis , IN , cat . no . 04693116001 ) , sonicated for 10 s , and spun at >20 , 000×g at 4°C for 10 min . The supernatants were then loaded and transferred onto nitrocellulose membranes for western blots . For fractionation experiments that extract Htt proteins at different fractions , the cell lysates are fractionated into P1 , P2 and S2 fractions according to the previously described protocol ( Kegel et al . , 2005 ) . Briefly , cell homogenates made in the homogenate buffer ( 10 mm HEPES , pH 7 . 4 , 250 mm sucrose , 1 mm EDTA plus 1× Complete Protease Inhibitor ) were centrifuged at 2000×g to obtain the crude nuclear pellet ( P1 ) and the postnuclear supernatant ( S1 ) . S1 was then centrifuged at 100 , 000×g to obtain the membrane pellet ( P2 ) and the cytosolic fraction ( S2 ) . The P1 and P2 pellets were then washed with the homogenate buffer and resuspended in 1× PBS buffer with 2% SDS by sonication on ice for 10 s . Equal amount ( 10–20 μg of total proteins ) of each fraction was loaded in each lane for western-blots . For detection of the Gpr52 protein , the cells are lysed in 2% Fos-choline-14 in 1× PBS to extract the membrane proteins . For mouse brain samples , all mice striatum and cortex tissues protein extracts were obtained in the following ice-cold extraction buffer: 50mM Tris . HCl pH 7 . 4 , 250mM NaCl , 5mM EDTA . Na2 , 1% Triton-100 with 1× Complete Protease Inhibitor , and then homogenized by electric homogenizer . The lysates were centrifuged at 20 , 000 g at 4°C for 20 min , and the supernatants were used for western-blots . The HTT antibodies 2B7 ( Weiss et al . , 2009 ) ( 1:1000 ) , ab1 ( Sapp et al . , 2012 ) ( 1:3000 ) and MW1 ( Ko et al . , 2001 ) ( 1:1000 ) have been described previously . The antibody S830 ( Sathasivam et al . , 2001 ) ( 1:10000 ) is a kind gift from Dr Gillian Bates . Commercially purchased antibodies include HTT antibody 2166 ( Millipore , cat . no . MAB2166 , 1:1000 ) , 3B5H10 ( Sigma , cat . no . P1874 , 1:1000 ) , 2050 ( AbD Serotec , Raleigh , NC , cat . no . MCA2050 ) and 2051 ( AbD Serotec , cat . no . MCA2051 ) , anti-â-tubulin ( Abcam , Cambridge , MA , #ab6046 , 1:5000 ) , anti-Calnexin ( Stressgen , San Diego , CA , cat . no . pAb-ADI-SPA-860 , 1:2000 ) , anti-Gpr52 ( GeneTex , Irvine , CA , cat . no . GTX108123 , 1:500 ) , anti-Ataxin3 ( Millipore , cat . no . MAB5360 ) and anti-HA ( Santa Cruz Biotechnology , Santa Cruz , CA , cat . no . sc-805 , 1:500 ) . Note that anti-Gpr52 antibody only detects human Gpr52 but not the mouse Gpr52 . We failed to generate anti-mouse Gpr52 antibody or mouse Gpr52 cDNA plasmids . Similar technical difficulties have been also reported by others ( Komatsu et al . , 2014 ) . All secondary antibodies were used at 1:5000 . For all the representative western-blots shown in the figures , three or more biological repeats have been performed showing consistent results . The HTRF assays were performed similarly to those previously described ( Lu et al . , 2013; Liang et al . , 2014 ) . For all the samples , the protein concentration ( by BCA , Life Technologies , cat . no . 23225 ) and/or the DNA content ( by Picogreen , Life Technologies , cat . no . P7589 , for lyse-in-well experiments ) were measured to correct the loadings . Different protein concentrations or cell numbers per well were tested to ensure that the signals were in the linear range . Background corrections were performed by subtracting the background signals from blank samples . Coverslip cultures were fixed in 4% paraformaldehyde for 15–20 min , washed with 1× PBS 10 min three times and incubated in a blocking buffer ( 10% donkey serum and 0 . 2% triton X-100 in PBS ) for 60 min and then incubated with primary antibodies: Tuj1 ( Covance , Princeton , NJ , USA , cat . no . 14971502 , 1:5000 ) , GABA ( Sigma , cat . no . A0310 , 1:200 ) , DARPP32 ( Millipore , cat . no . AB1656 , 1:1000 ) , Htt antibodies 2B7 ( Weiss et al . , 2009 ) ( 1:200 ) , 2050 or 2051 ( Bio-rad , MCA2050 or MCA2051 , 1:200 ) , Rab22A antibody ( abcam , cat . no . ab137093 ) , or Rab39B antibody ( abcam , cat . no . ab154826 ) overnight at 4°C . Fluorescence conjugated secondary antibodies were used to reveal the binding of primary antibodies ( Life Technology , cat . no . A21206 , A31572 , A21202 , A21203 , 1:1000 ) and nuclei were stained with DAPI ( Sigma , cat . no . D9542 , 1:1000 ) . Images were captured by Leica TCS SP8 confocol system . Co-localization was quantified using the Pearson's correlation coefficient and the overlap coefficient , which were calculated with Image-pro Plus software . Data represent the mean with standard deviation . One-way ANOVA experiments were performed to judge the statistical significance . mRNA levels were determined by qPCR . RNA from siRNA-transfected or compound treated cells was extracted using RNAprep Pure Cell/Bacteria Kit ( Tiangen , Beijing , China , cat . no . DP430 ) . Random-primed cDNA was obtained by reverse transcription using the FastQuant RT Kit ( Tiangen , cat . no . KR106 ) . DNase I was added to break down the genomic DNA . qPCR was then performed using SYBR Green Realtime PCR Master Mix ( Toyobo , Osaka , Japan , cat . no . QPK-201 ) . qPCR Primers used were as follows: Gpr52 forward: TTGCTTTATTGTTTGTTTACTTTATGC , Gpr52 reverse: GTGAAAGTAAGTGAAGCAGACAACC; Rabgap1l forward: GGAACTGGCACAGACCAAAC , Rabgap1l reverse: GCTCCTCTCTGATGCTCAAGTT; Hprt forward: GTCAACGGGGGACATAAAAG , Hprt reverse: CAACAATCAAGACATTCTTTCCA; Htt forward: CTGCACGGCATCCTCTATGT , Htt reverse: TGTTCACGCAGTGGGCTATT . All the primers were tested with standard curve , amplification efficiency was between 95%–105% , and the R2 for linear relationship is >0 . 999 . No reverse-transcriptase controls were used to ensure the specificity of the signals . The patient iPS-derived neurons exhibit HD-dependent phenotypes including elevated caspase-3 signals and neuronal loss upon BDNF removal . Similarly , the STHdh cells exhibit HD-dependent caspase 3 and/or 7 activity upon stress , such as serum removal . Briefly , patient and wild type iPS-derived neurons were cultured in NIM ( 1% N2 in DMEM:F12 ) for further 48 hr after transfected with siRNAs for 4 days . These phenotypes could be detected by the caspase activity assay as well as the neuronal loss assay . For the caspase activity assay , the NucView 488 caspase-3 dye ( Biotium , Hayward , CA , cat . no . 30029 ) was used for the caspase activity detection as based on manufacturer's protocols . The images of the caspase-3 dye and DAPI treated live cells were taken by Nikon ECLIPSE TE2000-S microscope . For the neuronal loss experiments , the Tuj1 confluence and shape were analyzed by ImageJ and the sholl analysis-plug in of ImageJ ( Sholl , 1953 ) . The images were analyzed blindly . Experiments were performed using 15 age-matched virgin females . We placed the flies in an empty vial and tapped them down . The percentage of flies that climbed past a 9-cm-high line after 18 s was recorded , and two replicates were tested in parallel for each genotype . The mean and S . E . M . of 10 observations is plotted for each day and data analyzed by ANOVA followed by Dunnett's post hoc test . Blinding was used both for carrying out the experiment and for analyzing the data . The nervous system driver line elav-GAL ( c155 ) as well as the classical loss-of-function alleles DopEcRMI02790 and GapcenAf01044 were obtained from the Bloomington Drosophila Stock Center at University of Indiana ( http://flystocks . bio . indiana . edu/ ) . The inducible shRNA lines DopEcRKK103494 and GAPcenAKK103588 were obtained from the Vienna Drosophila Resource Center ( http://stockcenter . vdrc . at/control/main/ ) . NT-Htt128Q flies express an N-terminal human Htt fragments comprising exons 1–4 ( first 336 amino acids including 128Q ) and have been previously described ( Al-Ramahi et al . , 2006 ) . FL-Htt200Q flies express the full length human Htt protein with 200Q and exhibit similar motor defects ( Video 1 ) . Statistical comparisons were conducted by the two-tailed unpaired Mann–Whitney U-test for comparing the average of each sample in the bar graphs . For comparison of sample averages with different doses of Gpr52 cDNA transfected vs the single control ( mock ) ( Figure 2C ) , one way ANOVA tests were performed followed by Dunnett's post hoc test . For comparisons between different groups over different time points , two-way ANOVA has been utilized ( Figure 3C–D ) . Significance was established at p < 0 . 05 . In all graphs , error bars , S . E . M . The biological replicate numbers are indicated on top of each bar and/or as the n numbers in the legends . The statistical powers for all analyses were calculated and confirmed to be >80% . For all the cellular assays , the cells were evenly suspended and then randomly allocated in each well tested . For the animal experiments , the littermates were allocated into each group based on their genotypes and thus randomization does not apply . The key experiments including detecting Htt level changes upon Gpr52 knock-down or knockout , the co-localization analysis , and the cellular phenotypic rescue analyses were confirmed by blind testing in which different investigators prepared/labeled the samples and the analyses were carried out without information exchange before obtaining the results . The key observations including the lowering of Htt , the epistasis between Gpr52 and Rabgap1l and the phenotypic rescue have been replicated by multiple independent investigators . Data were excluded when there were clear indications of artifact or experimental failures , such as contamination , transfection/infection failure , etc . Lentiviral particles for infection were produced in HEK293T cells transfected with the pLKO-Gpr52-shRNA or scrambled constructs with the three plasmid packaging mix ( pLP1 , pLP2 and pLPVSV-G ) . After 48 hr , the medium was collected , filtered through a 0 . 45-µm membrane and ultracentrifuged at 20 , 000 rpm for 90 min . The viral pellets were resuspended in sterile PBS with 1% BSA and stored at −80°C . Viral titers were determined using Abm's qPCR Lentivirus Titration Kit ( Abm , Richmond , Canada , cat . no . LV900 ) . The Gpr52 shRNA target sequences are: Gpr52_sh1: CTCCGCTGTTACACCATTATA; Gpr52_sh2: GTGGATCGATATCTTGCAATA . The viruses were applied at M . O . I = 3 for the primary neurons , and the knock-down of Gpr52 was confirmed by qPCR . | Huntington's disease is an inherited disorder of the central nervous system . Symptoms typically begin between the ages of 30 and 50 , and initially include clumsiness and uncontrollable movements , as well as personality changes and mood swings . Symptoms worsen over time and life expectancy is usually around 10 to 25 years following diagnosis . The disease is caused by a mutation in the ‘huntingtin’ gene , which leads to the production of an abnormal form of ‘huntingtin’ protein . This accumulates inside neurons in a region of the brain called the striatum , which is involved in the control of movement , and destroys them . However , it is not clear why other regions of the brain that also produce the mutant huntingtin protein are not affected . Yao , Cui , Al-Ramahi , Sun et al . have now identified a protein that could explain this phenomenon and open up new therapeutic possibilities for Huntington's disease . The protein , which is called Gpr52 , is a receptor located within the outer membrane of neurons , particularly those in the striatum . Reducing the levels of this protein reduced the amount of mutant huntingtin protein that was able to accumulate inside cells grown in culture . Moreover , mice that were genetically engineered to possess a mutant huntingtin gene , but only a single copy of the gene for Gpr52 , accumulated less mutant huntingtin in the striatum than mice with two copies of the Gpr52 gene . Further experiments revealed that Gpr52 protects mutant huntingtin from being broken down inside cells: it does this by activating a signaling pathway involving the cellular messenger cAMP . Encouragingly , when genetic techniques were used to reduce Gpr52 synthesis in a fruit fly model of Huntington's disease , the treated flies showed fewer movement impairments than flies that had not been treated . In addition , reduced levels of Gpr52 were observed to lead to dramatic protective effects in neurons derived from the stem cells of a patient with Huntington's disease . The fact that Gpr52 is located on the surface of neurons means that it might be possible to design drugs that can block its activity and thus reduce accumulation of mutant huntingtin . Such a treatment would be the first to target the causal mechanism behind Huntington's disease , rather than simply addressing the symptoms . The strategy could also be relevant to Alzheimer's disease , Parkinson's disease and other neurodegenerative disorders in which death of neurons is triggered by abnormal accumulation or aggregation of proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | A striatal-enriched intronic GPCR modulates huntingtin levels and toxicity |
FANCI:FANCD2 monoubiquitination is a critical event for replication fork stabilization by the Fanconi anemia ( FA ) DNA repair pathway . It has been proposed that at stalled replication forks , monoubiquitinated-FANCD2 serves to recruit DNA repair proteins that contain ubiquitin-binding motifs . Here , we have reconstituted the FA pathway in vitro to study functional consequences of FANCI:FANCD2 monoubiquitination . We report that monoubiquitination does not promote any specific exogenous protein:protein interactions , but instead stabilizes FANCI:FANCD2 heterodimers on dsDNA . This clamping requires monoubiquitination of only the FANCD2 subunit . We further show using electron microscopy that purified monoubiquitinated FANCI:FANCD2 forms filament-like arrays on long dsDNA . Our results reveal how monoubiquitinated FANCI:FANCD2 , defective in many cancer types and all cases of FA , is activated upon DNA binding .
Fanconi anemia ( FA ) is a devastating childhood syndrome that results in bone marrow failure , leukemia and head and neck cancers ( Gillio et al . , 1997; Butturini et al . , 1994 ) . FA is caused by inheritance of one of 22 dysfunctional FA genes ( FANCA-FANCW ) ( Tan and Deans , 2017 ) . Absence of any one member of the pathway causes genome instability during DNA replication , which results in mutagenic ( cancer-causing ) DNA damage and hypersensitivity to chemotherapeutic ( normal and cancer-killing ) DNA damage ( Deans and West , 2011 ) . Central to the FA pathway is the conjugation of ubiquitin to FANCI:FANCD2 ( ID2 ) complexes ( Walden and Deans , 2014 ) . ID2 monoubiquitination is critical to prevention of bone marrow failure in FA , but it is currently unknown how ID2-ub differs in its function to ID2 . Several proteins have been proposed to specifically bind FANCIUb or FANCD2Ub but not the un-ubiquitinated proteins ( Smogorzewska et al . , 2010; Lachaud et al . , 2014 ) . For example , FAN1 nuclease was proposed to interact with FANCD2Ub via its ubiquitin-binding domain ( UBZ ) ( Smogorzewska et al . , 2010 ) , whereas recruitment of SLX4 endonuclease to the interstrand crosslink site was shown to be dependent on FANCD2 ubiquitination ( Klein Douwel et al . , 2014 ) . However , support for these interactions is limited to analysis of ubiquitination-deficient ( K > R ) mutants , rather than evidence for direct ubiquitin-mediated protein interactions . The retention of FANCD2 in chromatin foci is dependent on its monoubiquitination by a ‘core complex’ of Fanconi anemia proteins ( Walden and Deans , 2014 ) . FANCI and the FA core complex are required to generate FANCD2-foci that mark the location of double strand breaks , stalled replication forks and R-loops ( Taniguchi et al . , 2002; Schwab et al . , 2015; Wienert et al . , 2019 ) in the nucleus , and protect nascent DNA at these sites from degradation by cellular nucleases ( Schlacher et al . , 2012 ) . The ubiquitinated form of FANCD2 , and also its ubiquitinated partner protein FANCI , become resistant to detergent and high-salt extraction from these foci ( Smogorzewska et al . , 2007; Montes de Oca et al . , 2005 ) , leading to speculation about the existence of a chromatin anchor or altered DNA binding specificity post-monoubiquitination ( Longerich et al . , 2014 ) . A recent electron microscopy study revealed a DNA interacting domain that is required for FANCI:FANCD2 binding to DNA ( Liang et al . , 2016 ) . The crystal structure of the non-ubiquitinated FANCI:FANCD2 shows that the monoubiquitination sites of FANCI:FANCD2 are buried and therefore inaccessible in the dimer interface of the complex ( Joo et al . , 2011 ) , suggesting that DNA binding might be required to expose the ubiquitin binding sites . Based on biochemical analyses , non-ubiquitinated FANCI and FANCD2 preferentially bind to branched DNA molecules which mimic DNA replication and repair intermediates ( Longerich et al . , 2014; Longerich et al . , 2009; Niraj et al . , 2017 ) ; however , how that activates monoubiquitination of FANCI:FANCD2 remains poorly understood . DNA is a cofactor for maximal ubiquitination ( Longerich et al . , 2014; van Twest et al . , 2017 ) , as is phosphorylation by the ATR kinase ( Tan et al . , 2020b; Ishiai et al . , 2008 ) . Here , we have reconstituted the FA pathway using recombinant FA core complex and fluorescently labeled DNA oligomer substrates . We show that once monoubiquitinated , FANCI:FANCD2 forms a tight interaction with double-stranded containing DNA . We report the successful purification of monoubiquitinated FANCI:FANCD2 complex bound to DNA using an Avi-ubiquitin construct , and show that the monoubiquitination does not promote any new protein:protein interactions with other factors in vitro . Instead , we reveal a new role of monoubiquitinated FANCI:FANCD2 in forming higher order structures and demonstrate how monoubiquitinated FANCI:FANCD2 interacts with DNA to initiate DNA repair . Our work uncovers the molecular function of the pathogenetic defect in most cases of FA .
Mono-ubiquitinated FANCI:FANCD2 ( henceforth IUbD2Ub ) is the active form of the complex in repair of DNA damage . Many previous studies have speculated about the existence of DNA repair proteins that specifically associate with IUbD2Ub . A summary of these proteins is presented in Table 1 . Using recombinant ID2 or IUbD2Ub prepared by Avi-ubiquitin purification method ( Tan et al . , 2020a ) , we sought to directly compare the binding of this panel of ID2-associated proteins . Each of the partner proteins was expressed using reticulocyte extracts ( Figure 1a ) , and the majority bound to the ID2 complex as predicted based on previously identified associations ( Figure 1b ) . The strongest binding proteins in terms of fraction of protein recovered were SLX4 , FAAP20 , SMARCAD , FANCJ , PSMD4 , SF3B1 , MCM5 and BRE . Although SMARCAD and SF3B1 appeared ‘sticky’ in control experiments , recovery of these proteins was still enriched by FANCD2:FANCI beads compared to background . Luciferase protein was used as a control 35S-labeled prey-protein . Surprisingly , our main observation was that none of the proteins showed any increased affinity for IUbD2Ub over ID2 ( Figure 1d ) . An alternative explanation for the observed increase in association between ID2 and its associated proteins after DNA damage is that IUbD2Ub has an increased affinity for DNA , which brings the protein into closer proximity to these partners . The majority of ID2 associated proteins are chromatin localized . In order to explore the stability of IUbD2Ub on DNA , we performed in vitro monoubiquitination reactions in the presence of IR-dye700 labeled 60 bp double-stranded DNA ( dsDNA ) . As previously characterized ( van Twest et al . , 2017 ) , we observed DNA-dependent appearance of monoubiquitinated forms of FANCD2 and FANCI when using recombinant FA core complex components ( Figure 2a–b ) . ID2 monoubiquitination readily lead to DNA mobility shifts using EMSA ( electromobility shift assay ) even at low concentrations , but this was not observed for the unmodified ( apo ) -ID2 complex in the absence of the enzymatically active FA core complex , or when monoubiquitination-defective K-to-R mutants of ID2 were used in the reaction ( Figure 2c , lanes 2–4 ) . Previously , we and others showed that various different dsDNA-containing structures could robustly stimulate ID2 monoubiquitination ( Longerich et al . , 2014; van Twest et al . , 2017; Sareen et al . , 2012 ) , but that single-stranded DNA ( ssDNA ) does not . To determine if monoubiquitinated ID2 had increased affinity for other dsDNA containing structures , we repeated the monoubiquitination reactions in the presence of different IR-dye700 labeled DNA structures . Interestingly , non-ubiquitinated ID2 also exhibited high affinity toward 3’ flap DNA structure ( similar to a replication fork stalled on the lagging strand ) , which has been previously observed ( Liang et al . , 2016 ) . The 3’-Flap structure , and each of the other dsDNA containing structures , led to increased ID2 monoubiquitination and increased retention of an EMSA shifted band ( Figure 3 ) . Conversely , ssDNA , which stimulates monoubiquitination more slowly ( van Twest et al . , 2017 ) but to similar levels at the long time point of this assay , did not cause an EMSA shift . Previous studies reported that monoubiquitination of ID2 complex may lead to dissociation of the heterodimer to its individual subunits , as measured by loss of co-immunoprecipitation of FANCI with FANCD2 ( 40 , 41 ) . In contrast , we did not observe any Ub-mediated dissociation of ID2 in vitro . First , western blotting of the EMSA gels confirmed that the gel shifted DNA band contains both FANCI and FANCD2 proteins ( Figure 4a ) . Second , FANCIUb still co-immunoprecipitated with FANCD2Ub at the plateau of the in vitro ubiquitination reaction ( Figure 4b ) . To determine the contribution of each of FANCD2Ub and FANCIUb to the clamping of IUbD2 Ub complex to DNA , we used ubiquitination-deficient ( KR ) mutants in the ubiquitination reaction . FANCIKR:FANCD2WT or FANCIWT:FANCD2KR mutant results in decrease in EMSA shift , and FANCIKR:FANCD2KR did not bind to DNA ( Figure 4c ) . However , this retention on DNA correlated with the extent of FANCD2 monoubiquitination retained by these mutant complexes . Western blotting the EMSA gels confirmed that both FANCD2 and FANCI are found in the EMSA shifted product , although in higher amounts when both proteins are capable of being monoubiquitinated ( Figure 4d ) . We postulated that the altered affinity for DNA induced by monoubiquitination must result from either a conformational change in the ID2 heterodimer after monoubiquitination , or participation of the conjugated ubiquitin directly in protein:DNA or protein:protein binding . To help distinguish these possibilities , we utilized mutants of ubiquitin that have previously been shown to mediate the known protein:ubiquitin or protein:DNA interactions in other ubiquitinated protein interactions ( Figure 5a; Husnjak and Dikic , 2012 ) . Each of these Ub mutants were conjugated to ID2 by the FA core complex with similar efficiency ( Figure 5b ) and their clamping onto DNA was then measured . Mutations in surface patch 1 ( F4A , D58A ) , surface patch 2 ( I44A , V70A ) , a DNA binding residue ( K11R ) or a tail mutant ( L73P ) had no apparent effect on DNA clamping ( Figure 5c ) . This result suggests that no canonical surface or region of ubiquitin is critical for DNA clamping of ID2 , and instead ubiquitin conjugation to ID2 probably induces a conformational rearrangement of the heterodimer . In order to examine the architecture of purified recombinant IUbD2Ub complex in the presence of dsDNA plasmid , we utilized a recombinant Avi-tag ubiquitin construct containing a 3C protease site between the biotinylated Avi-tag and the N-terminus of ubiquitin ( Tan et al . , 2020a; Figure 6a ) . This tagged ubiquitin is incorporated onto FANCI:FANCD2 by the FA core complex , allowing Avidin-Sepharose purification of monoubiquitinated ID2 that is then eluted by 3C protease cleavage . We recovered monoubiquitinated FANCI:FANCD2 complex only when FANCI is monoubiquitinated , suggesting that the N-terminus of D2-attached ubiquitin may be buried within the di-ubiquitinated complex , but the N-terminus of ubiquitin attached to FANCI is accessible for avidin binding ( Figure 6b ) . Using this purified protein , we compared FANCIub:FANCD2ub to unmodified FANCI:FANCD2 using electron microscopy ( EM ) . Surprisingly , we observed that FANCIub:FANCD2ub forms filament-like arrays when bound to dsDNA plasmid ( Figure 6c ) . Such arrays were not observed in the unmodified FANCI:FANCD2 protein preparation in the absence of presence of plasmid DNA , nor in previous investigations of human or Xenopus FANCI:FANCD2 complexes studied by EM ( Figure 6d–e and Swuec et al . , 2017; Liang et al . , 2016; Lopez-Martinez et al . , 2019 ) . When smaller DNA molecules were used as the substrate for ID2 binding , we either observed no filament-like structures ( 60 bp , Figure 7a ) or shorter filament-like structures ( 150 bp , Figure 7b ) compared to structures that were on average 7-8x longer than the characteristic double saxophone structure of ID2 heterodimer in the non-ubiquitinated state ( Figure 7c ) . The observation that array length correlated with the size of DNA available for ID2 binding strongly suggested that the association between heterodimer subunits in the array was DNA-mediated . To test whether the array of IubD2ub is also dependent upon binding to the same DNA molecule , we examined the plasmid-stimulated ubiquitination reaction products after treatment with the non-specific endonuclease , Benzonase . It is apparent from EM images that addition of Benzonase breaks the long arrays formed by IubD2ub complex into very short or heterodimer-sized units ( Figure 7d ) . This finding is consistent with Benzonase cleaving exposed DNA between IubD2ub units , leading to destabilization of the filamentous arrays . Together our results show that , in vitro , ubiquitination of ID2 leads to a ubiquitin- and DNA- stabilized filament-like structure . Due to variability in the length and shape of filament-like IubD2ub structures on longer DNA molecules we have not been able to uncover the shape or subunit rearrangement of the individual units of the arrays because attempts to class average arrays failed due to a lack of order . However , examination of IubD2ub purified together with short 60 bp DNA allowed us to collect sufficient images of individual particles for analysis . These particles were similar in size to non-ubiquitinated ID2 , but it is clear from individual molecule and class average views that the IubD2ub complex forms a distinct architecture from that of ID2 ( Figure 7e–f ) . In particular , the overall shape of individual particles and their class averages reveal a twisting that repositions the solenoid arms of one or both of the subunits bringing them into closer proximity . The conformational change induced appears to reduce the size of ID2 in one direction ( x vs y ) but not the other ( Figure 7g–h ) , similar to that predicted in a previously proposed model that placed DNA in a channel between FANCI and FANCD2 post DNA binding ( Longerich et al . , 2014 ) . These images support the view that monoubiquitination induces a conformational change in the ID2 complex that clamps it upon DNA .
In addition to a conformational change in ID2 induced by monoubiquitination ( that has also been concurrently discovered and reported by the Pavletich , Walden and Passmore labs Alcón et al . , 2020; Wang et al . , 2020; Rennie et al . , 2020 ) we found that monoubiquitinated IUbD2Ub formed large filament-like arrays when it was purified together with plasmid DNA , but not short 60 bp DNA fragments . Fourier Transformation of the EM images did not reveal clear evidence for layer lines , that are expected for fiber diffraction , so we expect that the IUbD2Ub is not a true filament . On average , the length of plasmid-associated structures is 7-8x ( but up to 40x ) that of that associated with 60 bp DNA . Larger or longer arrays may potentially be obscured from view because the purification strategy makes elution exponentially more difficult with increasing numbers of conjugated ubiquitin-molecules . Steps to remove ‘aggregates’ may have also inadvertently removed larger arrays . However , as the number of potential plasmid DNA binding sites for ID2 was in large excess the concentration of ID2 used to stimulate reaction , there appears to be some purpose to creation of these filamentous arrays . The modular nature of IUbD2Ub arrays suggests that IUbD2Ub binding to DNA is flexible and can adopt multiple conformations , akin to RPA binding and protecting ssDNA ( Yates et al . , 2018 ) . There is evidence that IUbD2Ub clamped in nucleoprotein filaments exist in cells . Antibodies against FANCD2 have long been used as a marker of double strand breaks , stalled replication forks and R-loops because the protein forms large , intensely stained foci during S-phase that are increased after treatment with DNA damaging agents ( Taniguchi et al . , 2002; Schwab et al . , 2015; Deans and West , 2009 ) . We suspect that these intense foci are due to coating of DNA around damaged forks , potentially in filamentous arrays similar to those we observed by EM . Support for the large size and extent of DNA binding reflective of filamentous arrays also comes from chromatin immunoprecipitation and sequencing ( ChIP-Seq ) using anti-FANCD2 ( 13 ) . FANCD2 , and two other damage markers MRE11 and γH2AX , showed no specific localization in a bulk population of cells , but strongly localized adjacent to a Cas9-induced site-specific DNA break . Both γH2AX and FANCD2 produced a broad peak centered at the target site kilobases ( kb ) to megabases ( mb ) in length . In contrast , MRE11 is located within a very tight peak within ~100 bp of the break . Chromatin within 1–2 kb of the DSB showed reduced occupancy by γH2AX , consistent with dechromatinization around break sites ( Iacovoni et al . , 2010 ) , but FANCD2 was present right up to the DSB . Accumulation of FANCD2 increases at the DSB early after cleavage , and accumulates more distant from the DSB progressively with time post-cleavage . This is suggestive of a polymerization of the FANCD2 signal away from the break site , as we hypothesize would occur for a protein that forms a growing array of molecules at broken DNA ( Wienert et al . , 2019 ) . The conserved function of FANCD2 as a histone chaperone ( Sato et al . , 2012; Higgs et al . , 2018 ) may even be directly linked to displacement of nucleosomes as filamentous arrays extend into break-adjacent chromatin . In this study , we also observed direct association of two ID2 heterodimers by co-immunopurification only after the protein becomes monoubiquitinated . This approach , if performed in cells , could be used to further delineate the mechanism and cellular factors required for the extension of IUbD2Ub arrays during fork protection . Of particular interest will be determining the role of BRCA1 in clamping and/or array extension . BRCA1:BARD1 was initially thought to be the E3 for FANCD2 monoubiquitination , because it co-immunoprecipitates FANCD2 , and FANCD2 does not form nuclear foci after damage in BRCA1-deficient cells ( Raghunandan et al . , 2015 ) . However , in various assays it was later shown that FANCD2 monoubiquitination does occur in BRCA1-deficient cells , but it is uncoupled from FANCD2 foci formation ( Jacquemont and Taniguchi , 2007; Moriel-Carretero et al . , 2017 ) . Filamentous structures on DNA play a genome protective role in prokaryotes: eg DAN protein forms a rigid collaborative filament that reduces accessibility during anoxia ( Lim et al . , 2013 ) , while the Vibrio cholera protein ParA2 forms protective filamentous structures on DNA during segregation ( Hui et al . , 2010 ) . Structural characterization has demonstrated how these filaments function and , in the case of ParA2 , can be targeted therapeutically ( Misra et al . , 2018 ) . The coating of ssDNA by RPA in eukaryotes , also protect DNA from the activity of nucleases , and directs the specific activity of others ( de Laat et al . , 1998; Chen et al . , 2013; Nguyen et al . , 2017 ) . We propose that a FANCIub:D2ub arrays may have a similar stabilizing role on newly synthesized dsDNA at a stalled replication fork . This property would explain why stalled forks are prone to degradation in FA and BRCA patient cells ( Schlacher et al . , 2012; Tian et al . , 2017 ) . In particular , we hypothesize that filamentous DNA-clamped Iub:D2ub could prevent access to DNA by MRE11 and DNA2 nucleases and prevent aberrant ligation of broken DNA to other parts of the genome by non-homologous end-joining . Second , the tight binding of FANCIub:D2ub to dsDNA , when localized to stalled replication forks , may also prevent the branch migration of replication forks and inhibit their spontaneous or helicase-mediated reversal ( Neelsen and Lopes , 2015 ) . Reversed forks are the substrate for degradation by DNA2 and WRN nuclease activities , providing a hypothetical link between the activities of FANCD2-monoubiquitination and the nuclease activity of DNA2 and WRN ( Thangavel et al . , 2015; Sidorova et al . , 2013 ) . Third , Iub:D2ub arrays may also locally suppress non-homologous end-joining ( NHEJ ) factors , and/or delineate the newly synthesized chromatin from unreplicated regions during the promotion of templated repair processes such as homologous recombination . FANCD2 , FANCI , and components of the FA core complex were identified amongst relatively few other factors , in a genome-wide screen for genes that promote templated repair over NHEJ ( Richardson et al . , 2017 ) . Stabilization of RAD51 filaments , required for HR , is also an in vitro property of ID2 ( Sato et al . , 2016 ) , suggesting Iub:D2ub filamentous arrays may exist adjacent to or coincident with RAD51 filaments in cells , in order to provide a polarity to the homologous recombination reaction without loss or gain of genomic sequences . Fanci and Fancd2 have common and distinct functions in mouse models of Fanconi anemia ( Dubois et al . , 2019 ) , while the double knockout of FANCI and FANCD2 has an unexpectedly distinct phenotype compared to single knockouts in human cells ( Thompson et al . , 2017 ) . But FANCIK523R expressing cells are less sensitive to DNA damage than FANCI knockout in human cells ( Smogorzewska et al . , 2007 ) , so what is the role of FANCI monoubiquitination ? Previous studies demonstrated that FANCI monoubiquitination is always subsequent to FANCD2 monoubiquitination , both in cells ( Sareen et al . , 2012 ) and in biochemical assays ( van Twest et al . , 2017 ) . FANCI also likely plays a role in recruiting the FA core complex to the substrate ( Castella et al . , 2015 ) . In this study , we show that FANCI-monoubiquitination is not necessary for clamping of the ID2 complex onto DNA ( Figure 4 ) . However , in vivo it is likely that FANCI monoubiquitination plays a critical role in regulating deubiquitination of the ID2 complex . FANCI recruits the deubiquitinating enzyme USP1:UAF1 ( Yang et al . , 2011 ) , which prevents trapping of monoubiquitinated FANCD2 at non-productive DNA damage sites , but only ID2Uband not IUbD2Ub is a substrate ( van Twest et al . , 2017 ) . It is also clear from our EM investigations that FANCI must play an important role in the structural integrity of IUbD2Ub filamentous arrays on DNA , possibly creating an asymmetry necessary for a specific polarity to array assembly . Onset of progressive bone marrow failure occurs at a median age of 7 in children with FA ( Butturini et al . , 1994 ) . Almost all these patients lack FANCD2 and FANCI monoubiquitination , due to mutation in either FANCD2 or FANCI or one of the nine other FANC proteins required for their monoubiquitination ( Walden and Deans , 2014 ) . The importance of the monoubiquitin signal is highlighted by the observation that up to 20% of patients acquire somatic reversion of the inherited mutation in a fraction of blood cells ( Soulier et al . , 2005 ) . These mutations restore monoubiquitination and prevent bone marrow failure . Our work suggests two potential strategies for treatment of FA: restoration of gene function , such as that which occurs in somatic revertants or , identification of novel mechanisms to stabilize an ID2:DNA-clamped complex for fork protection by ubiquitin-mediated or innovative means . New small molecule activators or inhibitors of ID2:DNA clamping could be therapeutics in FA or cancer-treatment . In vitro biochemistry has proven to be the most powerful tool in uncovering new functions of FANCD2-monoubiquination that had gone undiscovered for nearly 20 years . The approach is likely to be formidable in drugging the FA pathway in future studies .
Flag-FANCI and StrepII-FANCD2 were expressed using the pFastBac1 vector ( Life Technologies ) . For FANCI:FANCD2 complex , Hi5 cell pellets were resuspended in lysis buffer ( 50 mM Tris-HCl pH 8 . 0 , 0 . 1 M NaCl , 1 mM EDTA , 10% glycerol and 1X mammalian protease inhibitor ) , and sonicated . Lysates were clarified by centrifugation at 20 , 000 g and the supernatants were incubated with M2 anti-FLAG agarose resin for 2 hr . The resin was washed 5 × 5 min incubation with wash buffer ( 20 mM Tris-HCl pH 8 . 0 , 0 . 1 M NaCl , 10% glycerol ) , and the protein was eluted in the same buffer containing 100 μg/mL FLAG peptide . GST-UBE2T , Flag-BL100 , MBP-CEF were purified as described ( van Twest et al . , 2017 ) . Ubiquitin and His-UBE1 were purchased from Boston Biochem . His-Avi-ubiquitin was purified as described in Tan et al . ( 2020a ) . Standard ubiquitination reactions contained 10 μM recombinant human avidin-biotin-ubiquitin , 50 nM human recombinant UBE1 , 100 nM UBE2T , 100 nM PUC19 plasmid , 2 mM ATP , 100 nM FANCI:FANCD2 complex wild type ( WT ) or ubiquitination-deficient ( KR ) , in reaction buffer ( 50 mM Tris-HCl pH 7 . 4 , 2 . 5 mM MgCl2 , 150 mM NaCl , 0 . 01% Triton X-100 ) . 20 μL reactions were set up on ice and incubated at 25°C for 90 min . Reactions were stopped by adding 10 μL NuPage LDS sample buffer and heated at 80°C for 5 min . Reactions were loaded onto 4–12% SDS PAGE and run using NuPAGE MOPS buffer and assessed by western blot analysis using Flag ( Aviva Biosciences ) or StrepII ( Abcam ) antibody . Flag-tagged FANCI:FANCD2 and monoubiquitinated FANCI:FANCD2 was prepared by incubating purified FANCI:FANCD2 or monoubiquitinated FANCI:FANCD2 on Flag beads for 2 hr followed by extensive washes in buffer A ( 20 mM TEA pH 8 . 0 , 150 mM NaCl , 10% glycerol ) . 35S-labeled proteins containing UBZ or other ubiquitin domains ( Table 1 ) were generated using the TNT Quick Coupled T7 Transcription/Translation System ( Promega ) and 35S-labeled methionine ( Perkin Elmer ) . 10 μL of TNT product was incubated for 4 hr at 4°C in buffer A with 100 ng Flag-tagged FANCI:FANCD2 or monoubiquitinated FANCI:FANCD2 , 20 μL of Flag-beads ( Sigma-Aldrich ) in a 100 μL reaction . Beads were washed five times with buffer A and resuspended in LDS loading buffer . Proteins were separated by SDS-PAGE and visualized by autoradiography . Oligonucleotides used to create fluorescently labeled DNA were IRDYE-700-labeled X0m1 ( IDTDNA ) and other oligos with the sequences shown in Supplementary file 1 . Assembly of the different DNA structures was performed exactly as previously described ( Supplementary file 1; van Twest et al . , 2017 ) . 25 nM DNA substrates were incubated in 20 μL ubiquitination buffer containing 100 nM FANCI:FANCD2 , 100 nM BL100 , 100 nM CEF , 10 uM HA-ubiquitin ( Boston Biochem ) , 50 nM UBE1 ( Boston Biochem ) and 100 nM UBE2T at room temperature for 90 min to initiate ubiquitination . The reaction was resolved by electrophoresis through a 6% non-denaturing polyacrylamide gel in TBE ( 100 mM Tris , 90 mM boric acid , 1 mM EDTA ) buffer and visualized by Licor Odyssey system . Di-monoubiquitinated FANCI:FANCD2 complex was purified as described ( Tan et al . , 2020a ) . DNA molecules of 60 bp or 150 bp ( dsDNA from oligonucleotides ) or 2 . 6 kb ( circular plasmid DNA ) were used to stimulate the reaction for different experiments , as indicated . Gels containing monoubiquitinated FANCI and FANCD2 bands were excised and in-gel digested with trypsin and subjected to LC/MS analysis on ESI-FTICR mass spectrometer at Bio21 , The University of Melbourne . The analysis program MASCOT was used to identify ubiquitination sites on FANCI and FANCD2 . Freshly purified monoubiquitinated or non-ubiquitinated FANCI:FANCD2 complex was applied to glow-discharged , carbon/formvar grids and allowed to adsorb for 60 s . Specimen was then stained with 2% uranyl formate for 60 s . Specimen were imaged at a magnification of 73 , 000 x on a Ceta camera ( corresponding to a pixel size of 1 . 9 Å ) in Talos 120 kV . For FANCI:FANCD2 complex , 20 micrographs were analyzed and 4553 particles were picked for 2D classification . For monoubiquitinated FANCI:FANCD2 complex , 20 micrographs were analyzed and 4698 particles were picked for 2D classification . The length and width of 2D class average were measured using ImageJ . Monoubiquitinated or non-ubiquitinated FANCI:FANCD2 particles were semi-automatically picked using XMIPP3 ( 83 ) . The parameters of the contrast transfer function ( CTF ) for negative stained data was estimated on each micrograph using CTFFIND3 ( Mindell and Grigorieff , 2003 ) . Finally , reference free 2D alignment and averaging were executed using XMIPP3 . | Bone marrow is the spongy tissue inside bones that produces blood cells . Fanconi anemia is the most common form of inherited bone marrow death and affects children and young adults . In this disease , bone marrow cells cannot attach a protein tag called ubiquitin to another protein called FANCD2 . When DNA becomes damaged , FANCD2 helps cells to respond and repair the damage but without ubiquitin it cannot do this correctly . Without ubiquitin linked to FANCD2 bone marrow cells die from damaged DNA . Another protein , called FANCI , works in partnership with FANCD2 and also gets linked to ubiquitin . Tan et al . studied purified proteins in the laboratory to understand how linking ubiquitin changes the behavior of FANCD2 and FANCI . When the proteins have ubiquitin attached , they can form stable attachments to DNA . Without ubiquitin , however , the proteins only attach to DNA for short periods of time . Using electron microscopy , Tan et al . discovered that large numbers of the modified proteins become tightly attached to damaged DNA , helping to protect it and triggering DNA repair processes . Understanding the role of FANCD2 in Fanconi anemia could lead to new treatments . FANCD2 and FANCI have similar roles in other cells too . Stopping them from protecting damaged DNA in cancer cells could be used to enhance the success of chemotherapies and radiotherapies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2020 | Monoubiquitination by the human Fanconi anemia core complex clamps FANCI:FANCD2 on DNA in filamentous arrays |
DNA replication is a fundamental biological process . The initial step in eukaryotic DNA replication is the assembly of the pre-initiation complex , including the formation of two head-to-head hexameric helicases around the replication origin . How these hexameric helicases interact with their origin dsDNA remains unknown . Here , we report the co-crystal structure of the SV40 Large-T Antigen ( LT ) hexameric helicase bound to its origin dsDNA . The structure shows that the six subunits form a near-planar ring that interacts with the origin , so that each subunit makes unique contacts with the DNA . The origin dsDNA inside the narrower AAA+ domain channel shows partial melting due to the compression of the two phosphate backbones , forcing Watson-Crick base-pairs within the duplex to flip outward . This structure provides the first snapshot of a hexameric helicase binding to origin dsDNA , and suggests a possible mechanism of origin melting by LT during SV40 replication in eukaryotic cells .
DNA replication is essential in the inheritance of genetic information for living organisms . The initiation of DNA replication involves crucial steps in the assembly of replication proteins at the origin and in the subsequent melting of the origin DNA . For DNA replication in prokaryotic cells , the initiator protein , DnaA , first binds to the replication origin to initiate origin melting . The replicative DnaB-family of hexameric helicases are then loaded onto the fully melted single-stranded DNA ( ssDNA ) for replication fork unwinding [reviewed in O'Donnell et al . ( 2013 ) ] and references therein ) . For DNA replication in archaeal and eukaryotic cells , replicative helicases , such as minichromosome maintenance proteins ( MCM ) and SV40 Large T Antigen ( LT ) , load onto the origin dsDNA as head-head ( or N-N ) double hexamers [reviewed in Fanning and Zhao ( 2009 ) ; Gai et al . ( 2010 ) ; O'Donnell et al . ( 2013 ) ; Slaymaker and Chen ( 2012 ) and references therein] . The origin bound by the double hexameric helicases , together with other replication factors , will eventually melt the duplex and produce a pair of replication forks . Despite extensive efforts , the molecular interactions of these hexamer helicases in archaeal and eukaryotic replication systems with their origin dsDNA have yet to be defined . In SV40 DNA replication , the host eukaryotic cellular replication machinery is utilized , except for the MCM helicase , whose function is fulfilled by SV40 LT . As a result , the SV40 replication system serves as a model for studying the eukaryotic replication process [reviewed by Bullock and Simmons ( 1997 ) ; Fanning and Zhao ( 2009 ) and references therein] . To initiate DNA replication , MCM helicase proteins from archaeal and eukaryotic cells assemble a double hexamer at the origin dsDNA ( Evrin et al . , 2009; Fletcher et al . , 2003; Mendez and Stillman , 2003; Remus et al . , 2009; Ticau et al . , 2015 ) , a process that requires the cellular origin recognition complex ( Orc ) as well as additional protein factors . By contrast , the SV40 LT helicase alone is sufficient in recognizing its origin and assembles as a double hexamer at the origin dsDNA ( Borowiec et al . , 1990; Cuesta et al . , 2010; Valle et al . , 2006 ) . On linear blunt-ended dsDNA , containing the SV40 origin sequence , purified LT alone can generate bi-directional unwinding forks from the dsDNA origin region , producing 'rabbit ear'-like single-stranded DNA loops emanating out of the LT complex ( Wessel et al . , 1992b ) . These studies , together with other reports ( Borowiec et al . , 1990; Borowiec and Hurwitz , 1988; Joo et al . , 1998; Kim et al . , 1999; Valle et al . , 2006 ) , demonstrate that LT alone can sufficiently assemble as a double hexamer at the origin , melt the origin DNA , and generate bi-directional unwinding forks . Previous biochemical studies also demonstrate that assembly of the LT double hexamer on the origin dsDNA leads to local melting around the AT-rich sequences ( Borowiec et al . , 1990; Borowiec and Hurwitz , 1988 ) . Available structural and biochemical data suggest that the melted DNA region is located within the AAA+ domain of LT ( Kumar et al . , 2007; Shen et al . , 2005 ) . LT has three major functional domains ( Figure 1A ) : the Dna-J homology domain , the origin-binding domain ( OBD ) , and the helicase domain ( Bullock , 1997; Gai et al . , 2004a ) . LT helicase domain consists of a Zn-domain and an AAA+ ( ATPases Associated with diverse cellular Activities ) domain ( Figure 1A ) . The AAA+ domain has a fold of five β-strands at its core , with a few helices that function as a motor domain by coupling the energy of ATP binding and hydrolysis ( Neuwald et al . , 1999; Ogura and Wilkinson , 2001 ) . The core replication origin DNA of SV40 consists of 64 base pairs ( bps ) ( Figure 1B ) that can be divided into two halves , the AT-half and the early palindrome half ( EP-half ) , each containing two GAGGC penta-nucleotide sequences ( PEN ) plus a stretch of AT-rich sequence ( Dean et al . , 1987; Deb et al . , 1986; Sreekumar et al . , 2001 ) . An individual OBD alone has been shown to bind specifically to the PEN sequences ( Bochkareva et al . , 2006; Deb et al . , 1987; Meinke et al . , 2007 , 2011 , 2014 ) . However , when the LT construct containing the OBD-helicase domain in one polypeptide ( residues 131–627 or LT131 ) is used , the OBD ( residues 131–258 ) binds to a GC pseudo-PEN site in addition to the canonical PEN , and the AAA+ domain of the helicase binds to the AT-rich sequences of the origin ( Chang et al . , 2013 ) . Both the AT-half and EP-half of the origin can support the assembly of one LT hexamer ( Joo et al . , 1998; Kim et al . , 1999; Shen et al . , 2005 ) , and a full core origin supports the formation of two hexamers arranged in an N-to-N configuration ( Bullock et al . , 1997; Cuesta et al . , 2010; Valle et al . , 2006 ) . 10 . 7554/eLife . 18129 . 003Figure 1 . SV40 LT domain structure , core origin DNA ( oriDNA ) , and the co-crystal structure of the LT helicase domain bound to the AT-half oriDNA . ( A ) LT domain organization ( drawn to scale ) , with the LT131 used for the co-crystallization study indicated at the top . ( B ) The full 64-bp core replication oriDNA of SV40 . The four penta-nucleotides ( PEN ) ( GAGGC ( in red ) ) are flanked in the AT-half and in the EP-half origin . The AT-half origin DNA ( boxed in blue ) is co-crystallized with the LT hexameric helicase in this study . ( C ) The gel filtration profile ( Superose 6 ) of LT131 alone or in complex with the AT-half oriDNA in the presence of ADP or ATP without Mg2+ . LT131 alone equilibrates between hexamer and monomer peaks ( red ) , but will form a stable hexamer when ATP is present ( Gai et al . , 2004a; Li et al . , 2003 ) . In the presence of ATP ( blue ) or ADP ( green ) , LT131 forms stable hexamer-DNA complexes with the 32-bp AT-half oriDNA , which migrate slightly larger than the LT131 hexamer alone . The LT-oriDNA-ATP complex isolated from this peak showed no obvious dissociation after up to six hours , suggesting a very stable complex without ATP hydrolysis . ( D , E ) The preformed LT hexamer-oriDNA complexes ( made by first incubating LT131 with oriDNA and ATP , then adding Mg2+ last to initiate unwinding , see Materials and methods ) is active in unwinding the oriDNA ( D ) . For comparison , ( E ) shows the unwinding results obtained using LT131 with no preformed ( un-preformed ) hexamer-oriDNA complexes ( made by adding LT131 and Mg2+ last to initiate unwinding , see Materials and methods ) ;this shows unwinding efficiency similar to that obtained using the preformed LT-oriDNA complex ( D ) . This result suggests that the preformed stable LT-oriDNA complex is active in unwinding the oriDNA . Note that the unwound ssDNA product migrates slower than the dsDNA substrate because of hairpin formation at the palindrome origin sequence . ( F ) Overall view of the LT hexameric helicase bound to the AT-half origin DNA ( 32 bp ) in the central channel . ( G ) Superimposition of the DNA-free ( green ) and DNA-bound ( orange ) hexamer structures , both of which are in an ADP-bound state . For clarity , only two subunits of hexamer are shown so that it is possible to visualize the β-hairpins and DRF loops . The Zn-domain and the AAA+ domain of LT are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 00310 . 7554/eLife . 18129 . 004Figure 1—figure supplement 1 . Comparing the DNA register in the dimeric LT structure bound to the EP-half origin with that in the hexameric LT structure bound to the AT-half origin DNA . ( A ) The 64-bp full core origin of SV40 . The EP-half and AT-half origins are indicated . For the EP-half origin , the sequence that is in contact with OBD1 , OBD2 , and the helicase domain of the LT131 dimer , as revealed in the LT131 dimer-DNA co-crystal structure ( PDBid: 4GDF ) , is indicated . For the AT-half origin , the sequence that is in contact with LT helicase domain , as revealed in the LT helicase hexamer-DNA co-crystal structure , is indicated . ( B ) LT131 dimer-DNA co-crystal structure , showing how the two LT subunits bind to EP-half origin DNA ( PDBid: 4GDF ) . The EP-half origin sequence is positioned next to the complex structure with the same DNA register to indicate the corresponding DNA sequence bound by the different domains of LT131 . With OBD1 binding to PEN1 , OBD2 cannot reach PEN2 from subunit 2 because the linker is not long enough . Instead , OBD2 binds to pseudo-PEN1 , with an 180° rotation from OBD1 . ( C ) The LT helicase hexamer-DNA co-crystal structure , with the AT-half origin sequences positioned next to the complex structure with the same DNA register to indicate the corresponding DNA sequences bound by LT . For comparison , we tried to best align the register of bound DNA in this panel with that in ( B ) . The PEN1’ and pseudo-PEN’ positions are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 00410 . 7554/eLife . 18129 . 005Figure 1—figure supplement 2 . The full occupancy of ADP at all six nucleotide pockets of an LT hexamer and the arrangement of the β-hairpins ( β-HP ) and DRF loops of the AAA+ domain . ( A ) Omit map of ADP density from an LT hexamer-DNA complex structure ( one asymmetric unit ) , showing strong density of the ADP ( drawn at 5 sigma level ) at all six sites . Even though the electron densities for the six sites are not identical , they are all strong , indicating full occupancy in the DNA-bound hexamer structure , as is the case in all previously determined DNA-free LT hexamer structures . ( B ) A close-up view at one of the ADP densities , drawn at sigma contour level 5 . ( C ) The β-hairpins ( β-HP ) and DRF loops of the AAA+ domain in the central channel of a LT hexamer . The β-hairpin residues K512/H513 and the DRF residues D455 , R456 , F459 are shown as sticks . For clarity , only two opposing subunits of the hexamer are drawn . ( D ) The near-planar arrangement of the six β-hairpins ( β-HP ) and DRF loops in the AAA+ domain of the LT hexamer in the DNA complex structure , showing that the β-hairpins and loops are not absolutely planar , but slightly offset . The H513 on the β-hairpin tip is drawn in sticks . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 005 In order to understand how the LT hexameric helicase assembles at the origin and interacts with the origin DNA , we determined the structure of a LT hexameric helicase in complex with the AT-half of the origin to a resolution of 2 . 9 Å . This co-crystal structure , the first for a hexameric helicase bound to its origin dsDNA in its central channel , reveals detailed molecular interactions between the six subunits and the origin sequence . In addition , the structure also shows that a stretch of the AT-rich sequence bound within the narrow channel of the AAA+ domain is partially melted , most likely as a result of a 'squeezing' of the two phosphate backbones by the assembled hexameric helicase , which may be assisted by the specific interactions between residues of the LT AAA+ domain and the AT-rich origin sequence . The structural data reported here offer a mechanistic explanation for how LT accomplishes initial origin melting and may have implications for other related eukaryotic replicons .
In order to study the replicative helicase assembly around the origin DNA for SV40 replication initiation , we used an SV40 LT construct LT131-627 ( residues 131–627 , referred to as LT131 hereafter ) — which includes the OBD domain ( residues 131–258 ) and the helicase domain ( residues 260–627 ) ( Figure 1A ) — for co-crystallization with the AT-half or the EP-half of the origin dsDNA ( Figure 1B ) . Previously , using electron microscopy ( EM ) , we showed that a slightly longer LT108 construct can assemble as a double hexamer on dsDNA containing the full 64-bp core origin in the presence of ATP or ADP , and can support the replication of such DNA in HeLa cell extract ( Cuesta et al . , 2010 ) . To the best of our knowledge , LT is unique in that it can unwind blunt-ended dsDNA containing its origin sequence , whereas the other hexameric helicases characterized to date are not capable of unwinding blunt-ended dsDNA efficiently without the aid of additional protein component ( s ) . We previously also showed that purified LT131 alone , like the full-length LT , can efficiently unwind blunt-ended dsDNA containing the origin sequences in vitro ( Chang et al . , 2013 ) . LT131 also forms stable hexameric complexes bound to the 32-bp AT-half of the origin along with ADP or ATP ( Figure 1C ) . Interestingly , these stable pre-formed hexamers/double hexamers on origin dsDNA are also capable of unwinding the DNA when Mg2+ is supplied ( Figure 1D , E ) , which is consistent with previous reports for full-length LT ( Uhlmann-Schiffler et al . , 2002 ) . This result suggests that the LT double hexamers assembled at the origin sequence are fully competent in generating the initial local origin melting to promote the complete unwinding of the dsDNA . We obtained co-crystals of the complex containing LT131 ( Figure 1A ) and the 32-bp AT-half origin dsDNA ( bps 1–32 , plus 3’-T overhangs , Figure 1B ) in the presence of ADP and Mg2+ , and determined the structure to a resolution of 2 . 9 Å ( Table 1 ) . The structure shows one hexamer of the helicase domain ( residues 266–627 ) binding to the dsDNA in the hexamer channel in each asymmetric unit ( Figure 1F ) . All of the 32-bp AT-half origin dsDNA sequences were built into the DNA structure ( Figure 1B , Figure 1—figure supplement 1A–C ) . The hexamer structure contains the entire helicase domain ( Figure 1A ) , without the N-terminal OBD ( residues 131–265 ) built into the model . The OBDs are actually visible as six blobs of smeared electron density located N-terminal to the Zn domain , suggesting disordered OBDs , possibly due to the lack of interactions between any of the six OBDs with the PEN or pseudo-PEN origin sequences after the completion of the hexamer assembly . This is clearly different from the previously reported co-crystal structure of the dimeric LT131 in complex with the EP-half of the origin sequence ( Chang et al . , 2013 ) , in which one of the two OBDs binds to the canonical PEN in the same way that has been reported by others ( Bochkareva et al . , 2006; Meinke et al . , 2007 ) , and the other OBD binds to a newly identified pseudo-PEN ( GC ) sequence ( Figure 1—figure supplement 1B ) . 10 . 7554/eLife . 18129 . 006Table 1 . Crystallographic data and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 006Co-crystal of complex containing LT131 , AT-half origin dsDNA , and ADPData collectionSpace groupP1Wavelength ( λ ) 1 . 1159 ( Å ) Cell: a , b , c ( Å ) 107 . 485 , 107 . 587 , 107 . 870Angles: α , β , γ ( ° ) 107 . 133 , 106 . 954 , 106 . 865Resolution ( Å ) 50–2 . 90 ( 2 . 95–2 . 90 ) Rmerge ( % ) 5 . 3 ( 79 . 7 ) I / σI28 . 9 ( 1 . 52 ) Completeness ( % ) 97 . 7 ( 89 . 1 ) Redundancy3 . 1 ( 2 . 5 ) RefinementResolution ( Å ) 50 . 0–2 . 9 Unique reflections85 , 533Rwork / Rfree ( % ) 27 . 8/30 . 4No . atoms Protein17 , 598 DNA1354 Ion6 ADP162 r . m . s deviationsBond lengths ( Å ) 0 . 007Bond angles ( ° ) 1 . 194*Highest-resolution shell is shown in parentheses . All six nucleotide pockets of the LT hexamer are fully occupied by ADP , as shown by the strong electron density at all six pockets ( Figure 1—figure supplement 2A–B ) , which is consistent with the all-or-none nucleotide binding-mode on all the previously determined hexamer structures of LT ( Gai et al . , 2004b ) . This DNA-bound hexamer essentially adopts the same structure as the previously determined DNA-free ADP-bound hexamer ( Gai et al . , 2004b ) , with an Cα r . m . s deviation of superposition around 0 . 79 ( Figure 1G ) . Similar to all the DNA-free hexamer structures of the LT helicase domain ( Gai et al . , 2004b; Li et al . , 2003; Lilyestrom et al . , 2006; Zhou et al . , 2012 ) , the DNA-bound hexamer reveals that the six subunits are arranged with only minor offset to break symmetry ( Figure 1—figure supplement 2C , D ) , which can be considered to be a near-planar ring ( referred to as a planar ring hereafter for convenience ) . Within the AAA+ domain of LT , there is a β-hairpin bearing a H513 residue on the inner surface of the hexamer channel ( Figure 1—figure supplement 2C ) , and the β-hairpin displays movement of more than 17 Å along the channel during conformational changes triggered by ATP binding/hydrolysis and release ( Gai et al . , 2004b ) . C-terminal to the β-hairpin is the DRF loop that carries residues D455 , R456 , F459 ( DRF ) pointing into the channel ( Figure 1—figure supplement 2C ) . In the planar ring structure of the LT hexamer , the six β-hairpins and the DRF loops form a narrowest channel segment around the bound dsDNA ( Figure 1—figure supplement 2D ) . 10 . 7554/eLife . 18129 . 007Figure 2 . The structure of the LT hexamer helicase in complex with the AT-half origin . ( A ) The AT-half origin DNA sequence contains 32 bp with 3’-T overhangs bound by LT hexameric helicase in ( B ) . The side-bars indicate the DNA contacts by different regions within the LT helicase channel , including the Zn-domain , the central chamber , the β-hairpins ( β-HP ) and DRF loops ( DRF-LP ) . PEN1’ and pseudo-PEN1’ are not occupied by OBD in LT hexamer structure . The deformed region is indicated on the side . ( B ) The structure of LT131 hexamer bound to the 32-bp AT-half origin DNA . For clarity , only two subunits ( one on each side ) are shown . The positions of the PEN1’ and pseudo-PEN1’ where OBD1 and OBD2 would bind are indicated . ( C ) The structure of LT131 dimer bound to the 32-bp EP-half origin DNA ( PDBid: 4GDF ) , showing one of the two subunits using its OBD to bind PEN1 . The pseudo-PEN1’ is bound by the OBD of the second subunit of the dimer ( Figure 1—figure supplement 1A , B ) ; thehelicase domain of the dimer , including the Zn-domain and the AAA+ domain , binds to the AT-rich region of the EP-half origin . The structures in ( B ) and ( C ) are drawn to scale and aligned side-by-side for comparison between the hexamer-DNA and dimer-DNA complexes . ( D ) The dsDNA segment bound in the AAA+ channel is deformed and locally melted . ( E ) Top view of the LT hexamer ( each subunit in a discrete color ) from the N-terminal end , showing dsDNA in the central channel . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 00710 . 7554/eLife . 18129 . 008Figure 2—figure supplement 1 . Initial unbiased electron density inside the hexameric channel , and DNA orientation and register confirmed by Bromo-labeled dC ( Br-dC ) DNA . ( A ) A section of the initial electron density map , which was calculated on the basis of the MR LT hexamer model alone prior to any DNA building , showing the dsDNA density with clear nucleotide base steps running through the central channel . The map was calculated as follows ( also described in the Materials and methods session ) . The MR solution model of the LT hexamer was partially deleted on the β-hairpin and DRF loop residues facing the central channel to avoid model bias for the residues that potentially interact with DNA in the channel . The partially deleted MR LT hexamer model was used for calculations of the initial phases and for mask calculation for NCS averaging . Two-domain six-fold NCS averaging was then performed , using a tight mask so that a large central channel space where DNA is located was masked out during NCS averaging , which will avoid averaging the DNA density out in the central channel . The resulting NCS averaged map was drawn here at sigma level 1 , which shows the clear DNA base steps in the central channel . ( B , C ) The AT-half origin DNA inside the hexamer channel ( B ) , and the corresponding DNA sequence ( C ) , with the Bromo-labeled dC ( Br-dC at position 12 ) indicated ( colored in green ) . ( D , E ) Difference Fourier map of the anomalous data collected from the LT-DNA co-crystals containing the Bromo-labeled dC ( Br-dC ) DNA at the 12th dC position as shown in ( C ) . The map was drawn with contour sigma level 2 . 0 , showing a strong peak ( viewing from two angles ) at the expected location for the labeled dC , confirming the DNA orientation and register in the hexamer channel . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 008 The entire AT-half origin dsDNA sequence built into the hexamer channel of LT is shown in Figure 2A and B ( also Figure 1—figure supplement 1 ) , with one LT subunit bound to the 32-bp EP-half origin taken from the LT131 dimer-dsDNA structure shown in Figure 2C for comparison ( see Figure 1—figure supplement 1B for the dimer structure ) . The DNA orientation and register is further verified by using Bromo-labeled dC ( Br-dC ) DNA in the co-crystal structure ( Figure 2—figure supplement 1 ) . Bps 1–15 of the origin dsDNA are located outside the channel; bps 6–23 are within the Zn-domain channel; bps 24–26 are in the central chamber ( a wide chamber ) ; and bps 27–32 are within the AAA+ channel ( Figure 2A ) . The dsDNA segment bps 1–24 is in B-form , while a stretch of A-T base pairs , most of which are within the AAA+ channel lined by the β-hairpins and DRF loops , is severely deformed ( bps 25–32 , Figure 2A , B , D ) , disrupting the base paring to yield locally melted duplex DNA . The narrow parts of the hexamer channel in the AAA+ domain , lined by the β-hairpins and DRF loops , is between 14–19 Å ( Figure 1G ) , which is narrower than the diameter of the B-form DNA . The B-form DNA segment has approximately 19–20 Å distances between its two phosphate backbones . However , the backbone distances in the deformed DNA segment ( bps 25–32 ) are mostly in the range of 15–16 Å . The minor groove within this deformed stretch hasbackbone distances of around 4 . 5–5 . 0 Å , also much narrower than that of the B-form DNA segment , whichhas the distances around 7 . 2–10 . 0 Å . One major consequence of the closer phosphate backbone distance within the stretch of deformed DNA is the disruption of normal Watson-Crick base-paring within the duplex ( Figure 2D ) . Below , we describe the detailed structural features and the protein-DNA interactions along the central channel in the N- to C- direction . The LT hexamer channel around the N-terminal Zn-domain has an opening of approximately 21 Å between the side chains at its narrowest point , which widens greatly towards the central chamber . The entire channel length from the Zn-domain to the C-terminal DRF-loop is approximately 60 Å ( Figure 1G ) , encompassing a total of 17 bps of oriDNA bound within the helicase channel ( Figure 2A , B ) . The Zn-domain channel interacts with 8 bps of this oriDNA , all through the phosphate backbones of the DNA . Within the Zn-domain region , the dsDNA is in B-form and fits through the channel tightly , with its two phosphate backbones interacting with charged or polar amino acids in the channel surface ( Figure 3A–F ) . Interestingly , each of the two DNA backbones interact with nearly identical sets of amino acid residues ( N267 , K331 , N332 , K334 and T335 ) that come from different subunits ( subunits i-vi ) and that are arranged in different orientations ( Figure 3C–F , Supplementary file 1 ) . Specifically , the 5’→3’ ssDNA interacts with a set of residues from N→C domain of the helicase containing: N267 of subunit i and ii , K331 of i , ii , and iii , N332 of ii , iii , and iv , and T335 of iv and v ( Figure 3C–D ) . Its complementary strand ( also from 5’→3’ direction ) , interacts with a similar set of residues but from different subunits: K334 of i and ii , T335 of i and vi , N332 of v and vi , and N332 and K331 of v ( Figure 3E–F ) . Therefore , despite their opposite orientations , the two antiparallel ssDNA strands of the B-form DNA interact with two different sets of but chemically identical , amino acids along the channel surface of the Zn-domain . In other words , these two sets of amino acid residues make near identical bond interactions with DNA phosphate backbones running in either the 5’→3’ or 3’→5’ orientation . 10 . 7554/eLife . 18129 . 009Figure 3 . Detailed interactions of origin DNA with the Zn-domain of LT . ( A , B ) Two different views of dsDNA inside the LT hexameric channel , one with two complete subunits out of six ( A ) and the other with the DNA-binding residues on the Zn and AAA+ domains of all six subunits ( B ) . ( C , D ) Two views tshowing the detailed interactions of the 5’→3’ DNA strand with LT residues inside the Z-domain channel from the N→C direction . The six subunits are labeled as i , ii , iii , iv , v , vi . The residues contacting the DNA are shown as sticks . ( E , F ) Two views showing the detailed interactions of the 3’→5’ DNA strand with LT residues inside the Z-domain channel from the N→C direction . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 009 The central channel narrows down to a much smaller opening at the AAA+ domain where the β-hairpins and the DRF loops are arranged in near-planar rings ( Figure 1D , Figure 4—figure supplement 1A ) . The channel segment composed of β-hairpins and the DRF loops is approximately 22 Å in length , and binds 6 bps of deformed dsDNA ( bps 27–32 , plus the 3’-T overhang , Figure 2A ) . The planar arrangement of the β-hairpins and DRF loops of LT observed in this dsDNA-bound structure and the previously reported apo-structures ( Gai et al . , 2004b; Li et al . , 2003; Lilyestrom et al . , 2006; Zhou et al . , 2012 ) ( Figure 4—figure supplement 1A , B ) are in sharp contrast to the staircase β-hairpins reported for the E1 helicase in the ssDNA-bound ( Enemark and Joshua-Tor , 2006 ) and apo-structure ( Sanders et al . , 2007 ) ( Figure 4—figure supplement 1C–D ) . In the ssDNA-bound E1 structure , each of the six spirally arranged β-hairpins of the E1 hexamer has similar interactions with a spirally shaped six-nucleotide oligo-dT via the phosphate backbones of the ssDNA ( Figure 4—figure supplement 1C ) . The planar arrangement of the six β-hairpins and DRF loops in the dsDNA-bound hexamer structure of LT is nearly identical to that in the DNA-free LT hexamer structure , including the channel openings and the conformations of the β-hairpins and DRF loops ( Figure 1G ) . Because this stretch of channel is clearly narrower than the dimension of B-form dsDNA , the dsDNA bound inside has to be deformed in order to fit through the smaller channel . As a result , the two backbones in this segment are closer to each other than in standard B-form dsDNA; this deformation is a consequence of compression by two opposing subunits of the hexamer around the narrow channel . The deformation of DNA should also be assisted by its interactions with residues on the β-hairpins and DRF loops in the channel walls ( Figure 4A–F ) . When the two phosphate backbones are compressed by the β-hairpins and the DRF loops , the base-pairs within the duplex are forced to flip outward , leading to the local melting of the dsDNA ( Figure 4A–F ) . 10 . 7554/eLife . 18129 . 010Figure 4 . Detailed interactions of origin DNA with the AAA+ domain . ( A , C , E ) Detailed interactions of DNA with three pairs of subunits within the LT hexamer: subunit pair i & iv ( A ) , ii & v ( C ) , and iii & vi ( E ) . The β-hairpin residues K512/H513 and DRF loop residues D455/R456/F459 that bind to the DNA are labeled . ( B , D , F ) The detailed interactions of the oriDNA with the same three pairs of LT subunits shown in ( A ) [matching B] , ( C ) [matching D] , and ( E ) [matching F] . Dashed lines indicate hydrogen-bond interactions between LT residues and DNA ( in the stick model ) . Some of LT residues are 5–7 Å away from DNA , and would be suitable for the water-mediated formation of hydrogen bonding . In addition , a number of LT residues are within Van der Waals distance of the bound DNA , forming hydrophobic packing interactions; these residues include K512 of subunit i ( i-K512 ) , i-F459 , iv-F459 , ii-F459 , v-F459 , vi-F459 , and vi-H513 . The dislocated bases in the disrupted bases-pairs ( in the partially melted DNA region ) are drawn in thick sticks , with the DNA sequences of this region shown to the left of panel B . The structure shows clearly that identical residues on different subunits make unique contacts with the phosphate backbones and the major or minor gr oves of the DNA; this is possible due to the near-planar arrangement of the six subunits . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 01010 . 7554/eLife . 18129 . 011Figure 4—figure supplement 1 . The planar and staircase arrangements of β-hairpins in the hexamer structures of LT and E1 . ( A ) The near-planar arrangement of the six β-hairpins ( indicated by an arrow ) in the hexamer channel of LT . Such near-planar arrangement is maintained in all the structures of LT hexamer determined to date , regardless of their dsDNA-bound or DNA-free states . ( B ) A close-up view of the near-planar arrangement of the six LT β-hairpins from ( A ) , with H513 on the β-hairpin tip shown in sticks . The close-up view reveals that the six H513 and β-hairpins have no real six-fold symmetry . However , a near-planar arrangement is obvious when compared with the staircase arrangement shown in ( C ) . ( C ) The staircase arrangement of the six β-hairpins ( and the tip His residue ) in the E1 hexameric channel , showing a six-nucleotide ssDNA ( in thin sticks ) with each of the six nucleotides havingsimilar interactions with the six β-hairpins ( PDBid: 2GXA ) . ( D ) Alignment of the β-hairpins in the central channel of E1 hexamer structures in the ssDNA-bound state ( PDBid: 2GXA , in green ) and the DNA-free state ( PDBid: 2V9P , in orange ) , showing the same staircase arrangement for the β-hairpins , regardless of ssDNA-bound or ssDNA-free state . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 011 The major residues interacting with dsDNA in the AAA+ domain include K512 and H513 on the β-hairpins , and residues D455 , R456 , and F459 ( DRF ) on the DRF loops ( Figure 4 ) . All six subunits use these same residues to interact with DNA , but on each subunit , these residues contact DNA differently as a consequence of the near-planar arrangement of the β-hairpins and DRF loops ( Figure 4B , D , F ) . The β-hairpin residues K512/H513 from the subunit pairs ii + v ( Figure 4C , D ) and iii + vi ( Figure 4E , F ) more or less impinge on the two DNA backbones from opposite sides , while the DRF loop residues D455/R456/F459 from the subunit pairs i + iv ( Figure 4A , B ) and ii + v ( Figure 4C , D ) act on the phosphate backbones similarly , pushing the two phosphate backbones within this region closer to each other . This may be sufficient to destabilize the Watson-Crick base-pairing , forcing the bases to flip outward . Thus , these amino acids on the β-hairpins and DRF-loops from different LT subunits interact with the deformed DNA from the major and minor grooves as well as the phosphate backbones ( Figure 4A–F , Supplementary file 2 ) . 10 . 7554/eLife . 18129 . 012Table 2 . The central channel dimensions at the β-hairpins and DRF loops in the hexamer structures of various nucleotide-bound states determined previously and in the DNA-bound state reported here . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 012Between β-hairpin tip residues K512 and H513 ( Å ) Between DRF loop tip residues D455 and R456 ( Å ) dsDNA-bound ADP-hexamer ( this report ) 19 . 5 , 19 . 9 , 20 . 114 . 8 , 14 . 8 , 15 . 0dsDNA-free ADP-hexamer ( PDB:1SVL ) 11 . 6 , 13 . 1 , 14 . 614 . 0 , 16 . 0 , 18 . 3dsDNA-free ATP-hexamer ( PDB:1SVM ) 6 . 90 , 7 . 40 , 8 . 907 . 90 , 8 . 30 , 8 . 30dsDNA-free Empty hexamer ( PDB:1N25 ) 14 . 517 . 9
Interestingly , the DNA-bound and DNA-free forms of the LT hexameric helicase structures are largely the same . This includes both the small opening of the hexamer channel in the AAA+ domain and the near-planar arrangement of the six subunits , β-hairpins , and DRF loops in the central channel . Several DNA-free hexameric structures of the LT helicase domain have been reported previously , including the structures of three different nucleotide ( nt ) -bound states: the empty form alone or in complex with other proteins ( Li et al . , 2003; Lilyestrom et al . , 2006; Zhou et al . , 2012 ) ; the ADP-form , with all six nt pockets occupied by ADP , and the ATP-form , with all six nt pockets occupied by ADP-BeF3- ( Gai et al . , 2004b ) . These different DNA-free hexameric structures reveal substantial concerted conformational changes for the six β-hairpins and DRF loops within and along the central channel of LT hexamer , but the overall conformations of these hexamers are quite similar ( Gai et al . , 2004b; Li et al . , 2003; Lilyestrom et al . , 2006; Zhou et al . , 2012 ) . Subsequent comparisons of the dsDNA-bound and DNA-free LT hexameric helicase structures , both in the ADP-form , show that these structural features are essentially maintained regardless of DNA binding . This strongly suggests that the binding of dsDNA in the central channel , per se , has no major impact on the overall conformation of the protein hexamer . Obviously , it is the conformational state of the protein hexamer that determines the conformation of the bound dsDNA . By extension , the cyclic conformational changes of hexameric LT are expected to drive the remodeling of the bound dsDNA , as shown by the different nt-bound states: ATP-binding shrinks the central channel further than ADP-binding and triggers large β-hairpin movement inside the central channel ( Gai et al . , 2004b ) . By contrast , even though the E1 hexameric helicase shows a staircase arrangement of the six β-hairpins instead of a planar arrangement , the staircase conformation is observed in the E1 ssDNA-free structure and further maintained in the E1 ssDNA-bound structure . This evidence provides another example of how protein conformations in the DNA-free form may provide insight into how these proteins interact with DNA ( Figure 4—figure supplement 1C–D ) ( Enemark and Joshua-Tor , 2006; Sanders et al . , 2007 ) . Interestingly , the co-crystal structures of the prokaryotic helicase DnaB and the Rho RNA translocase , in complex with ssDNA and ssRNA , respectively , also reveal a spiral-like arrangement of their entire six subunits ( Itsathitphaisarn et al . , 2012; Thomsen and Berger , 2009 ) , suggesting that these hexameric helicases/translocases may bind single-stranded nucleic acids using a spiral arrangement . How then , are we able to reconcile the near-planar conformation of the β-hairpins observed in the LT hexameric helicase with the staircase conformations observed in E1 and other hexameric helicases ? One simple explanation could be that planar β-hairpins in the LT-dsDNA structure and staircase β-hairpins in the E1-ssDNA structure represent two distinct stages of DNA replication . The LT-dsDNA structure may correspond to a pre-initiation hexameric complex that is assembled at the origin before it propagates into full replication forks; whereas , the E1-ssDNA structure may represent a hexameric helicase encircling a single strand of DNA for later-stage replication fork unwinding ( Enemark and Joshua-Tor , 2008; Lee et al . , 2014 ) . This simple explanation would then imply an ability of LT and E1 to switch between the planar and staircase conformations , even though only one conformation has been consistently observed in structures of either protein determined to date . It is possible that the crystallographic packing somehow selectively traps the β-hairpins in the planar conformation for LT and the staircase conformation for E1 , despite the crystallization buffer condition . One other major difference between E1 and LT is that E1 requires the assistance of an E2 protein for origin assembly [reviewed by McBride ( 2013 ) and references therein] , whereas LT alone is sufficient for origin assembly . However , we cannot completely rule out the possibility of other potential mechanistic differences in origin assembly and origin melting by LT and by E1 helicases . In all determined hexameric helicase structures of LT , including the mutiple DNA-free structures reported previously ( Gai et al . , 2004b; Li et al . , 2003; Lilyestrom et al . , 2006; Zhou et al . , 2012 ) and the dsDNA-bound structure reported here , the channel opening within the N-terminal Zn-domain remains the same , whiereas the opening of the AAA+ domain channel around the β-hairpins and DRF loops vary by between 14–19Å ( Figure 1G ) ( Table 2 ) . It is apparent that the helicase conformations dictate the conformation of the bound DNA inside the hexameric channel . Specifically , when the origin dsDNA is bound inside the narrower segment of the channel lined by the β-hairpins and DRF loops of LT , the two phosphate backbones will be remodeled according to the size and shape of the hexameric channel . The remodeling process involves the 'squeezing' or compressing force of the β-hairpins and DRF loops pushing the two phosphate backbones closer to each other , forcing the Watson-Crick base-pairs between the two backbones to flip outwards , leading to base-pair disruption and base flipping in the AT-rich region ( Figure 4A–F ) . Therefore , the co-crystal structure suggests that these locally melted origin base-pairs are probably the consequence of a 'squeezing' by the narrower channel formed upon the completion of the hexamer assembly , rather than a more intuitive pulling or prying apart of the two strands . The K512/H513 residues on the β-hairpins and the D455/R456/F459 residues on the DRF loops of the AAA+ domain not only apply the 'squeezing' action to the two phosphate backbones directly , but also interact with the DNA from both the major and minor groove to generate the deformation and local melting ( Figure 4A–F , Supplementary file 2 ) . While the hexamer-dsDNA complex structure reported here has a near-planar ring of six subunits , the possibility of changing from spiral intermediate assembly to the final planar hexamer also exists . Such conformational conversion has previously been proposed to provide energy for origin melting ( Bochkareva et al . , 2006; Meinke et al . , 2007; Schuck and Stenlund , 2005 ) . No spiral assembly of hexamer/double-hexamer helicases on dsDNA has been reported , but spiral assembly of hexameric helicases on ssDNA has been reported for E1 , DnaB , and Rho ( Enemark and Joshua-Tor , 2006; Itsathitphaisarn et al . , 2012; Thomsen and Berger , 2009 ) . In the co-crystal structure of the LT131 dimer-oriDNA intermediate assembly ( Chang et al . , 2013 ) , the helicase domains of the two subunits are positioned side-by-side , whereas the two β-hairpin H513 of the dimer interact with the minor groove of the oriDNA with a partially spiral arrangement along the helical groove ( Figure 5A and inset ) . If additional LT subunits were to be added to the dimer without changing direction , it would generate a slightly offset hexamer that resembles a lock-washer ring more closely than the closed planar hexamer structure shown in this report . If such a partial helical assembly intermediate of the helicase domain were to be converted to a closed planar hexamer upon completion of a hexamer formation , additional snapping force would be generated , facilitating origin melting in the AT-rich sequence within the narrow channel . 10 . 7554/eLife . 18129 . 013Figure 5 . A model for the assembly pathway of the LT double hexamer complex on the viral origin DNA ( oriDNA ) . ( A ) The assembly of the intermediate complex containing LT dimer bound to each half of the oriDNA , as predicted on the basis of the co-crystal structure of the LT dimer and EP-oriDNA complex ( PDBid: 4GDF ) . The four PENs ( penta-nucleotides ) and the pseudo-PENs recognized by OBD are colored in red and blue . In this dimer intermediate , PEN2/PEN2’ near the center are not bound by the OBD from the bound LT dimer , probably because the linker to the helicase domain anchored to the AT-rich origin is too short for the OBD to reach PEN2 . Dimerization of LT is not through the ODB , but through the helicase domains . The H513 of the β-hairpin ( β-HP ) and R456/F459 of the DRF-loop contacting the oriDNA are shown in the boxed inset . The oriDNA is not distorted or meltedin this intermediate complex . ( B ) Assembly of two head-head hexamers from the dimer intermediate in ( A ) , which forms a near-planar closed-ring instead of the lock-washer open-ring , as predicted based on the hexamer-oriDNA structure here and the prior EM studies ( Cuesta et al . , 2010; Valle et al . , 2006 ) . The origin sequences at the bottom are aligned to the known binding locations on the LT double hexamer , in which the AT-rich sequences of the AT/EP halves are bound by the helicase domain . Red stars indicate the melted/distorted AT-rich sequences as identified biochemically upon assembly of LT double hexamer in the presence of ATP ( Borowiec et al . , 1990; Borowiec and Hurwitz , 1988 ) . The same studies also showed that the melting of the AT-half origin is much less extensive if ADP was used instead of ATP , even though both ADP and ATP induced similar melting on the EP-half origin . Green stars indicate the melted/distorted sequence in the co-crystal structure of the ADP-bound LT hexamer in complex with the AT-half origin ( with the two smaller stars indicating less distortion ) . It is expected that when the changes in LT conformation are coupled to the binding/hydrolysis of cyclic ATP , the extensively melted/distorted oriDNA will be propagated to form a pair of replication forks and thus to initiate DNA replication . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 01310 . 7554/eLife . 18129 . 014Figure 5—figure supplement 1 . The alignment of the structures of SV40-origin DNA bound to OBD alone ( LT131-259 ) and to the OBD-helicase domain ( LT131-627 ) of LT . The alignment is based on the origin DNA sequences in the structures with the 64-bp core origin sequence shown at the top . ( A ) The full core origin sequence of SV40 . The EP-half ( left ) and AT-half ( right ) origin are capable of supporting hexamer assembly . The four PENs ( in red and pink ) and the newly identified pseudo-PEN ( in blue ) are recognized by OBD . The AT-rich sequence interacts with the helicase domain . ( B ) Structures of the four PENs bound by individual OBDs , ( C ) the full EP-half origin bound by dimeric LT131-627 , and ( D ) the full AT-half origin bound by hexameric LT131-627 . For comparison between the dimeric and hexameric LT131-627 , the EP-half origin bound to the dimeric LT131-627 is aligned to the right side with the AT-half . LT131-627 ( containing the OBD-helicase domain ) is capable of forming a hexamer on each half origin and a double hexamer on a full-origin . Note that when a full hexamer is assembled on each half origin , only the helicase domain hexamerizes , the six OBDs are disordered in the structure , possibly because of the dynamic competition for binding to the three available binding sites ( two PENs and one pseudo-PEN ) . When the LT131-627 dimer is bound to an half origin , only the helicase domain dimerizes , and the two OBD domains that are bound to PEN1 and pseudo-PEN separately , with an 180° inversion , do not interact with each other . DOI: http://dx . doi . org/10 . 7554/eLife . 18129 . 014 Biochemistry studies have shown that the melting of the EP- and AT- halves of the origin can be induced by the assembly of LT onto the origin dsDNA in the presence of ATP or ADP , without the need for ATP hydrolysis ( Borowiec and Hurwitz , 1988 ) . Interestingly , even though the melting of the EP-half origin is similar in the presence of ATP or ADP , the melting of the AT-half origin is much less profound with ADP than it is with ATP ( Borowiec and Hurwitz , 1988 ) , indicating the intrinsic differences in melting between the two halves of the origin . The LT-origin dsDNA complex structure presented here corroborates these biochemical studies and provides direct structural data showing the locally melted region of the AT-half of the origin with ADP binding . We were never able to obtain dsDNA co-crystals in the presence of ATP , and so we were unable to compare AT-half origin melting with ATP versus that with ADP binding at a structural level . Because a previous report showed that the ATP-bound LT hexamer has a central channel that is about 3–5 Å narrower than the ADP-bound hexamer ( Gai et al . , 2004b ) , it is conceivable that ATP-binding would generate more extensive melting of the AT-rich origin . We did obtain co-crystals of hexameric LT with the EP-half origin dsDNA , but the diffraction quality was poor and we were unable to obtain high-resolution structural data to describe the melting EP-half in order to compare the intrinsic differences between the EP- and AT- halves of the origin in terms of hexamer formation and the associated remodeling of the origin . With the co-crystal structures of the LT dimer bound to the EP-half origin and the hexamer bound to the AT-half origin determined to date , we can begin to postulate a pathway for LT assembly at the origin ( Figure 5 ) . At the initial stage of LT assembly , a LT dimer binds to each half of the origin through initial OBD interactions with PEN1 and a newly identified pseudo-PEN site ( Figure 5A ) . This allows the AAA+ domain to contact , via β-hairpin and DRF loop residues , the AT-rich sequences of the AT- and EP-halves of the origin through charge and DNA shape readouts ( Figure 5A inset ) ( Chang et al . , 2013 ) . The importance of this interaction of origin DNA with β-hairpin residues H513 and K512 is consistent with previous studies showing that mutations in these residues abolish origin melting by LT ( Kumar et al . , 2007 ) . Furthermore , mutations of the E1 helicase equivalent His residue , H507 of E1 , also abolish origin melting by E1 ( Liu et al . , 2007; Schuck and Stenlund , 2011 ) . Even though no stable trimer or tetramer intermediates of LT-oriDNA were observed , the formation of the stable LT dimer-oriDNA intermediate complex does not cause origin DNA distortion or melting ( Chang et al . , 2013 ) . Interestingly , previous biochemical evidence has suggested that E1 helicase formed a trimer intermediate complex instead of a dimer on each half of the origin DNA , which melts the origin DNA before hexamer/double hexamer formation ( Liu et al . , 2007 ) . The LT dimer-oriDNA intermediate complex formed on the AT- and EP-halves of the origin can then act as a nucleation event that recruits new subunits to form the ring-shaped hexamer/double hexamer around the oriDNA ( Figure 5B ) . The helicase domains of each hexamer then bind to the AT-rich sequences , while the six OBDs take on dynamic conformations , loosely interacting with the limited number of PEN/pseudo-PEN sites . Upon completion of LT hexamer/double hexamer assembly , the helicase domains of the six LT subunits form a tight , near-planar ring , rather than a spiral conformation , around the dsDNA . Concurrent with the assembly of this ring with a narrow central channel around the AT-rich sequences , local origin melting within the helicase domain may be achieved as a result of the formation of the narrow channel , which consists of the β-hairpins and DRF loops . The narrow channel compresses ( or squeezes ) the two phosphate backbones of the dsDNA to distort and melt the origin DNA . It is possible that more helical subunits are added to adimer intermediate , which converts to a planar arrangement once a full hexamer is formed , a process that could provide further energy for distorting/melting the oriDNA . Models with spiral assembly by LT have been proposed previously ( Bochkareva et al . , 2006; Meinke et al . , 2007 ) ; in these models , which are based on structural studies using OBD alone binding to its PEN ( GAGGC ) recognition sequence , propose that the OBD domain assembles into a spiral conformation . Several co-crystal studies of the isolated OBD domain of LT ( or E1 ) have revealed how individual OBD recognizes the PEN DNA sequences ( Bochkareva et al . , 2006; Enemark et al . , 2002; Meinke et al . , 2007; Meinke et al . , 2013 ) : individual OBDs do not interact with each other , but bind tightly to the PENs using the same interactions ( Figure 5—figure supplement 1A , B ) . However , the OBDs in the hexameric LT131-oriDNA complex structure do not bind to the PENs and are disordered ( Figure 5—figure supplement 1D ) . The disordered OBDs in the hexamer complex also differ from the well-ordered OBDs of the LT131 dimer-oriDNA complex ( Chang et al . , 2013 ) , in which one OBD ( OBD1 ) binds to PEN1 and the other OBD ( OBD2 ) bind to a pseudo-PEN site with about 180° rotation ( Figure 1—figure supplement 1B , Figure 5—figure supplement 1C ) , with no interactions between the two OBDs . However , the helicase domains of the same two LT molecules bind to the AT-rich oriDNA as a tight dimer with about 60° rotation to each other . This structure of LT131 dimer-oriDNA suggests that , when OBD is linked to the helicase domain , the binding of two OBDs to PEN/pseudo-PEN probably functions to coordinate with the binding of the helicase domain to the AT-rich sequences where the initial melting occurs . Furthermore , when a complete closed hexameric ring is assembled around each half origin , it creates a problem of six OBDs with only three available binding sites ( PEN1 , PEN2 and one pseudo-PEN ) . The disordered OBDs in the hexamer-oriDNA structure indicate that OBD1 and OBD2 of the initial dimer-oriDNA intermediate no longer bind to oriDNA tightly , probably because of competition for limited binding sites among the six OBDs . This provides an explanation for the lack of well-defined electron density for the six OBDs in the hexamer structure , which is also consistent with our prior low-resolution EM studies showing the flexibility of the OBDs in the central part of the double hexamer structure ( Cuesta et al . , 2010; Valle et al . , 2006 ) . In summary , the complex structure of the LT hexamer assembled on its partially melted origin dsDNA provides the first view of the detailed atomic interactions between a hexameric initiator/helicase and oriDNA , and yields novel insights into a possible mechanism of origin melting . To convert this partially melted origin to a fully opened origin , represented by a pair of replication forks , the melted ssDNA must find a way to exit the hexameric channel , as LT has been shown to translocate preferentially along one ssDNA during unwinding ( Yardimci et al . , 2012 ) and E1 has been shown to bind ssDNA ( Enemark and Joshua-Tor , 2006 ) . But LT unwinding activity can bypass a polypeptide road-block ( similar to the size of a ssDNA ) cross-linked to DNA substrates , suggesting that LT hexamers have sufficient plasticity to alter their conformation around DNA and allow the passage of ssDNA without falling apart ( Yardimci et al . , 2012 ) . At present , we have no data regarding this process that might reveal how the initial melting of oriDNA propagates into fully open replication forks . Nevertheless , the partially melted AT-half origin within the LT helicase channel may represent a snapshot of the initially melted origin primed for further propagation into two bi-directional forks in the SV40 viral replication system and in other related eukaryotic replicons .
LT construct 131–627 ( LT131 ) ( Figure 1A ) was cloned in a pGEX-6p-1 vector as a GST fusion to the N-terminus of LT . The LT protein was purified as previously described ( Greenleaf et al . , 2008 ) . Briefly , the GST fusion protein was purified by glutathione affinity chromatography , followed by precision protease digestion to release LT from the GST . Gel filtration chromatography using Superdex-200 on FPLC was used to further purify the LT protein , which was concentrated to 10–20 mg/ml and stored at –80°C for crystallization . The AT-half origin dsDNA ( Figure 1B ) was prepared by annealing 5’-GGCCTCGGCC TCTGCATAAA TAAAAAAAAT TAT-3’ and 5’-TAATTTTTTT TATTTATGCA GAGGCCGAGG CCT-3’ as described below . The oligonucleotides ( Operon ) were purified through a Mono Q ion exchange column . The two purified oligonucleotides were mixed at a 1:1 molar ratio and annealed by heating to 90°C followed by slow cooling to room temperature over 30 min . Gel filtration chromatography using Superdex 75 was used to separate dsDNA from unannealed ssDNA . Purified dsDNA was concentrated to ~200 nM and stored at –20°C for co-crystallization . It was reported previously that the full length LT in a preformed complex with oriDNA is active in unwinding the oriDNA ( Uhlmann-Schiffler et al . , 2002 ) . In order to test whether the LT131-627 construct in a preformed complex with oriDNA is also active in unwinding oriDNA , we carried out helicase assays of the preformed LT131-oriDNA complex and the un-preformed LT131-oriDNA mixture . To obtain the preformed LT131-oriDNA complex , LT131 protein and a 146-bp dsDNA containing the SV40 core origin sequence was mixed in a buffer containing 20 mM Tris-Cl pH 7 . 5 , 50 mM NaCl , 0 . 1 mM EDTA , 0 . 1 mg/mL BSA and 1mM DTT , and the mixture was incubated for 60 min at 20°C to allow LT to bind to the oriDNA . Then , 4 mM ATP was added to the mixture for another 30 min incubation to allow ATP binding ( but no hydrolysis in the absence of Mg2+ ) locking in the LT hexamer . Such preformed LT-oriDNA complex is stable for up to six hours without obvious dissociation ( Figure 1C ) . To start the unwinding reaction , 10 mM MgCl2 was added to the preformed LT131 hexamer-oriDNA complex and the mixture was incubated at 37°C for 45 min . The reaction was terminated by adding 0 . 1% SDS , 25mM EDTA and 10% glycerol , and the reaction mixture was analyzed on 12% native polyacrylamide gel in 1X TBE buffer . As a control , the unwinding assay was also conducted side-by-side with un-preformed LT-oriDNA complex , in which reaction cocktails missing LT and Mg2+ were mixed on ice right before the unwinding assay ( to prevent formation of LT-oriDNA complex ) , with the MgCl2 and LT protein added last to initiate the reaction . LT131 was mixed with the purified AT-half or EP-half origin dsDNA at a molar ratio of 6:1 ( LT monomers: dsDNA ) in the presence of 1mM ADP or ATP . Crystals with the same space group and cell dimensions were obtained only in the presence of ADP using the hanging drop method in a buffer containing 70 mM Hepes pH 7 , 40 mM MgCl2 , 5 mM MnCl2 , and 10 mM DTT at 4°C . Crystals were transferred to the same buffer with the addition of 20% MPD for flash freezing with liquid nitrogen . Crystal diffraction data were collected at synchrotron beamlines at the Lawrence Berkeley National Laboratory and Argonne National Laboratory . The datasets were processed using HKL2000; the statistics are summarized in Table 1 . The initial phases of the structure were solved by molecular replacement ( MR ) using the LT helicase domain structure in the ADP-bound form ( PDB code 1SVL ) as the search model with the program PHASER , which located a hexamer in one asymmetric unit . The electron density map calculated using the MR solution model of LT revealed some DNA electron density in the central channel , which was not very well featured at this stage . We improved the electron density map through six-fold multi-domain NCS averaging ( i . e . separating the helicase domain into the Zn-domain and the AAA+ domain for two-domain six-fold NCS averaging ) with solvent flattening prior to any DNA model building , using the CCP4 program DMMULTI . For NCS averaging , the residues on these β-hairpins and the loops ( in particular , residues 330–340 of the Zn-domain , 509–518 of the β-hairpins , and 452–463 of the DRF loop ) facing the central channel were deleted from the MR model for two major purposes . The first purpose was to avoid model bias because these β-hairpin and loop residues may have different conformations due to interactions with DNA . The second purpose was to calculate a tight mask ( 1 . 0 Å radius from atoms ) covering the LT model with the central residues/loops deleted for NCS averaging . This was done so that the central channel space , where the DNA is located , would be completely masked out for NCS averaging , crucial for revealing the unbiased DNA density in the NCS averaged map . The mask was smoothed and islands inside the mask were removed using program MAMA in the RAVE suite ( Kleywegt and Jones , 1999 ) . The six-fold NCS averaging ( 50 cycles ) using the improved tight mask yielded a greatly improved electron density map in the central channel , which revealed clear base-pair steps in the helical nature of the DNA density in the central channel ( Figure 2—figure supplement 1A ) . This NCS averaged map was used to build the model ofthe DNA and the deleted residues/loops of LT in the central channel using program 'O' . However , for the fuzzy density of the N-terminal OBD , no significant improvement was achieved . The six OBD domains were only be visible as six blobs of density at low sigma level near , but not touching , the DNA density in the center . As a result , OBDs could not be built into the structure model . The initially built model was refined by simulated annealing using Phenix , and the refined model was used for further reiterative model building and refinement . While all 32-bp steps of the AT-half sequence were visible on the NCS averaged electron density map , we built all 32-bp , plus the two 3’-T overhang ( Figure 2A , B , Figure 1—figure supplement 1C ) . While all the DNA base pairs show strong electron densities , the density for the phosphate backbone is weaker , possibly because of the NCS-averaging effect of the six LT subunits . All the deleted LT residues in the central channel were rebuilt back into the model . The DNA register in the final model was confirmed using Bromo-dC labeled dsDNA ( Figure 2—figure supplement 1B–E ) . The DNA register is also consistent with both the complex structure of LT131 dimer bound to the EP-half origin ( Figure 1—figure supplement 1C ) ( Chang et al . , 2013 ) and the biochemistry data on origin DNA binding ( Joo et al . , 1998; Kim et al . , 1999; Sreekumar et al . , 2000 ) . The final model containing the LT helicase domain and a 32-bp origin DNA from the LT131 construct has very good refinement and geometry statistics , as shown in Table 1 . | When a cell divides to form a new cell , it must also copy its DNA . An important step for starting DNA replication is to break or “melt” the bonds between the two strands of DNA that make up the double helix . This happens at a specific site on the DNA called the replication origin , and allows double-stranded DNA to partially unwind into two single strands . Each strand acts as a template to form a new copy of its partner strand . In many organisms , ring-shaped proteins called helicases attach directly to double-stranded DNA to melt the bonds at the replication origin . Viruses also have helicases that they can use to hijack the cell’s replication machinery and get it to copy the viral DNA . One such helicase called Large T antigen is an important part of a tumor virus called simian virus 40 . Large T has a ‘hexameric’ structure , similar to the cell’s own helicases , and researchers often use it as a model for studying the replication process . However , key questions remain . How do hexameric helicases interact with the double-stranded DNA replication origin ? And what are the molecular mechanisms by which these helicases melt bonds in double-stranded DNA ? Gai et al . used a technique called X-ray crystallography to examine the molecular structure of the Large T helicase when it is attached to DNA . The resulting structures show that six subunits of Large T helicase assemble a ring around the replication origin , each making unique bonds with the DNA . The helicase ring squeezes the DNA , partially breaking the bonds between the DNA strands and causing local melting of the DNA . With the understanding of how the Large T hexamer helicase interacts with its DNA replication origin and how its assembly initiates DNA melting , the future challenge is to explain the process by which the partially melted origin opens up fully for replication to begin . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | The structure of SV40 large T hexameric helicase in complex with AT-rich origin DNA |
Sucrose’s sweet intensity is one attribute contributing to the overconsumption of high-energy palatable foods . However , it is not known how sucrose intensity is encoded and used to make perceptual decisions by neurons in taste-sensitive cortices . We trained rats in a sucrose intensity discrimination task and found that sucrose evoked a widespread response in neurons recorded in posterior-Insula ( pIC ) , anterior-Insula ( aIC ) , and Orbitofrontal cortex ( OFC ) . Remarkably , only a few Intensity-selective neurons conveyed the most information about sucrose’s intensity , indicating that for sweetness the gustatory system uses a compact and distributed code . Sucrose intensity was encoded in both firing-rates and spike-timing . The pIC , aIC , and OFC neurons tracked movement direction , with OFC neurons yielding the most robust response . aIC and OFC neurons encoded the subject’s choices , whereas all three regions tracked reward omission . Overall , these multimodal areas provide a neural representation of perceived sucrose intensity , and of task-related information underlying perceptual decision-making .
Chemical stimulation of taste receptor cells elicits signals that are transduced into neural representations of multiple attributes , such as taste quality , intensity ( the strength or concentration of a stimulus ) , and palatability ( hedonic value ) . These attributes form a single percept ( Accolla et al . , 2007; Breslin , 2013; Lemon , 2015 ) that informs the animal whether it is safe to ingest the food ( Tapper and Halpern , 1968 ) . Sucrose is the prototypical highly palatable tastant for sweet taste quality , and it provides a sensory cue predicting the presence of immediate energy sources . Although palatability and intensity usually change together , Wang et al . , 2018 found this is not always the case and suggested that they are two distinct representations . In rodents , palatability is measured by an increase in positive oromotor responses ( e . g . , licking ) elicited by increasing sucrose concentrations ( Spector and Smith , 1984 ) . In contrast , the intensity attribute cannot be directly measured by any licking response per se , as an animal must actively report the perceived concentration of sucrose , a process necessarily involving decision-making . Historically , the neural representation of sweet taste intensity has been characterized by firing rates ( spike counts ) that monotonically increase with sucrose concentration along the gustatory pathway from the periphery to primary ( IC ) and secondary ( OFC ) taste cortices ( Rolls et al . , 1990; Roussin et al . , 2012; Scott et al . , 1991; Thorpe et al . , 1983; Villavicencio et al . , 2018 ) . However , those responses were obtained in either anesthetized animals ( Barretto et al . , 2015; Wu et al . , 2015 ) , during passive intraoral delivery of tastants ( Maier and Katz , 2013; Scott et al . , 1991 ) , or in behavioral tasks where animals do not have to make any decision other than to lick ( Rosen and Di Lorenzo , 2012; Stapleton et al . , 2006; Villavicencio et al . , 2018 ) . Thus , the neural representation of the perceived intensity of sucrose that the animal actively reports has not presently been studied . Likewise , how this representation is transformed into perceptual decision-variables , such as choice , movement direction , and the presence or absence of reward remains to be elucidated . Here we trained rats in a sucrose intensity discrimination task and recorded electrophysiological responses in the posterior ( pIC ) , anterior ( aIC ) insular cortices , and the orbitofrontal cortex ( OFC ) , with the aim of elucidating how these cortices encode sucrose intensity and use this information to guide behavior . These three cortical areas are multimodal and chemosensitive and are involved in disgust ( pIC ) , tastant identification ( aIC ) , and subjective value and reward ( OFC ) ( Frank et al . , 2013; Gardner and Fontanini , 2014; Jezzini et al . , 2013; Jones et al . , 2006; Katz et al . , 2001; Kusumoto-Yoshida et al . , 2015; Maffei et al . , 2012; Maier and Katz , 2013; Verhagen et al . , 2004 ) . In rodents , the pIC has been shown to be involved in taste , disgust , expectancy , and aversive motivated behaviors ( Bermúdez-Rattoni , 2004; Chen et al . , 2011; Fletcher et al . , 2017; Gardner and Fontanini , 2014; Gutierrez et al . , 2010; Kusumoto-Yoshida et al . , 2015; Wang et al . , 2018 ) . In contrast , the aIC is involved in appetitive behaviors , and besides having neurons that respond selectively to sweet taste ( Chen et al . , 2011 ) , it also has neurons encoding reward probability and reward omission ( Jo and Jung , 2016 ) . Even though both pIC and aIC have roles in taste and decision-making , their contribution to sucrose intensity guided behavior remains unexplored . It is well known that OFC is involved in reward and subjective value ( Conen and Padoa-Schioppa , 2015; Jo and Jung , 2016; Kennerley and Wallis , 2009; Roesch et al . , 2006 ) , and it is a critical brain region for encoding decision-variables such as choice , movement direction , and reward omission ( Feierstein et al . , 2006; Hirokawa et al . , 2017; MacDonald et al . , 2009; Nogueira et al . , 2017 ) . However , it is not known whether OFC neurons encode decision-variables guided by sucrose intensity . Equally unknown is how these variables are encoded along the posterior-anterior axis of the Insula . To address these questions , we designed a novel sweet intensity discrimination task in which , to obtain a water reward , rats had to make a rightward or leftward movement based on the perceived intensity of sucrose ( Cue ) , while single-unit recordings in the pIC ( 1348 ) , or aIC ( 1169 ) , or OFC ( 1010 ) were performed . We found that stimulation with sucrose evoked a widespread response in these three cortical regions , indicating a distributed detection of taste/somatosensory information . 82% of the evoked responses showed no selectivity to sucrose intensity , whereas 18% could be labeled as sucrose intensity-selective . These selective neurons conveyed the most information about sucrose's sweet intensity . Analyses of the sucrose-evoked responses revealed that , in addition to firing rates , the spike timing of neurons contains additional information about sucrose’s intensity . Several differences and similarities were identified between the evoked pIC , aIC , and OFC responses . Overall , the three recorded areas similarly decoded sucrose concentration and equally tracked the outcome ( reward delivery or omission ) . A major difference among them was that the OFC neurons carry information about behavioral choice and movement direction , earlier and with higher quality than neurons in the Insula . In summary , these data show that the perceived intensity of sucrose is fully reconstructed from the firing rate and spike timing of a small population of neurons in the pIC , aIC , and OFC .
Twenty-eight rats were trained in a one-drop sucrose intensity discrimination task . The trial structure is depicted in Figure 1A . Briefly , trained rats initiate a trial by first visiting the central port ( Return ) . Licking at the central spout triggers the delivery of either 10 µL of 3 ( Low ) or 18 ( High ) wt% sucrose ( referred to as Cue-D; Stimulus ) . If a rat chooses correctly ( by moving to one of the two lateral ports; Response ) , three drops of water are delivered as a reward ( Outcome ) . Error trials were unrewarded . Subjects achieved the learning criterion ( ≥80% correct ) in about 25 sessions ( Figure 1B ) , and the implantation of an electrode array in one of the three cortical areas did not impair task performance ( paired t-test before vs . after surgery; t ( 27 ) = 0 . 95; p=0 . 35; Figure 1B ) . Once the animals learned the discrimination task , they were tested in a variant named generalization session ( Figure 1C ) . In these sessions that consisted of 20% of the trials , rats were required to classify 0 , 3 , 4 . 75 , 7 . 5 , 11 . 75 , or 18 wt% sucrose as either ‘Low’ or ‘High’ ( referred as Cue-G ) . In these trials , no reward was delivered ( Reward omission ) to avoid imposing an arbitrary Low/High threshold that could bias the behavioral report of the perceived sweetness intensity . In Cue-G trials the percentage of ‘High’ responses increased with increasing sucrose concentration ( Figure 1D ) , thus showing that the animals used sucrose intensity as a cue to solve the task ( since its quality is unchanged ( Pfaffmann et al . , 1979 ) ) . Surgery did not impair perceptual judgments based on sucrose intensity ( see Before vs . After surgery; Figure 1D ) . Other behavioral measurements , related to palatability ( Perez et al . , 2013; Spector et al . , 1998 ) , revealed that the latency to stop licking after High Cue-D delivery ( 18% sucrose ) was longer ( 0 . 74 ± 0 . 02 s ) than for the Low Cue-D ( 3% sucrose; 0 . 58 ± 0 . 01 s; p < 0 . 0001 ) . A similar trend was observed for generalization cues ( i . e . , Cue-G trials ) , in that rats exhibited a longer time to stop licking in trials where sucrose intensities were ≥to 4 . 75% relative to Low Cue-D ( One-way ANOVA: F ( 7 , 1360 ) = 17 . 10; p < 0 . 0001; Dunnett post-hoc; Figure 1E ) . Furthermore , we analyzed the relationship between licking and task performance and found that rats lick more rhythmically and similarly for both cues in sessions where their performance was better . This is reflected by a positive correlation between Low and High licking PSTHs and task performance ( r = 0 . 17 , p < 0 . 003; see Figure 1—figure supplement 1 ) . Thus , rats did not solve the task by licking differently for both cues . In the Return epoch , rightward movements ( left to center port direction ) were faster than leftward ( right to center port ) movements ( Figure 1F ) . In contrast , during the Response epoch , leftward or rightward movements were not significantly different ( Figure 1G ) , and therefore these movements were independent of the sucrose concentration . Interestingly , rats moved faster in the Response than in the Return period , perhaps a result of the water reward ( compare Figure 1F vs . Figure 1G ) . Finally , in the Outcome epoch , rats rapidly detected when the reward was omitted ( Cue-G trials and Cue-D error trials ) . That is , they spent more time licking when water was delivered than when it was omitted ( Figure 1H; see Supplementary file 1 for statistics ) . In sum , by using only a 10 µL drop of sensory stimulation , rats can make accurate perceptual decisions based on the perceived concentration of sucrose . A total of 1348 , 1169 , and 1010 single-units were recorded from pIC , aIC , and OFC , respectively ( see Figure 3—figure supplement 1A ) . Of these neuronal responses 480 , 403 , and 337 , respectively were recorded in generalization sessions and the rest in discrimination sessions ( with only cue-D trials ) . Recordings were performed unilaterally in the left hemisphere . Schematics and location of the recording sites are seen in Figure 1—figure supplement 2 . The temporal activation pattern of the neural responses in pIC , aIC , and OFC was classified as a function of the evoked response ( Cue-evoked or Non-evoked ) , modulation profile ( Phasic , Tonic , or Coherent; see Table 1 ) , and selectivity ( either Non-selective or Intensity-selective; see Table 2 ) . Most recorded neurons exhibited a statistically significant evoked response 90 . 6% ( 1221/1348 ) , 97 . 4% ( 1139/1169 ) , and 92 . 8% ( 937/1010 ) for the pIC , aIC , and OFC , respectively . The remaining neurons , named Non-evoked , were 9 . 4% ( 127/1348 ) , 2 . 6% ( 30/1169 ) , and 7 . 2% ( 73/1010 ) , respectively . Cue-evoked responses were then further classified according to five characteristic modulation profiles: Phasic , Tonic-Inactive ( Inact ) , Tonic-Active ( Act ) , Lick-coherent Inactive ( Coh-Inact ) , and Active ( Coh-Act ) ( Table 1 ) . Given that rats could use a drop of sucrose to make accurate perceptual decisions based on its intensity ( Figure 1 ) , we explored the neural correlates of these decisions in the pIC , aIC , and OFC . Figure 2 depicts the raster plots and corresponding peri-stimulus time histograms ( PSTHs ) of representative examples of Intensity-selective Cue-D evoked responses recorded in each of the three cortical regions . Examples of Non-selective Cue-D responses are shown in Figure 2—figure supplement 1 . Action potentials are depicted as black ticks and were aligned to Cue-D delivery ( time = 0 s ) . Trials were sorted as a function of Low ( 3% -green ) and High ( 18 wt% -red ) . The left column shows three different neurons that exhibited selective phasic responses to 3 wt% sucrose in the pIC , aIC , and OFC , respectively . Examples of the Tonic-Inactive ( second column ) revealed a selective inhibition for 3 wt% sucrose ( Low-preferred ) . After Cue-D delivery , the Tonic-Active neurons exhibited a sustained increase in firing rate ( third column; the upper and lower panels depict a Low-preferred neuron , whereas the middle panel a High-preferred response ) . The last two columns on the right-hand side display examples of neurons that fired synchronously with licking ( named Lick-coherent ) and , after Cue-D delivery , exhibited either a decrease ( Coh-Inact ) or an increase ( Coh-Act ) in their firing rate . Intensity-selective neurons were recorded in all three cortical regions and for all five classes of evoked responses , although with different proportions ( see Figure 3—figure supplement 1A and Table 2; see Supplementary file 2 for statistics ) . In general , pIC and aIC Intensity-selective neurons exhibited more similar responses between them than those found in the OFC ( see Table 2 ) . The only exception was that the aIC contained more Intensity-selective neurons with Tonic-Active and Coh-Act responses than the pIC . In contrast , the OFC had more Intensity-selective neurons exhibiting Tonic-Inactive and Active responses ( Table 2 ) . Overall , the percentage of Intensity-selective neurons were 14 . 8% , 17 . 9% , and 18 . 5% , in the pIC , aIC , and OFC , respectively ( see Table 2; Total , Inten-Sel ) . These data show that Intensity-selective neurons are found along the posterior-anterior taste neuroaxis . To determine , in fine-grain detail , the differences in licking and its impact upon neuronal responses , in Figure 2 , we also depicted the corresponding PSTHs of licking behavior and the times where the lick rate was significantly different between Low and High cues ( see dashed lines ) . We found that 45 . 1% of all Intensity-selective neurons have a ‘best-window’ ( interval with maximal discrimination between concentrations ) with no differences in licking ( see Figure 2 grey-line above the PSTHs ) . The remaining 54 . 9% of neurons have a lick rate difference inside the best-window , but most frequently they only covered a small fraction of the window ( Figure 2—figure supplement 2 ) . Specifically , the overlap of the lick rate differences covered 31 . 4% of the entire best-window ( Figure 2—figure supplement 2 ) . Thus , we conclude that is unlikely that most sucrose intensity representation can be attributed to differences in licking behavior . Figure 3A shows the color-coded population PSTH of the responses of all Intensity-selective neurons in each brain region , sorted as a function of the modulation profile and preferred concentration . What is clear in the figure is that diverse temporal patterns are evoked in response to the delivery of the Cue-D ( time = 0 s ) . The evoked responses can be transient , sustained , or oscillatory , with either increasing or decreasing firing rates . In the Stimulus epoch , the population responses revealed that the pIC and aIC were more excited , whereas the OFC was inhibited ( Figure 3—figure supplement 1B ) , suggesting an opposite interaction between the Insula and OFC during licking behavior . In agreement with the idea that in a default brain-network state , these two brain regions function out of phase ( Gutierrez-Barragan et al . , 2018 ) . In line with previous studies ( de Araujo et al . , 2006; Gutierrez et al . , 2010 ) , we found among these taste cortices that the pIC ( 60 . 3% ) and aIC ( 59 . 5% ) had a higher proportion of ( either increasing or decreasing ) lick-induced oscillatory responses than the OFC ( 27 . 6% ) . Likewise , we found that the coherence values of the OFC ( 0 . 24 ± 0 . 005 ) were significantly lower relative to pIC ( 0 . 26 ± 0 . 003 ) and aIC ( 0 . 26 ± 0 . 003 ) ( F ( 2 , 1672 ) =3 . 77; p = 0 . 02 ) ( Figure 3—figure supplement 3A ) . Therefore , the pIC and aIC had not only a higher proportion of Lick-coherent neurons than OFC , but also IC neurons were better entrained with rhythmic licking . More importantly , we also uncovered , for the first time , that the level of coherence was significantly higher in the Stimulus-epoch in comparison with the pre-Stimulus and the Outcome epochs ( all p’s < 0 . 0001 ) , suggesting that lick-spike coherence reflects more than oromotor responses , perhaps it prepares taste cortices to receive sensory inputs . For the neuronal populations of each brain region , a linear decoder was used to estimate the accuracy for discriminating Low and High sucrose trials ( Meyers , 2013; see Materials and methods ) . As seen in Figure 3B , the contribution of the Non-evoked responses ( grey bars ) was found to be at chance level ( 50% ) , indicating that they contained little , if any , information about sucrose intensity . In contrast , all Cue-D evoked neuronal responses ( All- black bars ) significantly decoded sucrose concentrations above chance level ( Figure 3B ) . Importantly , we found that the small population of Intensity-selective neurons ( red bars ) contained more information than the larger Non-selective population ( blue bars ) . Interestingly , the Non-selective group also decoded sucrose intensity significantly above chance level . One possibility is that they have subtle differences in firing rates that are not consistent enough across trials to produce a significant effect at single neuronal level . However , at the population level there is sufficient information about sucrose intensity . Alternatively and despite their similar firing rates ( spike counts ) evoked by Low and High Cue-D , these neurons could use spike timing to discriminate sucrose concentrations ( Gutierrez and Simon , 2013 ) . To test this hypothesis , the spikes of all Non-selective neurons were shuffled without changing their average firing rates . When the spike timing information was eliminated from these neuronal responses , their ability to decode among the sucrose’s intensities dropped to chance level ( Figure 3B; see the horizontal white lines across the blue bars ) . Thus , the additional information in the Non-selective population was likely conveyed by precise spike timing patterns of activity . The decoding algorithm also revealed that the Intensity-selective neurons in the three cortical regions decoded sucrose intensity better than Non-selective neurons ( Figure 3B; red bars ) . It is unlikely that these results were due to the differences in the population size since Intensity-selective neurons were always fewer in number than the Non-selective and All Cue-evoked neurons . Thus , the Intensity-selective population ( i . e . , less than 18% of neurons ) contained more information about sucrose intensity than the entire population . These data suggest the existence of a neuronal representation of sucrose concentration across these three gustatory cortical regions . That is , each taste cortex seems to contain a copy of this information . Finally , we note that by removing the spike timing information contained in the Intensity-selective neurons , their percent decoding dropped to nearly chance level , indicating that the neural representation of sucrose intensity is also conveyed in the spike timing of neurons . It has been reported that spike counts in a pair of simultaneously recorded neurons , elicited by a stimulus , can covary across the session , a phenomenon denominated as noise-correlation and these correlations are thought to covary with attentional , behavioral , and overall brain-state of the network ( Averbeck et al . , 2006 ) . Although the function of noise-correlations is not completely understood , it is well known that they could affect ( either increase or decrease ) population decoding ( Averbeck et al . , 2006; Averbeck and Lee , 2006; Carnevale et al . , 2013; Cohen and Kohn , 2011; Zohary et al . , 1994 ) . For this reason , we also determined the impact of removing the noise-correlations on the decoding accuracy of sucrose intensity . We found that pIC ( 0 . 21 ± 0 . 005 ) had a significantly higher noise-correlations in comparison to aIC ( 0 . 19 ± 0 . 006 ) and OFC ( 0 . 19 ± 0 . 007 ) ( F ( 2 , 386 ) = 7 . 85; p = 0 . 0005; Figure 3—figure supplement 2A ) . Nevertheless , removing noise-correlations by shuffling trials ( Figure 3—figure supplement 2B ) did not significantly affect decoding accuracy in any population or recorded brain region ( Figure 3B see the grey horizontal lines ) . Therefore , at least in these experiments , noise-correlations do not have a significant effect over decoding of sucrose intensity . We next determined which class of Cue-evoked responses contained sufficient information to decode sucrose’s intensity . To achieve this , we ran the neural classifier using a single neuronal population . In all three regions , the Coherent-Inactive and Coherent-Active had better percentage decoding accuracy than the Non-evoked control group ( Table 3 ) . Moreover , in the pIC , aIC , and OFC combining All-Coherent neurons ( All-Coh ) achieved the best sucrose decoding nearly matching that of the entire population ( All ) . Thus , Lick-coherent populations contained sufficient information in their responses to decode sucrose intensity . Finally , to further determine which population contained information necessary to decode sucrose intensity , we performed a dropped population analysis ( Gutierrez et al . , 2006 ) . In this analysis , only one population at a time was removed , and its decoding accuracy was compared against the decoding achieved by All the Cue-evoked populations combined ( Table 4; referred as ‘All’ ) . In the three cortical regions , the percent decoding accuracy was significantly reduced only when the two Lick-coherent groups were dropped from the entire population ( compare All-Coh vs . ‘All;’ Table 4 ) . In sum , both analyses suggest that the neural responses of the Lick-coherent neurons were both sufficient and necessary to decode sucrose intensity information . Having described the modulation profile evoked by two sucrose concentrations , we next characterized whether neurons in the three recorded cortices encode sucrose’s concentration-dependent information and decision-variables . We first describe neuronal responses that encode information about sucrose’s concentrations . This is followed by neuronal responses that correlate with animal choices ( ‘Choice neurons’ ) , Direction ( neurons with responses selective to either leftward or rightward movements ) , and finally , neurons that track the Outcome ( responses that indicate the presence or absence of reward ) . We also discuss the overlapping among these populations . To determine if there was a neuronal subpopulation that tracked sucrose concentrations among the Intensity-selective ( Cue-D ) neurons , we evaluated neural responses during generalization trials ( Cue-G: 3 , 4 . 75 , 7 . 5 , 11 . 75 , or 18% sucrose ) . In these sessions , we recorded a subpopulation of 480 , 403 , and 337 neurons from the pIC , aIC , and OFC , respectively . Similar to the Cue-D sessions , in the generalization sessions , we found that 94 . 1 ± 1 . 3% could be classified as Cue-evoked neurons and that from these , 83 . 3 ± 1 . 5% were Non-selective and 16 . 7 ± 1 . 5% were Intensity-selective . From the Intensity-selective population , the proportion that tracked the sucrose concentration ( either positively or negatively ) was 28 . 8% ( 19/66 ) , 36 . 1% ( 26/72 ) , and 32 . 3% ( 20/62 ) in the pIC , aIC , and OFC , respectively . Figure 4A shows raster plots and PSTHs of three representative neurons recorded in the pIC , aIC , and OFC whose responses increased with sucrose’s concentrations . That is , during the ‘best window’ ( cyan-shaded rectangle ) , these neurons responded with increasing activity to increasing sucrose concentrations ( see Insets ) . We also identified neurons with an activity that negatively correlated with sucrose concentrations ( Figure 4B ) . The population activity of all neurons with increasing ( red ) or decreasing ( blue ) responses was similar across the three cortical regions ( Figure 4C ) . Thus , all three of these cortical areas have neurons that track the sucrose concentration . Having demonstrated how sensory information about the sucrose concentration is encoded , we next identified neurons whose activity correlated with the animal’s behavioral choices . For this , in the Stimulus and Response epochs , we calculated the correlation between the neuronal activity and the perceptual intensity choices made by the animals on a trial-by-trial basis . To quantify the extent to which the neuronal responses could be underlying the behavioral decisions we compared psychometric and neurometric generalization curves ( see Materials and methods and de Lafuente and Romo , 2005 ) . Initially , we aligned the responses to the onset of the Stimulus epoch , but no significant neuronal responses were detected in this epoch ( data not shown ) . In contrast , Choice-related responses were found in the Response epoch in the aIC and the OFC . Since in the pIC only two Choice neurons were detected ( Figure 5—figure supplement 1B ) , no conclusions were drawn for this cortical area . The left panel of Figure 5A shows the PSTHs of a ‘Low-preferred’ aIC choice neuron ( left panel ) whose activity decreased with increasing sucrose concentrations . The right panel shows a ‘High-preferred’ OFC neuron that exhibits higher firing rates for trials ≥ 4 . 75% sucrose and that fired less for ≤3% sucrose ( Figure 5A , right panel ) . The cyan-shaded rectangle in the PSTHs depicts the window where neural responses best-predicted animal’s choices . It is seen that especially in OFC , the resulting neurometric function followed the psychometric function ( Figure 5A , insets ) . The averaged neurometric function of all 8 aIC ( of 403; 2 . 0% ) and 18 OFC ( of 337; 5 . 3% ) Choice-related neurons that covaried significantly with the behavioral psychometric function are shown in the left and right panel of Figure 5B , respectively . However , only in the OFC were the confidence intervals of the slopes overlapped , indicating that neuronal responses in this area better followed the behavioral choices , in comparison to the aIC . To determine the temporal dynamics of choice selectivity ( Low vs . High ) , we plot a ROC index across the Response epoch ( Figure 5C ) . We observed that aIC neurons encoded the choice only when the rat is responding ( time >0 s ) , while OFC Choice neurons discriminated between sucrose concentrations before subjects started to communicate their choice . That is , OFC neurons encoded the subject’s choice while the animals were still licking in the central port ( time <0 s ) . In sum , OFC neurons carry information about sucrose intensity judgment earlier than aIC neurons . Instead of using sucrose intensity , we note that the animals could be using palatability to generate their behavioral responses . To investigate this possibility , we used the sucrose-evoked lick rate to construct a palatability generalization function that we then compared with the psychometric generalization function . The results show that licking responses could be used to predict behavioral responses in only 1 out of 171 sessions . Moreover , no Choice-related neurons were recorded from this session ( data not shown ) . Therefore , we consider it is unlikely that rats guided their choices based on oromotor sucrose -evoked palatability responses but rather favor the idea that rats make decisions based on sucrose’s intensity . Previous studies have demonstrated the existence of neurons in the OFC that encoded information about movement direction ( Feierstein et al . , 2006; MacDonald et al . , 2009; Roesch et al . , 2006 ) . To both confirm and extend those studies we determined if there was a similar movement-direction coding in the pIC , aIC , and OFC . This was accomplished by employing a Receiver Operating Characteristic ( ROC ) curve ( Green and Swets , 1966 ) which determined how distinguishable were the firing rate distributions of two events ( i . e . , leftward vs . rightward movement ) . The area under the ROC curve was scaled from −1 to +1 , providing a Preference Index ( Pindex ) , where −1 means a complete preference for leftward direction , +1 a complete selectivity toward the rightward direction , and 0 indicates no preference . Then , using the firing rates , we computed Pindex’s for the Return and Response epochs . In the Response epoch , rats moved from the center port to a lateral port ( left/right ) whereas in the Return epoch go from a lateral port to central port ( left/right; see schematics Figures 1A and 6A ) . Thus , both epochs shared a similar , leftward or rightward , movement direction . We reasoned that Direction-selective neurons should fire for movements sharing a direction , but that may occur at different spatial locations . In the three cortices studied , we identified neurons that exhibited direction selectivity . Figure 6A shows three neuronal responses that exhibited either a Leftward-preferred selectivity ( upper and lower panel ) or a Rightward-preferred selectivity ( middle panel ) . That is , the Leftward-preferred responses increased for leftward movements and did not respond to rightward movements ( Figure 6A , cyan PSTHs ) . The middle panel shows a neuronal response from the aIC that fired better for a Rightward movement . The three panels in Figure 6B show , for all Direction-selective neurons in the three areas , the scatter plot of the Return’s Pindex relative to the Response’s Pindex . The black arrows indicate the Pindexes for the three representative neurons shown in Figure 6A . Note that Pindex values closer to the diagonal denote similar direction selectivity for the Return and Response epochs . We also found that Direction-selective neurons displayed similar responses for Correct and Error trials ( see raster plots ) , supporting the notion that movement direction was the primary feature modulating their firing rates . We note that the OFC had the best representation of direction selectivity tuning ( Response’s Pindex; One-way ANOVA: F ( 2 , 462 ) = 6 . 1; p < 0 . 0001 ) . A Bonferroni post hoc confirmed that OFC had a better representation of direction selectivity in comparison to pIC and aIC ( p < 0 . 05 and p < 0 . 01 , respectively ) . Another more complete example of direction selectivity can be seen in its population activity in both task epochs ( Figure 6C; also see the magnitude of Z-scores ) , with OFC yielding the greatest differences . To this point , a higher proportion of Direction-selective neurons was found in OFC ( 19 . 1% ) in comparison to pIC and aIC ( 10 . 8% , χ2 = 23 . 85 , p < 0 . 0001% and 10 . 8% , χ2 = 22 . 32 , p < 0 . 0001 , respectively; Figure 6B ) . Note that most of these neurons were Right-Selective neurons ( Insets Figure 6B ) perhaps because we recorded unilaterally in the left hemisphere . Overall , these data reveal that more OFC neurons tracked movement direction in comparison to both areas of the IC . Once the subjects are in the lateral goal-port , they would or would not receive water according to their choice ( Correct or Error ) and trial type ( discrimination ( Cue-D; water ) or generalization ( Cue-G; no water ) ) . Recall that in Cue-D trials , reward delivery depends upon task performance whereas , in Cue-G trials , no reward was delivered regardless of choice . Thus , for Cue-G trials of 3 and 18 wt% sucrose , rats could not predict if the reward would be delivered or omitted . Therefore , by analyzing all rewarded vs . unrewarded trials ( regardless of choice ) , we could disambiguate whether neurons tracked the outcome . In this regard , we identified a subpopulation of neurons that selectively fired for reward omission vs . reward delivery . Figure 7A displays the raster plots and PSTHs of three representative neurons . The pIC and OFC neurons did not respond to reward omission ( RWO-named the Inactive population ) , but they fired to rewarded trials ( RW- see dashed PSTHs ) . In contrast , the aIC neuron fired after reward omission ( named Active population ) , while no responses were observed during reward delivery ( Figure 7A , middle panel ) . Note that the pIC ( 57 . 1% vs . 17 . 9% ) and the aIC ( 45 . 7% vs . 28 . 3% ) had a higher proportion of neurons with Inactive than Active responses after reward omission ( χ2 = 72 . 88 , p<0 . 0001 and χ2 = 12 . 05 , p = 0 . 0005; respectively ) ; while the OFC the proportion was similar ( 36 . 5% vs . 42 . 4%; χ2 = 1 . 1 , n . s . ) , suggesting that pIC and aIC exhibited a bias toward having more neurons with Inactive responses after reward omission relative to OFC neurons . The population responses of both Inactive and Active Reward Omission neurons are seen in Figure 7B . Figure 7C depicts the lick rates during reward delivery ( dashed line ) and omission ( solid line ) . Note that the rats rapidly detected reward omission since they stopped licking faster when water was omitted ( the arrows indicate the second rewarded lick after delivery of the first water reward -time = 0 s- for RW trials; also see Figure 1H ) . Finally , to quantify , the temporal dynamics of behavioral and neural decoding of the outcome , we ran a population decoder analysis . The data presented in Figure 7D revealed that pIC , aIC , and OFC contain neurons that detect and provide more information about reward omission than licking behavior itself . That is , the decoding accuracy was better when the algorithm used spiking activity ( blue line ) instead of the licking rates ( black line ) . Our results suggest that all three of these cortical taste regions are highly sensitive to both reward delivery and reward omission . Given that neurons encoding sensory and decision-variables were detected in different task epochs , we tested if there were any overlapping populations . This was accomplished using a Fisher’s exact test to determine if the proportion of neurons that belong to two coding categories was above-expected chance levels . Figure 8A depicts a contingency table of the pairwise comparison of each coding profile category . For example , the left and middle quadrants indicates the number of neurons that encodes both Sensory and Direction ( the parenthesis indicates the corresponding percentage of overlapping ) . Also , since in the discrimination task the Low and High Cue-D were also associated with a left/right movement , it is possible that some Sensory responses recorded in the Stimulus epoch , besides discriminating sucrose intensity , could jointly encode movement direction ( in the Response epoch ) . We reasoned that if this were the case , then most Sensory neurons will also belong to the Direction population . This overlap was significant only in pIC ( Figure 8A , upper panel; in 10% of the neurons ) . Moreover , the same result was found when we used all the Intensity-selective neurons to compute the contingency matrix ( data not shown ) . Thus , it is unlikely that most Sensory neurons ( or Intensity-selective neurons ) jointly encoded sucrose intensity and movement direction . Instead , the results suggest that Sensory neurons play a more circumscribed role in chemosensory sucrose intensity processing . Other observations revealed that the OFC Direction population was significantly associated with Choice and Outcome ( Figure 8A , lower panel ) , suggesting that OFC neurons are capable of carrying , at multiple time periods , more than one spatiomotor variable related to performing the discrimination task . We also explored if neurons encoding decision-variables ( coding profile ) tend to exhibit a specific modulation profile ( i . e . , Phasic , Tonic , Coherent ) . It is important to note that all modulation profiles could , in principle , encode almost any of the sensory and decision-variables . However , only a few subpopulations exhibited a significant overlap . In general , no systematic overlapping pattern was shared across the three cortical regions , suggesting that by knowing the modulation profile of one neuron provides little , if any , information about what kind of decision-variables it might encode . That said , we observed that Lick-coherent neurons in the pIC and aIC had a higher likelihood of encoding decision-variables , except for Choice neurons in the aIC which were non-preferentially encoded by any modulation profile ( Figure 8B , upper and middle panels ) . In contrast , Tonic-Active neurons in the OFC jointly encoded Choice , Direction , and Outcome variables ( Figure 8B , lower panel ) . In sum , these data suggest a prominent role of the Lick-coherent neurons in encoding critical features of the task in the Insula , whereas in the OFC the tonic activity prevails in encoding decision-variables .
The neural representation of sweet taste intensity has been usually characterized by firing rates that monotonically ( or sigmoidally ) increase with sucrose concentration along the gustatory axis from the periphery to taste cortices ( Barretto et al . , 2015; Rolls et al . , 1990; Roussin et al . , 2012; Scott et al . , 1991; Thorpe et al . , 1983; Villavicencio et al . , 2018; Wu et al . , 2015 ) . These experiments usually have been performed in animals that do not have to make any other decision than to lick ( Stapleton et al . , 2006; Villavicencio et al . , 2018 ) or have the tastant passively delivered ( Katz et al . , 2002 ) . However , as noted , the intensity attribute of a tastant can only be measured in behaving animals that actively report the perceived concentrations of sucrose . To address this issue , we developed a sucrose intensity discrimination task ( see Figure 1 ) while recording from three cortical taste areas . We found that ~95% of recorded neurons were responsive to a single drop of sucrose ( Cue-D ) , but the majority of them were unable to distinguish between 3 and 18 wt% sucrose . We posit that such a massive number of responsive neurons , which includes Intensity-selective and Non-selective neurons ( see Figure 3—figure supplement 1 ) , could be the result of the arrival of a salient cue ( Cue-D ) to an over trained animal . This stimulus is highly relevant in the context of thirsty subjects whose internal state would motivate them to attend to the delivery of an stimuli whose accurate detection and identification will lead to obtaining water . These findings are in agreement with observations that a state of physiological need ( e . g . , hunger ) gates insular cortex responses to food cues ( Livneh et al . , 2017 ) and that caudolateral OFC neurons are sensitive to hunger ( Rolls et al . , 1989 ) . Likewise , in head-fixed trained mice , it was recently reported that odor stimulation also triggers a massive widespread cortical activation in mice performing a Go/No-Go goal-directed behavior ( Allen et al . , 2017 ) . Multiple pieces of evidence support the idea that a small population of cortical neurons could represent sucrose intensity . For example , electrophysiological studies in rodents and non-human primates have reported a low proportion ( ranging between 2–35% ) of Insular ( pIC ) and OFC neurons with selective responses to at least one taste quality ( Pritchard et al . , 2005; Rolls et al . , 1990; Rolls et al . , 1989; Scott et al . , 1991; Stapleton et al . , 2006; Thorpe et al . , 1983; Yamamoto et al . , 1989; Yaxley et al . , 1990 ) . In contrast , one recent study in anesthetized mice , using a calcium sensor ( GcAMP6 ) , reported almost 90% of taste responses in the pIC were tastant selective although only 26% were sucrose-best ( Fletcher et al . , 2017 ) . The differences in proportions could be explained by different data analysis and experimental preparations employed ( Chen et al . , 2011; Katz et al . , 2001 ) . Unfortunately , only a few studies have reported the proportion of cortical neurons tracking taste intensity . In this regard , in the monkey Insula cortex , Scott et al . ( 1991 ) found that less than 1 . 5% ( 24/1661 ) of the recorded neurons responded linearly to increasing glucose concentrations . To the best of our knowledge , our data is the first demonstration that a small subset of cortical neurons represents sucrose intensity better than the entire population . Although intensity-selective neurons comprise a ‘small’ population relative to Non-selective neurons ( Figure 3B and Table 2 ) , we note that 18% of neurons in a rat’s cortex would represent a large population . Further studies should investigate whether Intensity-selective neurons are sucrose-selective or broadly tuned ( Erickson , 2001 ) . In this regard , it is interesting that Intensity-selective neurons were present in the recorded three cortical regions . This result is somewhat surprising in light of the experiments in the mouse Insula cortex showing the existence of non-overlapping posterior ( aversive-pIC ) and anterior ( appetitive-aIC ) ‘hotspots’ ( Chen et al . , 2011 ) . Gain and loss of function experiments in mice have demonstrated that the ‘sweet hotspot’ is sufficient and necessary for sweet taste recognition ( Peng et al . , 2015 ) , and a similar topographic separation of disgust-appetitive in monkeys’ anterior Insula has been found , although in a different anatomical axis ( dorsal–appetitive and ventral-aversive ( Jezzini et al . , 2012 ) ) . One explanation for the distributed responses that we observed ( Figures 2 and 3 ) is that sucrose’s identity is initially encoded in the sweet hotspot ( located in aIC ) , but information about its perceived intensity is then distributed to other areas . Further experiments should involve the inactivation of one or more of these areas . Nevertheless , we found that each recorded cortical region decoded sucrose intensity equally well , including the pIC which according to ( Chen et al . , 2011 ) is where the ‘aversive hotspot’ is located . Thus , whatever the explanation , each of these three cortical regions contains information about sucrose intensity , revealing the distributed nature of taste intensity coding . However , the fact that all three areas decoded sucrose intensity equally well does not imply that they represent the same information , but rather they may encode different features of the sucrose intensity cues . In this regard , and despite that stimulation of the pIC elicits aversive behavioral responses ( Peng et al . , 2015 ) , we posit that the pIC should also plays a general role in gustation since it receives most inputs from the gustatory thalamus ( Cechetto and Saper , 1987 ) , which could rationalize why there is sucrose responses in this cortical area ( Fletcher et al . , 2017 ) . The aIC responses might be related to encoding the sweet percept , due to the ‘sweet’ hotspot , mentioned above , whose activation leads to appetitive behaviors ( Chen et al . , 2011; Peng et al . , 2015; Wang et al . , 2018 ) . Finally , the OFC responses during the Stimulus epoch could also signal the relative reward value of Low and High sucrose cues ( Rolls et al . , 1990; Tremblay and Schultz , 1999 ) . Our data suggest that the perceived intensity of sucrose is spatially distributed along taste cortices with a compact and distributed neural code , in the sense that a small subset of spatially disperse neurons contain more information , about sucrose intensity , than the entire population ( Field , 1994; Olshausen and Field , 2004; Stüttgen et al . , 2011 ) . The contribution of spike timing and spike count in taste identity coding has been extensively studied by Di Lorenzo and colleagues ( Di Lorenzo et al . , 2009; Di Lorenzo and Victor , 2003; Roussin et al . , 2012 ) . However , less is known about its contribution to the encoding of sweet intensity . In this regard , we found that additional information about sucrose‘s intensity was conveyed in the spike timing of neurons ( Figure 3B ) . A recent study in the olfactory system reported that piriform cortex neurons encode odor intensity by using only the spike timing , and not the spike count information ( Bolding and Franks , 2017 ) . Likewise , our results revealed that spike timing carries additional information about taste intensity . However , spike count is also a contributor since we found Sensory neurons that tracked the concentration of sucrose by increasing its firing rate ( Figure 4 ) . Thus , in the taste system , it seems that both spike count and spike timing information could be complementary codes for the perceived intensity of sucrose . Precise spike timing entrained by rhythmic licking serves as an internal clock , relevant for coordinating activity across brain regions ( Gutierrez et al . , 2010; Gutierrez et al . , 2006; Roussin et al . , 2012 ) . We found that the Lick-Coherent neurons were both sufficient and necessary to decode the perceived intensity of sucrose ( Tables 3 and 4 ) . More importantly , we also uncovered , for the first time , that the level of coherence was significantly higher in the Stimulus-epoch in comparison with the pre-Stimulus and the Outcome epochs ( Figure 3—figure supplement 3 ) . This result implies that lick-spike coherence not only reflects oromotor responses , but that it is also involved in gating the input of sensory and taste information that can be ‘read out’ across taste cortices in coordination with licking ( Buzsáki , 2010; Gutierrez et al . , 2010 ) . The best way to access the representation of the perceived intensity of sucrose is by allowing animals to make a decision about its intensity . Thus , it is important to determine the extent to which neuronal activity correlates with the animal’s behavioral choices . We found that a distinct subset of neurons exhibited Choice-related activity in aIC and OFC with responses that covaried with the subject’s choices ( Figure 5 ) . Furthermore , OFC ( but not aIC ) neurons tracked choice before a response was emitted . These findings are in agreement with behavioral observations which suggests the subjects have already made a decision before leaving the central port ( Perez et al . , 2013; Uchida and Mainen , 2003 ) . Our findings reveal a neural correlate of the perceived intensity of sucrose in the gustatory system . Spatial navigation is an essential behavior that allows organisms to explore the environment and direct their actions toward a goal ( Epstein et al . , 2017 ) . Spatial variables such as direction are essential to reach the desired outcome or to avoid punishment . Although spatial information is encoded in brain regions specialized for spatial processing , such as the hippocampus and entorhinal cortex , recently it has been found that other unexpected areas also contain spatial information ( Yin et al . , 2018 ) . In this regard , here we also found that OFC neurons robustly encoded movement direction . Likewise , neurons with direction selectivity in the OFC have been recorded in tasks involving two or four spatial locations ( Feierstein et al . , 2006; Lipton et al . , 1999; Roesch et al . , 2006 ) . Lesioning the OFC disrupts performance in an allocentric foraging task ( Corwin et al . , 1994 ) , and radial arm and Morris water maze ( Kolb et al . , 1983 ) . Moreover , the OFC also encodes head angle , spatial trajectory and movement speed in a spatial discrimination and reversal task in a plus maze ( Riceberg and Shapiro , 2017 ) . The latter evidence agrees with the high proportion of OFC Direction-selective neurons that we identified ( Figure 6 ) . In contrast , less is known about the participation of the Insular Cortex in encoding spatial navigation parameters; although , it is known that ablating either the pIC or the aIC results in a severe impairment of spatial navigation in a water maze ( Nerad et al . , 1996 ) . Here , for the first time , we found that neurons in the pIC and aIC tracked movement direction , probably through their connections with the entorhinal cortex ( Wang et al . , 2018 ) . However , according to the Pindex values , the encoding of direction was weaker in the IC in comparison to the OFC ( Figure 6 ) . Altogether , our data points to a dominant role for the OFC , and to a lesser extent the IC , in encoding movement direction; an essential feature of spatial navigation for goal-directed behaviors . The detection of either reward delivery or reward omission is essential for animals’ survival and for triggering learning based on reward prediction errors ( Schultz et al . , 1997 ) . Previous observations have shown that aIC and OFC neurons encode reward omission ( Feierstein et al . , 2006; Jo and Jung , 2016 ) and we found the pIC , aIC , and the OFC differentially respond to the presence and absence of reward ( Figure 7 ) ; suggesting a distributed tracking of reward omission . However , this is the first demonstration that pIC neurons could also encode reward omission . The pIC has a key role in updating the current outcome representation to guide action selection . This is because without affecting the execution of the instrumental responses its chemogenetic inhibition impairs the ability of subjects to adjust their actions based upon the outcome current value ( Parkes et al . , 2018; Parkes et al . , 2015 ) . Our results demonstrate a widespread representation of neural signals related to the Outcome , which is a crucial process for learning and adaptive behavior . As noted above , we identified several differences and similarities between the evoked pIC , aIC , and OFC responses . The main similarity among all three brain regions was that they decoded sucrose concentration equally well . A major difference was that OFC neurons carry information about decision-variables earlier and with higher quality than neurons in the Insula ( Figures 5–7 ) . That is , unlike the Insula , the OFC was the brain region with more neurons jointly encoding more than one decision-variable ( Choice , Direction , and Outcome; Figure 8A ) , indicating that the OFC has a complete representation of the most relevant task events . It follows that the OFC provides an up-to-date representation of task-related information that is required to yield the best outcome . In reinforcement learning , this information is named ‘state’ representation ( Schuck et al . , 2018; Stalnaker et al . , 2016; Sutton and Barton , 1998 ) . The OFC is also involved in encoding the subjective reward value of associated choices ( Conen and Padoa-Schioppa , 2015; Rolls , 2004; Tremblay and Schultz , 1999 ) . However , in our task correct actions ( choosing left/right ) led to the same reward ( i . e . , 3 drops of water ) , suggesting , in agreement with findings in an odor guided task ( Feierstein et al . , 2006 ) , that OFC neurons could encode spatiomotor variables , such as Choice and Movement direction , even for actions with the same reward value . Our results both confirm and extend these findings by further demonstrating that OFC neurons could represent decision-variables in a task guided by the intensity of sucrose . We posit that OFC may act as a hub that represents decision-variables regardless of the type of sensory input used to guide goal-directed behaviors . The OFC is a brain area well suited to perform this function since it receives connections from sensory areas related to olfactory , gustatory , visual , and somatosensory processing ( Cavada et al . , 2000 ) . We found evidence that in animals trained to identify sucrose intensity the taste system uses a compact and distributed code to represent its perceived intensity . Moreover , the perceived intensity of sucrose and the decision-variables associated with the discrimination task can be fully reconstructed from a small population of neurons in the pIC , aIC , and OFC .
Sucrose was reagent-grade chemical quality purchased from Sigma-Aldrich ( Mexico City , Mexico ) . It was dissolved in distilled water and used the following concentrations 3 , 4 . 75 , 7 . 5 , 11 . 75 , and 18 wt/vol% . Solutions were prepared fresh every other day . They were maintained under refrigeration , and they were used at room temperature . We used 28 male Sprague-Dawley rats weighing 300–320 g at the beginning of the experiment , and by the end of recordings , their weights were 412 . 3 ± 8 g . Animals were individually housed in standard laboratory cages in a temperature-controlled ( 22 ± 1°C ) room with a 12:12 h light-dark cycle ( lights were on 0700 and off at 1900 ) . All procedures were approved by the CINVESTAV Institutional Animal Care and Use Committee . During experiments , rats were given ad libitum access to tap water for 30 min after testing . Chow food ( PicoLab Rodent Diet 20 , St . Louis , MO , USA ) was always available in their homecage . All experiments were performed in the late light period from 1400 to 1900 h since at this period rats were more alert and motivated to work . Animals were trained in four identical standard operant conditioning chambers of internal dimensions 30 . 5 × 24 . 1×21 . 0 cm ( Med Associates Inc , VT , USA ) . The front panel of each chamber was equipped with one central and two laterals V-shaped licking ports with a photobeam sensor to register individual licks ( Med Associates Inc , VT , USA ) . Each port had a licking spout that consisted of either one ( for lateral ports ) or a bundle of up to 6 ( for the central port ) blunted needles ( 20-gauge ) that were carefully sanded and glued at the tip of a stainless-steel sipper tube . Each needle was connected to a solenoid valve ( Parker , Ohio , USA ) via a silicon tube . The volume of the drop was adjusted before each session and maintained by using an individual and constant air pressure system ( Perez et al . , 2013 ) . On the rear panel , there was an ambiance white noise amplifier with a speaker that was turned on in all the sessions . Chambers were enclosed in a ventilated sound-attenuating cubicle . Experimental events were controlled and registered by a computer via a Med Associates interface ( Med Associates Inc , VT , USA ) . All subjects were trained in a ‘Yes/No’ psychophysical task ( Stüttgen et al . , 2011 ) to emit a response by either going left or right based on the concentration of a 10 µL sucrose cue ( Low 3% or High 18 wt% ) . For trained animals , the task comprises four epochs: Return , Stimulus , Response , and Outcome . The outline of a trial is depicted in Figure 1A . A trial began when trained subjects moved from either lateral port to return the central spout; this epoch was named Return . Once in the central port , the rats were required to lick the empty spout a variable number of times ( between two or three ) to receive a 10 µL drop of either 3 or 18 wt% sucrose ( hereafter Cue-D ) . Rats could give additional empty licks after Cue-D delivery . These empty licks were used as a measure of the palatability oromotor responses elicited by sucrose ( Perez et al . , 2013 ) . The time elapsed from Cue-D delivery to the last lick in the central spout was designated as the Stimulus epoch . Subsequently , subjects had to move to either the Low or High sucrose-associated port ( Response epoch ) and emit , at least , one dry lick . If the response was correct , subsequent licks delivered three drops of water as a reward , while incorrect choices briefly turned off and on the lights during 50 ms ( at the second dry lick ) and subsequent licks were without a reward . The Outcome port comprises the interval where rats were licking in the lateral spout . The learning criterion was set at ≥80% correct responses during four consecutive sessions . Importantly , a drop of water was not delivered at the central port as a washout because in a pilot study we found that rats did not learn the task despite extensive training ( >50 sessions ) . We speculate that this was due to an imbalance in the reward value between the licking ports . Specifically , the reward value of one drop of water +one drop of sucrose ( 3 or 18 wt% ) at the central spout seems to be higher than the value of 3 drops of water delivered at the lateral spouts . The inclusion of a water washout failed to motivate rats and induced learning and thus the water washout , at central spout , was no longer used . Once the animals learned to discriminate between Low ( 3 wt% ) and High ( 18 wt% ) sucrose by getting at least 80% of the trials correct , the generalization sessions were introduced . Generalization sessions were composed of 20% of the trials ( 80% were of discrimination trials ) . These trials were like discrimination trials with the exception that after at least two discrimination trials subjects received a drop of either 0 , 3 , 4 . 75 , 7 . 5 , 11 . 75 , or 18 wt% sucrose . In these trials , no reward was delivered after choosing either lateral spout ( and in the second dry lick , the lights turned briefly on and off for 50 ms , signaling that no reward will be delivered ) . Discrimination and generalization sessions were interleaved , such that a generalization session occurred if at least one discrimination session with ≥80% correct responses took place the day before . This procedure avoids impairment of task performance . Since no statistical differences in task performance were found among groups , behavioral data were collapsed across subjects for the three brain regions recorded . For discrimination sessions , the percent correct responses were obtained by counting the number of trials for Low or High that subjects responded to the correct associated choice spout , divided by the total number of trials . To determine if performance was affected by electrode implantation , the average performance of the five sessions pre- and post-surgery were compared using a paired t-test ( Figure 1 ) . For generalization sessions , the percent responses given to the High concentration spout was plotted , and a sigmoid function was fitted to obtain the psychometric function . Likewise , surgery effects over generalization sessions were evaluated by comparing the average performance for all these sessions before and after surgery with a paired t-test . The time spent licking in the central port for each concentration ( Cue-D +Cue G trials ) during generalization sessions were collapsed and compared using a one-way ANOVA , and a Dunnett post hoc confirmed differences against sucrose 3 wt% Cue-D trials . As well , the time spent during the Return and Response epochs for each movement direction ( left or right ) , and during the Outcome epoch for reinforced and unreinforced trials , were collapsed and compared using an unpaired t-test . Once animals achieved the learning criterion and at least three consecutive generalization sessions were tested , then we proceeded to implant a custom-made 16 tungsten wires ( 35 µm diameter ) each arranged in a 4 × 4 ( 1 mm2 ) multielectrode array . The array was implanted in the posterior Insula ( pIC; n = 11 ) , in the anterior Insula ( aIC; n = 8 ) and the orbitofrontal cortex ( OFC , n = 9 ) . All subjects were anesthetized using ketamine ( 70 mg/kg , i . p . ) and xylazine ( 20 mg/kg , i . p . ) . The rats were put in a stereotaxic apparatus where a midline sagittal scalp incision was made to expose the skull and to put two holding screws . A third screw soldered to a silver wire that served as an electrical ground for recordings was inserted above the cerebellum ( Gutierrez et al . , 2010 ) . A craniotomy in the left hemisphere was made to implant an electrode array in one of the following sites: posterior IC ( AP: +1 . 0 to +1 . 4 mm , ML: +5 . 2 mm from bregma , DV: −4 . 4 to −4 . 7 mm ventral to dura ) , anterior IC ( AP: +1 . 6 to +2 . 3 mm , ML: +5 . 2 mm from bregma , DV: −4 . 6 to −4 . 7 mm ventral to dura ) or OFC ( AP: +3 . 5 mm , ML: +3 . 2 mm from bregma; DV: −4 . 4 mm ventral to dura ) . Dental acrylic was applied to cement the electrode array to the screws . The rats were given intraperitoneal enrofloxacin ( 0 . 4 ml/kg ) and ketoprofen ( 45 mg/kg ) for three days after surgery and were allowed to recover for one week . After the completion of the experiments , subjects were deeply anesthetized with an overdose of pentobarbital sodium ( 150 kg/mg , i . p . ) where they were transcardially perfused with PBS ( 1x ) followed by 4% paraformaldehyde . Brains were removed , stored for one day in 4% paraformaldehyde and posteriorly were changed to a 30 vol . /vol . % sucrose/PBS solution . Brains were sectioned in 40 µm coronal slices , and they were stained with cresyl violet to visualize the location of electrode tips . Neural activity was recorded using a Multichannel Acquisition Processor system ( Plexon , Dallas , TX ) interfaced with a Med Associates conditioning chamber to record behavioral events simultaneously . Extracellular voltage signals were first amplified x1 by an analog headstage ( Plexon HST/16o25-GEN2- 18P-2GP-G1 ) , then amplified ( x1000 ) and sampled at 40 kHz . Raw signals were band-pass filtered from 154 Hz to 8 . 8 kHz and digitalized at 12 bits resolution . Only single neurons with action potentials with a signal-to-noise ratio of ≥3:1 were analyzed ( Gutierrez et al . , 2010 ) . The action potentials were isolated on-line using voltage-time threshold windows and three principal components contour templates algorithm . A cluster of waveforms was assigned to a single unit if two criteria were met: Inter-Spike Intervals were larger than the refractory period set to 1 ms , and if it is formed a visible ellipsoid cloud composed of the 3-D projections of the first three principal component analysis of spike waveform shapes . Spikes were sorted using Offline Sorter software ( Plexon , Dallas , TX ) ( Gutierrez et al . , 2010 ) . Only time stamps from offline-sorted waveforms were analyzed . All data analysis was performed using MATLAB ( The MathWorks Inc . , Natick , MA ) and Graphpad Prism ( La Jolla , CA , USA ) . Unless otherwise indicated , we used the mean ± sem and the α level at 0 . 05 . | Imagine you wake up in the morning , and you pour yourself and your loved one coffee . They like it with two sugars but you only with one . Our ability to distinguish different sweet intensities allows us to detect how much sugar is in the coffee . It also helps us to predict the amount of energy present in foods and if it is safe to ingest . We can experience the sweet quality because our tongue contains sweet taste receptor cells that are switched on by sugar . This activates neurons across our taste system in the brain . However , we do not completely understand how these areas represent the intensity of sugar . Previous studies have only ‘passively’ measured different sugar concentrations , either using anesthetized animals or behavioral tasks that do not involve decision-making other than licking . But to accurately evaluate how animals perceive the intensity , active decision-making is required , such us ‘reporting’ the perceived concentration of sugar . Fonseca et al . set out to answer this question by training rats in a new sweet intensity discrimination task , in which the rats had to move to the left or right to obtain water as a reward . This way , the animals could ‘indicate’ how sweet they perceived the sugar water to be . At the same time , recordings from the three brain areas involved in taste responses were taken ( called the anterior and posterior insular cortices , and the orbitofrontal cortex ) to measure how the sugar intensity is processed in the brain . The results showed that a small group of neurons within all three areas contained more information about the sugar intensity than other neurons , suggesting the taste system uses a compact and distributed code to represent its intensity . The information about sugar intensity was contained in both the number of nerve impulses and in the precise timing with which these neurons fired . Many drinks and high-energy foods often contain large quantities of sugar , and their overconsumption contributes to the worldwide problems of obesity and its associated diseases . Therefore , a better understanding of the neurons that code information about the intensity of sugar could be a starting point for other studies to pinpoint the connections and areas in the brain involved in our irremediable attraction for sugar . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Sucrose intensity coding and decision-making in rat gustatory cortices |
Adipose tissue is crucial for the maintenance of energy and metabolic homeostasis and its deregulation can lead to obesity and type II diabetes ( T2D ) . Using gene disruption in the mouse , we discovered a function for a RhoA-specific guanine nucleotide exchange factor PDZ-RhoGEF ( Arhgef11 ) in white adipose tissue biology . While PDZ-RhoGEF was dispensable for a number of RhoA signaling-mediated processes in mouse embryonic fibroblasts , including stress fiber formation and cell migration , it's deletion led to a reduction in their proliferative potential . On a whole organism level , PDZ-RhoGEF deletion resulted in an acute increase in energy expenditure , selectively impaired early adipose tissue development and decreased adiposity in adults . PDZ-RhoGEF-deficient mice were protected from diet-induced obesity and T2D . Mechanistically , PDZ-RhoGEF enhanced insulin/IGF-1 signaling in adipose tissue by controlling ROCK-dependent phosphorylation of the insulin receptor substrate-1 ( IRS-1 ) . Our results demonstrate that PDZ-RhoGEF acts as a key determinant of mammalian metabolism and obesity-associated pathologies .
Obesity is a global health challenge . While it is generally recognized that obesity arises as a result of a complex interplay of genetic factors that govern predisposition for white adipose tissue expansion , as well as environmental factors , such as diet , the molecular mechanisms underlying this condition are not fully understood . Adipose tissue growth results from an increase in both adipocyte size and number and occurs as part of normal development , as well as during the onset of obesity ( Faust et al . , 1978; Prins and O'Rahilly , 1997 ) . Recent evidence shows that the number of adipocytes is set during childhood and adolescence and determines the adult fat mass ( Spalding et al . , 2008 ) . Adipogenesis , a differentiation program that gives rise to adipocytes , is regulated by a coordinated activation of a transcription factor network responding to the adipogenic and anti-adipogenic signals ( Cristancho and Lazar , 2011 ) . Insulin/IGF-1 signaling also has marked effects on adipogenesis . For instance , genetic disruption of the insulin receptor ( IR ) , the insulin receptor substrate ( IRS ) genes , as well as PKBα/Akt1 and PKBβ/Akt2 , respectively , the major mediators of insulin signaling , lead to impaired adipogenesis ( Miki et al . , 2001; Tseng et al . , 2004; Yun et al . , 2008 ) . Beyond soluble factors , changes in cell shape have also been implicated in influencing adipogenesis through activation of a small GTPase RhoA and its downstream effector Rho-kinase ( ROCK ) ( Dupont et al . , 2011; Kilian et al . , 2010; McBeath et al . , 2004 ) . RhoA-ROCK signaling is activated in mature adipocytes via mechanical stretch generated by obesity-related adipocyte hypertrophy ( Hara et al . , 2011 ) . Together with Rac and Cdc42 , RhoA constitutes a Rho branch of the Ras superfamily of small GTPases , which act in a variety of signaling pathways and regulate many cellular functions ( Ridley , 2001a ) . Initially identified as key regulators of actin cytoskeleton remodeling , Rho family members control many other cellular processes , including vesicular trafficking , cell cycle progression , cytokinesis , migration , apoptosis , gene transcription , and cell fate determination ( Coleman and Olson , 2002; Heasman and Ridley , 2008; McBeath et al . , 2004; Sordella et al . , 2002 ) . Rho GTPases are activated through a GTP-for-GDP nucleotide exchange reaction , catalyzed by the Rho GTPase-specific guanine nucleotide exchange factors ( RhoGEFs ) ( Ridley , 2001a ) . The Dbl family of RhoGEFs , with its 70 members in mammals , represents the largest class of Rho activators ( Rossman et al . , 2005; Schmidt and Hall , 2002 ) . Dbl family members contain at least one highly conserved Dbl homology ( DH ) catalytic domain , which facilitates the GDP/GTP exchange , and a pleckstrin homology ( PH ) domain N-terminal to the DH domain ( Zheng , 2001 ) . PDZ-RhoGEF ( ARHGEF11 ) , encodes a Rho guanine nucleotide exchange factor , with two transcript variants ARHGEF11 variant 1 and variant 2 in the human ( GeneBank accession # NM_014784 and NM_198236 ) and a single Arhgef11 transcript in the mouse ( Genebank accession NM_001003912 ) . Together with two other Dbl family members , the leukemia-associated RhoGEF ( LARG ) and p115RhoGEF , PDZ-RhoGEF constitutes a newly identified subfamily , the regulator of G protein signaling ( RGS ) domain-containing RhoGEFs ( RGS-RhoGEFs ) . Single nucleotide polymorphisms ( SNPs ) within the genes encoding RGS-RhoGEFs have also been informative in revealing their potential physiological functions . For instance , a non-synonymous SNP ( nsSNP ) mapping to the C-terminal region of LARG and leading to Y1306C ( Tyr1306Cys ) substitution has been implicated in increasing insulin sensitivity in non-diabetic Pima Indians ( Holzapfel et al . , 2007; Kovacs et al . , 2006 ) . Another nsSNP coding for the R1467H ( Arg1467His ) , a C-terminal variant of PDZ-RhoGEF was found in association with insulin resistance and type II diabetes in several populations , including Pima Indians , Old Order Amish , Caucasian , Korean and Chinese ( Bottcher et al . , 2008; Fu et al . , 2007; Jin et al . , 2010; Liu et al . , 2011; Ma et al . , 2007 ) . These findings suggest that LARG and PDZ-RhoGEF may impact insulin sensitivity and whole body metabolism . To elucidate the physiological function of PDZ-RhoGEF in vivo , we disrupted its gene in the mouse . While our data show that PDZ-RhoGEF is dispensable for many cellular processes that it had previously been linked to , we discovered its direct involvement in controlling adipose tissue homeostasis through regulation of adipocyte numbers and energy expenditure . Mechanistically , PDZ-RhoGEF was found to induce RhoA/ROCK-dependent phosphorylation of the insulin receptor substrate-1 ( IRS-1 ) and promote signaling via this adaptor protein in response to insulin/IGF-1 , controlling proliferation of adipogenic precursors and MEFs and whole body energy expenditure . Moreover , PDZ-RhoGEF-knockout ( KO ) mice were resistant to diet-induced obesity and insulin resistance in vivo . Our work firmly implicates PDZ-RhoGEF in the physiological control of adipose tissue development , obesity and insulin resistance .
To evaluate the physiological function ( s ) of PDZ-RhoGEF , we generated a genetically modified mouse line bearing a conditionally targeted PDZ-RhoGEF allele , Arhgef11 ( Figure 1—figure supplement 1A ) . By introducing CMV-driven global expression of cre recombinase , PDZ-RhoGEF heterozygous mice ( +/- ) were generated and full deletion of PDZ-RhoGEF allele was confirmed by Southern Blot analysis ( Figure 1—figure supplement 1B ) . We further verified the complete deletion of Arhgef11 ( -/- , KO ) by immunoblotting with a rabbit polyclonal antibody raised against the N-terminal fragment of PDZ-RhoGEF ( Figure 1—figure supplement 1D ) . PDZ-RhoGEF KO animals were born at the expected Mendelian ratio and were grossly indistinguishable from their wild type ( WT ) and heterozygous littermates . Although PDZ-RhoGEF KO mice fully developed , as they aged , they weighed less than their wild type counterparts . Most of the tissues showed no difference in weight between the two genotypes ( Figure 1—figure supplement 2A ) , however , the white adipose tissue mass was markedly reduced in mutant animals ( Figure 1A ) . The fat pads from the major fat depots , epididymal ( EWAT ) , knee ( KWAT ) , inguinal ( IWAT ) , and retroperitoneal white adipose tissue ( RWAT ) were significantly smaller in PDZ-RhoGEF KO compared to wild type mice , expressed both as total mass ( Figure 1B ) and as a percentage of body weight ( Figure 1—figure supplement 2B ) . Unlike WAT , brown adipose tissue mass ( BAT ) was slightly affected by PDZ-RhoGEF deletion ( Figure 1—figure supplement 2C , D ) . 10 . 7554/eLife . 06011 . 003Figure 1 . PDZ-RhoGEF controls body weight and adipose tissue mass . ( A ) Representative EWAT , RWAT , IWAT , and KWAT from NCD-fed 7-month old WT and PDZ-RhoGEF KO male mice . ( B ) Fat weight from 7-month old male mice on NCD ( n = 7–10 ) . ( C ) Body weight of mice on NCD ( n = 12–13 ) . ( D ) Body fat index determined by MRI analysis of wild type and PDZ-RhoGEF KO mice ( n = 7–9 ) maintained on a NCD for 16 weeks . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 00310 . 7554/eLife . 06011 . 004Figure 1—figure supplement 1 . PDZ-RhoGEF gene targeting in the mouse . ( A ) PDZ-RhoGEF targeting strategy ( i ) The 5’ genomic region of PDZ-RhoGEF ( wild-type , WT ) , ( ii ) the gene targeting vector , ( iii ) the targeted allele and ( iv ) the mutated allele ( mt ) . Short arm ( SA ) , long arm ( LA ) , flanking probe ( FP ) , neomycin ( Neo ) , phosphoglycerol kinase promoter ( PGK ) , cyclization recombinase ( cre ) , flipase recognition target ( FRT ) , and locus X over P1 ( loxP ) , are indicated . ( B ) Southern blot analysis of gene deletion following genetic introduction of Cre recombinase . ( C ) Confirmation of PDZ-RhoGEF deletion by immunoblotting with anti-PDZ-RhoGEF antibody ( TC1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 00410 . 7554/eLife . 06011 . 005Figure 1—figure supplement 2 . Tissue mass of WT and PDZ-RhoGEF KO mice . ( A ) Tissue mass from NCD-fed WT and PDZ-RhoGEF KO mice after being normalized by body weight ( n = 8 ) . ( B ) WAT mass from NCD-fed WT and PDZ-RhoGEF KO mice normalized by body weight ( n = 8 ) . ( C ) Total BAT mass from NCD-fed WT and PDZ-RhoGEF KO mice ( n = 15–2015–20 ) . ( D ) BAT mass from WT and PDZ-RhoGEF KO mice normalized by body weight as percent of body weight ( n = 15–20 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 00510 . 7554/eLife . 06011 . 006Figure 1—figure supplement 3 . The body weight , body size , and food intake of WT and PDZ-RhoGEF KO mice . ( A ) Body weight of female mice on NCD ( n = 12–13 ) . ( B ) The size of WT and PDZ-RhoGEF KO male mice determined by measuring the length from tip of nose to anus at 16 weeks ( n = 6–7 ) and 32 weeks of age ( n = 7–11 ) . ( C ) Food intake in WT and PDZ-RhoGEF KOmalemice ( n = 5 ) . Food intake per mouse was measured every two days over 15-day period . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 006 To further assess the effect of PDZ-RhoGEF deletion on body weight , an age-matched set of 6 week-old wild type and PDZ-RhoGEF KO mice were monitored for weight gain for 12 weeks while on a normal chow diet ( NCD ) ( 5% of total calories derived from fat , 200 kcal/kg ) . Throughout the monitoring period , both male and female PDZ-RhoGEF KO mice weighed less then their wild type counterparts ( Figure 1C , Figure 1—figure supplement 3A ) while showing no difference in body length ( Figure 1—figure supplement 3B ) or food consumption ( Figure 1—figure supplement 3C ) . Consistent with the anatomical analysis , MRI assessment of total fat of age matched 4-month old mice corrected for body weight , indicated that the body fat index for PDZ-RhoGEFKO mice was lower than that of the wild type ( Figure 1D ) . To investigate the changes underlying the weight reduction in PDZ-RhoGEF KO mice , we analyzed the brown adipose tissue ( BAT ) , which plays an important role in the maintenance of energy homeostasis . The relative expression of the uncoupling protein-1 ( UCP1 ) in BAT , the master regulator of BAT-mediated thermogenesis , was comparable between the genotypes ( Figure 2—figure supplement 1A ) , as was the expression of UCP-1 in the subcutaneous WAT ( IWAT ) ( a measure of the thermogenic UCP-1 positive brite/beige cells in WAT ) ( Figure 2—figure supplement 1B ) . Consistently , the animals from both genotypes showed no difference in core body temperature ( Figure 2—figure supplement 1C ) . We next examined the energy expenditure in our mice at different ages , over a 24 hr period , using indirect calorimetry , as well as the oxygen consumption rate ( OCR ) of isolated EWATs . Compared to wild type , the total oxygen consumption and the respiratory exchange rates ( RER ) in PDZ-RhoGEF KO mice were higher and particularly increased at time points towards the end of the light cycle ( revised Figure 2A , B , Figure 2—figure supplement 2A–C ) . Consistently , while the isolated EWATs from both genotypes displayed similar basal OCR , PDZ-RhoGEF-deficient EWATs reached a higher maximal OCR than wildtype upon the injection of carbonyl cyanide-4- ( trifluoromethoxy ) -phenylhydrazone ( FCCP ) , indicating their greater maximal respiratory capacity ( Figure 2C ) . PDZ-RhoGEF KO EWATsalso displayed a higher extracellular acidification rate ( ECAR ) , suggesting their preference for using glucose as a primary energy source ( new Figure 2D ) . 10 . 7554/eLife . 06011 . 007Figure 2 . The effect of PDZ-RhoGEF on energy expenditure . ( A ) PDZ-RhoGEF KO mice consumed more oxygen than the wild type mice ( n = 4 ) . ( B ) PDZ-RhoGEF KO mice have higher RER than wild type mice ( n = 4–6 ) . ( C ) Oxygen consumption rate ( OCR ) of EWAT determined in the presence of mitochondrial effectors , oligomycin A ( 10 μM ) , FCCP ( 10 μM ) , and rotenone ( 3 μM ) /antimycin A ( 12 μM ) . The values are presented as percent of the basal OCR measured at time 0 ( n = 3 ) . ( D ) Glycolysis was assessed by ECAR normalized to basal ECAR and measured at time 0 ( n = 3 ) . ( E-–G ) Analysis of physical activity , including total activity ( E ) , ambulatory ( F ) , and stereotype activity ( G ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 00710 . 7554/eLife . 06011 . 008Figure 2—figure supplement 1 . UCP 1 expression in BAT and core body temperature . ( A ) UCP1 expression in brown fat was determined by qPCR normalized to internal control GAPDH ( n = 4 ) . ( B ) UCP1 expression in IWAT was determined by qPCR normalized to internal control GAPDH ( n = 4–5 ) . ( C ) The core body temperature determined by rectal thermometer ( n = 5–6 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 00810 . 7554/eLife . 06011 . 009Figure 2—figure supplement 2 . Energy homeostasis of male wild type and PDZ-RhoGEF KO mice at different ages . ( A–C ) The time course of oxygen consumption rate ( VO2 ) measured during light and dark periods under normal food ( n = 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 009 Physical activity , a major determinant of total energy expenditure , was assessed by measuring total activity , including ambulatory and stereotype activity ( grooming , digging , etc . ) . PDZ-RhoGEF KO animals displayed high total activity ( Figure 2E ) . The locomotion , determined as ambulatory activity , was comparable between the genotypes ( Figure 2F ) , whereas a trend towards increased stereotype activity in PDZ-RhoGEF KO animals was detected ( Figure 2G ) , contributing to an upward trend in total activity ( Figure 2E ) . Considering that adipocyte number and adipocyte cell size can impact fat mass , we examined the adipose tissue properties of our mice . Morphometric analysis of fixed adipose tissue from the EWAT of wild type and PDZ-RhoGEF KO animals showed no discernable differences in adipocyte size between the genotypes ( Figure 3—figure supplement 1A ) . However , the number of mature adipocytes isolated from the EWAT of PDZ-RhoGEF KO mice was lower compared to that from wild type animals at various ages ( Figure 3A ) . Animals from both genotypes displayed comparable adipogenic potential , judged by the expression levels of the peroxisome proliferator-activated receptor γ ( PPARγ ) , a key determinant of adipocyte terminal differentiation , and adiponectin ( Acrp30 ) in EWAT ( Figure 3—figure supplement 1B ) . Moreover , PDZ-RhoGEF-deficient adipose tissue stromal cells ( ADSCs ) and MEFs retained the ability to differentiate into mature adipocytes , albeit at a lower rate ( Figure 3B , Figure 3—figure supplement 1C ) . 10 . 7554/eLife . 06011 . 010Figure 3 . Reduced adipocytes and progenitors in PDZ-RhoGEF KO mice . ( A ) Adipocyte numbers from 8- , 12- , 16-week old WT and PDZ-RhoGEF KO EWATs , ( B ) In vitro adipogenesis of EWAT ADSCs prepared from 16 week-old WT and PDZ-RhoGEF KO animals assessed by Oil Red-O staining . ( C ) DNA synthesis of EWAT-derived ADSCs upon insulin stimulation during MCE . ( D ) Adipogenitc progenitor cell proliferation presented as percent of BrdU + cells by FACS analysis under normal culture condition ( n = 3 ) . ( E ) Image and ImageJ quantification of BrdU-labeled EWATs from one-month old mice ( n = 6–9 ) . Bar scale = 100 μm ( 200X ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . ( F ) Stromal-vascular fraction from adult WT and PDZ-RhoGEF KO EWATs . EWAT was stained with adipogenic progenitor cells ( CD29+/CD34+/CD24+/Sca1+ ) ( n = 4 ) . ( G ) Preadipocytes in WT and PDZ-RhoGEF KO EWAT stromal-vascular fraction were stained with preadipocyte marker , pref-1 ( n = 3 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01010 . 7554/eLife . 06011 . 011Figure 3—figure supplement 1 . Cell size , adipogenic potential , and in vitro adipogenesis . ( A ) Morphometric analysis of epididymal adipocyte cell size of WT and PDZ-RhoGEF KO mice ( n = 5–7 ) . Scale bar = 100 μm ( 100X ) . ( B ) The protein levels of PPARγ and adiponectin ( Acrp30 ) in EWAT from 24- and 8-week old WT and PDZ-RhoGEF KO mice . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . ( C ) In vitro adipogenesis of MEFs ( E14 . 5 ) from WT and PDZ-RhoGEF KO assessed by Oil Red-O staining ( n = 3 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01110 . 7554/eLife . 06011 . 012Figure 3—figure supplement 2 . The role of PDZ-RhoGEF in cell proliferation . ( A ) Impaired proliferation of PDZ-RhoGEF KO MEFs grown in FCS . Cells were counted on the indicated days ( n = 3 ) . ( B ) [3H]-thymidine incorporation of WT and PDZ-RhoGEF KO MEFs in response to the indicated factors ( n = 3 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . ( C ) Impaired proliferation of 3T3-L1 preadipocytes with reduced PDZ-RhoGEF expression under normal culture condition . Cells were counted on the indicated days . Expression of PDZ-RhoGEF was determined by immunoblotting ( control: parental 3T3-L1 ) . ( D ) Impaired insulin-mediated cell proliferation in 3T3-L1 preadipocytes when PDZ-RhoGEF expression was reduced , judged by percent of BrdU + cells , representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01210 . 7554/eLife . 06011 . 013Figure 3—figure supplement 3 . PDZ-RhoGEF regulates clonal expansion during adipocyte differentiation in vitro . ( A ) Reduction of post-confluent mitosis determined by [3H}-thymidine incorporation and mitotic clonal expansion in PDZ-RhoGEF KO MEFs during in vitro adipogenesis , pulsed-chased by BrdU + ( Red ) and counterstained by DAPI ( Blue ) . ( B ) Overexpression of PDZ-RhoGEF resulted in increased DNA synthesis in 3T3-L1 cells after induction of adipogenesis . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01310 . 7554/eLife . 06011 . 014Figure 3—figure supplement 4 . Strategy of FACS-based purification of adipogenic progenitor cells . Adipogenic progenitor cells were sorted from the adipogenic progenitor specific maker-labeled SVF suspensions with AriaII ( Becton Dickson ) cell sorter , and the purity of sorted cells verified as >95% by rerunning the sorted population . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01410 . 7554/eLife . 06011 . 015Figure 3—supplement 5 . The proportion of resident/infiltrating adipose tissue macrophages and the expression of preadipocyte marker ( Pref-1 ) from NCD-fed WT and PDZ-RhoGEF KO mice . ( A ) Quantification of adipose tissue macrophages ( n = 6–9 ) from 4 weeks old WT and PDZ-RhoGEF KO mice . ( B ) The levels of Pref-1 mRNA in 16 weeks old WT and PDZ-RhoGEF KO EWAT SVF were determined by qPCR ( n = 4 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 015 PDZ-RhoGEF KO MEFs displayed impaired cell proliferation , both under normal growth conditions ( Figure 3—figure supplement 2A ) and in response to fetal calf serum ( FCS ) and Insulin-like Growth Factor-1 ( IGF-1 ) ( Figure 3—figure supplement 2B ) , as did the PDZ-RhoGEF KO ADSCs in response insulin ( Figure 3C ) . Similarly , lowered expression of PDZ-RhoGEF in a preadipocyte cell line 3T3-L1 reduced their proliferative capacity under normal culture conditions , as well as in response to insulin ( Figure 3—figure supplement 2C , D ) IGF-1/insulin signaling has been implicated in the control of the last proliferative phase of adipogenesis , also known as mitotic clonal expansion ( MCE ) ( Avram et al . , 2007 ) . To further address the effect of PDZ-RhoGEF on MCE , post-confluent mitosis and clonal expansion were monitored by [3H]-thymidine uptake and pulse-chase bromodeoxyuridine ( BrdU ) labeling in MEFs after induction of adipogenesis . PDZ-RhoGEF KO MEFs displayed greatly diminished DNA synthesis and cell proliferation compared to wild type ( Figure 3—figure supplement 3A ) . Finally , ectopic expression of PDZ-RhoGEF promoted DNA synthesis during clonal expansion of 3T3-L1 cells ( Figure 3—figure supplement 3B ) . Taken together , these data implicate PDZ-RhoGEF in coupling insulin/IGF-1 signaling to the proliferative phases of adipose tissue development . The adipogenic progenitor cells exist within the stromal vasculature fraction ( SVF ) of WAT ( Rodeheffer et al . , 2008 ) and are represented by the Lin-/CD29+/CD34+/Sca1+/CD24+ ( CD24+ ) and the Lin-/CD29+/CD34+/Sca1+/CD24- ( CD24- ) populations , which we isolated from our mice ( Figure 3—figure supplement 4 ) . Judged by BrdU incorporation , deletion of PDZ-RhoGEF led to a reduction of proliferative capacity in cultured adipogenic progenitors ( Figure 3D ) . To probe the proliferative capacity of the adipose tissue in vivo during neonatal development , wild type and PDZ-RhoGEF KO postnatal day 6 . 5 mice were injected with BrdU at 12 hr intervals for 3 . 5 days . At one-month of age , mice were sacrificed and the BrdU-retaining cells in the adipose tissue counted . PDZ-RhoGEF KO EWAT displayed fewer BrdU-retainingcells than the wild type EWAT , indicative of impaired proliferation ( Figure 3E ) . Of note , EWAT from both genotypes displayed the same proportion of infiltrating macrophages , determined by staining for the macrophage marker F4/80 ( Figure 3—figure supplement 5A ) , eliminating the possibility that the differences in the BrdU signal arose from variable infiltrates . To pursue the possible role of PDZ-RhoGEF in regulation of adipose precursor population , we quantified the adipocyte progenitor cell ( Lin-/CD29+/CD34+/Sca1+/CD24+ ) and preadipocyte ( Pref-1+ ) populations within the SVF of the adult mouse fat pads . Both the adipogenic progenitor and preadipocyte populations were reduced in PDZ-RhoGEF KO animals ( Figure 3F , G ) , accompanied by a proportional decrease in mRNA expression of Pref-1 ( Figure 3—figure supplement 5B ) . We explored a possible general function of PDZ-RhoGEF in controlling proliferative capacity . The defective proliferation of PDZ-RhoGEF-deficient MEFs ( Figure 3—figure supplement 3A , B ) was paralleled by reduced Rho activation upon FCS and IGF-1 treatment ( Figure 4A ) . Of interest , the LPA-dependent Rho stimulation was comparable between the genotypes , reinforcing that PDZ-RhoGEF is not required for LPA-mediated RhoA activation ( Figure 4A ) . Moreover , RhoA-mediated actin cytoskeleton reorganization , stress fiber formation and migration upon exposure to LPA , FCS , and IGF-1 were unaffected by PDZ-RhoGEFloss ( Figure 4—figure supplement 1A–C ) . 10 . 7554/eLife . 06011 . 016Figure 4 . PDZ-RhoGEF is required for the optimal IGF-1 signaling output in MEFs and adipogenic progenitor cells . ( A ) RhoA activation was determined by the RBD-binding assay in response to the indicated factors ( FCS [10%] , IGF-1 [100 ng/ml] , LPA [20 μM] ) . ( B ) Differential PKB/Akt phosphorylation in WT and PDZ-RhoGEF KO MEFs in response to IGF-1 . ( C ) Differential PKB/Akt phosphorylation and PDZ-RhoGEF expression in WT and PDZ-RhoGEF KO adipogenic progenitor cells in response to IGF-1 . ( D ) IGF-1 response of PDZ-RhoGEF KO MEFs is rescued by ectopic expression of myc-tagged PDZ-RhoGEF . ( E ) IGF-1 stimulated tyrosine phosphorylation of IR . ( F ) Serine phosphorylation of IRS1 and interaction between IRS1 and PI3K in WT and PDZ-RhoGEF KO MEFs in response to IGF-1 . ( G ) Phosphorylation of PKB/Akt S473 in response to IGF-1 is dependent on ROCK activity . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01610 . 7554/eLife . 06011 . 017Figure 4—figure supplement 1 . Stress fiber formation and mitogen-induced cell migration . ( A ) Stress fiber formation 15 min after mitogen stimulation ( FCS [10%] , IGF-1 [100 ng/ml] , LPA [20 μM] ) was determined by phaloidin-rhodamine staining ( B ) Twenty-four hours after mitogen stimulation ( as in [A] ) , cell migration of MEFs with the indicated genotypes was determined by a wound-healing assay . ( C ) Six hours after plating , IGF-1-induced migration of MEFs with the indicated genotypes was determined by the transwell migration assay . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 017 To probe the function of PDZ-RhoGEF in mediating IGF-1 sensitivity , we tested the effect of PDZ-RhoGEF loss on PKB/Akt signaling . PDZ-RhoGEF KO MEFs , as well as the cultured FACS-sorted adipogenic progenitor cells , exhibited markedly lower PKB/Akt activation in response to IGF-1 ( Figure 4B , C ) . Re-expression of PDZ-RhoGEF in the PDZ-RhoGEF-deficient MEFs rescued the IGF-1-induced PKB/Akt phosphorylation ( Figure 4D ) . However , activation-specific tyrosine phosphorylation of the IGF-1 receptor was comparable between the two genotypes ( Figure 4E ) , indicating that diminution of PKB/Akt phosphorylation in PDZ-RhoGEF KO MEFs stems from an impaired PDZ-RhoGEF-dependent step downstream of the receptor . Insulin receptor substrate-1 ( IRS-1 ) acts immediately downstream of the insulin/IGF-1 receptors to propagate their signals ( Taniguchi et al . , 2006 ) . Furthermore , IRS-1 S632/635 have been implicated as substrates of ROCKI/II ( Furukawa et al . , 2005; Lee et al . , 2009 ) . In IGF-1-stimulated PDZ-RhoGEF KO MEFs , both tyrosine and IRS1 serine 632/635 ( S632/635 ) phosphorylation were greatly reduced compared to wild type MEFs , as was its direct association with the p85α and p110α subunits of PI3K ( Figure 4F ) . In contrast , phosphorylation of IRS-1 at S612 , another putative Rho kinase ( ROCKI/II ) targeted residue ( Sordella et al . , 2003 ) , was not affected by PDZ-RhoGEF loss ( Figure 4F ) , indicating that phosphorylation of this residue in MEFs can be uncoupled from that of S632/635 . When wild type MEFs were pretreated with Y27632 , a specific inhibitor of ROCKI and ROCKII , IGF-1-dependent PKB/Akt phosphorylation was reduced to a level comparable to that found in PDZ-RhoGEF KO MEFs ( Figure 4G ) , suggesting that these kinases mediate the PDZ-RhoGEF-dependent IGF-1 response . Thus , PDZ-RhoGEF modulates IGF-1-mediated cell proliferation in MEFs via regulation of ROCK-dependent phosphorylation of IRS-1 at S632/635 . As one of the peripheral insulin targeted tissues , fat mass is associated with insulin sensitivity and metabolic homeostasis ( Guilherme et al . , 2008 ) . Upon IP injection of insulin , activation-specific tyrosine phosphorylation of the insulin receptor ( IR ) was comparable between wild type and PDZ-RhoGEF KO EWAT ( Figure 5A ) . However , activation of PKB/Akt in EWAT from PDZ-RhoGEF KO animals was considerably reduced ( Figure 5A ) , consistent with a role for PDZ-RhoGEF in the transmission of the insulin signal downstream of IR and reminiscent of the MEFs response to IGF-1 ( Figure 4B , E ) . Further , both tyrosine and S632/635 phosphorylation of IRS-1 in PDZ-RhoGEF KO EWAT were decreased in response to insulin ( Figure 5B ) , as was the direct association of IRS-1 with the p85α and p110α subunits of PI3K ( Figure 5B ) , indicative of impeded insulin signaling throughput via IRS-1 ( Furukawa et al . , 2005; Lee et al . , 2009 ) . In accordance with the MEF data , phosphorylation of S612 of IRS1 remained similar between the genotypes ( Figure 5B ) . Of note , PDZ-RhoGEF loss did not affect insulin-induced PKB/Akt phosphorylation in two other peripheral insulin target tissues , the liver ( Figure 5C ) , where PDZ-RhoGEF is abundant ( Figure 5—figure supplement 1 ) , or the skeletal muscle ( Figure 5D ) , where PDZ-RhoGEF expression is relatively low ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 06011 . 018Figure 5 . PDZ-RhoGEF is required for optimal IGF-1 signaling output in EWAT . ( A ) IR phosphorylation and differential PKB/Akt phosphorylation in WT and PDZ-RhoGEF KO EWAT in response to insulin . ( B ) Serine phosphorylation of IRS1 and interaction between IRS1 and PI3K in WT and PDZ-RhoGEF KO EWAT in response to insulin . ( C ) IR phosphorylation and differential PKB/Akt phosphorylation in WT and PDZ-RhoGEF KO liver in response to insulin . ( D ) IR phosphorylation and differential PKB/Akt phosphorylation in WT and PDZ-RhoGEF KO skeletal muscle in response to insulin . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 01810 . 7554/eLife . 06011 . 019Figure 5—figure supplement 1 . The tissue expression profile of PDZ-RhoGEF . The indicated tissue lysates were immunoblotted with the anti-PDZ-RhoGEF antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 019 Despite the reduction of fat mass , insulin-mediated glucose excursion was not affected in PDZ-RhoGEF KO mice , as determined by the glucose tolerance test ( GTT ) ( Figure 6—figure supplement 1A ) , nor was the insulin-dependent glucose transport in the soleus and EDL ( extensor digitorum longus ) skeletal muscles ( Figure 6—figure supplement 1B ) . To further elucidate the effect of PDZ-RhoGEF on glucose utilization in the liver , we monitored the glucose-dependent gene expression of the carbohydrate response binding protein ( ChREBP ) and the glucose transporter 2 ( Glut2 ) , and found no difference between the wild type and PDZ-RhoGEF KO livers ( Figure 6—figure supplement 1C ) . Furthermore , there was no lipid accumulation in the liver ( Figure 6—figure supplement 1D ) . To probe if the effect of PDZ-RhoGEF on adipocyte numbers and energy expenditure can impact the onset of diet-induced obesity and T2D , two age groups of mice ( 6-week old and 16-week old ) were transferred to a high fat diet ( HFD ) ( Supplementary file 1 ) for 14 to 16 weeks . Wild type and PDZ-RhoGEF KO mice from both age groups gained weight at similar rates upon switching to HFD ( Figure 6A , Figure 6—figure supplement 2A ) . Nevertheless , the wild type mice from both age groups retained a proportionally higher body weight and higher adiposity ( Figure 6—figure supplement 2B , C ) , despite consuming equal amounts of food ( Figure 6—figure supplement 2D ) . Notably , the animals from both genotypes showed no difference in total oxygen consumption , RER or total locomotor activity ( Figure 6—figure supplement 3A–C ) . Importantly , wild type but not PDZ-RhoGEF KO mice , developed glucose intolerance ( Figure 6B , Figure 6—figure supplement 4A ) and insulin resistance ( Figure 6C , Figure 6—figure supplement 4B ) , as well as impaired insulin-dependent glucose transport in the soleus and EDL ( Figure 6D ) . Further , wild type EWAT exhibited higher frequency of larger adipocytes ( Figure 6E ) , indicative of adipocyte hypertrophy , which is associated with a decrease in adiponectin secretion ( Table 1 ) and increased local inflammation with macrophage infiltration , detected as crown-like structures ( CLS ) ( Figure 6E ) . 10 . 7554/eLife . 06011 . 020Figure 6 . PDZ-RhoGEF deficiency protects animals from diet-induced obesity , glucose intolerance and insulin resistance . ( A ) The body weight of mice switched to HFD at 18 weeks of age was monitored for 14 weeks ( n = 8 ) . ( B ) GTT on overnight fasted 32-wk old mice after 14 weeks on HFD ( n = 4 ) . ( C ) Insulin-mediated glucose excursion in mice described in ( B ) determined by ITT ( n = 8 ) . ( D ) Ex vivo glucose transport in isolated skeletal muscles ( Soleus and EDL ) from HFD-fed animals ( n = 4 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . ( E ) Morphometric analysis of epididymal adipocyte cell size and crown-like structure ( arrows ) in WT and PDZ-RhoGEF KO mice ( n = 4–6 ) . Scale bar = 100 μm ( 100X ) . ( F ) Histological analysis of representative liver sections from WT and PDZ-RhoGEF KO mice . Scale bar = 100 μm ( 100X ) . ( G ) The mRNA expression of DGAT1 and G-6-Pase in liver from HFD-fed animals ( n = 4 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 02010 . 7554/eLife . 06011 . 021Figure 6—figure supplement 1 . Glucose metabolism in WT and PDZ-RhoGEF KO mice under normal diet . ( A ) Glucose tolerance test in NCD-fed WT and PDZ-RhoGEF KO mice ( n = 8 ) . ( B ) Glucose transport in isolated skeletal muscles from mice under NCD ( n = 4 ) . ( C ) The level of mRNA of ChREBP and Glut2 in the liver ( n = 4 ) . ( D ) Histological analysis of representative liver H&E stained sections from NCD-fed WT and PDZ-RhoGEF KO mice . Scale bar = 100 μm ( 100X ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 02110 . 7554/eLife . 06011 . 022Figure 6—figure supplement 2 . Adiposity and food consumption in WT and PDZ-RhoGEF KO mice under high-fat diet . ( A ) . The body weight of mice switched to HFD at 6-weeks of age was monitored for 16 weeks ( n = 4 ) . ( B ) Anatomical quantification of fat mass from mice of late-onset group after 24 weeks under HFD ( n = 8 ) . ( C ) Anatomical quantification of fat mass from early-onset group after 16 weeks under HFD ( n = 4 ) . ( D ) Food intake in WT and PDZ-RhoGEF KO mice ( n = 4 ) . Food intake per mouse was measured every week over 12-week period . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 02210 . 7554/eLife . 06011 . 023Figure 6—figure supplement 3 . Energy expenditure of WT and PDZ-RhoGEF KO mice under HFD . ( A ) The total oxygen consumption rate ( VO2 ) . ( B ) total RER . ( C ) total physical activity measured under HFD ( n = 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 02310 . 7554/eLife . 06011 . 024Figure 6—figure supplement 4 . Glucose metabolism in WT and PDZ-RhoGEF KO mice . ( A ) Glucose tolerance test on overnight fasted 22-week old WT and PDZ-RhoGEF KO mice , after 16-week high-fat feeding ( n = 4 ) . ( B ) Insulin tolerance test on overnight fasted 22-week old WT and PDZ-RhoGEF KO mice , after 16-week high-fat feeding ( n = 4 ) . ( C ) Glucose tolerance test on overnight fasted PDZ-RhoGEF KO mice , NCD ( before HFD ) and HFD ( after HFD ) ( n = 8 ) . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 024 While PDZ-RhoGEF KO livers remained normal , wild type mice developed non-alcoholic fatty liver ( NAFL ) ( Figure 6F ) , a pathological consequence of ectopic fat deposition due to adipose hypertrophy . In wild type , but not PDZ-RhoGEF KO mice on HFD , this was accompanied by increased mRNA expression of diglyceride acyltransferase 1 ( DGAT1 ) , an enzyme that participates in triglyceride synthesis ( Figure 6G ) , Unlike PDZ-RhoGEF KO animals , wt mice displayed elevated mRNA levels of glucose-6-phosphatase ( G-6-Pase ) , a key enzymes involved in at final step of gluconeogenesis and glycogenolysis in liver ( Figure 6G ) , suggesting that increased glucose level in wild type mice may be partly due to elevated hepatic glucose production ( Table 1 ) ( Elam et al . , 2010; Higuchi et al . , 2008; Pettinelli et al . , 2009; Villanueva et al . , 2009 ) . Eventually , wild type mice developed T2D with elevated fasting glucose and insulin ( Table 1 ) , while the PDZ-RhoGEF KO animals maintained normal glucose homeostasis ( Figure 6—figure supplement 4C ) . 10 . 7554/eLife . 06011 . 025Table 1 . Quantification of fasting serum metabolites of wild type and PDZ-RhoGEF KO mice . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 025Wild typePDZ-RhoGEF KOTriglycerides ( mg/ml ) 0 . 63 +±0 . 020 . 43 + ± 0 . 01*Glucose ( mmol/l ) 12 . 13 + ± 0 . 537 . 62 + ± 0 . 16*Insulin ( ng/ml ) 1 . 80 + ± 0 . 200 . 51 + ± 0 . 07*Adiponectin ( ng/ml ) 0 . 29 + ± 0 . 010 . 36 + ± 0 . 01*Values are from 8-month-old male mice of the indicated genotype fed with high fate diet after overnight fasting . Data represent mean + ± s . e . m . *p<0 . 05 compared with wild type ( n = 8 ) . We further investigated the molecular mechanism of how PDZ-RhoGEF contributes to HFD-induced insulin resistance and T2D . In accordance with the impaired whole body insulin resistance , peripheral insulin-target tissues from wild type mice displayed diminished PKB/Akt phosphorylation on HFD ( Figure 7A ) . Nevertheless , total IR levels and its tyrosine phosphorylation remained similar in livers and EWAT from both genotypes on HFD ( Figure 7—figure supplement 1 ) , suggesting that the insulin resistance in these tissues is not due to IR desensitization or degradation . 10 . 7554/eLife . 06011 . 026Figure 7 . PDZ-RhoGEF contributes to the HFD-induced insulin resistance through a S6K1-mediated negative feedback . ( A ) Western blot analysis of insulin-mediated PKB/Akt phosphorylation in peripheral tissues: EWAT , liver , and skeletal muscles ( EDL and soleus ) . ( B ) Phosphorylation status of IRS1 S632/635 , S307 , JNK1 T183/185 , and p70S6K1 T389 in EWATs from HFD-fed WT and PDZ-RhoGEF KO mice . ( C ) Phosphorylation of p70S6K1 T389 in EWATs from NCD-fed WT and PDZ-RhoGEF KO mice after in vivo insulin stimulation . ( D ) Differential protein levels of PDZ-RhoGEF in EWAT of NCD- and HFD-fed wild type mice . ( E ) PDZ-RhoGEF modulates insulin/IGF-1 signaling to impact adipose tissue homeostasis ( I ) and susceptibility to dietary-induced obesity and T2D ( II ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 02610 . 7554/eLife . 06011 . 027Figure 7—figure supplement 1 . EWAT insulin sensitivity determined by IR tyrosine phosphorylation . After 14-week of HFD , IR from both WT and PDZ-RhoGEF KO EWAT and liver remained sensitive to insulin , determined by IR tyrosine phosphorylation . DOI: http://dx . doi . org/10 . 7554/eLife . 06011 . 027 We next explored inhibitory serine phosphorylation of IRS1 as the source of insulin resistance downstream of IR in the white adipose tissue ( Griffin et al . , 1999; Hotamisligil et al . , 1996 ) . Under HFD , phosphorylation of IRS1 S632/635 and S307 , targets of c-Jun N-terminus kinase 1 ( JNK1 ) and mTOR/S6K1 , were elevated in wild type EWAT ( Figure 7B ) . While JNK1 activation appeared unaffected by PDZ-RhoGEF deletion , p70S6K1 activation was considerably higher in wild type EWAT under HFD ( Figure 7B ) . Even on normal food , insulin-mediated p70S6K1 activation was reduced in PDZ-RhoGEF KO EWAT ( Figure 7C ) . Intriguingly , prolonged HFD not only aggravated the activation of p70S6K signaling but also led to an elevation of PDZ-RhoGEF protein levels in all peripheral insulin-targeted tissues , EWAT , skeletal muscle ( EDL ) , and liver ( Figure 7D ) , suggesting that high levels of PDZ-RhoGEF accompany HFD-induced insulin resistance in wild type animals .
In search of a molecular mechanism for PDZ-RhoGEF action , we discovered differential signaling at the level of IRS-1 as the likely cause for the suboptimal proliferative signaling in the absence of PDZ-RhoGEF . Differential S632/635 IRS-1 phosphorylation and association with PI3K , as well as PKB/Akt activation in MEFs ( Figure 4B , E ) and EWAT ( Figure 5A , B ) , points to a mechanism of PDZ-RhoGEF/RhoA impact on insulin/IGF-1 signaling involving ROCKs . Indeed , inhibition of ROCKs in wild type MEFs impaired their IGF-1 response ( Figure 4F ) . With 182 serine and 60 threonine residues , IRS-1 has been shown to be a target of many kinases , including ROCKs , JNK ( c-Jun NH2-terminal kinase ) , mTOR ( mammalian target of rapamycin ) , p70S6K , PKCs ( protein kinase C qθ , zζ , lλ ) , IKKβ ( inhibitor of nuclear factor κB kinase β ) , and MAPK ( Gual et al . , 2005; Taniguchi et al . , 2006 ) . Consistent with our results , ROCK-mediated phosphorylation of IRS-1 at S632/635 was previously found to promote signaling throughput via IRS-1 and increased glucose uptake in adipocytes and mouse skeletal muscle ( Furukawa et al . , 2005; Lee et al . , 2009 ) . The uncoupling of IRS-1 S632/635 and S612 phosphorylation in MEFs and EWAT upon disruption of PDZ-RhoGEF may represent an additional means of achieving specificity in RhoA signaling . Intriguingly , PDZ-RhoGEF appears capable of directing ROCK activity towards IRS-1 S632/635 in fibroblasts and adipose tissue , whereas other GEFs may participate in modulating its activity towards S612 in other tissues . Alternatively , the distinct RhoGEFs may differentially activate ROCK family members ( ROCK1/β and ROCKII/α ) , which in turn may harbor subtle , but important differences in specificity towards IRS-1 . The lack of difference in the BAT mass , food consumption , UCP-1 expression , in BAT and IWAT , as well as the core body temperature in PDZ-RhoGEF KO mice , indicate that the adaptive thermogenesis is not dependent on PDZ-RhoGEF . Nevertheless , PDZ-RhoGEF deletion increased total oxygen consumption and total activity , largely due to an increase in the stereotype activity , but not the ambulatory activity . RER analysis indicates that while both wild type and PDZ-RhoGEF KO mice utilize carbohydrates and fat as fuel ( 0 . 8<RER<1 . 0 ) , higher RER values in PDZ-RhoGEF KO mice suggest preferential use of carbohydrates as a fuel source in their skeletal muscles . Indeed , an increase in physical activity found in PDZ-RhoGEF KO mice may impact exercise- and contractility-mediated glucose transport in the skeletal muscle ( Richter and Hargreaves , 2013 ) and in turn increase the use of carbohydrates as energy source . Remarkably , despite its reduced PI3K pathway activation in response to insulin , PDZ-RhoGEF KO EWAT displayed an increase in ECAR , revealing of the preferential use of glucose as an energy source in this tissue . This may in part be due to the compensatory , PI3K-independent mechanisms for glucose utilization in EWAT ( Chang et al . , 2007 ) and contribute to the maintenance of glucose homeostasis in PDZ-RhoGEF KO animals and their lean and metabolically healthy phenotype . The phenotype of PDZ-RhoGEF-deficient mice partially resembles that of mice lacking other insulin signaling pathway components . For instance , IRS-1- or PKBα/Akt1-deficiency leads to decreased body fat ( Miki et al . , 2001; Wan et al . , 2012 ) . Similarly , adipose tissue-specific disruption of the insulin receptor ( FIRKO ) displayed reduced adiposity . Unlike the proliferative defect found in PDZ-RhoGEF KO adipose tissue development , the reduced adipose tissue size in FIRKO mice results from decreases in adipocyte size in the mutant adult fat pads ( Bluher et al . , 2002 ) . Further confounding the comparison between the phenotypes arising from IR and PDZ-RhoGEF deletion is the contribution of PDZ-RhoGEF to signaling via the IGF-1 receptor ( IGF-1R ) . IGF-1R is the dominant IR/IGF-1R family member expressed in the stromal-vascular cells of the developing adipose tissue , including progenitor cells and pre-adipocytes ( Entingh-Pearsall and Kahn , 2004; Nougues et al . , 1993; Wright and Hausman , 1995 ) . Thus , in vivo , PDZ-RhoGEF impact on the proliferative capacity of adipogenic precursors likely relates to its function in signaling via the related IGF-1R , whereas in mature adipose tissue it may also affect IR signaling ( Figure 5A , B ) . Notably , dual deletion of IGF-1 and insulin receptors in adipose tissues ( FIGIRKO ) results in reduced adiposity due to a decrease in adipocyte abundance ( Boucher et al . , 2012 ) . Taken together , the correlation between full body deletion of PDZ-RhoGEF and deletion of IR/IGF-1R in adipose tissue further reinforces that modification of IR/IGF-1R signaling in adipose tissue plays a central role in controlling adipose tissue homeostasis . A mechanistic insight into the contribution of PDZ-RhoGEF-mediated control of energy expenditure and adipocyte numbers to reduced adiposity may require tissue-specific deletion of PDZ-RhoGEF in various mouse tissues . Deletion of PDZ-RhoGEF had no effect on food consumption when fed with either normal or high-fat diet . Under a normal dietary setting , PDZ-RhoGEF , via RhoA/ROCK activation , sensitized IRS1/PI3K signaling to insulin/IGF-1 , thereby maintaining the proliferative potential of adipogenic progenitors and supporting adipose tissue expansion during development ( Figure 7E–I ) . Of note , PDZ-RhoGEF may also act on mature adipocytes or other cell types to influence adipose tissue development and homeostasis . Adipocyte number is a key determinant of fat mass in humans , whereby individuals with higher number of adipocytes are prone to obesity ( Spalding et al . , 2008 ) . Likewise , wild type mice , which displayed greater total adipocyte numbers , became obese and developed T2D when on prolonged HFD . Reduced adipose tissue expansion in PDZ-RhoGEF KO animals prevented these pathological consequences of the HFD . Of interest , under conditions of energy overload , PDZ-RhoGEF levels were elevated in the peripheral insulin-target tissues and accompanied by increased downstream signaling to p70S6K1 and enhanced negative feedback loop input into the insulin/insulin receptor signaling cascade ( Figure 7E–II ) . Our data firmly support the involvement of PDZ-RhoGEF in the genetics of fat mass regulation and predisposition to obesity and T2D , previously suggested by the association of the PDZ-RhoGEF SNP R1467H with the epidemiology of T2D ( Bottcher et al . , 2008; Fu et al . , 2007; Jin et al . , 2010; Liu et al . , 2011; Ma et al . , 2007 ) . Further work on determining the nature of the PDZ-RhoGEF H1467R SNP variant , using functional reconstitution of PDZ-RhoGEF KO cells and mice , as well as biochemical characterization of its activity towards RhoA will be needed to fully understand its action in obesity and T2D predisposition .
All animal work was conducted according to the Policies and Guidelines of the Canadian Council on Animal Care and the Province of Ontario’s Animals for Research Act . Wild type and PDZ-RhoGEF KO male mice were weaned at 4 weeks of age and placed on normal chow diet ( NCD ) ( 5% fat , TD LM-485 , Harlan ) until 17 weeks of age . Body weight was measured at 6-weeks of age . For the early-onset obesity group , mice were fed with high fat diet ( HFD ) ( 45% fat , TD-01435 , Harlan , Table S1 ) beginning at 6 weeks of age for 16 weeks . Animals from the late-onset group were fed with HFD beginning at 18-weeks of age for 14 weeks . Body weights were recorded weekly . White adipose tissues ( WATs ) masses were determined anatomically from various depots , epididymal ( EWAT ) , retroperitoneal ( RWAT ) , inguinal ( IWAT ) , knee ( KWAT ) , from HFD and NCD fed male mice . The final tissue weight was determined with or without normalization to the total body weight ( n = 8–12 ) . Food intake per mouse was determined over 15 days ( n = 4–5 ) . Magnetic resonance imaging of abdominal adiposity used a 7 T Biospec 70/30 USR ( Bruker Corporation , Ettlingen , DE ) , equipped with a B-GA12 gradient coil insert and 7 . 2 cm inner diameter cylindrical quadrature RF volume coil . Mice were anaesthetized and maintained at 1 . 8% isoflurane , delivered through the bite bar of a dedicated slider bed ( Bruker Corporation , Ettlingen , DE ) with mice in prone orientation . A strong positive fat signal was generated using a respiratory-gated T2-weighted RARE pulse sequence , applied as a stack of 20–24 2D coronal images with full body coverage ( echo time 36 ms , field-of-view 90 x 40 mm , matrix size 360 x 160 , in-plane resolution 250 μm , slice thickness 1 mm ) . The pulse sequence repetition time and the total data acquisition time were determined by respiratory rate ( scan time of approx . 6 min at respiratory rate of 30 breathes per minute ) . Fat pad volumes were assessed with medical image processing , analysis , and visualization ( MIPAV ) software ( National Institute of Health ( NIH ) Center for Information Technology ) . Body fat indices were calculated by dividing adipose tissue volume by body weight . The maximal oxygen consumption rate ( VO2 ) , heat production , respiratory exchange rate ( RER VO2/VCO2 ) , heat production ( Kcal/hr ) , and activity in 9 to 14 month-old male WT and PDZ-RhoGEF KO mice were determined using the Oxymax system ( Columbus Instruments ) . The method of indirect calorimetry scores counts as ambulation when the animal traverses the cage , breaking a series of IR beams in sequence . Repeated interruptions of the same IR beam do not incur ambulatory counts . All beam interruptions are scored as Total Activity ( X TOT ) . Subtraction of ambulatory counts ( X AMB ) from the Total counts provides counts associated with stereotypy ( grooming , scratching , etc . ) . Samples were prepared as described ( Dunham-Snary , et al . , 2014 ) . In brief , Tissues were dissected from 8-week old mice and washed with Seahorse basal medium ( DMEM ) with 25 mM Glucose and 25 mM HEPES . Each tissue was sampled six times using a 2 mm UniCore Harris Punch ( Whatman , Sigma-Aldrich ) , resulting in 2 mg tissue samples . The tissue punches ( 4 mg ) were transferred to XF24 Islet Capture Microplate ( Seahorse Bioscience , North Billerica , MA ) and a screen was loaded on to each well . Each well was washed with running medium ( DMEM/25 mM Glucose and without HEPES ) . OCR and ECAR were determined in running medium with Seahorse XFe24 analyzer ( Seahorse Bioscience ) . PDZ-RhoGEF was introduced to MEFs using a retroviral packaging system . The viral particles containing myc-PDZ-RhoGEF was generated by Transfection of the Phoenix E retroviral packaging line and MEFs ( passage 1 ) were infected and expanded for characterization . Fasting glucose tolerance test ( GTT ) and insulin tolerance test ( ITT ) were performed on age matched ( 8 month-old ) wild type and PDZ-RhoGEF KO mice . For GTT , PDZ-RhoGEF wild type mice ( n = 8 ) and KO mice ( n = 8 ) were starved for 16 hr before intraperitoneal ( IP ) challenge with 1 g D-glucose/kg of body weight and their blood glucose monitored from the tail vein using an automated glucose monitor ( Actusoft , Roche ) . ITT was performed following a three hour fast ( n = 8 ) . Human insulin ( 1 U/kg of body weight , Novo Nordisk ) was given to the animals by intraperitoneal injection and the blood glucose level was measured every 15 min up to one hour as described above . Serum was collected from wild type ( n = 6 ) and PDZ-RhoGEF KO mice ( n = 8 ) after 16 hr of starvation for determination of fasting glucose , insulin , triglycerides , and glycerol . Glucose levels were measured by the automated glucose monitor ( Actusoft , Roche ) . Insulin levels were determined by Ultrasensitive insulin ELISA kit ( Crystal Chem Inc . , Downers Grove , IL ) in triplicates . The levels of fasting triglycerides were determined by the serum triglyceride determination kit ( Sigma-Aldrich , St . Louis , MO ) . Whole extensor digitorum longus ( EDL ) and soleus muscles were isolated and incubated in the presence or absence of 100 nM insulin , and the uptake of 2-deoxyglucose ( 2-DG ) was measured as described previously ( Rudich et al . , 2003 ) Epididymal white adipose tissues were prepared from NCD-fed 28-week old wild type and PDZ-RhoGEF KO male mice and 42-week old mice from both genotypes with the least 24 weeks on HFD . Paraffin embedded tissues were sectioned in 4-μm-thick sections and stained with H&E . Adipocyte cell size was determined with Image J software ( US National Institutes of Health ) . At least 100 cells were measured from each genotype ( n = 4 for NCD; n = 6 for HFD ) . Mature adipocytes and stromal vascular fraction were isolated from epidydimal white adipose tissue as described ( Tang et al . , 2008 ) . In brief , adipose tissues were harvested and minced in adipocyte isolation buffer ( 100 mM HEPES pH 7 . 4/120 mM NaCl/50 mM KCl/5 mM D-glucose/1 mM CaCl2/1 . 5% fatty acid-free BSA ) containing 0 . 8 mg/ml ( 125 U/ml ) collagenase III ( Worthington ) at 37°C with shaking for 2 hr . Digestion was terminated by adding DMEM:F12 ( 1:1 ) /10% FCS/2 mM L-glutamine/1 mM sodium pyruvate/50 mM bβ-mercaptoethanol and digested fat tissues were passed through a 180-μm nylon mesh filter ( Millipore ) . Stromal vascular fraction was collected using a 30 ml syringe with 14 G ( 80 mm long ) needle to avoid the floating adipocytes , followed by centrifugation to collect stromal vascular fraction . Isolated stromal vascular fraction was subjected to FACS analysis to determine the adipocyte progenitor cell population . Adipocyte numbers were determined from approximately 100 mg of adipose tissues digested in 1 ml of adipocyte isolation buffer at 37°C with shaking at 125 rpm for 2 hr . After digestion , adipocytes in 10 μl suspension were counted using hematocytmeter . At least 200 adipocytes from each preparation were counted . Total number of adipocytes for each fad pad was calculated by converting total number per ml into total number per fad pad . Post-confluent MEFs were subjected to the adipocyte differentiation induction medium ( complete medium/1 mM dexamethasone/5 mg/ml insulin/0 . 5 mM isobutylmethylxanthine [IBMX] ) or insulin ( 5 μg/ml ) alone . Cell cycle re-entry and DNA synthesis was determined by [3H]-thymidine incorporation at the indicated times . BrdU labeling was performed to monitor mitotic clonal expansion . MEFs were plated onto glass coverslips and grown to confluence . 15 hr after induction of differentiation , cells were pulse-labeled for 3 hr with BrdU ( 10 μM ) and then transferred to BrdU-free induction medium . 72 hr later , cells were fixed and BrdU positive cells were detected by phycoerythrin-conjugated anti-BrdU antibody ( Molecular Probe , Invitrogen ) and counter stained with 4' , 6-diamidino-2-phenylindole ( DAPI ) . For in vitro differentiation , post-confluent cells were maintained in induction medium for 2 days , and then were fed with complete medium supplement with 5 mg/ml of insulin every other day . At the end of the monitoring period ( day 8 ) , cells were fixed with 10% phosphate buffered formalin . Oil droplets were stained with Oil Red-O as described ( Tang et al . , 2003 ) . Relative lipid content was determined by absorbance at 510 nm by spectrophotometer ( Beckman/Coulter , DU800 ) . Post-confluent EWAT ADSCs ( from 16-week old male mice ) were stimulated with insulin ( 5 ng/ml ) and DNA synthesis was assessed by [3H]-thymidine incorporation at each indicated time point . Isolated stromal-vascular fraction was resuspended in FACS media ( 2% FCS/PBS/0 . 01% NaN3 ) and cells were stained with markers for adipose stem cell and preadipocytes ( Rodeheffer et al . , 2008 ) . Antibody incubation was performed on ice for 20 min and fixed with 1% paraformaldehyde overnight at 4°C . Samples were acquired on Canto II flow cytometry ( BD Biosciences ) and analyzed with FlowJo software ( Tree Star , Inc . , Ashland , OR ) . For all ex vivo analysis , PDZ-RhoGEF expression , progenitor cell proliferation , and PI3K signaling , adipogenic progenitor cells were sorted from adipogenic progenitor specific maker-labeled SVF suspensions with AriaII ( Becton Dickson ) cell sorter , and the purity of sorted cells was verified as >95% by rerunning the sorted population . Antibodies were purchased from eBioscience unless otherwise stated , including the following: biotin-lineage cocktail ( ) , Sca1-Pacific Blue ( Biolegend , San Diego , CA ) , Sca1-APC , CD34-Alexa-Fluro-700 ( Biolegend ) , CD24-PE , CD31-PE-Cy7 , Terr-119-APC-eFluro-780 , biotin-CD45/streptavidin-PerCp , CD29-FITC ( BD Biosciences ) , streptavidin-PE-Cy7 , and anti-Dlk/Pref-1 ( MBL ) . Sorted CD34+/CD29+/Sca1+ cell were expanded under hypoxic conditions ( 3% O2 , 5% CO2 ) with MesenCult medium according to the manufacture’s protocol ( Stem cell Technology ) . The cells at passage 1–3 were used for differentiation , colony-forming assay ( CFU-F ) , insulin stimulated BrdU uptake , and PI3K signaling experiments under normaxia condition in DMEM/10% FCS . In vivo BrdU labeling was done as described ( Staszkiewicz et al . , 2009 ) . To quantify cell proliferation during early adipose tissue development , 6 . 5-day-old neonates were injected intraperitoneally with BrdU ( Sigma-Aldrich ) at a concentration of 50 μg/g of body weight , twice daily at 12 hr interval for 3 . 5 consecutive days . Animals were perfused through the left ventricle with physiological saline , and then 4% paraformaldehyde after a 3-week chase period and BrdU-labeled tissues were harvested for immunohistochemistry . BrdU-retaining cell population was quantified as percent of cells positive for BrdU ( n = 6–9 ) . Cells and tissues were disrupted in lysis buffer buffer ( 20 mM Tris ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 1% Triton X-100 , 2 . 5 mM Na-pyrophosphate , 1 mM b-glycerophosphate , 1 mM Na3VO4 , 1 mg/ml Leupeptin , 1 mM PMSF ) . Tissues were excised and lysed . Equal amounts of protein amounts were fractionated by NuPAGE ( Novex ) and analyzed using antibodies specific for phospho-Akt/PKBS473 , PKB/Akt , and IR ( Cell signaling , Danvers , MA ) , phospho-IRbY1150/1151and β-tubulin ( Upstate ) . IRS-1 phosphorylation and protein-protein interaction analysis were performed by immunoprecipitating total IRS-1 with the anti-IRS-1 antibody ( Cell Signaling ) and immunoblotting with antibodies specific for phospho-IRS-1S632/635 , phospho-IRS-1S612 , IRS-1 , p110α ( Millipore ) . ( Cell Signaling ) , phospho-tyrosine , p85α ( Milliopore ) . RhoA activity was assessed by the RBD assay as described ( Diekmann and Hall , 1995 ) using a monoclonal antibody against RhoA ( Santa Cruz Biotechnology ) . For cell proliferation data , all results are presented as: mean + ± S . D of at least three independent experiments . For animal data , all results are presented as: mean ± S . E . M . . Statistical significance was calculated by a Student’s t test using Excel software ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) .
The gene-targeting construct was generated using the FRT/loxP system . A replacement vector ( pSPUC ) was generated in which an FRT-flanked neo expression cassette was introduced into the intron between exon 1 and 2 of Arhgef11 ( GeneBank accession # NM_001003912 ) and two loxP sites were engineered to flank the targeted exon 2 , which created a null allele of PDZ-RhoGEF when Cre recombinase was introduced . The targeting construct was electroporated into embryonic stem cells ( ES ) from 129J/HSD strain . The homologous recombination and specific integration were confirmed by genomic Southern Blot . Subsequently , the homologous targeted ES clones were micro-injected into blastocysts isolated from C57/BL6 strain and transferred into the 129J/HSD foster mothers . The germline transmission was confirmed by genomic Southern Blot . PDZ-RhoGEF heterozygous and null animals were generated by crossing with a Cre transgenic line , B6 . C-Tg ( CMV-cre ) 1Cgn/J ( Jackson laboratory ) that expresses Cre recombinase globally under the transcriptional control of a human cytomegalovirus ( CMV ) promoter . The deletion of exon 2 was confirmed by genomic PCR with specific primers , Southern Blot with flanking probe , and Western Blot with anti-PDZ-RhoGEF antibody . The congenic C57BL6 PDZ-RhoGEF animals were generated by backcrossing C57/BL6:129J/HSD six to ten generation with C57/BL6 strain . The N-terminal fragment of human PDZ-RhoGEF ( aa1-735 , PDZ-RhoGEF∆C ) was subcloned into a bacterial expression vector pGEX-4T . 1 ( Amersham Pharmacia Biotech ) to generate a recombinant fusion protein GST ( Glutathione-S-Transferase ) -PDZ-RhoGEF∆C . Purified recombinant fusion protein was used to make a polyclonal rabbit anti-PDZ-RhoGEF antibody ( Department of Comparative Medicine , University of Toronto ) . The anti-PDZ-RhoGEF antibody was further purified with GST-sepharose affinity gel filtration . MEFs were derived from congenic C57BL6 mouse embryos at day 14 . 5 ( E14 . 5 ) . All the experiments were carried out using early passage of primary MEFs ( passage 3 ) derived from male PDZ-RhoGEF KO embryos and the littermate PDZ-RhoGEF+/+ embryos ( WT ) as control . Genotype and gender were determined by genomic PCR . For assessment of mitogen-stimulated cell proliferation , cells were plated and starved of serum for 48 hr and proliferation scored by measuring [3H]-thymidine uptake in response to mitogens , IGF-1 ( 80 ng/ml ) , LPA ( 20 mM ) or FCS ( 10% ) . Post-confluent ADSCs were stimulated with insulin ( 5 ng/ml ) and DNA synthesis was assessed by [3H]-thymidine incorporation at each indicated time point . Cell proliferation was also determined by cell counting ( Coulter counter , Beckman Coutler ) . Cell migration was evaluated using both wound healing and transwell migration assays . Briefly , for the wound healing assay , MEFs were plated and grown on a 24-well plate coated with 0 . 1 mg/ml of poly-D-lysine to form a confluent monolayer and serum starved for 48 hr . Wound was generated by scraping the monolayer with a pipette tip in the presence 10 mM of mitomycin C and migration was induced by adding serum-free medium , FCS ( 10% ) , IGF-1 ( 100 ng/ml ) , LPA ( 20 μM ) in the presence with mitocycin C . Cells were monitored for a 24 hr period , then fixed in 4% paraformaldehyde and stained with 0 . 1% crystal violet . For transwell assay , MEFs in a serum-free medium were plated onto the upper chamber of transwell plares ( Corning , New York ) , and IGF-1 ( 100 ng/ml ) , and was added to the lower chamber . After 6 hr , the cells from the inside of chamber were removed with cotton swab and the cells on the bottom of the filter were fixed with 4% paraformaldehyde and cells stained with 2 μg/ml DAPI in 0 . 1% Triton X-100/PBS . To assess stress fiber formation , MEFs were plated on to coverslips in 24-well plates and serum starved for 48 hr . Cells were fixed with 4% paraformaldehyde 5 minutes after addition of FCS ( 10% ) , IGF-1 ( 100 ng/ml ) and LPA ( 20 μM ) . Fixed cells were permeabilized in 0 . 1% Triton X-100/PBS , and stress fiber was stained with Rhodamine-Phalloidin ( 1 U/ml ) in 0 . 1% Triton X-100 PBS . Stained cells were analyzed by confocal microscopy ( Zeiss LSM510 ) . Total RNA was extracted from frozen tissue samples using Trizol ( Invitrogen ) , followed by RNeasy kit ( Qiagen ) to further purify RNA . Complementary DNA was synthesized from total RNA with the SuperScript transcriptase III and random hexamer primers ( Invitrogen , Carlsbad , CA ) . The real-time polymerase chain reaction ( PCR ) measurement of individual cDNAs was performed using SYBR green dye to measure duplex DNA formation with ABI 7500 Real-Time PCR System ( Applied Biosciences , Waltham , MA ) . The expression was normalized to the expression of either 18S ribosomal , glyceraldehyde-3-phosphate dehydrogenase ( GAPDH ) RNA , or mouse TATA box binding protein 1 ( mTBP1 ) . The primers used in the real time RT-PCR were listed as following: Pref-1 forward: 5’-GACCCACCCTGTGACCCC-3’ reverse: 5’-CAGGCAGCTCGTGCACCCC-3’ UCP-1 forward: 5’-AGGCTTCCAGTACCATTAGGT-3’ reverse: 5’-CTGAGTGAGGCAAAGCTGATT-3’ 18S forward: 5’AAACGGCTACCACATCCAAG3’ reverse: 5’CCTCCAATGGATCCTCGTTA3’ GAPDH forward: 5’-AACTTTGGCATTGTGGAAGG-3’ reverse: 5’-ACACATTGGGGGTAGGAACA-3’ G-6-Pase forward: 5’-GAAGGCCAAGAGATGGTGTGA-3’ reverse: 5’-TGCAGCTCTTGCGGTACATG-3’ DGAT-1 forward: 5’-GGCCTGCCCCATGCGTGATTAT-3’ reverse: 5’-CCCCACTGACCTTCTTCCCTGTAGA-3’ GLUT2 forward: 5’-GGCTAATTTCAGGACTGGTT-3’ reverse: 5’-TTTCTTTGCCCTGACTTCCT-3’ ChREBP forward: 5’-CTG GAC CGA CCC CTC TCC TT-3’ reverse: 5’-TTG TGG GTG CAG GAA GCG TA-3’ mTBP1 forward: 5’-GGCCTCTCAGAAGCATCACTA-3’ reverse: 5’-GCCAAGCCCTGAGCATAA-3’ To quantify macrophages at early adipose tissue development , BrdU-labeled tissue sections were with an anti-F4/80 monoclonal antibody . For each section from each individual mouse , three different high-power fields from individual mouse were analyzed . The total number of adipocytes and the number of nuclei of F4/80-expressing cells were counted for each field using Image J ( NIH ) . The fraction of F4/80-expressing cells for each sample was calculated as the sum of the number of nuclei of F4/80-expressing cells divided by the total number of adipocytes in each section of a sample . Dissected livers and EWATs were fixed in formalin , embedded in paraffin , sectioned in 4 μm slices , and stained with hematoxylin-eosin ( H&E ) . PDZ-RhoGEF expression was determined by immunoblotting . Protein lysates were prepared from various tissues from wild type male mice , subjected to 3~8% Tris-Acetate NuPAGE gel ( Invitrogen ) and immunoblotted with anti-PDZ-RhoGEF and anti-actin antibodies . Post-confluent MEFs were subjected to the adipocyte differentiation induction medium ( complete medium/1 mM dexamethasone/5 mg/ml insulin/0 . 5 mM isobutylmethylxanthine ( IBMX ) ) or insulin ( 5 mg/ml ) alone . Cell cycle re-entry and DNA synthesis was determined by [3H]-thymidine incorporation at the indicated times . BrdU labeling was performed to monitor mitotic clonal expansion . MEFs were plated onto glass converslips and grown to confluence . 15 hr after induction of differentiation , cells were pulse-labeled for 3 hr with BrdU ( 10 mM ) and then transferred to BrdU-free induction medium . 72-hr later , cells were fixed and BrdU positive cells were detected by phycoerythrin-conjugated anti-BrdU antibody ( Molecular Probe , Invitrogen ) and counter stained with 4' , 6-diamidino-2-phenylindole ( DAPI ) . For in vitro differentiation , post-confluent cells were maintained in induction medium for 2 days , and then were fed with complete medium supplement with 5 mg/ml of insulin every other day . At the end of the monitoring period ( day 8 ) , cells were fixed with 10% phosphate buffered formalin . Oil droplets were stained with Oil Red-O as described ( Tang et al . , 2003 ) . Relative lipid content was determined by absorbance at 510 nm by spectrophotometer ( Beckman/Coulter , DU800 ) . Post-confluent ADSCs were stimulated with insulin ( 5 ng/ml ) and DNA synthesis was assessed by [3H]-thymidine incorporation at each indicated time point . The number of mature adipocytes derived from MEFs was averaged of three independent fields after Oil Red-O staining ( n = 3-–4 ) . | Obesity is a growing public health concern around the world , and can lead to the development of type 2 diabetes , heart disease and cancer . Both genetics and environmental factors such as diet contribute to obesity . Fat cells are essential to good health , but the excess accumulation of fat cells in obese people involves a complex process that is regulated by interactions between numerous genes , cellular messengers and mechanical forces . Learning more about these factors could help prevent or treat obesity . One mutation in the gene encoding a protein called PDZ-RhoGEF has been linked to both obesity and type 2 diabetes . People with mutations in this gene are not responsive enough to insulin , a hormone important for sugar metabolism . This can interfere with the body’s ability to burn energy in food or lead to a dangerous build up of sugar in the blood as seen in type 2 diabetes . But exactly what PDZ-RhoGEF normally does to prevent this is unclear . Chang et al . now show that PDZ-RhoGEF controls fat cell production and the body’s ability to release the energy contained in food . First , mice that had been genetically engineered to lack PDZ-RhoGEF were compared to typical mice . The mice without PDZ-RhoGEF had fewer fat cells than the typical mice , and they burned more energy . The mutant mice walked around about as much as the typical mice but they were more likely to have repetitive movements , the mouse equivalent of human nervous ticks . Insulin normally stimulates the production of fat cells . But the mutant mice were less able to produce fat cells as they developed into adults . When fed a high fat food diet , the normal mice became fatter and insensitive to insulin and developed other health problems linked to excess fat in the body . The mutant mice on the same diet , however , stayed thin and avoided these health issues . The experiments show that PDZ-RhoGEF helps relay insulin’s message within the body , and as such it plays a critical role in regulating metabolism , sugar levels and fat accumulation . Future work should ask how PDZ-RhoGEF affects other complications linked to obesity , and explore the possibility of developing treatments for obesity based on the biology of this molecule . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods",
"Supplementary",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2015 | The Rho-guanine nucleotide exchange factor PDZ-RhoGEF governs susceptibility to diet-induced obesity and type 2 diabetes |
Histamine-dependent and -independent itch is conveyed by parallel peripheral neural pathways that express gastrin-releasing peptide ( GRP ) and neuromedin B ( NMB ) , respectively , to the spinal cord of mice . B-type natriuretic peptide ( BNP ) has been proposed to transmit both types of itch via its receptor NPRA encoded by Npr1 . However , BNP also binds to its cognate receptor , NPRC encoded by Npr3 with equal potency . Moreover , natriuretic peptides ( NP ) signal through the Gi-couped inhibitory cGMP pathway that is supposed to inhibit neuronal activity , raising the question of how BNP may transmit itch information . Here , we report that Npr3 expression in laminae I-II of the dorsal horn partially overlaps with NMB receptor ( NMBR ) that transmits histaminergic itch via Gq-couped PLCβ-Ca2+ signaling pathway . Functional studies indicate that NPRC is required for itch evoked by histamine but not chloroquine ( CQ ) , a nonhistaminergic pruritogen . Importantly , BNP significantly facilitates scratching behaviors mediated by NMB , but not GRP . Consistently , BNP evoked Ca2+ responses in NMBR/NPRC HEK 293 cells and NMBR/NPRC dorsal horn neurons . These results reveal a previously unknown mechanism by which BNP facilitates NMB-encoded itch through a novel NPRC-NMBR cross-signaling in mice . Our studies uncover distinct modes of action for neuropeptides in transmission and modulation of itch in mice .
How itch and pain information is encoded and transmitted has been subjected to numerous studies for more than a century ( Chen , 2021 ) . There is increasing evidence indicating the pivotal roles of neuropeptides in the coding of itch information in primary sensory neurons ( Chen , 2021 ) . A pruritogenic stimulus activates skin , immune , and nerve cells , or an inflammatory response , which provokes the release of itch-specific neuropeptides from primary afferents to activate G-protein-coupled receptors ( GPCRs ) in the spinal cord ( Chen , 2021; Wang and Kim , 2020 ) . Notably , gastrin-releasing peptide ( GRP ) and neuromedin B ( NMB ) , two mammalian neuropeptides , have been shown to encode nonhistaminergic itch and histaminergic itch , respectively ( Akiyama et al . , 2014; Barry et al . , 2020; Sun and Chen , 2007; Wan et al . , 2017; Zhao et al . , 2014b ) . Moreover , murine GRP-GRPR signaling is important for the development of contact dermatitis-induced itch ( Chen et al . , 2020; Liu et al . , 2020; Shiratori-Hayashi et al . , 2015; Zhao et al . , 2013 ) . These findings are in accordance with human and animal studies showing that histaminergic and nonhistaminergic itch is transmitted through parallel primary afferent pathways ( Chen , 2021; Johanek et al . , 2007; Namer et al . , 2008; Roberson et al . , 2013; Wilson et al . , 2011 ) . B-type or brain natriuretic peptide ( BNP ) , encoded by the gene Nppb , has been implicated in itch at discrete regions , including skin cells , sensory neurons , and spinal cord ( Liu et al . , 2020; Meng et al . , 2018; Mishra and Hoon , 2013; Solinski et al . , 2019 ) . The natriuretic peptide ( NP ) family also consists of atrial ( ANP ) and C-type natriuretic peptides ( CNP ) ( Potter et al . , 2006 ) . BNP binds to both NPRA and NPRC , encoded by Npr1 and Npr3 , respectively , with equality affinity , but not NPRB , while ANP also binds NPRA directly , resulting in the elevation of the second message cyclic GMP concentration ( Figure 1A; Potter et al . , 2006 ) . Although NPRC is considered to function as a clearance or silent receptor ( Maack et al . , 1987 ) , it can also mediate guanylyl cyclase ( GC ) receptor-coupled Gαisignaling under certain physiological conditions ( Anand-Srivastava , 2005 ) . BNP-NPRA signaling was initially proposed as an itch-specific pathway responsible for transmitting both histamine- and CQ-evoked itch that acts upstream of GRP-GRPR signaling ( Huang et al . , 2018; Mishra and Hoon , 2013 ) . However , that BNP transmits all types of itch is at odds with the fact that GRP is required only for nonhistaminergic itch . Further , genetic ablation of spinal Grp neurons fails to impact itch behaviors ( Barry et al . , 2020 ) , indicating that spinal Grp neurons do not constitute a functional circuit for itch . Recent studies have shown that BNP-NPRA signaling is involved in histaminergic itch as well as chronic itch in mice which comprises the histaminergic component ( Liu et al . , 2020; Solinski et al . , 2019 ) . Given that BNP can bind both NPRA and NPRC , two cognate receptors for BNP ( Figure 1A ) , the relationship between NPRA/NPRC and NMBR/GRPR and the role of NPRC in itch transmission remains undefined . Considering that the GC-cGMP signal transduction pathway mediated by BNP is inhibitory ( Potter et al . , 2006 ) and that BNP-NPRA/NPRC signaling may exert an inhibitory rather than excitatory function , analogous to GαI protein-coupled signaling , it is paradoxical that BNP would transmit rather than inhibiting itch information . In the present study , we have examined these open questions using a combination of RNA-scope ISH , genetic knockout ( KO ) mice , spinal siRNA knockdown , cell ablation , calcium imaging , pharmacological and optogenetic approaches . We found that NPRC , rather than NPRA , is a major functional receptor for BNP in the spinal cord , and BNP facilitates histamine-evoked itch through NPRC-NMBR crosstalk . Importantly , our studies confirmed that BNP is an inhibitory neuropeptide that alone fails to evoke Ca2+ response and itch-related scratching behavior , in contrast to GRP and NMB; However , BNP becomes excitatory by facilitating NMB-mediated itch transmission . Thus . distinct modes of action for neuropeptides coordinate itch transmission in the spinal cord .
As ANP also binds NPRA at high affinity , we tested whether intrathecal injection ( i . t . ) of ANP could induce scratching behavior and found that ANP failed to induce robust scratching behaviors at the dose of 1–20 μg ( equivalent to 6–120 µM , Figure 1—figure supplement 1A ) . Among three NPs , interestingly , only BNP facilitates histamine itch ( Figure 1—figure supplement 1B ) . BNP evoked dose-related scratching behavior ( 1–5 µg that is equivalent to 30–150 µM ) with a peak scratching number of 74 ± 16 . 2 ( Figure 1B ) , consistent with previous reports ( Kiguchi et al . , 2016; Liu et al . , 2014 ) . However , these doses are much higher than endogenous concentrations of ligands that should be within the nanomolar range , implying that scratching behaviors evoked by BNP reflect a non-specific pharmacological artifact . Time-course analysis showed that scratching behavior was delayed by 20–30 min after BNP injections as described without isoflurane treatment ( Kiguchi et al . , 2016; Liu et al . , 2014; Figure 1C ) , which is distinct from the rapid onset of scratching response evoked by i . t . GRP or NMB ( Figure 1—figure supplement 2H ) . The use of isoflurane for anesthesia could result in complex effects on neural circuits , especially inhibitory circuits ( Constantinides and Murphy , 2016 ) . We compared the effect of i . t . BNP on awake and isoflurane-anesthetized animals and found that isoflurane pretreatment significantly enhanced BNP-induced scratching behaviors in the first 10 min ( Figure 1—figure supplement 2G ) . This suggests that induced scratching behavior is an indirect rather than direct effect of BNP with isoflurane treatment , consistent with the fact that NPRA/NPRC are inhibitory receptors . Recent studies showed that Npr1 is widespread in the dorsal horn ( Fatima et al . , 2019 ) , in marked contrast to lamina II specific Grp expression ( Barry et al . , 2020 ) . Consistently , single nucleus RNA sequencing ( snRNA-seq ) from isolated spinal cord neurons found only a partial overlap ( 20 ~ 30% ) of Npr1 and Grp ( Sathyamurthy et al . , 2018 ) . To visualize the distribution of NP receptors in the spinal cord , we performed RNAscope in situ hybridization ( ISH ) and found that all three NP receptors are expressed in the dorsal horn of the spinal cord ( Figure 1D-N and Figure 5A-C ) . Consistent with a previous ISH study ( Barry et al . , 2020; Fatima et al . , 2019; Mishra and Hoon , 2013 ) , Npr1+ neurons are distributed in a gradient manner with higher intensity throughout laminae I-IV and are rather sparse in the deep dorsal horn ( Figure 1D and 5A ) . Remarkably , Npr3+ + are predominantly restricted to lamina I-II ( Figure 1H , J , L , N , 5C ) with 64 . 6% being excitatory neurons and 32 . 0% being inhibitory neurons ( Figure 1J–M ) . In contrast , Npr2 expression is homogenous throughout the spinal cord dorsal horn , implying that Npr2 lacks modality-specific function ( Figure 5B ) . RNAscope ISH showed minimal overlapping expression between Npr1 and Grpr ( Figure 1D and E ) . However , approximately 30% ( 31/103 ) of Npr1+ + in laminae I-II of the dorsal horn express Grp ( Figure 1—figure supplement 1E , F ) , and this number is further reduced to 23% ( 31/132 ) ( data not shown ) when all Npr1+ neurons in the dorsal horn are counted . Moreover , Npr1 and Nmbr minimally overlap in the dorsal horn ( Figure 1F and G ) , excluding the likelihood of NPRA-NMBR crosstalk . Npr1 and Npr3 also showed minimal overlapping expression in the spinal cord ( Figure 1N and O ) . By contrast , approximately 47 . 8% of Npr3+ + express Nmbr ( Figure 1H , I ) , raising the possibility that NPRC is involved in crosstalk with NMBR . To examine the role of NP receptors in acute itch behavior , we first analyzed the phenotype of Npr1 knockout ( KO ) mice ( Oliver et al . , 1997 ) . A previous study showing that CNP-NPRB is essential for axonal bifurcation of DRG neurons in the developing spinal cord ( Schmidt et al . , 2009 ) prompted us to evaluate the innervation of primary afferents in the spinal cord of Npr1 KO mice . We found that innervations of peptidergic CGRP+ and non-peptidergic IB4+ primary afferents in the superficial dorsal horn of Npr1 KO mice are comparable with wild-type ( WT ) littermates ( Figure 1—figure supplement 2A ) . The innervations of TRPV1+ , GRP+ , and SP+ primary afferents are also comparable between WT and Npr1 KO mice ( Figure 1—figure supplement 2B-D ) , indicating that NPRA is dispensable for the innervation of primary afferents . GRP and NMB have been implicated in nonhistaminergic and histaminergic itch , respectively ( Sun et al . , 2009; Wan et al . , 2017; Zhao et al . , 2014b ) . To examine whether GRPR and NMBR function normally in the absence of NPRA , we compared the scratching behaviors between Npr1 KO mice and their WT littermates after i . t . injection of GRP or NMB and found no significant differences in their responses to either GRP or NMB between the groups ( Figure 2A ) . However , Npr1 KO mice showed significantly impaired scratching responses to intradermal ( i . d . ) injection of histamine and chloroquine ( CQ ) , archetypal pruritogens for histaminergic and nonhistaminergic itch , respectively ( Sun and Chen , 2007 ) , as compared with WT littermates ( Figure 2B ) . The highly restricted expression of NPRC in lamina II and its high binding affinity to BNP prompted us to examine the role of NPRC in itch . However , Npr3 KO mice showed severe skeletal abnormalities , resulting in the failure of most KO mice to survive to adult stage for behavioral analysis ( Matsukawa et al . , 1999 ) . To determine whether the impaired scratching response of the global Npr1 KO mice could have resulted from the Npr1 deficiency in the spinal cord , DRGs where Npr1 is also expressed ( Zhang et al . , 2010 ) , or skin cells ( Meng et al . , 2018 ) , we knocked down Npr1-3 either individually or in combination in C57BI/6 mice using sequence-specific siRNA . I . t . Npr1 siRNA treatment significantly attenuated the scratching behavior evoked by histamine and CQ ( Figure 2C and D ) , whereas Npr3 siRNA treatment selectively attenuated histamine , but not CQ itch ( Figure 2C and D ) . Npr2 siRNA had no effect on CQ and histamine itch , making it unlikely to be involved in itch transmission ( Figure 2C and D ) . The effect of the knock-down of target mRNA in the spinal cord and DRGs was verified using real-time RT-PCR ( Figure 2E and F ) . These results revealed that Npr1 and Npr3 are differentially required for acute itch behavior at the spinal level . Further , we infer that there is little functional compensation among the three NP receptors . The slow onset of scratching behavior elicited by BNP , even at a high dose ( 150 µM ) , in the first 30 min contrasts sharply with rapid onset of GRP/NMB-induced scratching behavior ( Figure 1—figure supplement 2H ) , implying that direct activation of NPRA/NPRC itself is insufficient to initiate scratching response . This raises the question as to the specific role BNP may play in the first phase , which is physiologically relevant to the histamine and CQ itch that usually occurs within this period . We suspected that BNP may play a modulatory function in acute itch behavior in a manner resembling the role of serotonin in itch modulation ( Zhao et al . , 2014a ) . To test this , we pre-treated mice with BNP at a lower dose ( 30 µM , i . t . ) followed by i . d . histamine at a dose of 100 µg that is insufficient to induce robust scratching behaviors . The low-dose effect was similar to the saline control , which enabled us to examine the facilitatory effect , rather than the additive effect ( Figure 3A–D ) . At 30 µM , BNP failed to induce scratching behaviors ( Figure 1B ) . Strikingly , histamine-induced scratching responses were significantly enhanced with BNP pretreatment compared with saline control ( Figure 3A ) . BNP also produced similar potentiating effects on CQ-induced scratching responses ( Figure 3B ) . Since NMB is required for histamine itch via NMBR exclusively ( Wan et al . , 2017; Zhao et al . , 2014b ) , we tested the possibility that BNP may facilitate histamine itch by modulating NMBR function . At 0 . 05 nmol , i . t . NMB itself could not induce significant scratching behavior ( Figure 3C ) . However , co-injection of BNP ( 30 µM ) and NMB ( 0 . 05 nmol ) markedly increased NMB-induced scratching behavior compared with that of mice receiving only NMB ( Figure 3C ) . Importantly , BNP failed to potentiate scratching behaviors induced by GRP ( 0 . 01 nmol , i . t . ) ( Figure 3D ) . Next we assessed whether BNP may function upstream or independently of GRPR to modulate itch by comparing the facilitatory effect of BNP on histamine itch between Grpr KO and WT mice . If BNP acts upstream of or depends on GRPR , BNP may fail to potentiate histamine itch in Grpr KO mice . We found that BNP similarly potentiated histamine itch in Grpr KO mice ( Figure 3E ) , consistent with the findings that GRP-GRPR signaling is not required for histamine itch ( Sun et al . , 2009 ) . Together , these results show that the role of BNP signaling in the spinal cord is dependent on NMB-NMBR signaling and independent of GRP-GRPR signaling . Next , we evaluated whether NMB has a modulatory function that resembles BNP in histamine itch . Mice pretreated with NMB ( 0 . 05 nmol , i . t . ) did not exhibit enhanced scratching behaviors evoked by either histamine or CQ ( Figure 3F and G ) , ruling out the possibility that NMB would function as a modulator . To determine whether NPRC facilitates histamine itch , we tested the effect of ANP-4–23 , a selective NPRC receptor agonist ( Maack et al . , 1987 ) , on NMB-evoked scratching behavior . Although ANP-4–23 or NMB at the low dose could not induce substantial scratching behavior individually , their combined administration ( i . t . ) evoked robust scratching behavior ( Figure 3H ) , demonstrating that ANP-4–23 could potentiate NMB action . To further determine the role of NPRC in itch , we pharmacologically inhibited NPRC with AP 811 ( i . t ) , a highly selective NPRC antagonist ( Koyama et al . , 1994 ) , and found that AP 811 significantly reduced BNP facilitated histamine itch ( Figure 3I ) . Given that NMBR acts through the canonical Gq-PLC-Ca2+ signaling in histamine itch ( Wan et al . , 2017; Zhao et al . , 2014b ) , we also tested whether NPRC is coupled to NMBR to facilitate histamine itch . Indeed , BNP-facilitated histamine itch was markedly reduced by U 73122 treatment , a selective PLC inhibitor ( Figure 3I ) , suggesting an intracellular coupling between NPRC and NMBR . We previously showed that NMB exerts its role exclusively through NMBR in the spinal cord , as NMB is a functional antagonist for GRPR in spite of its cross-binding activity with GRPR in the spinal cord ( Zhao et al . , 2014b ) . Double RNAscope ISH showed that ~ 60% of Nppb neurons co-expressed Nmb in the DRG ( Figure 3J and K ) , whereas Grp neurons in DRGs showed minimal co-expression ( ~19% ) with Nppb ( Figure 3L and M ) . This prompted us to test whether BNP may facilitate NMB/histamine itch signaling through crosstalk between NMBR , which is required for histamine itch , and NPRA or NPRC , two receptors that bind to BNP . To test whether BNP can potentiate NMBR function , we took advantage of the fact that NMB exclusively activates NMBR neurons in the spinal cord ( Zhao et al . , 2014b ) and examined the response of NMBR neurons to NMB using Ca2+ imaging of dorsal horn neurons ( Munanairi et al . , 2018 ) . NMBR functions via the canonic Gq coupled PLCβ-Ca2+ signaling analogous to GRPR ( Liu et al . , 2011; Zhao et al . , 2014a ) . Using a protocol for investigating facilitating effect ( Zhao et al . , 2014a ) , we found that NMB at 20 nM , but not at 10 nM , was able to induce Ca2+ transient in perspective NMBR neurons identified with the first application ( Figure 4A and B ) . BNP by itself rarely induce Ca2+ transient , regardless of the dose , in these neurons . However , when BNP ( 200 nM ) was co-applied with a subthreshold concentration of NMB ( 10 nM ) , it dramatically potentiated Ca2+ transients in response to the second NMB application ( Figure 4B ) . Overall , of 1513 neurons analyzed , 100 responded to 20 nM NMB ( 6 . 6 % ) and 16 to BNP ( 200 nM , 1 . 1 % ) . Thus were classified as NMBR neurons . From which , 16 responded to co-application of BNP ( 200 nM , 1 . 1 % ) . From 33 NMBR neurons used in this study , 8 ( 24% ) were potentiated by co-application of NMB ( 10 nM ) and BNP . Notably , the percentage of NMBR neurons that responded to both NMB and BNP is largely consistent with the finding that 29% of which express Npr3 . To further evaluate the response specificity of NMBR neurons to NMB , we analyzed the response of the dorsal horn neurons lacking Nmbr to NMB ( up to 20 nM ) . Importantly , none of NMBR KO neurons showed Ca2+ transients ( n = 196 neurons ) , albeit they all responded to KCL ( Figure 4C ) . Together , these results confirmed the identity as well as response specificity of NMBR neurons to NMB ( Figure 4C ) . To probe the possibility of NPRC-NMBR crosstalk , we took advantage of the fact that HEK 293 cells express endogenous Npr1 and Npr3 as shown by qRT-PCR ( Figure 4—figure supplement 1 ) . In HEK 293 cells stably expressing NMBR , application of neither BNP ( 1 µM ) nor NMB ( 1 pM ) induced Ca2+ response , whereas their co-application evoked robust Ca2+ transients ( Figure 4D ) . Importantly , the effect of BNP was greatly attenuated by Npr3 siRNA treatment ( Figure 4D ) , indicating that BNP facilitates NMB/NMBR signaling through NPRC . NPRC has been linked to the inhibition of adenylate cyclase ( AC ) /cAMP signaling and can activate pertussis toxin ( PTX ) -sensitive Gαi/βγ signaling pathway ( Anand-Srivastava et al . , 1990 ) . To examine whether the Gαi/βγ pathway is involved in the facilitatory effect of BNP/NPRC , HEK 293 cells were treated with pertussis toxin ( PTX ) to inactivate Gαi protein ( Murayama and Ui , 1983 ) . Subsequent incubation of BNP and NMB induced much smaller Ca2+ spikes compared to control cells ( Figure 4E ) . The amplitude of intracellular Ca2+ concentrations ( [Ca2+]i ) was significantly reduced by PTX treatment ( Figure 4F ) . Pre-incubation of gallein , a small molecule Gβγ inhibitor , also blocked the facilitation effect of BNP on NMB-induced calcium spikes ( Figure 4E and F ) . Consistently , i . t . gallein markedly attenuated the facilitatory effect of BNP on histamine itch ( Figure 4G ) . Collectively , these results indicate that BNP-NPRC signaling potentiates NMB/NMBR signaling via Gαi/βγ signaling . BNP-saporin ( BNP-sap ) has been used to ablate NPRA neurons in the spinal cord ( Mishra and Hoon , 2013 ) . Nevertheless , the expression of NPRC in the dorsal horn has raised the question of whether BNP-sap may additionally ablate neurons expressing NPRC ( Figure 1H–N ) . Because BNP-sap at 5 µg , as described previously ( Mishra and Hoon , 2013 ) , resulted in the lethality of WT mice , we reduced the dose to 2 . 5 µg so that enough animals could survive for behavioral and molecular analysis . RNAscope ISH showed that the number of Npr1+ + were reduced to ~50% ( 82 . 9 ± 3 . 3 in control vs . 39 . 8 ± 2 . 2 in BNP-sap ) ( Figure 5A , F ) , whereas Npr2+ neurons were not affected ( 288 . 0 ± 18 . 2 in control vs . 300 . 5 ± 7 . 8 in BNP-sap ) ( Figure 5B and F ) . Moreover , BNP-sap ablated ~67% of Npr3+ neurons ( 41 . 8 ± 1 . 4 in control vs . 13 . 8 ± 1 . 8 in BNP-sap ) as well as ~37% of Grp+ + ( 44 . 6 ± 1 . 9 in control vs . 22 . 6 ± 2 . 0 in BNP-sap ) ( Figure 5C , D and F ) . As expected , the number of Nmbr+ neurons was also significantly reduced after BNP-sap injection , likely due to Npr3 expression in these neurons ( Figure 5E and F ) . A premise for ablation of neurons with peptide-conjugated saporin approach is the internalization of the receptor upon binding to the saporin , resulting in cell death ( Wiley and Lappi , 2003 ) . To test whether BNP can also internalize NPRB and NPRC , HEK 293 cells were transfected with Npr1 , 2 , and 3 cDNA tagged with mCherry ( mCh ) separately . Indeed , BNP internalized NPRA and NPRC , but not NPRB , in HEK 293 cells ( Figure 5G ) , indicating that BNP-sap could ablate both NPRA and NPRC cells . Interestingly , behavioral studies showed that histamine itch was significantly reduced in BNP-sap mice ( Figure 5H ) , whereas CQ itch was not affected ( Figure 5I ) . These results suggest that NPRA and NPRC neurons in the spinal cord play an important role in histamine itch , which can be attributed to partial ablation of NMBR neurons . To explore whether NPRA is important for the development of chronic itch mediated by nonhistaminergic mechanisms , we used the mouse dry skin model deprived of histaminergic component ( Miyamoto et al . , 2002 ) . From day 8 , Npr1 KO mice appeared to show a tendency toward reduced scratching behavior relative to their WT littermates . The differences , however , were not statistically significant ( Figure 6—figure supplement 1 A ) . The normal dry skin itch in Npr1 KO mice prompted us to examine Nppb expression in DRGs of dry skin mice . Real-time RT-PCR results showed that the levels of Nppb mRNA in dry-skin mice were either significantly reduced or not changed depending on the day examined ( Figure 6—figure supplement 1B , data not shown ) . Similarly , dry-skin mice also showed unchanged expression of somatostatin ( Sst ) , a peptide largely co-localized with Nppb in DRG neurons ( Huang et al . , 2018; Stantcheva et al . , 2016 ) and Tac1 ( Figure 6—figure supplement 1B-D ) . In contrast , Grp and Nmb expression levels in mice with dry skin were increased by 447% and 87% , respectively ( Figure 6—figure supplement 1B ) . These findings suggest that BNP-NPRA signaling is not required for the development of dry skin itch . SST type two receptor ( SST2R ) is expressed in GABAergic neurons in the spinal cord and has been considered to be a sole receptor for SST ( Kardon et al . , 2014; Polgár et al . , 2013 ) . To test whether SST2R neurons may inhibit both itch and pain transmission , we pharmacologically inhibit these neurons by i . t . SST injection followed by validating the nature of evoked scratching behavior since the injection onto the nape may induce itch- , pain-related or undefined scratching behavior ( Shimada and LaMotte , 2008 ) . I . t . SST-evoked scratching was markedly reduced but not abolished by intraperitoneal injection ( i . p . ) of morphine ( Figure 6A ) , a method used to evaluate whether i . t . induced scratching/biting behavior reflects pain ( Hylden and Wilcox , 1981 ) . In addition , scratching behaviors evoked by i . t . injection of SST and octreotide ( OCT ) , a selective SST2R agonist , were significantly attenuated on mice with bombesin-saporin ( BB-sap ) treatment , which can completely block nonhistaminergic itch transmission ( Sun et al . , 2009; Figure 6B and C ) . These results suggest that the pharmacological inhibition of SST2R neurons could result in disinhibition of both itch and pain transmission . We then evaluated the expression of Nppb and Sst in neuropathic itch using BRAFNav1 . 8 mice that developed spontaneous scratching behavior resulting from enhanced expression of itch-sensing peptides/receptors in sensory neurons ( Zhao et al . , 2013 ) . Interestingly , Nppb was dramatically down-regulated , whereas Sst was barely detectable in DRG neurons of BRAFNav1 . 8 mice ( Figure 6—figure supplement 1C , D , E ) . The reduced Sst expression may reflect the dampening effect of the dorsal horn GABAergic neuronal activity under neuropathic itch conditions . Thus , BNP and SST are not involved in the development of dry skin and neuropathic itch in mice . If a subset of primary afferents exclusively expresses itch- but little pain-related neuropeptides , it can be predicted that cutaneous activation of these afferents would evoke itch-related scratching behavior . For example , optical activation of the skin innervated by Grp primary afferents evoked frequency-dependent itch-related scratching behavior ( Barry et al . , 2020 ) . To further explore whether BNP-expressing afferents are itch-specific , we used Sst-Cre mice as a surrogate to perform optical stimulation of skin-innervating Sst-expressing fibers . Sst-Cre mice were crossed with Ai32 mice to generate Sst-ChR2 mice that express channelrhodopsin-2/EYFP or ChR2/EYFP in the Sst locus , as confirmed by highly overlapping expression of YFP and Sst ( Figure 6F ) . Sst-ChR2 or Sst-cre mice were stimulated with 473 nm blue light with a fiber optic held just above the nape skin ( 15 mW power from fiber tip ) at 1 , 5 , 10 , or 20 Hz with a 3 s On-Off cycle for 5 minutes ( Figure 6D and Figure 6—video 1 ) . Stimulation at all frequencies failed to evoke significant scratching behaviors in Sst-ChR2 mice compared to Sst-cre mice ( Figure 6E ) . At last , we analyzed the expression of Sst-ChR2 in DRGs . Consistent with previous studies ( Stantcheva et al . , 2016 ) , we found that Sst-ChR2 sensory neurons do not co-express the peptidergic marker calcitonin gene-related peptide ( CGRP ) , nor do they show the non-peptidergic Isolectin B4 ( IB4 ) -binding ( Figure 6G ) . However , some Sst-ChR2 sensory neurons do co-express the myelinated marker neurofilament heavy ( NF-H ) and transient receptor potential cation channel subfamily V member 1 ( TRPV1 ) ( Figure 6G ) . Examination of the hairy nape skin revealed that expression of Sst-ChR2 in the epidermis as well as expression in some hair follicles of the dermis within apparent lanceolate endings ( Figure 6H ) .
In this study , we demonstrate that in sensory neurons , BNP can function as a neuromodulator , rather than a neurotransmitter , to facilitate itch transmission . We also show that NPRC in the spinal cord is crucial for mediating BNP-facilitated histaminergic itch . Ca2+ signaling is a hallmark feature of neuronal activation ( Berridge , 1998 ) . By demonstrating that BNP alone could not activate Ca2+ transients either in HEK293 cells or in the dorsal horn neurons , we verify that BNP is an inhibitory neuropeptide , in line with the observation that i . t . BNP , even at very high doses ( 2 . 5–5 µg ) , fails to induce the rapid onset of scratching behavior , typical for ligand-mediated acute activation of excitatory itch receptors in the spinal cord ( Sun and Chen , 2007; Wan et al . , 2017; Zhao et al . , 2014b ) . Such high doses therefore most likely fall outside the range of its endogenous concentration ( e . g . at picomolar concentrations ) required for mediating acute itch transmission directly ( Wan et al . , 2017 ) . Although Npr1 siRNA knockdown in both spinal cord and DRGs makes it difficult to ascribe itch deficits to either region , the observation that BNP-sap treatment , in spite of partial ablation , attenuates histamine but not CQ itch suggests that NPRA in DRGs , rather than in the spinal cord , is involved in CQ itch . Whether NPRA may facilitate histaminergic itch in DRGs awaits further clarification . Given a very small fraction of spinal NPRA neurons also express NMBR , the possibility that BNP-NPRA signaling may marginally contribute to histaminergic itch cannot be excluded . Coupled with previous studies suggesting the inhibitory effect of BNP on nociceptive neurons ( Liu et al . , 2014; Zhang et al . , 2010 ) , it remains plausible that NPRA in DRGs may exert opposing functions by facilitating itch while inhibiting certain types of inflammatory pain . Nonetheless , the observation that most Nppb neurons express little GRP implies that BNP/NMB and GRP released from distinct types of primary afferents may be responsible for the activation of the dorsal horn neurons expressing NPRC/NMBR and GRPR , respectively . Perhaps , the most unexpected finding is that the BNP-NPRC signaling facilitates histamine itch through NPRC-NMBR crosstalk . Given that PTX acts on Gαi directly in the Gαiβγ heterotrimer ( Smrcka , 2008 ) and gallein acts on Gβγ directly , the findings that inhibition of Gαi and Gβγ signaling similarly attenuates the facilitatory effect of BNP on NMB-induced Ca2+ response in NPRC/NMBR cells and BNP-facilitated histamine itch suggest an intracellular coupling of the Gαi/βγ signaling pathway to PLCβ signaling downstream of NMBR ( Figure 7A ) . Numerous studies have shown that inhibitory GI-coupled receptors enhance or facilitate the activation of the canonical PLCβ−Ca2+ signaling transduction pathway downstream of Gq-coupled receptors ( Prezeau et al . , 2010; Werry et al . , 2003 ) . For example , we have shown that inhibitory receptors such as 5-HT1A and μ-opioid receptor isoform MOR1D in mice or MOR1Y in humans could facilitate or activate Ca2+ signaling downstream of GRPR via intracellular Gi-Gq crosstalk ( Liu et al . , 2019; Liu et al . , 2011; Zhao et al . , 2014a ) . Nevertheless , NPRC-NMBR represents the first example of the crosstalk between a non-GPCR inhibitory receptor and an excitatory GPCR . The attenuation of NPRC-GRPR crosstalk by pharmacological inhibition of either Gβγ or PLCβ signaling suggests that , irrespective of the receptor type , Gi-Gq coupling leading to Ca2+ mobilization are versatile and universal mechanisms . In congruence with this , even the rare Gi-Gβγ-PLCβ-Ca2+ signaling pathway that was considered to be a stand-alone paradigm was recently found to be dependent on Gq signaling ( Pfeil et al . , 2020 ) . On the other hand , partial attenuation of BNP-facilitated histaminergic itch by PTX and gallein implies that Gβγ-independent signaling mechanisms may also have a role in NPRC-NMBR crosstalk . It is intriguing that ANP may not play a role in facilitating histamine itch , even though it can bind to NPRC with equal potency as BNP . It is possible that ANP may interact with the receptor at different binding sites ( Savoie et al . , 1995 ) . Alternatively , it could be due to a much faster clearance rate of ANP ( 0 . 5–4 min ) than BNP ( four to more than 20 min ) or CNP ( Potter , 2011 ) . Our study clarifies the role of SST and Nppb/Sst fibers in itch . The finding that cutaneous activation of BNP/SST fibers failed to evoke scratching behavior suggests that activation of these fibers does not transmit itch information as a functional entity . These results may seem surprising in light of robust scratching behavior evoked by i . t . SST . However , they are consistent with the observation that DRG-specific deletion of Sst did not influence itch transmission ( Huang et al . , 2018 ) . Although diminished itch behavior of Wnt1Cre/Sstf/f mice was used to argue for a role of SST in itch disinhibition in sensory neurons ( Huang et al . , 2018 ) , the deficit can be ascribed to deletion of Sst in the brain regions , where Wnt1 is expressed in numerous neural precursors ( Lewis et al . , 2013 ) . It is possible that cutaneous activation of Nppb/Sst fibers may provoke the release of BNP/SST onto the spinal cord; However , their endogenous release may not be sufficient to evoke scratching behavior . Together , these loss- and gain-of-function studies indicate that itch-related scratching behavior evoked by i . t . SST represents a pharmacological artifact . Interestingly , conditional knockout of Sst in DRGs attenuated thermal and mechanical pain ( Huang et al . , 2018 ) , suggesting that SST release may contribute to certain types of nociceptive transmission . Consistently , the downregulation of Sst in mice with chronic itch may reflect a dampening effect of spinal inhibitory neuronal activity for nociceptive transmission under pathological itch conditions . Thus , the role of SST in sensory neurons is limited to the disinhibition of certain types of pain by inhibiting SST2R inhibitory neurons gating nociceptive transmission ( Figure 7—figure supplement 1 ) . This interpretation is in line with recent studies suggesting that spinal inhibitory neurons that gate itch and pain , respectively , are anatomically segregated ( Chen , 2021 ) . Most human chronic itch conditions are resistant to antihistamines , which is a major focus of current research due to clinical implications . However , current mouse models of chronic itch , including allergic contact dermatitis ( ACD ) and atopic dermatitis ( AD ) , are additionally mediated by histamine-dependent mechanisms , as they are chemically induced allergic inflammatory responses that involve histamine release from mast cells ( Shim and Oh , 2008; Wang and Kim , 2020 ) . In these mouse models , BNP may be upregulated in DRGs as a result of mast cell degranulation and may facilitate histaminergic itch ( Liu et al . , 2020; Solinski et al . , 2019 ) . However , it is either downregulated or not altered in a dry skin itch model , which is mediated exclusively through GRP-dependent nonhistaminergic mechanism ( Akiyama et al . , 2010; Miyamoto et al . , 2002; Zhao et al . , 2013 ) . Consistently , NPRA is dispensable for the development of dry skin itch and Nppb is downregulated in a neuropathic itch model as shown in the present study . In summary , we demonstrate a novel BNP-NPRC-NMBR crosstalk in the modulation of itch transmission ( Figure 7A ) . Considering that BNP potently enhances NMB function , but not vice versa , these studies suggest that neuropeptides in sensory neurons either encode or modulate itch information ( Figure 7B ) . Our studies reveal an unexpected role of NPRC in facilitating NMB-mediated histaminergic itch transmission ( Figure S4 ) . These results delineate distinct functions of GRP , NMB , and BNP in histaminergic and nonhistaminergic itch and highlight the different modes of action for neuropeptides in the coding and modulating of itch ( Figure 7B; Chen , 2021 ) .
Male mice between 7 and 12 weeks of age were used for behavioral experiments . C57BI/6 mice were purchased from The Jackson Laboratory ( http://jaxmice . jax . org/strain/013636 . html ) . Npr1 KO mice ( Oliver et al . , 1997 ) , Grpr KO mice ( Hampton et al . , 1998 ) , Nmbr KO mice ( Ohki-Hamazaki et al . , 1999 ) , and their wild-type ( WT ) littermates were used . We cross SstCre mice ( Taniguchi et al . , 2011 ) with a flox-stop channel rhodopsin-eYFP ( ChR2-eYFP ) line ( Ai32 ) ( Madisen et al . , 2012 ) to generate mice with ChR2-eYFP expression in Sst neurons ( Sst-ChR2 ) . All experiments were performed in accordance with the guidelines of the National Institutes of Health and the International Association for the Study of Pain and were approved by the Animal Studies Committee at Washington University School of Medicine . The dose of drugs and injection routes are indicated in figure legends . ANP , BNP , CNP , SST , and OCT were supplied from GenScript USA Inc ( Piscataway , NJ ) . GRP18-27 , NMB and ANP-4–23 were from Bachem ( King of Prussia , PA ) . Histamine , chloroquine , Npr1 siRNA , Npr2 siRNA , Npr3 siRNA , and scrambled control siRNA were purchased from Sigma ( St . Louis , MO ) . BNP-saporin ( BNP-sap ) and blank-saporin were made by Advanced Targeting Systems . Pertussis toxin ( PTX ) and gallein were from R&D Systems ( Minneapolis , MN ) or Selleck ( Houston , TX ) . AP 811 was from Tocris ( Minneapolis , MN ) . U73122 was from Selleck ( Houston , TX ) . Behavioral tests were videotaped ( HDR-CX190 camera , Sony ) from a side angle . The videos were played back on the computer and the quantification of mice behaviors was done by persons blinded to the treatments and genotypes . Itch behaviors were performed as previously described ( Sun and Chen , 2007; Zhao et al . , 2014b ) . Briefly , mice were given 30 min to acclimate to the plastic arenas ( 10 × 10 . 5 × 15 cm ) . Mice were then briefly removed from the chamber for drug injections . Injection volume was 10 µL for i . t . injection and 50 µL for i . d . injection . Doses of drugs are indicated in figure legends . The number of scratching responses was counted for 30 min at 5 min intervals . One scratching bout is defined as a lifting of the hind limb toward the body and then a replacing of the limb back to the floor or the mouth , regardless of how many scratching strokes take place between those two movements . Scratching toward the injection site was counted after i . d . injection , and all scratching bouts were counted after i . t . injection . Dry Skin ( Xerosis ) : The dry skin model was set up as described ( Akiyama et al . , 2010; Miyamoto et al . , 2002 ) . Briefly , the nape of mice was shaved , and a mixture of acetone and diethyl ether ( 1:1 ) was painted on the neck skin for 15 s , followed by 30 s of distilled water application ( AEW ) . This regimen was administrated twice daily for 9 days . Spontaneous scratching behaviors were recorded for 60 min on the morning before AEW treatment . BRAFNav1 . 8 mice were generated as described previously ( Zhao et al . , 2013 ) . RNAscope ISH was performed as described ( Munanairi et al . , 2018; Wang et al . , 2012 ) . Briefly , mice were anesthetized with a ketamine/xylazine cocktail ( ketamine , 100 mg/kg and xylazine , 15 mg/kg ) and perfused intracardially with 0 . 01 M PBS , pH 7 . 4 , and 4% paraformaldehyde ( PFA ) . The spinal cord was dissected , post-fixed in 4% PFA for 16 hr , and cryoprotected in 20% sucrose overnight at 4 °C . Tissues were subsequently cut into 18-μm-thick sections , adhered to Superfrost Plus slides ( Fisher Scientific ) , and frozen at −20 °C . Samples were processed according to the manufacturer’s instructions in the RNAscope Fluorescent Multiplex Assay v2 manual for fixed frozen tissue ( Advanced Cell Diagnostics ) , and coverslipped with Fluoromount-G antifade reagent ( Southern Biotech ) with DAPI ( Molecular Probes ) . The following probes , purchased from Advanced Cell Diagnostics were used: Nppb ( nucleotide target region 4–777; accession number NM_008726 . 5 ) , Sst ( nucleotide target region 18–407; accession number NM_009215 . 1 ) , Npr1 ( nucleotide target region 941–1882; accession number NM_008727 . 5 ) , Npr2 ( nucleotide target region 1162–2281; accession number NM_173788 . 3 ) , Npr3 ( nucleotide target region 919–1888; accession number NM_008728 . 2 ) , Grp ( nucleotide target region 22–825; accession number NM_175012 . 2 ) , Grpr ( nucleotide target region 463–1596; accession number - NM_008177 . 2 ) , Nmbr ( nucleotide target region 25–1131; accession number NM_008703 . 2 ) , Vgat ( Slc32a1 , nucleotide target region 894–2037; accession number NM_009508 . 2 ) , and Vglut2 ( Slc17a6 , nucleotide target region 1986–2998; accession number NM_080853 . 3 ) . Sections were subsequently imaged on a Nikon C2+ confocal microscope ( Nikon Instruments , Inc ) in three channels with a 20 X objective lens . The criterion for including cells as positive for a gene expression detected by RNAscope ISH: A cell was included as positive if two punctate dots were present in the nucleus and/or cytoplasm . For co-localization studies , dots associated with single DAPI stained nuclei were assessed as being co-localized . Cell counting was done by a person who was blinded to the experimental design . ISH was performed using digoxigenin-labeled cRNA probes as previously described ( Chen et al . , 2001 ) . Briefly , mice were anesthetized with an overdose of a ketamine/xylazine cocktail and fixed by intracardiac perfusion of cold 0 . 01 M PBS , pH 7 . 4 , and 4% paraformaldehyde . The spinal cord , DRG , and hairy nape skin tissues were immediately removed , post-fixed in the same fixative overnight at 4 °C , and cryoprotected in 20% sucrose solution . DRGs , lumbar spinal regions and hairy nape skin were frozen in OCT and sectioned at 20 µm thickness on a cryostat . Immunohistochemical ( IHC ) staining was performed as described ( Zhao et al . , 2007 ) . Spinal cord and DRG tissues were sectioned at 20 µm thickness . Hairy nape skin was sectioned at 30 µm thickness . Free-floating sections were incubated in a blocking solution containing 2% donkey serum and 0 . 1% Triton X-100 in PBS ( PBS-T ) for 2 h at room temperature . The sections were incubated with primary antibodies overnight at 4 °C , washed three times in PBS , incubated with the secondary antibodies for 2 hr at room temperature , and washed three times . Sections were mounted on slides with Fluoromount G ( Southern Biotech ) and coverslips . Fluorescein isothiocyanate ( FITC ) -conjugated Isolectin B4 from Griffonia simplicifolia ( IB4 , 10 µg/mL; L2895 , Sigma ) , IB4-AlexaFluor 568 conjugate ( 2 μg/mL , ThermoFisher Scientific ) or the following primary antibodies were used: rabbit anti-CGRPα ( 1:3000; AB1971 , Millipore ) , guinea pig anti-Substance P ( 1:1000; ab10353 , Abcam ) , guinea pig anti-TRPV1 ( 1:1000; GP14100 , Neuromics ) , chicken anti-NF-H ( 1:2000 , EnCor Biotechnology , CPCA-NF-H ) , rabbit anti-βIII-Tubulin ( 1:2000 , Biolegend , 802001 ) , rabbit anti-GFP ( 1:1000 , Molecular Probes , A11122 ) , chicken anti-GFP ( 1:500 , Aves Labs , GFP-1020 ) . The secondary antibodies were purchased from Jackson ImmunoResearch Laboratories , including Cyanine 3 ( Cy3 ) , Cyanine 5 ( Cy5 ) - or FITC conjugated donkey anti-rabbit , anti-mouse , anti-chicken or anti-guinea pig IgG ( Cy3 , 0 . 5 µg/ml; FITC , 1 . 25 µg/mL ) , biotin-SP ( long-spacer ) -conjugated donkey anti-rabbit IgG ( 1 µg/mL ) and avidin-conjugated Alexa Fluor 488 ( 0 . 33 µg/mL ) . Images were taken using a Nikon Eclipse Ti-U microscope with Cool Snap HQ Fluorescent Camera and DS-U3 Brightfield Camera controlled by Nikon Elements Software ( Nikon ) or a Leica TCS SPE confocal microscope with Leica LAS AF Software ( Leica Microsystems ) . The staining was quantified by a person blinded to the genotype using ImageJ ( version 1 . 34e , NIH Image ) as previously described ( Zhao et al . , 2013 ) . Images were taken using a Nikon C2+ confocal microscope system ( Nikon Instruments , Inc ) and analysis of images was performed using ImageJ software from NIH Image ( version 1 . 34e ) . At least 3 mice per group and 10 lumbar sections across each group were included for statistical comparison . Negative control siRNA ( SIC001 ) and selective siRNA duplex for mouse Npr1 ( SASI_Mm01_00106966 ) , mouse Npr2 ( SASI_Mm01_00201357 ) , and mouse Npr3 ( SASI_Mm01_00036567 ) were purchased from Sigma . RNA was dissolved in diethyl pyrocarbonate-treated PBS and prepared immediately prior to administration by mixing the RNA solution with a transfection reagent , RVG-9R ( Genscript ) . The final concentration of RNA was 2 µg/10 µL . siRNA was delivered to the lumbar region of the spinal cord . The injection was given once daily for 6 consecutive days as described previously with some modifications ( Liu et al . , 2011; Luo et al . , 2005; Tan et al . , 2005 ) . Behavior testing was carried out 24 hr after the last injection . Mice were treated with one-time i . t . injection of BNP-sap ( 2 . 5 µg/mouse ) as previously described ( Mishra and Hoon , 2013 ) with the reduced dose , due to the lethal effect of BNP-sap ( 5 µg ) . Behavioral tests were performed 2 weeks after BNP-sap injection . After behavioral tests , the spinal cord/DRGs of mice were processed for real-time RT-PCR and ISH . Real-time RT-PCR was performed as previously described with Fast-Start Universal SYBR Green Master ( Roche Applied Science ) ( Liu et al . , 2011; Liu et al . , 2014 ) . All samples were assayed in duplicate ( heating at 95 °C for 10 s and at 60 °C for 30 s ) . Data were analyzed using the Comparative CT Method ( StepOne Software version 2 . 2 . 2 . ) , and the expression of target mRNA was normalized to the expression of Actb and Gapdh . The following primers were used: Nppb ( NM_008726 . 4 ) : 5’- gtcagtcgtttgggctgtaac-3’ , 5’- agacccaggcagagtcagaa-3’; amplicon size , 89 bp; Sst ( NM_009215 . 1 ) : 5’- CCCAGACTCCGTCAGTTTCT –3’ , 5’- CAGCAGCTCTGCCAAGAAGT –3’ , amplicon size , 87 bp; Npr1 ( NM_008727 . 5 ) : 5’- tggagacacagtcaacacagc-3’ , 5’- cgaagacaagtggatcctgag-3’; amplicon size , 70 bp; Npr2 ( NM_173788 . 3 ) : 5’- tgagcaagccacccactt-3’ , 5’- agggggccgcagatatac-3’ , amplicon size , 60 bp; Npr3 ( NM_008728 . 2 ) : 5’- TGCACACGTCTGCCTACAAT-3’ , 5’- GCACCGCCAACATGATTCTC –3’ , amplicon size , 138 bp; Grpr ( NM_008177 . 2 ) : 5’-TGATTCAGAGTGCCTACAATCTTC-3’ , 5’-TTCCGGGATTCGATCTG-3’; amplicon size , 71 bp; Nmbr ( NM_008703 . 2 ) : 5’- gggggtttctgtgttcactc –3’ , 5’- catggggttcacgatagctc –3’ , amplicon size , 67 bp; Actb ( NM_007393 . 3 ) : 5’-TGTTACCAACTGGGACGACA-3’ , 5’-GGGGTGTTGAAGGTCTCAAA-3’; amplicon size , 166 bp; and Gapdh ( NM_008084 . 2 ) : 5’-CCCAGCAAGGACACTGAGCAA-3’ , 5’-TTATGGGGGTCTGGGATGGAAA-3’; amplicon size , 93 bp . Primary culture of spinal dorsal horn neurons was prepared from 5- to 7-day-old C57BI/6 mice ( Zhao et al . , 2014b ) and NMBR WT and KO mice ( Zhao et al . , 2014b ) . The protocol is essentially the same as previously described ( Munanairi et al . , 2018 ) . After decapitation , laminectomy was performed , and the dorsal horn of the spinal cord was dissected out with a razor blade and incubated in Neurobasal-A Medium ( Gibco ) containing 30 μL papain ( Worthington ) at 37 °C for 20 min . Enzymatic digestion was stopped by replacing it with 1 ml Neurobasal-A medium . After washing with the same medium three times , gentle trituration was performed using a flame polished glass pipette until the solution became cloudy . The homogenate was centrifuged at 1500 rpm for 5 min , and the supernatant was discarded . Cell pellet was re-suspended in culture medium composed of Neurobasal-A medium ( Gibco , 92% vol/vol ) , fetal bovine serum ( Invitrogen , 2% vol/vol ) , HI Horse Serum ( Invitrogen , 2% vol/vol ) , GlutaMax ( 2 mM , Invitrogen , 1% vol/vol ) , B27 ( Invitrogen , 2% vol/vol ) , Penicillin ( 100 μg/mL ) and Streptomycin ( 100 μg/mL ) and plated onto 12 mm coverslips coated with poly-D-lysine . After three days of culture , neurons were used for calcium imaging as described previously ( Zhao et al . , 2014b ) . HEK 293 cells were purchased from ATCC ( Cat . CRL-1573 ) with a certificate of analysis confirming cell line identity by STR profiling and confirming lack of mycoplasma contamination . All experiments were carried out on cells cultured for less than ten passages from the purchased stock . HEK 293 cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum in a humidified atmosphere containing 5% CO2 . Stable HEK293 cell lines were made as described previously ( Liu et al . , 2011 ) . Briefly , cells were transfected with pcDNA3 . 1/NMBR , pcDNA3 . 1/NPR1-mCherry , pcDNA3 . 1/NPR2-mCherry , or pcDNA3 . 1/NPR3-mCherry by electroporation ( GenePulser Xcell , Bio-Rad ) . Stable transfectants were selected in the presence of 500 µg/ml G418 ( Invitrogen ) . For internalization assay , HEK 293 cells expressing mCherry-tagged receptors were plated on glass bottom dishes coated with poly-D-lysine overnight and imaged every 10 min for 30 min using a Nikon C2+ confocal microscope system ( Nikon Instruments , Inc ) in the presence of 10 µM BNP . For calcium imaging assay , HEK 293 cells expressing NMBR were plated onto 12 mm coverslips coated with poly-D-lysine . Overnight cell cultures were loaded with Fura 2-acetomethoxy ester ( Molecular Probes ) for 30 min at 37 °C . After washing , neurons or HEK 293-NMBR cells were imaged at 340 and 380 nm excitation to detect intracellular free calcium ( Zhao et al . , 2014b ) . PTX ( 200 ng/ml ) , gallein ( 0 . 1 mM or 2 mM ) , or AP 811 ( 0 . 1 µM ) were pre-incubated with NMBR cells for 10 min to block Gαi , Gβγ , or NPRC , respectively . Sst-ChR2 mice and wild-type littermates ( Sst-cre ) were used for optical skin stimulation experiments . The nape skin was shaved 3 days prior to stimulation in all mice tested . One day prior to the experiments , each mouse was placed in a plastic arena ( 10 × 11 X 15 cm ) for 30 min to acclimate . For blue light skin stimulation , a fiber optic cable was attached to a fiber-coupled 473 nm blue laser ( BL473T8-150FC , Shanghai Laser and Optics Co . ) with an ADR-800A adjustable power supply . Laser power output from the fiber optic cable was measured using a photometer ( Thor Labs ) and set to 15 mW from the fiber tip . An Arduino UNO Rev three circuit board ( Arduino ) was programmed and attached to the laser via a BNC input to control the frequency and timing of the stimulation ( 1 , 5 , 10 , or 20 Hz with 10ms on-pulse and 3 s On – 3 s off cycle for 5 min ) . During stimulation , the mouse was traced manually by a fiber optic cable with a ferrule tip that was placed 1–2 cm above the nape skin . Videos were played back on a computer for scratching behavior assessments by observers blinded to the animal groups and genotypes . Values are reported as the mean ± standard error of the mean ( SEM ) . Statistical analyses were performed using Prism 6 ( v6 . 0e , GraphPad , San Diego , CA ) . For comparison between two or more groups , unpaired , paired two-tailed t-test , one-way ANOVA followed by Tukey post hoc tests , or two-way repeated-measures ANOVA followed by Sidak’s post hoc analysis , was used . Normality and equal variance tests were performed for all statistical analyses . p < 0 . 05 was considered statistically significant . | An itch is a common sensation that makes us want to scratch . Most short-term itches are caused by histamine , a chemical that is released by immune cells following an infection or in response to an allergic reaction . Chronic itching , on the other hand , is not usually triggered by histamine , and is typically the result of neurological or skin disorders , such as atopic dermatitis . The sensation of itching is generated by signals that travel from the skin to nerve cells in the spinal cord . Studies in mice have shown that the neuropeptides responsible for delivering these signals differ depending on whether or not the itch involves histamine: GRPs ( short for gastrin-releasing proteins ) convey histamine-independent itches , while NMBs ( short for neuromedin B ) convey histamine-dependent itches . It has been proposed that another neuropeptide called BNP ( short for B-type natriuretic peptide ) is able to transmit both types of itch signals to the spinal cord . But it remains unclear how this signaling molecule is able to do this . To investigate , Meng , Liu , Liu , Liu et al . carried out a combination of behavioral , molecular and pharmacological experiments in mice and nerve cells cultured in a laboratory . The experiments showed that BNP alone cannot transmit the sensation of itching , but it can boost itching signals that are triggered by histamine . It is widely believed that BNP activates a receptor protein called NPRA . However , Meng et al . found that the BNP actually binds to another protein which alters the function of the receptor activated by NMBs . These findings suggest that BNP modulates rather than initiates histamine-dependent itching by enhancing the interaction between NMBs and their receptor . Understanding how itch signals travel from the skin to neurons in the spinal cord is crucial for designing new treatments for chronic itching . The work by Meng et al . suggests that treatments targeting NPRA , which was thought to be a key itch receptor , may not be effective against chronic itching , and that other drug targets need to be explored . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2021 | BNP facilitates NMB-encoded histaminergic itch via NPRC-NMBR crosstalk |
Expansion of biliary epithelial cells ( BECs ) during ductular reaction ( DR ) is observed in liver diseases including cystic fibrosis ( CF ) , and associated with inflammation and fibrosis , albeit without complete understanding of underlying mechanism . Using two different genetic mouse knockouts of β-catenin , one with β-catenin loss is hepatocytes and BECs ( KO1 ) , and another with loss in only hepatocytes ( KO2 ) , we demonstrate disparate long-term repair after an initial injury by 2-week choline-deficient ethionine-supplemented diet . KO2 show gradual liver repopulation with BEC-derived β-catenin-positive hepatocytes and resolution of injury . KO1 showed persistent loss of β-catenin , NF-κB activation in BECs , progressive DR and fibrosis , reminiscent of CF histology . We identify interactions of β-catenin , NFκB , and CF transmembranous conductance regulator ( CFTR ) in BECs . Loss of CFTR or β-catenin led to NF-κB activation , DR , and inflammation . Thus , we report a novel β-catenin-NFκB-CFTR interactome in BECs , and its disruption may contribute to hepatic pathology of CF .
The liver possesses unique regenerative potential . During chronic liver injury , however , liver fibrosis accompanies regeneration and can progress to cirrhosis , which can then progress to end-stage liver disease ( ESLD ) or hepatocellular cancer ( HCC ) ( Pellicoro et al . , 2014 ) . Currently , cirrhosis is the 11th leading cause of death globally , and the incidence of liver disease continues to rise as conditions such as non-alcoholic fatty liver disease ( NAFLD ) and alcoholic liver disease continue to prevail ( Asrani et al . , 2019 ) . Thus , there has been great interest in studying mechanisms of injury , inflammation , and fibrosis during liver injury in order to effectively develop novel therapies . The role of hepatic epithelial cells ( referred henceforth as ‘hepithelial’ cells ) , which include both hepatocytes and cholangiocytes or biliary epithelial cells ( BECs ) , in regulating microenvironment is beginning to be appreciated . Loss of hepatocyte differentiation in chronic liver diseases and ESLD , either due to much needed hepithelial proliferation for repair , or as an adaptation to escape injury , seems to contribute to not only loss of key hepatic functions , but is also causally associated with increased immune response and hepatic fibrosis ( Argemi , 2019; Nishikawa , 2015 ) . However , how hepithelial cells may modulate hepatic immune microenvironment is unclear . As an important hepithelial cell type , BECs are known to undergo proliferation to replace dying BECs in cholangiopathies or cystic liver diseases , as well as can under phenotypic switch to generate de novo hepatocytes when hepatocytes are chronically injured and/or are unable to optimally proliferate , phenomena termed as ductular reaction ( DR ) ( Sato et al . , 2019; Wilson and Rudnick , 2019 ) . Reactive ductules , however , can secrete pro-inflammatory and pro-fibrotic cytokines to induce inflammation , activate myofibroblasts , and induce fibrosis . The extent of DR correlates with fibrosis in many types of liver injuries ( Lowes et al . , 1999; Richardson , 2007; Zhao et al . , 2018 ) . The molecular underpinnings of reactive DR are incompletely understood although molecules like Yes-associated protein-1 ( YAP1 ) have been implicated ( Planas-Paz , 2019 ) . β-Catenin , the major downstream effector of the Wnt signaling , is a well-known mediator of hepatocyte proliferation . Liver-specific ( hepatocyte and BECs ) β-catenin knockout ( KO1 ) mice generated by breeding β-catenin-floxed and albumin-cre mice show delayed liver regeneration ( LR ) after partial hepatectomy or after toxicant-induced liver injury ( Apte , 2009; Tan et al . , 2006 ) . When KO1 were administered choline-deficient , ethionine-supplemented ( CDE ) diet , it triggered greater steatosis , cell death , DR , inflammation , and fibrosis than wild-type ( WT1 ) , and upon switching to normal diet for 2 weeks for recovery , continued to show greater injury due to an impairment of hepatocyte proliferation ( Akhurst , 2001; Russell , 2019 ) . Similar greater injury , fibrosis , and DR were observed in CDE-fed hepatocyte-only β-catenin KO ( KO2 ) , generated by delivering adeno-associated virus serotype 8 carrying a plasmid encoding Cre recombinase under a hepatocyte-specific thyroxine-binding globulin ( TBG ) promoter ( AAV8-TBG-Cre ) into the β-catenin-floxed mice ( Russell , 2019 ) . Intriguingly , labeling BECs for fate-tracing showed the liver repair to occur through BEC-to-hepatocyte transdifferentiation upon recovery for 2 weeks and up to 6 months on normal diet , although long-term impact on injury resolution , inflammation , DR , and fibrosis was not studied in either model ( Russell , 2019 ) . In the current study , we investigate hepatic injury and repair in KO2 and KO1 mice challenged for 2 weeks with CDE diet and allowed to recover on normal diet for 2 weeks , 3 months , and 6 months . Intriguingly , we observed highly divergent injury-repair responses in the two models . KO2 mice showed progressive repair through expansion of BEC-derived β-catenin-positive hepatocytes and resolution of inflammation , DR , and fibrosis . However , KO1 display progressive and peculiar DR composed of numerous small luminal structures lined by a single layer of BECs , even after being on normal diet for 6 months , which is associated with fibrosis and inflammation , and is reminiscent of cystic fibrosis ( CF ) -like morphology . We identify a unique interactome of β-catenin , p65 subunit of NF-κB and cystic fibrosis transmembranous conductance regulator ( CFTR ) in BECs and show perturbations in these interactions , leading to excessive NF-κB activation and inflammation in BECs in both KO1 and CF patients .
We previously showed CDE diet for 2 weeks led to enhanced injury , fibrosis , and DR in mice lacking β-catenin in hepatocytes only ( KO2 ) , generated by delivering AAV8-TBG-Cre into Ctnnb1flox/flox; Rosa-stopflox/flox-EYFP mice , as compared to WT2 mice , generated by injecting AAV8-TBG-Cre into Ctnnb1+/+; Rosa-stopflox/flox-EYFP mice ( Russell , 2019 ) . And that upon switching to normal diet , liver repair occurred via hepatocyte proliferation in WT2 but through BEC-to-hepatocyte transdifferentiation in KO2 ( Russell , 2019 ) . To specifically investigate durability of repair especially after the increased injury observed in the KO2 mice at 2 weeks of CDE diet , we fed CDE diet to KO2 and WT2 mice for 2 weeks and switched to normal diet for 2 weeks , 3 months , or 6 months ( Figure 1A ) . KO2 mice had elevated serum alanine aminotransferase ( ALT ) and total bilirubin ( BR ) levels than WT2 at 2 weeks of CDE diet , but returned to normal at 2 weeks onwards after switching to normal diet , similar to WT2 , although BR levels tended to be higher in KO2 up to 3 months of recovery ( Figure 1B ) . Alkaline phosphatase ( ALP ) was increased in WT2 and KO2 after 2 weeks of CDE injury but returned to normal at 2 weeks of recovery in both groups ( Figure 1B ) . Previously by fate-tracing , we observed a BEC transdifferentiated to hepatocytes and expanded in KO2 ( Russell , 2019 ) . Likewise , by immunohistochemistry ( IHC ) for β-catenin that is only present in BECs in KO2 at baseline , there were increased numbers of β-catenin-positive hepatocytes at 3 months and 6 months of recovery ( Figure 1C , Figure 1—figure supplement 1 ) . RT-PCR for β-catenin gene ( Ctnnb1 ) expression showed increasing expression in KO2 livers over time , becoming comparable to WT2 at 6 months ( Figure 1D ) . Since glutamine synthetase ( GS ) is a β-catenin-specific target , we also examined its protein by IHC in WT2 and KO2 at the 6-month recovery time . Indeed , like in WT2 , zone 3 GS-positive cells are present in KO2 showing β-catenin presence and activation at this time point ( Figure 1—figure supplement 2 ) . Sirius Red staining for fibrosis showed greater collagen deposition in KO2 than WT2 at 2 weeks of CDE diet and persisted at 2 weeks of recovery ( Figure 2A and B ) . Interestingly , despite being on normal diet and lack of any ongoing injury , KO2 continued to show fibrosis at 3 months , eventually resolving at 6 months ( Figure 2A and B ) . Likewise , expression of Col1a1 and Tgfβ2 tended to be higher in KO2 compared to WT2 mice at 3-month recovery , but were comparable to WT2 at 6 months ( Figure 2C ) . Since increased DR was observed in KO2 after CDE diet-induced injury , and fibrosis can be associated with DR , we next performed IHC for pan-cytokeratin ( PanCK; Figure 2D ) . There was robust DR in KO2 mice and WT2 mice after 2-week CDE diet and was also evident at 2 weeks of recovery although it appeared to be more pronounced in KO2 . At 3 months of recovery , normal bile ducts are seen in WT2 , whereas DR composed of flattened , non-luminal , and single or few cell clusters is evident throughout liver lobule in KO2 ( Figure 2D ) . At 6 months , there was no DR in either group ( Figure 2D ) . Quantification of PanCK staining was done using NIH Imager as described in Methods and showed higher trends but insignificant differences in areas covered by PanCK-positive DR in KO2 than WT2 at all times ( Figure 2—figure supplement 1 ) . Gene expression of BEC markers Krt19 and Epcam supported these observations ( Figure 2E ) . Expression of the gene encoding tissue inhibitor of metalloproteinase 1a ( Timp1 ) , a well-known inhibitor of matrix metalloproteinases , known for a role in extracellular matrix degradation , was determined next as a possible mechanism of fibrosis resolution ( Yoshiji , 2000; Yoshiji , 2002 ) . Higher expression of Timp1 persisted in KO2 as compared to WT2 at all recovery times except 6 months , coinciding with resolution of fibrosis and DR ( Figure 2F ) . Taken together , these results suggest that higher DR is associated with greater fibrosis in KO2 , and resolution of the DR and fibrosis took longer in KO2 than WT2 , which correlated with enhanced repopulation of the KO2 liver with β-catenin-positive hepatocytes and normalization of Timp1 levels . Next , we placed Albumin-Cre+/- Ctnnb1flox/flox ( KO1 ) mice lacking β-catenin in hepatocytes and BECs , and their wild-type littermates ( WT1 ) on CDE diet for 2 weeks and allowed recovery on normal diet for 2 weeks , 3 months , and 6 months ( Figure 3A ) . We observed severe liver injury in KO1 mice , shown by significantly higher serum ALT and total BR after 2 weeks of CDE diet compared to WT1 . During recovery , serum ALT levels in KO1 and WT1 mice decreased to normal levels ( Figure 3B ) , while BR remained mildly elevated in KO1 mice up to 3 months of recovery as compared to WT1 , which returned to normal at 2 weeks of recovery ( Figure 3B ) . Serum ALP levels were comparably increased in WT1 and KO1 at 2 weeks of CDE injury and returned to normal levels at 2-week recovery ( Figure 3B ) . Since β-catenin is lacking in hepithelial cells in the KO1 livers , IHC for β-catenin and RT-PCR for Ctnnb1 showed continued absence in KO1 and not WT1 at 3-month and 6-month recovery on normal diet ( Figure 3C and D , Figure 3—figure supplement 1 ) . As a surrogate for β-catenin presence and activity , we also assessed GS by IHC in WT1 and KO1 at the 6-month recovery time . Unlike WT1 , there is a complete absence of GS in zone 3 hepatocytes in KO1 showing continued absence of β-catenin at this time point ( Figure 3—figure supplement 2 ) . We previously reported increased fibrosis in KO1 mice compared to WT1 littermates after 2-week CDE diet ( Russell , 2019 ) . Here , we evaluated fibrosis during recovery on normal diet in both WT1 and KO1 . Despite normalization of serum transaminases during recovery , we observed continued fibrosis especially in the periportal area in KO1 and especially at 3 months and 6 months by Sirius Red staining , whereas WT1 mice displayed resolution of fibrosis as early as 2-week recovery ( Figure 4A ) . Quantification verified significant increases in fibrosis in KO1 at all time points compared to WT1 ( Figure 4B ) . Additionally , expression of Col1a1 tended to be higher in KO1 mice during recovery ( Figure 4C ) . DR was next assessed by IHC for PanCK . While there was a dramatic decrease in DR overtime in WT1 mice , a profound DR was observed in KO1 mice at all times , which was even more pronounced at 6 months of recovery ( Figure 4D ) . Quantification of PanCK staining showed significant differences in areas covered by PanCK-positive DR in KO1 than WT1 at all times of recovery from CDE diet and even significant and progressive increase from 3 months to 6 months in KO1 ( Figure 4—figure supplement 1 ) . Furthermore , the DR was peculiar and composed of numerous small luminal structures lined by a single layer of PanCK-positive columnar cells at 3 months and 6 months , rather than more flattened and invasive DR without lumen seen at earlier stages of CDE injury and recovery in both WT1 , KO1 , and even KO2 ( Figures 4D and 2D ) . Enhanced gene expression for Krt19 and Epcam was simultaneously evident in KO1 at these times ( Figure 4E ) . To determine if the continued DR was due to ongoing BEC proliferation , we co-stained KO1 livers from 3-month recovery with PanCK and proliferating cell nuclear antigen ( PCNA ) ( Figure 4—figure supplement 2A ) . Significantly more BECs were proliferating in KO1 compared to WT1 mice ( Figure 4—figure supplement 2B ) . A subset of BECs in DR were also positive for phospho-Erk1/2 ( p-Erk1/2 ) , known for regulating BEC proliferation ( Figure 4—figure supplement 2C; Pepe-Mooney , 2019 ) . Since decreased gene expression of Timp1 correlated with reduced fibrosis in KO2 at 6 months of recovery , we next investigated its levels in KO1 and WT1 . Timp1 tended to be upregulated in KO1 mice at all times but significantly at 6-month recovery time ( Figure 4F ) . Taken together , these results suggest that KO1 mice that continue to lack β-catenin in hepithelial cells show persistent Timp1 and fibrosis , and display continued and morphologically distinct DR associated with increased BEC proliferation at all times after the initial 2-week CDE diet injury , despite lack of active insult . To discern the basis of disparate DR and fibrosis between the two models , we first focused on investigating differences in specific injury processes between KO2 and WT2 during CDE injury and recovery . Increased hepatic bile acids have been implicated in hepatic injury and repair ( Fickert and Wagner , 2017 ) , and our lab has previously reported altered hepatic bile acids ( BAs ) in KO1 mice after methionine-choline-deficient ( MCD ) diet ( Behari , 2010; Thompson , 2018 ) . However , CDE diet-fed KO2 or WT2 mice showed no significant increase in hepatic bile acids at any time , suggesting these to not be driving DR or fibrosis in this model ( Figure 4—figure supplement 3A ) . We also investigated cell senescence as a possible driver of DR and fibrosis ( Wei-Yu et al . , 2015 ) . However , no significant hepatocyte senescence was observed by p21 IHC in WT2 or KO2 mice during long-term recovery from CDE diet ( Figure 4—figure supplement 3B ) . Although ALTs were not elevated , we wanted to directly address any ongoing injury in recovering WT1 and KO1 mice . Cleaved caspase 3 staining showed minimal cell death in both WT1 and KO1 mice at 3 months of recovery ( Figure 4—figure supplement 3C ) . Thus , BA alterations , cellular senescence , and cell death are not the basis of DR and fibrosis in CDE injury , and hence cannot explain differences in recovery between KO1 and KO2 mice . Next , we assessed if continued absence of β-catenin in KO1 but not in KO2 at 6 months of recovery could be affecting adherens junctions ( AJs ) integrity and could explain differences in DR and fibrosis between KO2 and KO1 mice . Temporal disruption of cell-cell junctions has been associated with pathologies in hepatobiliary injury including after CDE diet ( Pradhan-Sundd , 2018a ) . However , at 6 months of recovery , immunoprecipitation ( IP ) with E-cadherin showed an association of E-cadherin to β-catenin in KO2 and with γ-catenin in KO1 , as has been shown in β-catenin-deficient livers by us previously ( Figure 4—figure supplement 3D; Pradhan-Sundd , 2018b ) . Thus , intact AJs are present in both models during recovery and cannot be the basis of sustained DR and fibrosis in KO1 . Previously , the presence of immune cell infiltration has been shown to be essential in the development of DR and fibrosis . Mice lacking Th1 immune signaling or interferon-γ showed impaired DR and decreased fibrosis after CDE diet ( Knight , 2007 ) . To address inflammation , we performed staining for CD45 , a pan-leukocyte marker . There were high numbers of CD45-positive cells in WT2 and KO2 mice after 2-week CDE diet , which declined in WT2 after 2 weeks of recovery , decreased overall but persisted pan-lobularly in KO2 up to 3 months , and normalized to WT2 levels by 6 months ( Figure 5A ) . CD45-positive cells were present in high numbers in both WT1 and KO1 at 2 weeks of CDE diet , and while these numbers gradually returned to baseline in WT2 at 2 weeks of recovery , a more intense periportal-appearing infiltration was seen in KO1 especially at 3 months and 6 months ( Figure 5B ) . Quantification of the area occupied by CD45-positive cells at all time points revealed that CD45-positive cell infiltration significantly resolved at 3 months and 6 months of recovery from CDE diet in both WT2 and KO2 ( Figure 5—figure supplement 1A ) . However , CD45-positive cells in KO1 always tended to be higher than WT1 at all respective time points analyzed . Statistically significant and higher inflammation was observed in KO1 at 6 months as compared to WT1 at all time points ( Figure 5—figure supplement 1B ) . To determine if these inflammatory cells were close to DR , we performed triple immunofluorescence ( IF ) for PanCK , CD45 , and myofibroblast marker α-smooth muscle actin ( αSMA ) ( Figure 5—figure supplement 2A ) . At 2-week CDE diet , both WT2 and KO2 livers exhibited inflammatory cells and αSMA-positive cells close to BECs . After 2 weeks of recovery , αSMA-positive cells were no longer associated with BECs in WT2 and no immune cells were seen . In KO2 mice , BECs , αSMA-positive cells , and leukocytes were seen in close proximity to each other up to 3 months of recovery and returned to WT2 state at 6 months ( Figure 5—figure supplement 2A ) . KO1 livers showed closely associated CD45- and PanCK-positive cells at all times in contrast to WT1 , which lacked CD45 cells at all times after 2 weeks of recovery ( Figure 5—figure supplement 2A ) . Intriguingly , even in KO1 , αSMA-positive cells were not observed at any time after 2 weeks of recovery . Overall , the presence of CD45-positive cells close to PanCK-positive cells was clear in KO1 but not in KO2 at 6 months ( Figure 5C ) . Analysis in whole livers for expression of Adgre1 , gene encoding macrophage marker F4/80 , showed significant increase in KO1 compared to WT1 mice at 3-month recovery ( Figure 5D ) , and these macrophages were located close to the DR ( Figure 5—figure supplement 2B ) . Bone marrow monocyte-derived macrophages express high levels of Itgam ( CD11b ) , and infiltrate liver during injury , express pro-inflammatory cytokines , and are involved in both progression and recovery phases of fibrosis ( Guillot and Tacke , 2019 ) . Itgam expression was significantly higher in KO1 at 3-month and 6-month recovery times ( Figure 5D ) . There was also increased staining for Ly6G , a marker for monocytes , granulocytes , and neutrophils , in KO1 at 3-month and 6-month recovery times ( Figure 5E ) . Together , these results show resolution of DR and fibrosis in KO2 correlated with reduced inflammation , whereas persistent β-catenin-negative DR and continuing fibrosis in KO1 mice at late recovery stages from CDE injury were associated with persistent periportal inflammation . To address the mechanism of enhanced periportal inflammation , we next assessed the status of NF-κB , the master regulator of immune cell response . We have previously shown an inhibitory interaction of β-catenin with p65 subunit of NF-κB in hepatocytes and the absence of β-catenin in KO1 led to increased NF-κB activation in response to lipopolysaccharide ( LPS ) or tumor necrosis factor-α challenge ( Nejak-Bowen et al . , 2013 ) . Further , while immune cells are essential for DR and fibrosis , reactive ductules are a well-known source of pro-inflammatory and pro-fibrogenic cytokines and thus this cross-cellular signaling perpetuates overall injury ( Fava et al . , 2005; Pinto et al . , 2018 ) . Since inflammatory cells were specifically enriched in KO1 in the periportal region and associated closely to the DR , we next assessed NF-κB status along with β-catenin in BECs in KO2 and KO1 . At baseline in KO2 , p65 subunit of NF-κB was present in the cytosol of the CK19-positive cells , as was β-catenin ( Figure 6A ) . In KO1 , at baseline , CK19-positive BECs lacked β-catenin and p65 was still evident in cytosol ( Figure 6A ) . At 2 weeks of CDE diet , when DR and inflammation is ongoing in both KO2 and KO1 livers , a subset of CK19-positive BECs showed comparable nuclear p65 by confocal microscopy in both groups ( Figure 6B and C ) . At 6 months in KO2 , β-catenin-positive , CK19-positive BECs showed only cytosolic p65 similar to baseline ( Figure 6B ) . However , unlike at baseline , KO1 livers at 6-month recovery from CDE injury showed profound nuclear translocation of p65 in almost all CK19-positive BECs , which continued to lack β-catenin ( Figure 6B ) . Quantification showed significant difference in nuclear p65 in BECs in KO1 versus KO2 at 6-month recovery time point ( Figure 6C ) . To verify if nuclear p65 indicated NF-κB activation , 84 downstream target genes were checked by RT-PCR array ( fold-change threshold = >2 , p-value threshold = 0 . 05 ) . Relative to WT1 , we found a striking upregulation in the expression of 44% of genes ( 37/84 ) in KO1 , whereas 92% of genes ( 77/84 ) in KO2 livers were unchanged at 6 months of recovery ( Figure 6E ) . Clustergram showed KO2s were indistinguishable from WT1 , but KO1 clearly separated from both groups ( Figure 6—figure supplement 1 ) . Altogether , these data suggest a pronounced and prolonged NF-κB activation in BECs lacking β-catenin along with periportal inflammation during recovery phase from CDE injury , whereas presence of β-catenin dampened NF-κB activation in BECs to curbed inflammation and assist in recovery from the same injury . To more conclusively address the relationship between β-catenin and NF-κB in BECs directly , we utilized immortalized mouse small cholangiocyte cells ( SMCCs ) , which were transfected with control- or β-catenin siRNA together with either β-catenin-TCF TOPFlash reporter or p65 luciferase reporter . Knockdown of β-catenin in SMCCs , shown by a significant decrease in TOPFlash activity , induced p65 luciferase activity , which was three times greater than caused by 100 ng/ml LPS stimulation ( Figure 7A ) . Conversely , expression of stable S45Y-β-catenin or T41A-β-catenin ( not shown ) in SMCCs led to increased TOPFlash activity but significantly decreased p65 reporter activity with or without LPS ( Figure 7B ) . Similar negative regulation was also observed in MzChA and HuCCT1 , two independent human CCA cell lines ( Figure 7—figure supplement 1A and B ) . To address how β-catenin regulates NF-κB activity , SMCCs transfected with control- or Ctnnb1-siRNA were cultured with or without LPS , and subjected to cell fractionation to isolate nuclear- and cytoplasmic-enriched fractions . By WB , at baseline , notably higher levels of both p65 and β-catenin were evident in the cytoplasmic and not nuclear fraction . Upon β-catenin silencing , significantly higher levels of p65 were found in the nuclear compartment versus the controls , both without LPS , but more so following LPS treatment ( Figure 7C and D ) . Next , we modulated β-catenin activity in SMCCs to determine changes in global gene expression . Bulk RNA-seq was performed on β-catenin-silenced or control SMCCs , and on SMCCs transfected with S45Y-β-catenin or eGFP . Using a cutoff p-value ≤ 0 . 05 and abs ( log2FC ) ≥ 1 . 5 , we found 335 β-catenin-regulated genes in BECs . Specifically , β-catenin-silenced SMCCs showed 75 upregulated and 122 downregulated , and β-catenin-active SMCCs showed 76 upregulated and 69 downregulated genes ( Figure 7—figure supplement 1C–E ) . While there was a minimal overlap ( Figure 7—figure supplement 1E ) , JASPAR was queried to identify transcription factor ( TF ) binding profiles in the 335 differentially expressed genes ( DEGs ) . RELA ( p65 ) was identified among the top TFs ( ranking by p-value ) , with 9 . 6% of DEGs showing known RELA regulation ( 32/335 , p=0 . 015 ) and 15 . 2% ( 51/335 , p=0 . 088 ) showing NF-κB binding sites ( Figure 7E ) . Interestingly , from RNA-seq and by qPCR , we found modest but significant increase in CCL2 , and a more pronounced and significant increase in CXCL5 expression , in the β-catenin-silenced SMCCs ( Figure 7F ) . After 6-month recovery from CDE diet , a significant induction in CCL2 and CXCL5 expression was noted in KO1 but not in WT1 and KO2 ( Figure 7G ) . Next , we investigated if β-catenin interacts with p65 in BECs . Whole-cell lysates from SMCCs and normal mouse liver were used to pull down p65 . We identified robust β-catenin association with p65 in SMCCs ( Figure 7H1 ) . A fainter but positive β-catenin-p65 association was evident in whole livers , likely due to low BEC representation in protein lysates from whole livers ( Figure 7I ) . To further verify presence of β-catenin-p65 complex in BECs in vivo , we examined β-catenin and p65 localization using confocal microscopy . In WT1 liver at baseline , a notable colocalization of p65 and β-catenin was evident in the cytoplasm of BECs , which was expectedly absent in KO1 ( Figure 7—figure supplement 1F ) . Quantification of colocalization using ImageJ showed that about 35% of p65 is associated with β-catenin , which was significantly greater than KO1 ( Figure 7J ) . Altogether , biochemical and IF studies identify a heretofore undescribed β-catenin-p65 complex in the cytoplasm of BECs . Further , β-catenin seems to prevent p65 nuclear translocation and NF-κB activation , and may be important in shutting off NF-κB activation when its signaling is no longer required . Absence of β-catenin in BECs enhances nuclear translocation of p65 at baseline , and more so in the presence of known positive regulators of NF-κB signaling such as LPS , and contributes to enhanced expression of genes involved in proliferation , inflammation , and fibrosis ( Kim et al . , 2015a; Luedde and Schwabe , 2011 ) . Since persistent DR observed in KO1 at 3 months and 6 months after recovery from CDE injury showed unique morphology consisting of numerous small luminal structures lined by a single layer of PanCK-positive cells , we decided to interrogate livers from clinical cases exhibiting DR including alcoholic hepatitis ( AH ) , polycystic liver disease ( PLD ) , and CF . As seen by panCK IF , AH cases displayed DR with variable morphology including areas of luminal small DR ( shown in representative case ) , whereas PLD showed large cysts lined by flattened BECs ( Figure 8A ) . DR in CF was more homogeneous and appeared uniformly reminiscent of what we observed in 3-month and 6-month recovery times in KO1 ( Figures 4D and 8A ) . Intriguingly , strongest and significant nuclear p65 was consistently observed in DR seen in CF cases followed by PLD with only a very small subset of cells in DR showing nuclear p65 in AH ( Figure 8B ) . While β-catenin seem to be unaltered by IF staining in all three pathologies ( Figure 8A ) , we observed a decrease in total β-catenin in a single CF case from whom two independent frozen liver samples were available ( Figure 8C ) . CF cases typically have varying loss-of-function ( LOF ) mutations in CFTR gene . Since β-catenin-p65 interactions were observed in SMCCs , we next assessed if CFTR is interacting with this complex . We observed concomitant pulldown of both B-band ( faster migrating core glycosylated immature form ) as well as slowly migrating C-band ( complex glycosylated form ) ( Chang , 2008 ) , along with p65 , when we immunoprecipitated β-catenin in SMCCs ( Figure 8D ) . To mimic LOF of CFTR seen in CF , we next silenced Cftr in SMCCs . Knockdown of Cftr led to a pronounced increase in p65 reporter activity ( Figure 8E ) . Likewise , the expression of NF-κB target chemokines Ccl2 and Cxcl5 was significantly induced upon Cftr knockdown ( Figure 8F ) . To query impact of β-catenin modulation , we next coexpressed stable-β-catenin ( S45Y-mutant ) in control and Cftr-siRNA-transfected SMCCs . Stabilization of β-catenin significantly decreased CFTR knockdown-induced p65 activation ( Figure 8E ) . To query if β-catenin loss in BECs had any impact on CFTR levels , we examined total levels of CFTR in KO1 and WT1 at baseline and at 6 months of recovery from CDE diet . We did not observe any differences in the levels of total CFTR protein , suggesting that CFTR may be an upstream effector of β-catenin while β-catenin does not regulate CFTR levels in BECs ( Figure 8—figure supplement 1 ) . Thus , we identify important interactions between β-catenin , p65 , and CFTR in BECs , and LOF of CFTR leads to destabilization of β-catenin , which in turn allows p65/NF-κB activation more readily in response to the available cues including the presence of cytokines . Also , β-catenin stabilization in BECs could tamper such enhanced p65 activation and may have potential therapeutic benefit in controlling unchecked NF-κB activation , inflammation , and fibrosis in reactive ductular cells .
DR is a common hallmark of many chronic liver pathologies , although its morphology is heterogeneous ranging from isolated invasive ductular cells , luminal phenotype , and sometimes purely cystic ( Sato et al . , 2019; Wilson and Rudnick , 2019; Nejak-Bowen , 2020 ) . The significance of DR remains controversial and has been associated with both repair and disease progression ( Sato et al . , 2019; Kamimoto et al . , 2020 ) . Its role as a source of de novo hepatocytes through the process of transdifferentiation is indisputable in preclinical models shown by many fate-tracing studies ( Russell , 2019; Wei-Yu et al . , 2015; Nejak-Bowen , 2020; Raven , 2017 ) . At the same time , DR can induce fibrosis by secreting pro-inflammatory and pro-fibrogenic factors to contribute to the disease process ( Lowes et al . , 1999; Richardson , 2007; Zhao et al . , 2018; Kim et al . , 2015a; Aguilar-Bravo , 2019 ) . What drives the pro-inflammatory and pro-fibrogenic phenotype of these reactive BECs and what reverts these cells back to normal is poorly understood . Previously , we and others have described β-catenin-p65 complex in hepatocytes , breast and colon cancer cells , which could inhibit NF-κB activation ( Nejak-Bowen et al . , 2013; Deng , 2002 ) . Indeed , being a ‘sticky’ protein , β-catenin interacts with many proteins in a cell to modulate their activities ( Russell and Monga , 2017 ) . The exact biological significance of β-catenin-p65 interaction is not well understood and likely context dependent . We identify the existence of this complex in normal BECs . We show β-catenin-p65 complex to be present in the cytoplasm in the BECs to prevent nuclear translocation and activation of p65 and in turn keep NF-κB activity in check . While knockdown of β-catenin in BECs in vitro led to some baseline nuclear translocation , it allowed more profound nuclear translocation and activation of p65 in the presence of LPS , a well-known NF-κB activator . Conversely , β-catenin stabilization in BECs dampened NF-κB activation basally or in the presence of LPS . β-Catenin activation is observed in BECs within the DR during chronic liver injuries ( Apte et al . , 2008; Hu , 2007; Okabe , 2016 ) . However , unlike in hepatocytes , β-catenin activation in BECs does not play a role in their proliferation ( Russell , 2019; Okabe , 2016 ) . Our current study shows that β-catenin stabilization in BECs may in fact be ‘mopping’ up p65 to dampen and eventually shutoff NF-κB activation , reverting BECs to their quiescence . NF-κB activation has been shown to be important in BEC proliferation and DR by regulating Jagged/Notch signaling ( Kim et al . , 2015a ) . And NF-κB in BECs has been suggested to play a role in inducing pro-inflammatory and pro-fibrogenic milieu as well through regulating expression of many of its target genes including Ccl2 and Cxcl5 ( O’Hara et al . , 2013 ) . In summary , absence of β-catenin in BECs as seen in KO1 prevents β-catenin-p65 complex formation , leading to sustained NF-κB activation after injury , whereas this complex reforms in KO2 especially due to known β-catenin increase in BECs after injury , thus preventing chronic NF-κB activation . Molecular underpinnings of β-catenin stabilization in BECs especially during cholestatic liver injuries remain under investigation , although portal fibroblasts , macrophages , hepatocytes , and BECs have all been shown to secrete ligands like Wnt7a , Wnt7b , Wnt10a , and Wnt5a ( Hu , 2007; Okabe , 2016; Wilson , 2020 ) . Another intriguing observation was the distinct morphology of DR evident in the recovery phase from the CDE injury in β-catenin-deficient livers , which was reminiscent of the histology of the CF cases ( Sakiani et al . , 2019 ) . Surprisingly , very few BECs in AH showed nuclear p65 , while PLD cases showed variable but increased nuclear p65 in cells lining the cysts . However , DR in CF cases exhibited strongest and consistent nuclear p65 . This led us to investigate interactions between β-catenin-p65 and CFTR , whose gene is mutated in CF . CFTR protein is present in BECs only in the liver ( Cohn et al . , 1993 ) . We observed a pulldown of CFTR ( B-band and C-band ) with β-catenin in a BEC line . To mimic LOF , which is the common end result of CFTR mutations in CF patients , we silenced CFTR in SMCC line , which led to a profound p65 activation that was decreased upon β-catenin stabilization . This suggests an important tripartite regulatory interaction between these proteins . Interestingly , in human lung epithelial cells , a proteomic screen identified interaction of β-catenin with WT CFTR but not with ∆F508 CFTR , a major site of LOF mutation in the CFTR ( Pankow , 2015 ) . Additionally , in mouse intestine , CFTR was shown to stabilize β-catenin and prevent its degradation ( Liu et al . , 2016 ) . The same study showed that ∆F508 CFTR is unable to interact with β-catenin , leading to β-catenin degradation and eventually resulting in activation of NF-κB-mediated inflammatory cascade . Our results show an existence of a tripartite interaction between β-catenin , p65 , and CFTR in cholangiocytes . LOF of CFTR or its reduced levels led to decrease in β-catenin protein in BECs both in vitro ( SMCC ) and in vivo ( CF patient liver lysate ) , leading to nuclear translocation of p65 , NF-κB activation , and increased expression of its pro-inflammatory chemokine targets . Absence of β-catenin in BECs however did not alter CFTR levels , suggesting CFTR to upstream effector of β-catenin while not being dependent on β-catenin for its own stability . We believe that the classical pathology of liver disease in CF including periductal inflammation , DR , periportal fibrosis , and focal biliary cirrhosis may be explained by our observations , in addition to previously described mechanisms such as Rous sarcoma oncogene cellular homolog ( Src ) -dependent Toll-like receptor-4 activation ( Fiorotto , 2016 ) . Since β-catenin activation in BECs inhibited NF-κB activation occurring due to CFTR silencing , this strategy may have therapeutic implications in controlling CF disease progression in the liver and elsewhere and future studies will directly address this novelty . NF-κB activation has been shown to be important in BEC proliferation and DR by regulating Jagged/Notch signaling ( Kim et al . , 2015a ) . The same study showed NF-κB activation to be regulated by cysteine-rich protein 61 ( CYR61 ) , whose knockdown reduced DR . Incidentally , CFTR was also significantly reduced in that study and might have been a mechanism of DR through disruption of CTFR-p65-β-catenin interactions . The mechanism of NF-κB activation and whether these tripartite interactions are playing any role in other pathologies such as PLD or subset of AH cases requires further investigation . There is very little understanding of the process of resolution of DR along with its associated fibrosis , although these are strongly linked to inflammation ( Ko et al . , 2019 ) . Our study provides novel insight into not only the molecular underpinnings of a reactive cholangiocyte , but also sheds light on how BECs regulate the immune microenvironment . Levels of Timp1 correlate with fibrosis , and several groups have investigated the use of TIMP1 levels as a biomarker for fibrosis in hepatitis C patients ( Leroy , 2004; Boeker et al . , 2002 ) . Mice with overexpression of TIMP1 developed dramatically more fibrosis after CCl4 treatment ( Yoshiji , 2000 ) , and TIMP1 transgenic mice showed impaired fibrosis resolution after cessation of CCl4 ( Yoshiji , 2002 ) . Our data is consistent since expression of Timp1 was associated with fibrosis , and levels of Timp1 normalized upon resolution of fibrosis .
All animals are housed in temperature and light-controlled facilities and are maintained in accordance with the Guide for Care and Use of Laboratory Animals and the Animal Welfare Act . Generation of Albumin-Cre;Ctnnb1flox/flox mice and wild-type littermates has been described previously ( Tan et al . , 2006 ) . Generation of Ctnnb1flox/flox;Rosa-stopflox/flox-EYFP reporter mice was also described previously ( Russell , 2019 ) . In brief , 23–25-day-old Ctnnb1flox/flox;Rosa-stopflox/flox-EYFP mice were injected intraperitoneally with 1 × 1012 genome copies ( GCs ) of adeno-associated virus serotype 8 encoding Cre recombinase under the hepatocyte-specific thyroid binding globulin promoter ( AAV8-TBG-Cre; Addgene ) followed by a 12-day washout period . The same protocol was utilized on Ctnnb1+/+;Rosa-stopflox/flox-EYFP mice to generate WT2 mice . When mice were 4–6 weeks old , choline-deficient diet ( Envigo Teklad Diets ) supplemented with 0 . 15% ethionine drinking water ( Acros Organics , 146170100 ) was administered for 2 weeks . For recovery time points , animals were switched back to normal chow diet for up to 6 months . Serum biochemistry analysis was performed by automated methods at the University of Pittsburgh Medical Center clinical chemistry laboratory . All studies were performed according to the guidelines of the National Institutes of Health and the University of Pittsburgh Institutional Animal Use and Care Committee . All patient tissue sections were provided by the Pittsburgh Liver Research Center’s ( PLRC’s ) Clinical Biospecimen Repository and Processing Core ( CBPRC ) , supported by P30DK120531 . Sections from 5 patients with healthy liver , 12 patients with DR from AH ( n = 10 ) and/or NASH ( n = 2 ) , 5 patients with DR associated with PLD , and 6 patients with DR in CF cases were triple stained with CK19 , p65 , and β-catenin for further analysis . Patient information from these groups of cases is listed in Supplementary file 1 . Two pieces of frozen livers from one CF patient ( TP10-P531 ) were provided by Pitt Biospecimen Core and used for WB . Information on this case is also included in Supplementary file 1 . The IHC protocols have been described previously ( Russell , 2019 ) . In brief , liver tissue was fixed in 10% buffered formalin for 48 hr prior to paraffin embedding . Blocks were cut into 4-µm sections , deparaffinized , and washed with PBS . For antigen retrieval , samples were microwaved for 12 min in pH 6 sodium citrate buffer ( PanCK , CD45 , p-Erk1/2 , cleaved caspase 3 ) or Tris-EDTA buffer ( p21 ) , were pressure cooked for 20 min in pH 6 sodium citrate buffer ( β-catenin ) , or were incubated with Proteinase K ( Agilent Dako , S302030-2 ) for 10 min ( F4/80 ) . Samples were then placed in 3% H2O2 for 10 min to quench endogenous peroxide activity . After washing with PBS , slides were blocked with Super Block ( ScyTek Laboratories , AAA500 ) for 10 min or 10% goat serum in PBS for 10 min ( p21 ) . The primary antibodies were incubated at the following concentrations in antibody diluent: PBS + 1% BSA ( Fisher BioReagents , BP1605-100 ) with 0 . 1% Tween 20 ( Fisher BioReagents , BP337-500 ) : PanCK ( Dako , Z0622 , 1:200 ) , cleaved caspase 3 ( Cell Signaling , 9664 , 1:100 ) , p-Erk1/2 ( Cell Signaling , 4370 , 1:100 ) , F4/80 ( Bio-Rad , MCA497A488 , 1:100 ) for 1 hr at room temperature or at 4°C overnight: β-catenin ( Abcam , ab32572 , 1:50 ) and p21 ( Santa Cruz , sc-471 , 1:25 ) . Samples were washed with PBS three times and incubated with the appropriate biotinylated secondary antibody ( Vector Laboratories ) diluted 1:500 in antibody diluent for 30 min at room temperature . Samples were washed with PBS three times and sensitized with the Vectastain ABC kit ( Vector Laboratories , PK-6101 ) for 30 min . Following three washes with PBS , color was developed with DAB Peroxidase Substrate Kit ( Vector Laboratories , SK-4100 ) , followed by quenching in distilled water for 5 min . Slides were counterstained with hematoxylin ( Thermo Scientific , 7211 ) , dehydrated to xylene , and coverslips applied with Cytoseal XYL ( Thermo Scientific , 8312-4 ) . For Sirius Red staining , samples were deparaffinized and incubated for 1 hr in Picro-Sirius Red Stain ( American MasterTech , STPSRPT ) , washed twice in 0 . 5% acetic acid water , dehydrated to xylene , and coverslipped . Images were taken on a Zeiss Axioskop 40 inverted brightfield microscope . Liver tissue was fixed in 10% buffered formalin overnight , cryopreserved with 30% sucrose in PBS overnight , frozen in OCT compound ( Sakura , 4583 ) , and stored at –80°C . OCT-embedded samples were cut into 5-µm sections , allowed to air-dry , and then washed in PBS . Antigen retrieval was performed through microwaving in pH 6 sodium citrate buffer . Slides were washed with PBS and permeabilized with 0 . 1% Triton X-100 in PBS for 20 min at room temperature . Samples were washed three times with PBS and then blocked with 2% donkey serum in 0 . 1% Tween 20 in PBS ( antibody diluent ) for 30 min at room temperature . Antibodies were diluted as follows: PanCK ( Dako Z0622 , 1:200 ) , PCNA ( Santa Cruz Biotechnology , sc-56 , 1:1000 ) , p65 ( Santa Cruz Biotechnology , sc-372 , 1:500 ) , β-catenin ( BD Biosciences , 610154 , 1:500 ) , CK-19 ( DSHB , TROMA-III , 1:10 ) in antibody diluent , and incubated at 4°C overnight . Ly6G ( clone: RB6-8C5 ) antibody was purchased from Thermo Fisher Scientific , Waltham , MA . Samples were washed three times in PBS and incubated with the proper fluorescent secondary antibody ( AlexaFluor 488/555/647 , Invitrogen ) diluted 1:400 in antibody diluent for 2 hr at room temperature . The αSMA antibody is directly conjugated to Cy3 and requires no secondary antibody . Samples were washed three times with PBS and incubated with DAPI ( Sigma , B2883 ) for 1 min . Samples were washed three times with PBS and mounted with fluoromount ( SouthernBiotech ) or ProLong Gold antifade reagent ( Invitrogen , P10144 ) . Images were taken on a Nikon Eclipse Ti epifluorescence microscope or a Zeiss LSM700 confocal microscope . To determine BEC proliferation , for each sample seven images at 200× magnification of periportal regions were taken , and in each image the number of PanCK+/PCNA+ cells was manually counted . For quantification of Sirius Red , PanCK , CD45 staining , staining intensity was measured in ImageJ for 3–5 images at 100× magnification per mice sample . Positive area to whole area ratio was calculated as percent-positive area using the NIH ImageJ software . To determine p65 nuclear-positive cholangiocytes , four images from each animal or patient at 200× magnification of DR regions were counted . Whole liver was homogenized in TRIzol ( Thermo Scientific , 15596026 ) and nucleic acid was isolated through phenol-chloroform extraction . Cellular DNA was digested with DNA-free Kit ( Ambion , AM1906 ) , and RNA was reverse-transcribed into cDNA using SuperScript III ( Invitrogen , 18080-044 ) . Real-time PCR was performed in technical triplicate on a StepOnePlus Real-Time PCR System ( Applied Biosystems , 4376600 ) or on a Bio-Rad CFX96 Real-Time System using the Power SYBR Green PCR Master Mix ( Applied Biosystems , 4367660 ) . Target gene expression was normalized to the average of two housekeeping genes ( Gapdh and Rn18s ) , and fold-change was calculated utilizing the ΔΔ-Ct method . For RT-PCR arrays , 2 μl of cDNA , 7 . 5 μl of Power SYBR Green PCR Master Mix , and 4 . 5 μl of nuclease-free water were premixed and added to each well of RT² Profiler PCR Array Mouse NFκB Signaling Pathway ( Qiagen , PAMM-025Z ) for target gene qPCR . RT-PCR arrays data were analyzed at GeneGlobe ( https://geneglobe . qiagen . com/us/analyze/ ) . The average of three housekeeping genes ( Rn18s , Actb , and Gapdh ) was used for normalization . Volcano plot and clustergram were generated by the data analysis web portal mentioned above . Whole liver tissue was homogenized in RIPA buffer premixed with fresh protease and phosphatase inhibitor cocktails . Cytoplasmic and nuclear extracts were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Reagents ( Thermo Fisher Scientific , 78835 ) . The concentration of the protein was determined by the bicinchoninic acid assay . For IP , 1 mg of SMCC lysate was precleared with 40 μl of Protein A/G PLUS-Agarose ( Santa Cruz Biotechnology , sc-2003 ) for 2 hr at 4°C . After centrifugation ( 3000 rpm , 1 min ) , the supernatant was incubated with 2 μg of p65 antibody ( Santa Cruz Biotechnology , sc-8008 , 1:100 ) , 2 μg of β-catenin antibody ( BD Biosciences , 610154 , 1:100 ) , or control IgG overnight at 4°C . The next day , samples were incubated with 40 μl of Protein A/G PLUS-Agarose for 1 hr at 4°C . The pellet was collected , washed with RIPA buffer for three times , resuspended in 10 μl of loading buffer , and subjected to electrophoresis . Protein lysate was separated on pre-cast 7 . 5% or 4–20% polyacrylamide gels ( Bio-Rad ) and transferred to the PVDF membrane using the Trans-Blot Turbo Transfer System ( Bio-Rad ) . Membranes were blocked for 75 min with 5% skim milk ( Lab Scientific , M0841 ) or 5% BSA in Blotto buffer ( 0 . 15 M NaCl , 0 . 02 M Tris pH 7 . 5 , 0 . 1% Tween in dH2O ) , and incubated with primary antibodies at 4°C overnight at the following concentrations: β-catenin ( Cell Signaling , 8480 , 1:1 , 000 in 1% BSA ) , p65 ( Cell Signaling , 8242 , 1:1000 in 1% BSA ) , CFTR ( Alomone Labs , ACL-006 , 1:250 in 1% BSA ) , GAPDH ( Cell Signaling , 5174 , 1:10 , 000 in 1% milk ) , and histone H3 ( Cell Signaling , 9715 , 1:1 , 000 in 1% milk ) . Membranes were washed in Blotto buffer and incubated with the appropriate HRP-conjugated secondary antibody for 75 min at room temperature . Membranes were washed with Blotto buffer , and bands were developed utilizing SuperSignal West Pico Chemiluminescent Substrate ( Thermo Scientific , 34080 ) and visualized by autoradiography . SMCCs ( obtained from Dr . Kari Nejak-Bowen , University of Pittsburgh , PA ) , and human CCA cell lines MzChA and HuCCT1 ( obtained from Dr . Gregory J . Gores , Mayo Clinic , Rochester , MN ) were seeded on six-well plates in a humidity-saturated incubator with 5% CO2 maintained at 37°C . Cells were mycoplasma negative tested with MycoAlert Mycoplasma Detection Kit ( Lonaz , LT07-218 ) . For p65 reporter assay , cells were transfected with 2 μg of p65 reporter and 0 . 2 μg of Renilla ( internal control ) together with 2 μg of either eGFP ( control ) or S45Y to overexpress constitutively active β-catenin using Lipofectamine 3000 Transfection Reagent ( Invitrogen , L3000008 ) , or together with si-Control ( Cell Signaling , 6568 ) and si-β-catenin ( Cell Signaling , 6225 ) to knock down β-catenin using Lipofectamine RNAiMAX Transfection Reagent ( Invitrogen , 13778150 ) . For TOPFlash reporter assay , p65 reporter above was replaced by TOPFlash plasmid . Cells were treated with 100 ng/ml of LPS 6 hr before harvest . si-CFTR ( Santa Cruz Biotech , sc-35053 ) was used to knock down CFTR in SMCCs . Cells were harvested at 48 hr , and luciferase signals were got using Dual-Luciferase Reporter Assay System ( Progema , E1910 ) and normalized to the value of Renilla . Total hepatic bile acids were measured using the Mouse Total Bile Acids Assay Kit from Crystal Chem ( Downers Grove , IL ) , as per the manufacturer’s instructions . To isolate total bile acids from liver , 50–100 mg frozen liver tissue was homogenized in 70% ethanol at room temperature , then samples were incubated in capped glass tubes at 50°C for 2 hr . The homogenates were centrifuged at 6000 g for 10 min to collect the supernatant . Total bile acid concentrations were determined using the calibration curve from the standard provided in the kit and the mean change in absorbance value for each sample . Twelve SMCC RNA samples were measured: three for CTNNB1 activation ( SMCC-S45Y ) , three for CTNNB1 activation control ( SMCC-eGFP ) , three for CTNNB1 silencing ( SMCC-si-β-catenin ) , and three for CTNNB1 silencing control ( SMCC-si-Control ) . In total , 12 RNA-seq libraries were sequenced . For each library , quality control was performed to each raw sequencing data by tool FastQC . Based on the QC results , low-quality reads and adapter sequences were filtered out by tool Trimmomatic ( Bolger et al . , 2014 ) . Surviving reads were then aligned to mouse reference genome mm10 by aligner Hisat2 ( Kim et al . , 2015b ) . HTSeq tool ( Anders et al . , 2015 ) was then applied to the aligned file for gene quantification . Based on the gene count , differential expression analysis was applied to compare CTNNB1 activation samples with their corresponding controls , and to compare CTNNB1 silencing samples with their corresponding controls , respectively . R package DESeq2 ( Love et al . , 2014 ) was employed to perform the test , and DEGs were defined as genes with fold-change higher than 1 . 5-fold and p-value ( or adjusted p-value ) smaller than 0 . 05 . These DEGs were further used to detect common upstream TFs based on the JASPAR database ( Fornes , 2020 ) . Opposite regulation directions of the activation and silencing models were finally compared in terms of DEGs . All statistical analyses were performed by R programming . Raw RNA-seq data and gene count quantification were submitted to NCBI GEO database with accession ID GSE155981 ( https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE155981 ) . For analysis of serum biochemistry between two groups , a two-tailed t-test was performed . For analysis of cell counts , such as proliferating BECs , a Mann–Whitney U test was performed . A p<0 . 05 was considered significant , and plots are mean ± SD . Detailed statistic information for each assay is given in the figure legends . All statistical analysis and graph generation were performed using GraphPad Prism software . | The liver has an incredible capacity to repair itself or ‘regenerate’ – that is , it has the ability to replace damaged tissue with new tissue . In order to do this , the organ relies on hepatocytes ( the cells that form the liver ) and bile duct cells ( the cells that form the biliary ducts ) dividing and transforming into each other to repair and replace damaged tissue , in case the insult is dire . During long-lasting or chronic liver injury , bile duct cells undergo a process called ‘ductular reaction’ , which causes the cells to multiply and produce proteins that stimulate inflammation , and can lead to liver scarring ( fibrosis ) . Ductular reaction is a hallmark of severe liver disease , and different diseases exhibit ductular reactions with distinct features . For example , in cystic fibrosis , a unique type of ductular reaction occurs at late stages , accompanied by both inflammation and fibrosis . Despite the role that ductular reaction plays in liver disease , it is not well understood how it works at the molecular level . Hu et al . set out to investigate how a protein called β-catenin – which can cause many types of cells to proliferate – is involved in ductular reaction . They used three types of mice for their experiments: wild-type mice , which were not genetically modified; and two strains of genetically modified mice . One of these mutant mice did not produce β-catenin in biliary duct cells , while the other lacked β-catenin both in biliary duct cells and in hepatocytes . After a short liver injury – which Hu et al . caused by feeding the mice a specific diet – the wild-type mice were able to regenerate and repair the liver without exhibiting any ductular reaction . The mutant mice that lacked β-catenin in hepatocytes showed a temporary ductular reaction , and ultimately repaired their livers by turning bile duct cells into hepatocytes . On the other hand , the mutant mice lacking β-catenin in both hepatocytes and bile duct cells displayed sustained ductular reactions , inflammation and fibrosis , which looked like that seen in patients with liver disease associated to cystic fibrosis . Further probing showed that β-catenin interacts with a protein called CTFR , which is involved in cystic fibrosis . When bile duct cells lack either of these proteins , another protein called NF-B gets activated , which causes the ductular reaction , leading to inflammation and fibrosis . The findings of Hu et al . shed light on the role of β-catenin in ductular reaction . Further , the results show a previously unknown interaction between β-catenin , CTFR and NF-B , which could lead to better treatments for cystic fibrosis in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"medicine"
] | 2021 | β-Catenin-NF-κB-CFTR interactions in cholangiocytes regulate inflammation and fibrosis during ductular reaction |
The axolotl can regenerate multiple organs , including the brain . It remains , however , unclear whether neuronal diversity , intricate tissue architecture , and axonal connectivity can be regenerated; yet , this is critical for recovery of function and a central aim of cell replacement strategies in the mammalian central nervous system . Here , we demonstrate that , upon mechanical injury to the adult pallium , axolotls can regenerate several of the populations of neurons present before injury . Notably , regenerated neurons acquire functional electrophysiological traits and respond appropriately to afferent inputs . Despite the ability to regenerate specific , molecularly-defined neuronal subtypes , we also uncovered previously unappreciated limitations by showing that newborn neurons organize within altered tissue architecture and fail to re-establish the long-distance axonal tracts and circuit physiology present before injury . The data provide a direct demonstration that diverse , electrophysiologically functional neurons can be regenerated in axolotls , but challenge prior assumptions of functional brain repair in regenerative species .
Under physiological conditions , the neurogenic capacity of the adult mammalian brain is largely restricted to two neurogenic niches , the subventricular zone of the lateral ventricle , which gives rise to interneurons of the olfactory bulb and the subgranular zone of the dentate gyrus , which generates granule cells of the hippocampus ( Ming and Song , 2011 ) . Neurons in other brain regions are only generated during embryonic development and are not replaced postnatally . In contrast to mammals , other vertebrates are endowed with superior capacity to regenerate multiple organs , including parts of the central nervous system ( CNS ) . Among these , urodele amphibians like the axolotl ( Ambystoma mexicanum ) are endowed with the capacity to add new neurons to the brain throughout life ( Maden et al . , 2013 ) and can regenerate the spinal cord and parts of the brain after mechanical injury ( Burr , 1916; Kirsche and Kirsche , 1964a; Butler and Ward , 1967; Piatt , 1955 ) . Resection of the middle one-third of one hemisphere , but not the whole hemisphere , in the axolotl telencephalon results in reconstruction of the injured hemisphere to a similar length as the contralateral , uninjured side ( Kirsche and Kirsche , 1964a; Kirsche and Kirsche , 1964b; Winkelmann and Winkelmann , 1970 ) . Similarly , after mechanical excision of the newt optic tectum , new tissue fills the space produced by injury ( Okamoto et al . , 2007 ) . Interestingly , in the newt , selective chemical ablation of dopaminergic neurons within a largely intact midbrain triggers regeneration of the ablated pool of neurons ( Berg et al . , 2010; Parish et al . , 2007 ) . In addition to urodeles , teleost fish have also been extensively studied for their capacity to regenerate the CNS and have led to the identification of some of the molecular signals involved in the regenerative process ( Kizil et al . , 2012 ) . These studies highlight the value of regenerative organisms as models to understand the mechanisms that govern brain regeneration for possible application to the mammalian brain . However , the mammalian CNS is notoriously complex , and its ability to compute high-level functions , like those of the mammalian cerebral cortex , relies on the presence of a great diversity of neuronal subtypes integrated in specific long-distance and local circuits and working within a defined tissue architecture . Disruption of brain structure , connectivity , and neuronal composition is often associated with behavioral deficits , as observed in models of neurodevelopmental , neuropsychiatric , and neurodegenerative disease . It is therefore likely that functional regeneration of higher-order CNS structures will entail the regeneration of a great diversity of neuronal subtypes , the rebuilding of original connectivity , and the synaptic integration of newborn neurons in the pre-existing tissue . It is not known to what extent even regenerative species can accomplish these complex tasks , beyond their broad ability to generate new neurons and to rebuild gross brain morphology . It remains therefore debated whether any vertebrates are capable of true functional brain regeneration . Using the adult axolotl pallium as the model system , we have investigated whether a diverse array of neuronal subtypes can regenerate and whether their tissue-level organization , connectivity , and functional properties can also regenerate after mechanical injury . In contrast to the teleostean pallium , the everted nature of which makes linking distinct regions to their mammalian counterparts difficult ( Northcutt , 2008 ) , the gross neuroanatomy of the axolotl pallium , organized around two ventricles , shows clear similarities to that of the mammalian telencephalon . In addition , while the evolutionary origin of the mammalian cerebral cortex remains controversial ( Molnár , 2011 ) , it is likely that the axolotl pallium contains a basic representation of several of the neuronal subtypes found in the mammalian cerebral cortex and thus may serve as a good model for investigating regeneration of neuronal heterogeneity and complex circuit function . Here , we demonstrate that both pre- and post-metamorphosis adult axolotls are able to regenerate a diversity of neurons upon localized injury to the dorsal pallium . This process occurs through specific regenerative steps that we defined in live animals using non-invasive magnetic resonance imaging ( MRI ) . Strikingly , newborn neurons can acquire mature electrophysiological properties and respond to local afferent inputs . However , they unexpectedly fail to rebuild long-distance circuit and the original tissue architecture . The data provide the first proof for the precision with which axolotls regenerate a diverse set of neurons , which in turn become electrophysiologically active and receive local afferent inputs . Notably , however , our results also challenge prior assumptions of functional brain regeneration in salamanders by uncovering unappreciated limitations in the capacity of adult axolotls to fully rebuild original long-distance connectivity and tissue organization , a finding that redefines expectations for brain regeneration in mammals .
In order to investigate whether axolotls can reconstruct neuronal diversity , we have started by building a molecular map of the neuronal populations present in the axolotl pallium . The neuronal composition of the axolotl pallium is largely unknown , with the urodele pallium being classically divided into medial pallium ( MP ) , dorsal pallium ( DP ) , and lateral pallium ( LP ) based on the size and location of cell nuclei , as well as rudimentary connectivity ( Herrick , 1948; Kokoros and Northcutt , 1977; Westhoff and Roth , 2002 ) ( Figure 1a ) . We first used Nissl staining of serial sections to build an anatomical atlas of the pallium , which we have subsequently used to precisely match the rostro-caudal location of sections among all animals compared in this study ( Figure 1—figure supplement 1 ) . 10 . 7554/eLife . 13998 . 003Figure 1 . The axolotl pallium contains molecularly diverse neuronal subpopulations . ( a ) Schematic representation of the axolotl pallium showing ependymoglia cells ( red ) lining the ventricle , neurons ( black ) positioned in the dorsal pallium close to the ventricle , and fiber tracts ( green ) occupying the region closer to the pia . ( b ) Schematic representation of cortical neuronal subtypes in the mouse neocortex . Major classes of murine cortical projection neuron subtypes as well as selected classes of cortical interneurons are depicted . ( c ) In situ hybridization and immunohistochemistry for selected cortical projection neuron and interneuron markers on coronal sections of the adult axolotl pallium . Insets , high-magnification images of regions of marker expression in the dorsal pallium ( top panels ) and the medial/lateral pallium ( bottom panels ) . ( d ) Schematic map of the localization of molecularly distinct neuronal subtypes within the adult axolotl pallium . ( e ) Satb2HI regions in the dorso-lateral pallium are distinctly separate from regions of Ctip2HI and Tle4HI expression , showing a molecular boundary . ( f ) Satb2HI cells largely coexpress Calb2 . A , Anterior; P , Posterior; D , Dorsal; V , Ventral; MP , Medial Pallium; DP , Dorsal Pallium; LP , Lateral Pallium; S , Septum; Str , Striatum; Ctx , Cortex; IN , cortical Interneurons; PN , cortical Projection Neurons; SCPN , Subcerebral Projection Neurons; CThPN , CorticoThalamic Projection Neurons; CPN , Callosal Projection Neurons; L4N , Layer 4 Neurons; SST , Somatostatin; V , Ventricle . Scale bars; 500 μm ( c , top panels ) , 200 μm ( c , insets; e and f , left panels ) , 100 μm ( e and f , right panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 00310 . 7554/eLife . 13998 . 004Figure 1—figure supplement 1 . Anatomical map of the axolotl pallium . ( a ) Top view of the axolotl brain , with an indication of the section numbering system in the pallium . Inset: Whole-body image of a representative axolotl used in the study . ( b ) Nissl staining of representative coronal sections spaced at 300 μm . ( c ) Immunohistochemistry for NeuN or Hu-C/D and neuronal subtype markers , Ctip2 , Satb2 , Tle4 , and Calb2 shows that the majority of neuronal subtype markers colocalize with either NeuN or Hu-C/D . Tle4 is also expressed at low levels in NeuN- ependymoglia cells . Th , Thalamus; OT , Optic Tectum; SC , Spinal Cord . Scale Bars , 500 μm ( b ) , 50 μm ( c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 004 Next , we built a molecular map of neurons present in the pallium . We selected 19 genes that , in the mouse , are known to mark specific subtypes of either excitatory pyramidal neurons or inhibitory interneurons of the cerebral cortex ( Figure 1b ) and performed in situ hybridization on coronal sections of the adult axolotl brain , using riboprobes designed on axolotl cDNA sequences . Among all markers tested , expression of nine genes could be detected reliably in the axolotl pallium: ctip2 , satb2 , tle4 , rorb , er81 , fezf2 , gad2 , sst , and calb2 . In the mouse neocortex , Fezf2 and Ctip2 are specifically expressed by subcerebral projection neurons ( ScPNs ) in layer Vb at high levels and by corticothalamic projection neurons ( CThPNs ) in layer VI at low levels ( Arlotta et al . , 2005; Chen et al . , 2005a; Chen et al . , 2005b; Molyneaux et al . , 2005 ) . Within these two neuronal populations , Er81 labels ScPNs ( Yoneshima et al . , 2006 ) , but not CThPNs , which selectively express Tle4 ( Molyneaux et al . , 2007 ) . Callosal projection neurons ( CPNs ) across all layers distinctly express Satb2 and are negative for Ctip2 , Fezf2 , and Tle4 ( Alcamo et al . , 2008; Britanova et al . , 2008 ) . Finally , Rorb labels locally connected excitatory interneurons of layer IV ( Jabaudon et al . , 2012 ) . Among Gad2+ cortical GABAergic interneurons , Sst and Calb2 mark specific subtypes ( Bartolini et al . , 2013 ) , although Calb2 has been shown to also be transiently expressed in pyramidal neurons residing in layer Va ( Liu et al . , 2014 ) . We found that , similarly to the mouse brain , these markers labeled specific regions of the axolotl pallium . Ctip2 was highly expressed in DP with lower expression in MP and LP; Satb2 , detected by immunohistochemistry , was expressed in a restricted region of DP and in scattered cells within MP; tle4 marked a domain at the border of the dorso-lateral pallium and was expressed at lower levels across DP and MP; rorb was expressed in a distinct domain in the dorso-lateral pallium; er81 was expressed in MP at high levels and in DP at lower levels; fezf2 was expressed mostly in the ventral MP and in scattered cells within LP ( Figure 1c ) . Among the murine cortical interneuron markers , gad2 was expressed in scattered cells within the dorso-medial pallium and MP , sst was expressed in scattered cells across DP and MP , and calb2 defined a circumscribed domain in DP , as well as sparse cells in MP ( Figure 1c ) . While all genes tested showed unique distributions , the data indicated that markers of subpopulations of mouse pyramidal neurons , which are normally grouped into defined neocortical layers , cluster within distinct domains in the axolotl pallium ( Figure 1d ) . In order to have reliable molecular reference points after regeneration , we selected , for further analysis , four of the nine genes tested - ctip2 , satb2 , tle4 , and calb2 - because of their highly circumscribed and robust profiles of expression within defined regions of the dorsal and dorso-lateral pallium . We used immunohistochemistry to define , at the single cell level , the colocalization of these markers . We found that the dorso-lateral pallium contains Satb2HI , Ctip2LO , Tle4HI cells and Satb2LO , Ctip2HI , Tle4HI cells ( Figure 1e ) . Additionally , Satb2HI cells in the dorso-lateral pallium also express Calb2 ( Figure 1f ) . Cells labeled with these markers also colocalized with NeuN or Hu-C/D , indicating that they are neurons ( Figure 1—figure supplement 1 ) . Together , the data provide the first map of molecularly distinct populations of neurons within the adult axolotl pallium to enable injury of defined neuronal groups with fidelity . We established a stereotactic injury model to remove a reproducible portion of the dorsal pallium such that the ctip2 , satb2 , tle4 , and calb2 domains were ablated ( Figure 2a ) . To define the time course of regeneration , we first investigated the overall pallial morphology of fixed whole brain preparations from adult axolotls sacrificed at 1 , 2 , 4 , and 11 weeks post injury ( wpi ) . We found that the wound closes by 4wpi , and is largely undetectable by 11wpi ( Figure 2b ) . 10 . 7554/eLife . 13998 . 005Figure 2 . Temporal dynamics of successful pallial regeneration after acute mechanical injury . ( a ) Location of the injury site in the dorsal pallium , left hemisphere . ( b ) Representative stereoscope photographs of injured brains at 1 , 2 , 4 , and 11wpi . ( c ) 3D renderings of in vivo MRI images ( top panels ) and MRI coronal cross-sections within the injury site ( bottom panels ) of a representative axolotl brain during a 28-day time course post-injury . ( d ) Enlarged insets ( from the red boxes in c ) show three distinct stages of wound closure: stump formation , protrusion ( red arrow ) , and closure . Red dotted lines outline the tissue . ( e ) The sizes of the injury site ( in mm3 , see Materials and methods ) remain unchanged for the first 14 days , but decrease rapidly within a subsequent four-day period ( n=3 ) . ( f ) Schematic of in vivo BrdU labeling in the early post-injury phase . ( g ) Time course of BrdU and NeuN immunohistochemistry on coronal sections at 1 , 2 , and 4wpi shows limited cell proliferation within the first 2 weeks and increased proliferation by 4 wpi . Insets , magnified images at 4wpi ( panel g’ ) . ( h ) Schematic of in vivo BrdU labeling at mid post-injury phases . ( i ) Immunohistochemistry for BrdU and NeuN on coronal sections at 11wpi shows that a large subset of BrdU+ cells is composed of NeuN+ postmitotic neurons . Insets , magnified images at 11wpi ( panel i’ ) . ( j ) Model of the regenerative process during the first 11 weeks after acute mechanical injury . Wound closure proceeds in 3 distinct stages during the first 4 wpi: stump generation , stump protrusion , and wound closure . Only subsequently , newly proliferated BrdU+ cells populate , in large number , the injured region . By 11wpi , newborn neurons are generated . OB , Olfactory Bulb; Th , Thalamus; OT , Optic Tectum; wpi , weeks post injury; D , Day; LP , Lateral Pallium; MP , Medial Pallium; Str , Striatum; S , Septum . Scale bars; 500 μm ( g and i ) , 100 μm ( g’ and i’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 00510 . 7554/eLife . 13998 . 006Figure 2—figure supplement 1 . 4wpi proliferative cells express markers of ependymoglia cells . ( a–h ) Immunohistochemistry against Sox2 ( a and c ) , BrdU ( b , d , f , and h ) , and Gfap ( e and g ) on representative coronal sections of 4wpi pre-metamorphosis animals . This analysis shows that the majority of the BrdU+ nuclei in the injured hemisphere are Sox2+ and Gfap+ . ( i–l ) Immunohistochemistry against Gfap ( i ) , BrdU ( j and l ) , and Sox2 ( k ) on representative coronal sections of 4wpi post-metamorphosis animals . BrdU+ nuclei in the injured hemisphere are Sox2+ and Gfap+ even after metamorphosis . V , Ventricle . Scale Bars , 500 μm ( whole brain images ) 200 μm ( insets ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 00610 . 7554/eLife . 13998 . 007Figure 2—figure supplement 2 . Newborn neurons populate distant , uninjured regions of the brain . ( a , b ) An increase in the numbers of BrdU-labeled cells and PH3+ mitotic cells is observed upon injury both at the injury site ( dorsal pallium , a ) and in regions distal to the injury ( rostral telencephalon , b ) . Upper panels , schematics of the telencephalic regions analyzed in 4wpi injured and sham axolotls ( pallium in a , rostral telencephalon in b ) . Middle panels , representative images used for cell counting . Lower panels , quantification of BrdU+ nuclei and PH3+ nuclei in rostral-caudal matched coronal sections of injured ( n=7 ) and sham ( n=5 ) animals . ( c , d ) Regeneration of BrdU+/NeuN+ newborn neurons upon injury at both the injury site ( dorsal pallium , c ) and in regions away from the injury ( rostral telencephalon , d ) at 11wpi . Upper panels , schematics of the telencephalic regions analyzed in 11wpi injured and sham axolotls ( pallium in c , rostral telencephalon in d ) . Middle panels , representative immunohistochemistry images used for cell counting . Lower panels , quantification of BrdU+/NeuN+ newborn neurons in the injured pallium ( c ) and rostral telencephalon ( d ) ( n=6 ) compared to the sham ( n=4 ) . V , Ventricle . Scale Bars; 500 μm ( a-b , left panels ) , 50 μm ( a-b , right panels , c-d ) . All results are expressed as the mean ± SEM . *p<0 . 05 , **p<0 . 01 , unpaired , two-tailed Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 007 The tissue dynamics that accompany brain regeneration in vivo have never been investigated . To gain information on the tissue-level morphological changes that take place over the course of the regenerative process in individual animals and to account for animal-to-animal variability , we performed in vivo MRI of live axolotls undergoing pallial regeneration . This method demonstrated that the wound closure process occurs rapidly , between 2 and 4wpi ( Figure 2c–e and Video 1 ) . Notably , the lateral pallium formed a stump , which subsequently thinned out before protruding towards the medial pallium and closing the gap generated by the injury ( Figure 2d ) . This indicates that wound healing occurs largely by dynamic tissue remodeling that occurs before cellular proliferation starts . 10 . 7554/eLife . 13998 . 008Video 1 . in vivo MRI reveals temporal dynamics of the first 4 weeks of pallial regeneration . The 3D surface rendering of MRI images ( left panel ) shows the gradual wound closure process . The raw MRI images ( right panel , red dotted lines outline the tissue ) show that within the first 2 weeks , the tissue does not change in shape or trajectory . During the next three time points , three stages of wound closure are observed: stump formation , protrusion , and wound closure , which occur rapidly . The time course of regeneration is in days ( bottom panel ) . D , Day; LP , Lateral Pallium; DP , Dorsal Pallium; MP , Medial Pallium . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 008 To assess the cellular dynamics of the first eleven weeks post-injury and to map the time course of neurogenesis , animals were pulsed with Bromodeoxyuridine ( BrdU; see Materials and methods ) . At 4wpi , a large number of BrdU+/NeuN- cells populated the injured hemisphere , indicating that accumulation of proliferative cells precedes neuronal differentiation ( Figure 2g ) . Of note , BrdU+ cells expressed Sox2 and Gfap , two ependymoglia cell markers ( Kirkham et al . , 2014 ) ( Figure 2—figure supplement 1 ) . By 11wpi , the majority of BrdU+ cells expressed NeuN ( 69 . 5% , n=6 animals ) ( Figure 2i ) indicating that neuronal differentiation occurs between 4 and 11wpi . Interestingly , we observed an increased number of BrdU+ and PH3+ proliferative cells in regions rostral to the injury site , at the level of the rostral telencephalon , despite the fact that this region was not injured ( Figure 2—figure supplement 2 ) . To determine whether this distal proliferative response leads to regeneration of new neurons , we quantified the number of BrdU+/NeuN+ cells in the rostral telencephalon at 11wpi in injured and sham controls , in which the skin and skull were opened but the brain was not injured . Quantification revealed that there were significantly more BrdU+/NeuN+ nuclei in the rostral telencephalon of injured animals compared to sham animals , indicating that an injury to the pallium leads to generation of newborn neurons distally ( Figure 2—figure supplement 2 ) . Taken together , these data demonstrate that regeneration of the pallium initiates with distinct morphological changes to the injury edge ( stump ) , which thins out and closes the wound . This process precedes rapid proliferation of progenitors and neurogenesis ( Figure 2j ) . In addition , the progenitor proliferative response induced by injury is not restricted to the region immediately adjacent to the injury site but induces a more global response in distal uninjured regions . To assess the axolotl’s ability to regenerate distinct neuronal populations in distinct regions of the pallium , we quantified the number of NeuN+ , Ctip2+ , Satb2+ , Tle4+ , and Calb2+ neurons at 11wpi in the dorsal pallium of contralateral , injured , and sham hemispheres . Cells were counted in four consecutive sections , sampled every 300 μm , over a total rostral-caudal length of 900 μm spanning the injury site . The sum of all cells counted in all sections was taken for each sample ( Figure 3a ) . From this analysis , we found no significant differences between the number of cells present in the contralateral , injured ( regenerated ) , and sham hemispheres for any of the markers quantified ( Figure 3b–f ) . The data indicate that the regenerated pallium contains each neuronal population in numbers comparable to control ( contralateral and sham ) pallium . 10 . 7554/eLife . 13998 . 009Figure 3 . Molecularly diverse neuronal subtypes regenerate in the axolotl pallium . ( a ) Schematic of the sections used for quantification ( panels b–f and k ) . Images from 4 consecutive coronal sections , 300 μm apart , and spanning the injury site from contralateral , injured , and sham hemispheres were used for cell counting . ( b–f ) Quantification of NeuN+ , Ctip2+ , Satb2+ , Tle4+ , and Calb2+ nuclei in the region of interest shows no significant differences among contralateral , injured , and sham hemispheres . ( g–j ) Newborn BrdU+ cells in injured coronal sections co-express Ctip2 , Satb2 , Tle4 , and Calb2 , showing that neuronal subtype diversity is replenished upon regeneration; insets , high-resolution confocal images of co-labeled nuclei . ( k ) Quantification of BrdU+/neuronal subtype-marker+ nuclei in the dorsal pallium of contralateral , injured , and sham hemispheres . LP , Lateral Pallium; DP , Dorsal Pallium; MP , Medial Pallium; V , Ventricle . Scale Bars; 200 μm ( g–j , top panels ) , 20 μm ( g–j , bottom panels ) . All results are expressed as the mean ± SEM . **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001; one-way ANOVA with post-hoc Tukey’s multiple comparison test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 00910 . 7554/eLife . 13998 . 010Figure 3—figure supplement 1 . Neuronal subtypes in uninjured regions of the pallium remain unchanged in number at 11wpi . ( a , b ) Immunohistochemistry against Ctip2 , Satb2 , Tle4 , Calb2 , and NeuN on representative coronal pallium sections of injured 11wpi axolotls , showing the medial pallium ( a , left panels ) and the lateral pallium ( b , left panels ) . Quantification of Ctip2+- , Satb2+- , Tle4+- , Calb2+- , and NeuN+-nuclei numbers from four sections spaced at 300 μm in the medial pallium ( a , left panels ) and the lateral pallium ( b , right panels ) . Scale Bars , 50 μm . All results are expressed as the mean ± SEM . One-way ANOVA . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 01010 . 7554/eLife . 13998 . 011Figure 3—figure supplement 2 . Neuronal diversity in the dorsal pallium is regenerated in post-metamorphic axolotl brain . ( a ) Representative images of adult axolotls before and after L-thyroxine-induced metamorphosis . Left , pre-metamorphosis axolotl , inset shows the presence of a tail fin ( left panel ) . Right , post-metamorphosis axolotl , inset shows the lack of a tail fin . Post-metamorphosis axolotls also lack external gills ( red arrowhead ) . ( b ) Stereoscope images showing the injury sites of the post-metamorphosis axolotls ( left hemisphere , red dotted rectangle ) at 4wpi and 11wpi . ( c ) Representative immunohistochemistry images of coronal section show similar localization pattern of BrdU+/PH3+/NeuN- proliferating cells to that of the pre-metamorphosis axolotl in the dorsal pallium . ( d ) Regenerated cells labeled with BrdU colocalize with neuronal markers NeuN , Ctip2 , Satb2 , Tle4 , and Calb2 as shown in representative injured coronal sections; insets , magnification of co-labeled nuclei ( arrowheads ) . LP , Lateral Pallium; DP , Dorsal Pallium; MP , Medial Pallium; V , Ventricle . Scale Bars; 500 μm ( c , top panels ) , 200 μm ( c , bottom panels , d , top panels ) , 20 μm ( d , bottom panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 011 To determine whether there was an increase in the number of newborn neuronal subtypes in the injured pallium compared to controls , we quantified the number of cells that showed colocalization of each neuronal subtype-specific marker with BrdU at 11wpi . This analysis showed that the injury significantly increases the number of newborn neurons within each molecularly defined neuronal subpopulation , compared to controls ( n=6 contralateral; n=6 injured; n=4 sham ) ( Figure 3g–k and Table 1 ) , indicating that distinct neuronal populations are regenerated upon injury . 10 . 7554/eLife . 13998 . 012Table 1 . Quantification of BrdU+/Neuronal Subtype Marker+ nuclei in contralateral , injured ( 11wpi ) , and sham hemispheres . The numbers represent the sum of BrdU+/Neuronal Subtype Marker+ nuclei from four consecutive sections spanning the injury site . All results are expressed as the mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 012Ctip2+/BrdU+Satb2+/BrdU+Tle4+/BrdU+Calb2+/BrdU+Contralateral ( n=6 ) 12 . 83 ± 3 . 1145 . 167 ± 1 . 53721 . 00 ± 15 . 352 . 000 ± 1 . 26511wpi Injured ( n=6 ) 199 . 2 ± 37 . 9370 . 33 ± 16 . 32310 . 2 ± 76 . 54101 . 7 ± 11 . 10Sham ( n=4 ) 4 . 250 ± 3 . 0656 . 000 ± 4 . 37810 . 25 ± 17 . 212 . 250 ± 1 . 652 These numbers likely underestimate the number of neurons within each population that are newly born due to the difficulty of labeling by BrdU the entirety of the dividing cell pool . It is therefore not possible to conclude whether some of the neurons in the regenerate are newborn ( yet not labeled by BrdU ) or are endogenous neurons that have migrated from neighboring territories . If migration of neurons into the injured region were a major contributor to the regenerative process , this could cause a decrease in the number of certain neuronal subtypes in regions adjacent to the injury site . We thus quantified the number of neuronal subtypes in MP and LP . We found no significant differences between contralateral , injured , and sham hemispheres for any of the neuronal subtype markers tested ( Figure 3—figure supplement 1 ) , suggesting , albeit not fully excluding , the possibility that no major waves of migration of endogenous neurons reconstitute neuronal diversity in the regenerated dorsal pallium . It has been previously reported that the brain regeneration capability of the axolotl decreases upon metamorphosis ( Kirsche et al . , 1965 ) . To address this issue directly , we induced metamorphosis in adult , sexually mature axolotls by administration of L-thyroxine exogenously via immersion ( Page and Voss , 2009 ) ( Figure 3—figure supplement 2 ) . Upon metamorphosis , we injured the dorsal pallium of post-metamorphosis axolotls and found that morphological regeneration takes place within a similar time frame as the pre-metamorphosis axolotls ( Figure 3—figure supplement 2 ) . At 11wpi , we found that the post-metamorphic injured pallium contains BrdU+/NeuN+ , BrdU+/Ctip2+ , BrdU+/Satb2+ , BrdU+/Tle4+ , and BrdU+/Calb2+ nuclei ( Figure 3—figure supplement 2 ) . Together , these data show that adult pre- and post-metamorphosis axolotls are able to regenerate neuronal subtypes within the dorsal pallium . To determine whether the regenerated neurons in the injured pallium mature into functional neurons , we sought to characterize their electrophysiological properties by using whole-cell patch clamp recordings ( Figure 4a ) . We compared neurons identified as EdU+ by post hoc staining in the dorsal pallium of injured animals ( 11-15wpi; n=14 neurons ) and neurons at matched locations in uninjured control animals ( n=9 neurons ) ( Figure 4b ) . Injected , depolarizing current steps evoked action potentials in all cells recorded in the regenerated pallium ( Figure 4c–d ) . We found that regenerated neurons exhibited passive and active membrane properties that were comparable to those of control neurons ( Figure 4e and Table 2 ) . However , in some cases ( 2/14 neurons ) , when compared with uninjured neurons , newly regenerated neurons displayed passive and active membrane properties indicative of a slightly immature state , such as a larger Rm , a smaller Cm , more depolarized resting membrane potential , and broader and shorter action potentials ( Figure 4—figure supplement 1 and data not shown ) . These results indicate that the regenerated pallium contained neurons of varying maturity , but a large number of regenerated neurons were able to acquire mature electrophysiological traits comparable to endogenous control neurons . 10 . 7554/eLife . 13998 . 013Figure 4 . Regenerated neurons show electrophysiological features similar to uninjured pallium neurons and receive afferent input . ( a ) Schematic representation of whole-cell patch clamp recording and subsequent biocytin filling . ( b ) Representative images of EdU+/Biocytin+ neurons in the injured pallium . ( c ) Representative traces of the membrane voltage changes in response to depolarizing and hyperpolarizing current steps injections of neurons from uninjured and injured animals . ( d ) Zoom-in of single action potential evoked by a depolarizing current step recorded in neurons of dorsal pallium . ( e ) Summary of passive and active electrophysiological properties: membrane capacitance ( Cm ) , membrane resistance ( Rm ) , resting membrane potential ( RMP ) , action potential ( AP ) threshold , AP height and AP half width . ( f ) Sample traces of spontaneous postsynaptic currents ( sPSCs ) recorded under voltage clamp ( Vh = −70 mV ) . ( g ) Summary of sPSCs features in neurons from uninjured and injured animals . All results are expressed as the mean ± SD . *p<0 . 05; unpaired , two-tailed Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 01310 . 7554/eLife . 13998 . 014Figure 4—figure supplement 1 . EdU+ neurons in the injured pallium show immature electrophysiological properties . ( a ) Changes of the membrane voltage in response to depolarizing and hyperpolarizing current steps injections from an EdU+ neuron recorded in the injured pallium . ( b ) Zoom-in of single action potential evoked by a depolarizing current step . Broad and short action potential is typical of immature neurons . ( c ) Sample trace of spontaneous postsynaptic currents ( sPSCs ) recorded under voltage clamp ( Vh=−70 mV ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 01410 . 7554/eLife . 13998 . 015Figure 4—figure supplement 2 . Most sPSCs are abolished by administration of AMPA receptor blocker . ( a ) Sample trace of spontaneous postsynaptic currents ( sPSCs ) recorded under voltage clamp ( Vh=−70 mV ) . ( b ) Sample trace of the same neuron after application of AMPA receptor blocker , NBQX [10 µM] , which abolishes most sPSCs . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 01510 . 7554/eLife . 13998 . 016Figure 4—figure supplement 3 . Spontaneous calcium transients in neurons do not differ between injured and control dorsal pallium . ( a ) Schematic of ex vivo calcium imaging . ( b ) Representative image of Fluo-4 staining in the dorsal pallium after dye loading ( left panel ) . Region of interest ( ROI ) segmentation for individual neurons ( middle panel ) and merge ( right panel ) . ( c ) Representative traces of spontaneously generated calcium transients in labeled neurons within dorsal pallium . ( d ) Raster plot of peaks in traces from panel c . ( e ) Quantification of the peak amplitude ( left panel ) , rise time ( middle panel ) , and fall time ( right panel ) in individual neurons shows no significant difference between injured and control dorsal pallium . V , ventricle . All results are expressed as the mean ± SEM . Unpaired , two-tailed Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 01610 . 7554/eLife . 13998 . 017Table 2 . Parameters of electrophysiological properties of neurons in uninjured and injured brains . All results are expressed as the mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 017Cm ( pF ) Rm ( mOhms ) RMP ( mV ) Threshold ( mV ) AP Height ( mV ) Half Width ( ms ) Uninjured ( n=9 ) 49 . 3 ± 16 . 4294 . 1 ± 155 . 8-59 ± 4-36 . 5 ± 361 . 9 ± 52 . 3 ± 0 . 4Injured ( n=14 ) 45 . 1 ± 21 . 8426 . 7 ± 361 . 6-59 ± 5-30 ± 668 . 7 ± 14 . 32 . 7 ± 0 . 7 To determine whether the regenerated neurons received synaptic input , we recorded spontaneous postsynaptic currents ( sPCSs ) . We found that all EdU+ , regenerated neurons in the injured pallium received postsynaptic currents ( Figure 4f ) and displayed no differences in postsynaptic response compared to control neurons in the uninjured pallium ( Figure 4g ) . Most of the spontaneous currents were blocked by the AMPA receptor antagonist , NBQX , suggesting that AMPA receptors were involved in the synaptic transmission ( Figure 4—figure supplement 2 ) . To understand the kinetics of large-scale neuronal activity , we performed ex vivo calcium imaging with single-cell resolution ( Figure 4—figure supplement 3 ) . Measuring the key characteristics of spontaneously generated calcium transients , including the peak amplitude , rise time , and fall time , we detected no significant differences between uninjured ( n=5 animals ) and injured ( 11-15wpi; n=4 animals ) dorsal pallium neurons ( Figure 4—figure supplement 3 and Video 2–3 ) . 10 . 7554/eLife . 13998 . 018Video 2 . Calcium imaging in the uninjured dorsal pallium reveals spontaneous calcium transients during homeostasis . The individual cells labeled by Fluo-4 in the control pallium show spontaneous , transient increase in fluorescence . V , ventricle; DP , Dorsal Pallium . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 01810 . 7554/eLife . 13998 . 019Video 3 . Calcium imaging in the injured dorsal pallium reveals spontaneous calcium transients . The individual cells labeled by Fluo-4 in the injured pallium show spontaneous , transient increase in fluorescence . V , ventricle; DP , Dorsal Pallium . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 019 Together , these results indicate that the regenerated neurons are able to mature into electrophysiologically functional neurons and to receive and respond locally to afferent inputs from other neurons . It is unclear whether , beyond generating the diversity of neuronal constituents , the original topography of neurons and fiber tracts in the pallium are also rebuilt upon injury , yet this may be required for proper pallial function . To quantify the radial organization of neurons in regenerated pallium , we radially subdivided the dorsal pallium into three equally sized bins and quantified the number of NeuN+ neurons in bin 3 ( closest to pia ) normalized to the area of each individual bin ( disorganization index; DI ) . The sum of DI from four consecutive sections spanning the injury site was taken for each group . In uninjured animals , bin 3 is mostly populated by neuronal processes and does not contain cell bodies . In agreement , we found that virtually no NeuN+ neurons are present within this bin in the contralateral and sham hemispheres . However , in the regenerated pallium at 11wpi , we consistently found neurons in bin 3 ( Contra n=7 DI 19 . 46 ± 3 . 272; Injured n=7 DI 212 . 4 ± 65 . 85; Sham n=5 DI 22 . 95 ± 8 . 508 ) , indicating that the radial organization of neurons in the dorsal pallium at 11wpi is disrupted compared to controls ( Figure 5a–b ) . To confirm that the observed radial disorganization is not temporary , we performed the same analysis at 20wpi . Similar to the 11wpi time point , neurons were consistently positioned in bin 3 in the injured hemisphere , but not in the control ( Contra n=5 DI 40 . 87 ± 9 . 746; Injured n=5 DI 548 . 6 ± 104 . 1 ) ( Figure 5a–b ) , indicating that neuronal disorganization remains present at 20wpi and is a stable feature of the regenerated pallium . 10 . 7554/eLife . 13998 . 020Figure 5 . Tissue architecture is disrupted in the regenerated pallium . ( a ) The radial distribution of NeuN+ cells within the dorsal pallium is altered after regeneration . Representative immunohistochemistry images of coronal sections of contralateral , injured , and sham hemispheres at 11wpi ( left ) and 20wpi ( right ) . NeuN+ neurons populate the bin closest to the pia ( bin 3 ) in the injured hemispheres at both time points , in contrast to controls . ( b ) Quantification of the disorganization index in four consecutive sections of contralateral ( 11wpi and 20wpi ) , injured ( 11wpi and 20wpi ) , and sham ( 11wpi ) hemispheres . ( c ) The topography of neuronal subtypes is altered upon regeneration . Representative immunohistochemistry images of coronal sections of contralateral ( top panels ) and injured ( bottom panels ) hemispheres at 20wpi . ( d ) Immunohistochemistry for Tubb3 shows atypical organization of neuronal processes in the 11wpi injured pallium compared to uninjured contralateral and sham controls . LP , Lateral Pallium; MP , Medial Pallium; V , Ventricle . Scale Bars; 400 μm ( a ) , 50 μm ( c , d ) . All results are expressed as the mean ± SEM . *p<0 . 05 , **p<0 . 01; one-way ANOVA with post-hoc Tukey’s multiple comparison test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 020 The data prompted the question of whether the topography of neuronal subtypes is disorganized in the regenerate . We thus performed immunohistochemistry for Ctip2 , Satb2 , Tle4 , and Calb2 at 20wpi . We found that , while the uninjured contralateral hemisphere showed stereotyped organization , in which neuronal subtypes cluster within distinct domains ( Figure 5c , top panels ) , the dorsal pallium of the injured hemisphere contained neuronal subtypes localized outside of their stereotyped domains ( Figure 5c , bottom panels ) . In agreement , immunohistochemistry for Tubb3 in the dorsal pallium of 11wpi contralateral , injured , and sham hemispheres showed that the injured pallium contained scattered neuronal processes that lacked the stereotypical , circumscribed organization along the pia seen in the uninjured pallium ( Figure 5d ) . Taken together , these results indicate that the molecular diversity of neuronal subtypes is rebuilt within an altered pallial architecture . Given the abnormal tissue architecture of the regenerated pallium , we sought to determine whether regenerated neurons are able to remake long-distance axonal projections to original targets . In amphibians , tracing studies have shown that neurons in the dorsal pallium project to the medial pallium ( MP ) , lateral pallium ( LP ) , and olfactory bulb ( OB ) ( Westhoff and Roth , 2002; Northcutt et al . , 1980; Neary et al . , 1990 ) . We selected OB-projecting neurons for further analysis because their cell bodies ( in DP ) are distinctly separated from their axon targets ( in OB ) . We injected red fluorescent microspheres ( Retrobeads ) in the OB of control , uninjured ( n=5 ) and regenerated axolotls at both 11wpi ( n=6 ) and 42wpi ( n=3 ) . All animals were sacrificed one week post-injection . In each sample , we quantified the total number of Retrobead+ neurons present in the dorsal pallium , across four consecutive sections spanning the injury site . We found that a reduced number of neurons were retrogradely traced from the OB in the regenerated versus control axolotls ( Uninjured n=5 81 . 60 ± 13 . 90 neurons; 11wpi n=6 24 . 83 ± 5 . 326 neurons; 42wpi n=3 19 . 33 ± 2 . 603 neurons ) ( Figure 6a–b ) . 10 . 7554/eLife . 13998 . 021Figure 6 . Long-distance projections from the regenerated dorsal pallium are significantly reduced . ( a ) Retrograde tracing from the olfactory bulb shows reduced number of labeled cells in the dorsal pallium and lateral pallium at 11wpi ( middle panels ) and 42wpi ( right panels ) compared to uninjured control ( left panels ) . Insets , high-magnification images of DAPI ( green ) and Retrobeads ( red ) . ( b ) Quantification of the number of Retrobead+ cells in the dorsal and lateral pallium at different time points after injury . ( c ) Representative immunohistochemistry images of BrdU+/Retrobead+ cells in the regenerated dorsal pallium at 11wpi ( left panel ) and 42wpi ( right panel ) . ( d ) ex vivo DTI of 11wpi brains show disruption of fiber tracts ( arrowheads ) in the injured hemisphere . White dotted lines ( left panels ) represent regions of interest in injured and contralateral hemispheres . OB , Olfactory Bulb; CP , Choroid Plexus; Th , Thalamus; V , Ventricle; DP , Dorsal Pallium; MP , Medial Pallium; LP , Lateral Pallium . Scale Bars; 50 μm ( b ) , 20 μm ( d ) . All results are expressed as the mean ± SEM . *p<0 . 05 , **p<0 . 01; one-way ANOVA with post-hoc Tukey’s multiple comparison test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 02110 . 7554/eLife . 13998 . 022Figure 6—figure supplement 1 . The neuronal topography of the lateral pallium is not disorganized . ( a ) Representative immunohistochemistry images of the lateral pallium from contralateral , injured ( 11wpi ) , and sham brains , labeled for NeuN ( top panels ) and Tle4 ( bottom panels ) . ( b ) Representative in situ hybridization images of rorb on 11wpi injured , contralateral , and uninjured lateral pallium sections . LP , Lateral Pallium; V , Ventricle . Scale Bars; 50 μm ( a ) , 200 μm ( b ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 022 Notably , regenerated BrdU+/Retrobead+ neurons accounted for only 10 . 2% and 6 . 6% of the total number of OB-projecting neurons located in the DP at 11wpi and 42wpi , respectively , suggesting that the majority of the axons extending from the regenerated pallium to the OB originate from pre-existing BrdU- neurons that were spared by the injury ( Figure 6c , arrowheads ) . Given that the number of Retrobead+ neurons remains significantly reduced at 42wpi , it is unlikely that this is a matter of neuronal maturity or low speed of axon growth . Interestingly , we also found a reduction in the number of Retrobead+ neurons in the lateral pallium , despite the fact that this region was not mechanically injured ( Uninjured n=5 118 . 2 ± 38 . 96 neurons; 11wpi n=6 21 . 83 ± 4 . 629 neurons; 42wpi n=3 29 . 67 ± 10 . 35 neurons ) ( Figure 6a–b ) . However , this was not accompanied by disorganization of the lateral pallium at 11wpi , as detected by NeuN , Tle4 , and rorb labeling ( Figure 6—figure supplement 1 ) . Potential severance of the tract that connects the lateral pallium to the olfactory bulb during the dorsal pallium injury may account for this reduction . As expected , we found no significant difference between groups in the retrogradely labeled neurons that project to the olfactory bulb from the medial pallium ( Figure 6a and data not shown ) . To further investigate the failure of regenerated neurons to regenerate long-distance projections , we imaged the axon bundles connecting the pallium to the olfactory bulb using diffusion tensor imaging ( DTI ) in both control and regenerated animals at 11wpi ( n=2 ) . In agreement with our retrograde tracing results , magnetic resonance tractography revealed distinct three-dimensional fiber tracts connecting the pallium and the olfactory bulb in the control , contralateral hemisphere . In striking contrast , we observed significant truncation of the same bundles in the regenerated hemisphere ( Figure 6d , arrowheads ) . These results show that regeneration of the pallium does not remake long-distance projections present in the original brain . To investigate whether regenerated neurons could excite physiological targets that were more closely located , we stimulated the dorsal pallium and performed extracellular field recordings in the lateral and medial pallium . We used ex vivo brain slices of uninjured and injured ( 11–15wpi ) axolotls ( Figure 7a–b ) . We found that , in both DP-to-LP and DP-to-MP connections , the amplitude of the pre-synaptic fiber volley ( a measure of the number of axons being activated by an electrical stimulus; Negative Peak 1 or NP1 ) was significantly reduced in the injured pallium compared to controls ( DP-to-LP: Uninjured n=5 animals; Injured n=3 animals; DP-to-MP: Uninjured n=4 animals; Injured n=3 animals ) ( Figure 7c–d ) , indicating that the activity of presynaptic axons is altered upon regeneration due to a change in number or firing properties of axons . In accordance , we also observed that the field population spike ( fEPSP ) of both LP and MP was significantly decreased in injured axolotls compared to controls upon dorsal pallium stimulation ( Figure 7c–d ) , indicating that the MP and LP are not functionally activated to the same extent in the regenerated brain versus controls . 10 . 7554/eLife . 13998 . 023Figure 7 . Reduced activation of medial and lateral pallium by regenerated neurons in the dorsal pallium . ( a ) Schematic of tissue preparation for slice electrophysiology . ( b ) Schematic of experimental details showing the location of the stimulation ( DP ) and the field recording ( MP and LP ) . ( c ) Dorsal-lateral NP1 and field pop spike amplitude is decreased in injured pallium ( n=15 slices; black circles ) compared to control pallium ( n=15 slices; white circles ) as measured by extracellular field recordings . Inset , representative traces for uninjured and injured pallium recordings . ( d ) Dorsal-medial NP1 and field pop spike amplitude is decreased in injured pallium ( n=9 slices; black circles ) compared to control pallium ( n=9 slices; white circles ) as measured by extracellular field recordings . DP , Dorsal Pallium; LP , Lateral Pallium; MP , Medial Pallium; V , Ventricle . All results are expressed as the mean ± SEM . **p<0 . 005 , ***p<0 . 001; Two-way ANOVA repeated measures with Bonferroni post hoc test . DOI: http://dx . doi . org/10 . 7554/eLife . 13998 . 023 The data indicate that despite the striking ability of axolotl to regenerate a diversity of neuronal subtypes that are electrophysiologically mature and receive local afferent inputs , newborn neurons cannot rebuild long-distance circuit , as was previously assumed . This demonstrates an unknown obstacle to functional brain repair even in a highly regenerative species .
The function of the mammalian brain relies on the processing power of an outstanding diversity of neuronal subtypes , which are integrated into distinct networks necessary for the execution of specific functions . One central goal of regenerative medicine in the CNS is therefore to rebuild not only the original heterogeneity of neurons but also their specific patterns of connectivity within the endogenous tissue . Prior work suggested that the axolotl might be a good model to understand the mechanisms of complete CNS regeneration because of its capacity to grossly regenerate large portions of the brain when mechanically injured ( Maden et al . , 2013; Burr , 1916; Kirsche and Kirsche , 1964; Winkelmann and Winkelmann , 1970 ) . However , it remains unclear to what extent brain regeneration in the axolotl leads to reformation of functional tissue . Here , we found that , upon a large mechanical injury to the pallium , the adult axolotl can regenerate the original neuronal diversity . Notably , newborn neurons acquire intrinsic electrophysiological properties and process afferent input in a manner that is indistinguishable from the endogenous neurons in uninjured brains . It is also interesting that this capacity is not lost after metamorphosis , challenging the theory that , in this species , regeneration is partly linked to the maintenance of a paedomorphic state in adulthood ( Kirsche et al . , 1965 ) . These results indicate that , beyond instructing the birth of new neurons , the adult axolotl brain is capable of regenerating a diversity of neurons that in turn are electrophysiologically functional , even when a large region of the brain is removed . An outstanding question in the field of brain regeneration remains whether , beyond rebuilding cellular complexity and local connectivity , new connections to distant targets can be restored upon brain regeneration . It is similarly unknown whether tissue architecture and neuronal topography can be regenerated . We found that axolotls possess only limited capacity to rebuild original tissue architecture after large deletions of the pallium . In addition , we uncovered an unexpected limitation to the capacity of newborn neurons to extend axons to original targets in distal brain regions . It is possible that the observed limitations relate to a lack of instructive signals from the environment rather than intrinsic limitations of the regenerated neurons . For example , in order to reach the OB , newborn neurons of the pallium will have to extend axons through uninjured territory that may lack developmental ligands necessary for axon guidance to the OB . The processes and mechanisms that initially trigger wound closure and subsequently sustain brain repair in regenerative vertebrate species are not known . Using in vivo MRI , we were able to observe , for the first time , the dynamic changes in tissue morphology that occur in live animals over the course of brain regeneration . We found that early steps of wound closure involve the generation of thinner processes from a stump that directionally grow towards each other before fusing . This strategy resembles that observed in the limb , in which closure of the wound by the wound epithelium after amputation is necessary for subsequent regenerative steps and may serve as a source of signals to govern downstream regenerative events ( Mescher , 1976; Thornton , 1957 ) . It is possible that , much like the limb , morphogenetic movements act as a signaling event to trigger cellular proliferation and accumulation of newborn , differentiated cells at the site of injury . In support of this possibility , we did not observe major waves of neuronal migration from regions adjacent to the injury site , and the topography of neurons in these regions remains properly organized upon regeneration . In the future , it will be interesting to determine the cellular source of the newborn neurons . Activated ependymoglia cells are one of the major sources of injury-induced newborn neurons in the newt and zebrafish ( Berg et al . , 2010; Kroehne et al . , 2011 ) , and these cells may similarly be the main origin of newborn neurons in the axolotl pallium . An alternative , yet not exclusive , possibility is that reprogramming of endogenous differentiated cells may contribute to the generation of new neurons . Such reprogramming events were reported , for example , in the zebrafish retina , where Mueller glia appears capable of generating new neurons upon retinal injury ( Bernardos et al . , 2007; Fausett and Goldman , 2006; Ramachandran et al . , 2010; Thummel et al . , 2008 ) . Together , our findings establish the axolotl pallium as a model to further investigate how diverse populations of electrophysiologically functional neurons are mechanistically regenerated . In addition , the results highlight previously unappreciated limits for what vertebrates are able to achieve when faced with the need to regenerate the most complex tissues . This should help define concrete goals and expectations for cellular replacement outcomes in the central nervous system of mammals .
Wildtype axolotls ( Ambystoma mexicanum ) were obtained from the Ambystoma Genetic Stock Center ( University of Kentucky , Lexington , KY ) and were housed in standard conditions at 16°C in Holtfreter’s solution . Adult animals , 15–18 cm in length , were used for all experiments . Anesthesia was achieved by immersion in 0 . 03% ethyl-p-aminobenzoate ( benzocaine; Sigma-Aldrich , St . Louis , MO ) or 0 . 1% ethyl 3-aminobenzoate methanesulfonate ( tricaine; Sigma-Aldrich ) . All animal experiments performed were approved by the Harvard University IACUC and were in accordance with institutional and federal guidelines . To access the telencephalon of anesthetized axolotls , two rectangular scalp skin flaps followed by cranial flaps were created dorsally with spring scissors . Incisions outlining a 0 . 8 mm X 1 mm rectangular injury site were made in the left dorsal pallium with a microknife ( Fine Science Tools , Foster City , CA ) secured onto a micromanipulator . The caudal-medial corner of the rectangular injury site is defined as 1 mm lateral from and 2 mm rostral from the choroid plexus . The injury site tissue was removed using No . 5 forceps . After the injury , the cranial flaps were replaced , and the skins flaps were secured with non-absorbable silk sutures ( Myco Medical , Cary , NC ) . In sham injury , the telencephalon was similarly exposed but the dorsal pallium was left intact . 250 μl ( 8 mg/ml in PBS ) of BrdU ( Sigma-Aldrich ) or EdU ( Thermo Fisher Scientific , Waltham , MA ) was injected intraperitoneally . For animals sacrificed at 4wpi , BrdU/EdU was injected three times a week while for animals sacrificed at 11wpi , BrdU/EdU was injected once at 3 , 4 , and 5wpi due to the toxic effect of more frequent regimens of BrdU/EdU administration . BrdU incorporation was analyzed by immunohistochemical staining ( see below ) . EdU incorporation was detected with the Click-it EdU Alexa Fluor 488 Imaging Kit ( Thermo Fisher Scientific ) . Uninjured brains were dissected and fixed overnight in 4% paraformaldehyde ( PFA ) at 4°C . Injured and sham-injured brains were fixed overnight with the cranium attached , dissected from the cranium , and re-fixed overnight in 4% PFA at 4°C . The extra fixation step is necessary to preserve the morphological integrity of the injured tissue . After fixation , the brains were equilibrated in 30% sucrose solution at 4°C , embedded in OCT ( Sakura Finetek , Torrance , CA ) , and cryosectioned coronally at 30 μm thickness . 10 series ( 120 sections ) were collected from each brain onto Vistavision microscope slides ( VWR , Radnor , PA ) and stored at -80°C . For BrdU detection , sections were incubated in 2M HCl for 20 min at 37°C , before incubation in blocking buffer ( 0 . 3% BSA , 8% serum , 0 . 3% Triton X-100 ) for 2 hr at room temperature ( RT ) . Sections were incubated overnight at 4°C in primary antibodies , which were diluted in blocking buffer . Primary antibodies used in this study were as follows: rat anti-CTIP2 antibody 1:100 ( Abcam , Cambridge , MA ) , mouse anti-SATB2 1:50 ( Abcam ) , rabbit anti-TLE4 1:3000 ( gift of Stefano Stifani ) , rabbit anti-CALB2 1:300 ( Swant , Marly , Switzerland ) , rat anti-BrdU 1:300 ( Accurate Chemical , Westbury , NY ) , mouse anti-BrdU 1:300 ( BD Biosciences , San Jose , CA ) , mouse anti-NeuN 1:300 ( Millipore , Billerica , MA ) , rabbit anti-Hu-C/D 1:300 ( Genetex , Irvine , CA ) , rabbit anti-PH3 1:300 ( Millipore ) , rabbit anti-Sox2 1:100 ( Abcam ) , mouse anti-GFAP 1:300 ( Sigma-Aldrich ) , mouse anti-Tubb3 1:300 ( Covance , Princeton , NJ ) . Sections were incubated in appropriate secondary antibodies from the Molecular Probes Alexa Fluor series ( Thermo Fisher Scientific ) at 1:750 dilutions for 2 hr at RT . After washing , sections were incubated in DAPI 1:50 , 000 ( Thermo Fisher Scientific ) for 10 min at RT . Sections were coverslipped with Fluoromount-G ( SouthernBiotech , Birmingham , AL ) before imaging . Tissue sections were imaged using a Nikon 90i fluorescence microscope equipped with a Retiga Exi camera ( Q-Imaging , Surrey , Canada ) and analyzed with Volocity image analysis software v4 . 0 . 1 ( PerkinElmer , Waltham , MA ) . Confocal images were acquired with Zeiss LSM 700 confocal microscope and analyzed with the ZEN Black software . In situ hybridization was performed as described previously ( Lodato et al . , 2014 ) . Riboprobes were generated as previously described ( Arlotta et al . , 2005 ) . Riboprobes for in situ hybridization were generated from axolotl cDNA using the following primers: fezf2 , F: AAAGCGACAGCAAACTCAGC R: GTTTCCTTTCTGGTGGAAGC; ctip2 , F: CCGTTCAGTCTTTTGCGAAT R: TAGCTGCCTTCCATCAATCC; tle4 , F: AACAAACAGGCGGAAATTGT R: CTCTAAAGCTTGCCGATGGA; rorb , F: GAAATTTGGCAGGATGTCCA R: ACGTCTCCAGGTGGGATTTA; er81 , F: ATGACCAGCAAGTGCCTTTT R: AGGTTTGACTGCTGGCATTG; sst , F: CAGACAAGCAAGCAGCAGAG R: TCTGTGGGAACTGCAGAGTG; calb2 , F: GGTCGAGTGCCGAGTTTATG R: TTCACCCTCTCCCCAAATAA; gad2 , F: GCGAGCAAAGGGTACAGAAG R: GGCAATACATTTTTCCTTCAAAA Nissl staining was performed as previously described ( Macklis , 1993 ) . All primary data from immunohistochemistry and in situ hybridization experiments were repeated at least three times and analyzed by one investigator , then confirmed by a second , independent investigator who was blinded to experimental conditions . For quantification of proliferating cells and newborn neurons , anatomically matched sections were processed to detect BrdU/PH3 or BrdU/NeuN , respectively . For quantification of the disorganization index , boxes of 400 μm in width and spanning the thickness of the pallium were superimposed at matched locations on each section and divided into three equally sized bins . For quantification of Retrobead+ cells , anatomically matched sections were processed to detect Retrobead+/DAPI+ cells . All counts were performed by an investigator who was blinded to the condition . Unpaired , two-tailed t-test , one-way ANOVA , or two-way ANOVA was used for statistical analysis . Adult axolotls of 15–18 cm in length were induced to undergo metamorphosis as described previously ( Page and Voss , 2009 ) . The animals were treated with L-thyroxine ( 50 nM , Sigma-Aldrich ) for 6 months while kept in 50% water , 50% land habitat . Upon complete gill absorption and exhibition of land walking capabilities , they were transferred to a peat moss habitat and switched to a cricket diet . No surgery was performed before animals were able to feed independently . For retrograde tracing , Retrobeads ( Lumafluor , Durham , NC ) were injected ( 69 nl/injection , 3 injections per site ) into regions of interest in the adult axolotl brain in vivo . One week after injection , animals were sacrificed and the brains were processed as described above . The dorsal pallium and individual neurons were visualized and identified with a microscope equipped with Nomarski optics and infrared illumination ( BX-51WI , Olympus , Shinjuku , Japan ) . Whole-cell patch clamp recordings were obtained from axolotl pallium of injured and uninjured animals using recording pipettes ( Glass type 8250 , King Precision Glass , Claremont , CA ) pulled in a horizontal pipette puller ( P-87 , Sutter Instruments , Novato , CA ) to a resistance of 3–4 MΩ , and filled with internal solution containing ( in mM ) : 117 . 0 K-gluconate , 13 . 0 KCl , 1 . 0 , MgCl2 , 0 . 07 CaCl2 , 0 . 1 EGTA , 10 . 0 HEPES , 2 . 0 Na-ATP , 0 . 4 Na-GTP , and 0 . 3% biocytin , pH adjusted to 7 . 3 with KOH and osmolarity adjusted to 298–300 mOsm with 15 mM K2SO4 . For data acquisition and analysis , a Multiclamp 700B amplifier ( Molecular Devices Corporation , Sunnyvale , CA ) and digidata 1440A were used to acquire whole cell signals . The signals were acquired at 20 KHz and filter at 2 KHz . Conventional characterization of neurons was made in voltage and current clamp configurations . Access resistances were continuously monitored and experiments with changes over 20% were aborted . Analyses were performed using Origin ( version 8 . 6 , Microcal , Malvern , UK ) and MiniAnalysis ( Synaptosoft , Decatur , GA ) . Ex vivo 300 μm thick slices were prepared by sectioning uninjured and injured ( 11–15wpi ) brains on the vibratome . The extracellular field recording was performed as described ( Peça et al . , 2011 ) . Briefly , a slice was placed into recording chamber ( Warner Instruments , Hamden , CT ) and constantly perfused with oxygenated aCSF at room temperature at a speed of 2 . 0 ml/min . A platinum iridium concentric bipolar electrode ( FHC ) was positioned on the dorsal pallium to stimulate the neurons either in uninjured control or injured slice . A borosilicate glass recording electrodes filled with 2M NaCl was placed onto the lateral or medial pallium approximately 600 μm away from the stimulating electrode . Dorsal-lateral or dorsal-medial field population spikes were elicited by delivery step depolarization ( 0 . 15 ms duration with 0 . 1 mA intensity at a frequency of 0 . 1 Hz ) . Stable baseline response of pop spike for at least 5 min from individual slice was ensured before moving to input-output assay . Input-output curves were determined for both the negative peak 1 ( NP1; fiber volley ) and pop spike amplitude by delivery of three consecutive stimulations from 0 to 1 mA with 0 . 1 mA increments . Recordings were performed at room temperature and data were sampled using pCLAMP 10 software ( Molecular Devices ) . Ex vivo 300 μm thick slices were prepared by sectioning uninjured and injured ( 11–15wpi ) brains . Staining of the slices in Fluo-4 ( Thermo Fisher Scientific ) was performed in accordance to protocol ( Dawitz et al . , 2011 ) . Slices were imaged on Zeiss Cell Observer for 10 min with a frame rate of 3 Hz . The resulting movie was analyzed using the FluoroSNAPP software ( Patel et al . , 2015 ) . Animals for imaging were anesthetized by immersion in 0 . 1% tricaine ( Western Chemical , Ferndale , WA ) . Once anesthesia was achieved , animals were transferred to an MRI-compatible bed where anesthesia was maintained by partial immersion in tricaine with exposed skin covered by a damp towel . Animals were imaged using a 9 . 4T horizontal bore animal imaging system ( Biospec 94/20 , Bruker Corporation , Billerica , MA ) . A circular , 20 mm receive-only surface coil was placed over the head . The animals were then situated with their heads near the isocenter of the magnet within an 84 mm quadrature transmit/receive volume coil . T2-weighted images were acquired using a RARE ( rapid acquisition with relaxation enhancement ) sequence ( Hennig et al . , 1986 ) . All images had RARE factor 8 and echo time TE=33 ms , and a sufficient number of slices were acquired to cover the entire brain . Sagittal images were acquired with repetition time TR=2181 ms , 1 mm slice thickness , 158 µm × 188 µm in-plane resolution , and 1 signal average . Axial images were acquired with TR=2000 ms , 500 µm slice thickness , 150 µm × 150 µm in-plane resolution , and 16 signal averages . Coronal images were acquired with TR=3817 ms , 500 µm slice thickness , 158 µm × 188 µm in-plane resolution , and 4 signal averages . 3D surface rendering of the MRI images were created using OsiriX ( Rosset et al . , 2004 ) . Regions of interest ( ROI ) were manually drawn on consecutive coronal T2-weighted images to outline the contours of the brain . The pixel values within and outside of the ROIs were changed to 65 , 000 and -3024 , respectively . 3D surface rendering was performed with these settings: highest resolution , 100 iterations , and 1000 pixel values . The sizes of the injury were determined by manually drawing ROIs for regions that are lacking continuous closure of the ventricle . The volumes of the ROIs were calculated using the Compute Volume algorithm . The brains of anesthetized axolotls were extracted and fixed in 4% PFA containing 1 mM gadolinium ( Gd-DTPA ) MRI contrast agent to reduce the T1 relaxation time while ensuring that enough T2-weighted signal remained . For MR image acquisition , the brains were placed in formalin . Brains were scanned on a 9 . 4T Bruker Biospec MR system . The pulse sequence used for image acquisition was a 3D diffusion-weighted spin-echo echo-planar imaging sequence , TR=330 msec , TE=31 . 23 msec , number of segments 4 , with an imaging matrix of 232 × 96 × 96 pixels . Spatial resolution was 50 × 50 × 50 µm . Sixty diffusion-weighted measurements and one non-diffusion weighted ( b=0 ) measurement were acquired , corresponding to a cubic lattice in Q-space at b = 12 , 000 sec/mm2 with δ = 12 . 0 msec , Δ = 24 . 2 msec , with 8 averages . The total acquisition time was 17 hr and 11 min for each imaging session . Diffusion Toolkit and TrackVis ( http://trackvis . org ) were used to reconstruct and visualize tractography pathways . Trajectories were propagated by consistently pursuing the orientation vector of least curvature . The color-coding of tractography pathways is based on a standard RGB code , applied to the vector between the end-points of each fiber ( red: left-right , green: dorsal-ventral , and blue: anterior-posterior directions ) . We terminated tracking when the angle between two consecutive orientation vectors was greater than the given threshold ( 60° ) for each specimen . ROIs were placed at the injury site and anatomically matched region in the contralateral hemisphere . Representative 3D fiber tracts that pass through the ROIs and a single coronal plane are shown . | Humans and other mammals have a very limited ability to regenerate new neurons in the brain to replace those that have been injured or damaged . In striking contrast , some animals like fish and salamanders are capable of filling in injured brain regions with new neurons . This is a complex task , as the brain is composed of many different types of neurons that are connected to each other in a highly organized manner across both short and long distances . The extent to which even the most regenerative species can build new brain regions was not known . Understanding any limitations will help to set realistic expectations for the success of potential treatments that aim to replace neurons in mammals . Amamoto et al . found that the brain of the axolotl , a species of salamander , could selectively regenerate the specific types of neurons that were damaged . This finding suggests that the brain is able to somehow sense which types of neurons are injured . The new neurons were able to mature into functional neurons , but they were limited in their ability to reconnect to their original , distant target neurons . More research is now needed to investigate how the axolotl brain recognizes which types of neurons have been damaged . It will also be important to understand which cells respond to the injury to give rise to the new neurons that fill the injury site , and to uncover the molecules that are important for governing this regenerative process . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"neuroscience"
] | 2016 | Adult axolotls can regenerate original neuronal diversity in response to brain injury |
Many animals use visual signals to estimate motion . Canonical models suppose that animals estimate motion by cross-correlating pairs of spatiotemporally separated visual signals , but recent experiments indicate that humans and flies perceive motion from higher-order correlations that signify motion in natural environments . Here we show how biologically plausible processing motifs in neural circuits could be tuned to extract this information . We emphasize how known aspects of Drosophila's visual circuitry could embody this tuning and predict fly behavior . We find that segregating motion signals into ON/OFF channels can enhance estimation accuracy by accounting for natural light/dark asymmetries . Furthermore , a diversity of inputs to motion detecting neurons can provide access to more complex higher-order correlations . Collectively , these results illustrate how non-canonical computations improve motion estimation with naturalistic inputs . This argues that the complexity of the fly's motion computations , implemented in its elaborate circuits , represents a valuable feature of its visual motion estimator .
A major goal in neuroscience is to understand how the brain computes behaviorally relevant stimulus properties from streams of incoming sensory data ( Sejnowski et al . , 1988 ) . Visual motion guides behaviors across the animal kingdom . To navigate , many vertebrates and invertebrates use visual data to estimate the velocity of full field motion , and they use that estimate to judge their motion with respect to their environment ( Sperry , 1950; Kalmus , 1964; Reichardt and Poggio , 1976; Orger et al . , 2000 ) . Spatially localized motion perception ( Hubel and Wiesel , 1962; Barlow and Hill , 1963; Buchner , 1976 ) is also important , as it can indicate the presence of predators or prey in the environment ( Reichardt et al . , 1983; Gabbiani et al . , 1999; Nordström et al . , 2006; Zhang et al . , 2012 ) , and spatial velocity gradients allow animals to judge relative distances ( Rogers and Graham , 1979; Srinivasan et al . , 1991; Kral , 2003; Pick and Strauss , 2005 ) . In principle , different algorithms could be used to estimate different types of motion . However , data suggest that many animals compute local motion over an array of spatially localized elementary motion detectors , or EMDs , and then differentially pool those signals for use in different behaviors and neural operations ( Hubel and Wiesel , 1962; Barlow and Hill , 1963; Buchner , 1976; Britten et al . , 1992; Gabbiani et al . , 1999; Franz and Krapp , 2000; Rust et al . , 2006 ) . The Hassenstein-Reichardt correlator ( HRC ) was introduced nearly sixty years ago to model the EMD underlying the beetle's optomotor response ( Hassenstein and Reichardt , 1956 ) . It has since provided numerous insights into motion-guided behaviors across a variety of insect species . The HRC's successes are most striking in flies , where the HRC accurately predicts a wide variety of behavioral and neural responses ( Götz , 1968; Buchner , 1976; Egelhaaf and Borst , 1989; Haag et al . , 2004 ) , and even adaptation to stimulus statistics ( Borst et al . , 2005 ) . The HRC's importance also extends to primates and vertebrates , where the EMDs are often described in terms of the motion energy model ( Adelson and Bergen , 1985 ) . In particular , although the HRC and motion energy models differ in terms of their intuition and neuronal bases , both models rely on the same mathematical fact about moving visual stimuli—motion causes pairs of spatially separated points to become correlated when a stimulus moves from one location towards the other ( Adelson and Bergen , 1985; van Santen and Sperling , 1985 ) . A simple algebraic identity shows that the HRC and motion energy models are computationally equivalent . Research on Drosophila's motion detection system has progressed quickly in recent years . With an influx of novel genetic , anatomical , and physiological tools , Drosophila researchers are able to perform experiments that have revealed an intricate neural circuit whose details were not anticipated . For example , separate pathways process the motion of light and dark moving edges ( Joesch et al . , 2010; Clark et al . , 2011; Maisak et al . , 2013 ) , and different neurons within these pathways coordinate the motion response depending on the velocity of motion ( Ammer et al . , 2015 ) . Furthermore , connectomic analysis has revealed that more spatial and temporal channels converge onto the fly's motion computing neurons than had been predicted by the HRC's two-input architecture ( Takemura et al . , 2013 ) . Going forward , it is critical that the field discovers which of these circuit details are computationally relevant and which are not . Since many of these details go beyond the HRC's premise , we must consider alternate theories if we hope to understand how circuit details contribute to motion estimation . A large body of theoretical and experimental work supports the hypothesis that visual systems are tailored for functionality in the animal's natural behavioral context ( Simoncelli and Olshausen , 2001 ) . For example , photoreceptors adapt effectively across the ecological range of light levels ( Juusola and Hardie , 2001 ) , the excess number of OFF vs ON retinal ganglion cells matches the excess information of dark vs light contrasts in natural images ( Ratliff et al . , 2010 ) , and several learning algorithms predict receptive fields similar to early cortical neurons when applied to natural images ( Olshausen and Field , 1996; Bell and Sejnowski , 1997 ) . These examples are special cases of the general hypothesis that the early visual system provides an efficient code for the natural visual environment , and recent research suggests that efficient coding accounts for certain aspects of higher-level coding and perception as well ( Tkačik et al . , 2010; Hermundstad et al . , 2014; Yu et al . , 2015 ) . Several recent studies have established connections between the biological algorithms used for visual motion estimation and the statistical demands of naturalistic motion estimation . Natural stimuli are intricately structured and light–dark asymmetric ( Geisler , 2008 ) , and a variety of low and high order correlations characterize motion in such an environment . Although the HRC and motion energy models only respond to pairwise correlations in their inputs ( Adelson and Bergen , 1985; van Santen and Sperling , 1985 ) , the Bayes optimal visual motion estimator also incorporates a variety of higher-order correlations of both even and odd order ( Potters and Bialek , 1994; Fitzgerald et al . , 2011 ) . Accordingly , certain visual stimuli that contain only higher-order correlations induce motion percepts in both vertebrates and insects ( Chubb and Sperling , 1988; Quenzer and Zanker , 1991; Zanker , 1993; Orger et al . , 2000; Hu and Victor , 2010; Clark et al . , 2014 ) , and theoretical work shows that the correlations that characterize these stimuli can also improve motion estimation in natural environments ( Clark et al . , 2014 ) . This demonstrates that neither the HRC nor the motion energy model can account for the totality of experimentally observed motion percepts and suggests that departures from these canonical models might improve motion estimation accuracy . Relatively little is known about the neural basis of these higher-order motion percepts , although several studies have suggested intriguing commonalities across insect and primate species ( Clark et al . , 2014; Nitzany et al . , 2014 ) . Here we investigate whether the computational demands imposed by accurate motion estimation in natural environments can illuminate the unexpected details of Drosophila's motion estimation circuit or account for non-Reichardtian motion perception in flies . We study a sequence of five computational models , each of which considers a conceptually new aspect of the motion estimation problem . Since each model succeeds in improving estimation accuracy , these results provide a range of nonlinear circuit mechanisms that flies and other animals might incorporate into their motion estimators . We describe how observed elements of Drosophila's motion estimation circuitry could support such computations ( Table 1 ) . Importantly , four of the five models also predict the signs and approximate magnitudes of known non-Reichardtian motion percepts in flies . Since the models were tuned exclusively for estimation accuracy , these results support the view that non-Reichardtian motion percepts probe ethologically relevant aspects of biological motion estimators . More generally , our results posit normative interpretations for some unexpected aspects of the fly's motion estimation circuit and behavior and suggest that non-Reichardtian aspects of fly circuitry and behavior might be closely linked through the statistics of natural scenes . 10 . 7554/eLife . 09123 . 003Table 1 . The different models used in this paper , experimental results that support each model , and references for those resultsDOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 003ModelSupporting evidenceFront-end nonlinearity• Photoreceptors show nonlinear responses to contrast changes ( Laughlin , 1989; Juusola and Hardie , 2001; Juusola and Hardie , 2001; van Hateren and Snippe , 2006 ) • Some neurons in the early visual system have nonlinear responses that make their output signals nearly uniform ( Laughlin , 1981 ) Weighted 4-quadrant model• Visual processing is divided early into ON and OFF channels ( Joesch et al . , 2010; Clark et al . , 2011; Behnia et al . , 2014; Meier et al . , 2014; Strother et al . , 2014 ) • The two output channels ( T4/T5 ) are sensitive to light and dark edges ( Maisak et al . , 2013 ) , but their inputs are incompletely rectified ( Behnia et al . , 2014 ) • Stimuli targeting the four quadrants are differentially represented in neural substrates ( Clark et al . , 2011; Joesch et al . , 2013 ) Non-multiplicative nonlinearity• Pure multiplication is not a trivial neural operation ( Koch , 2004 ) • Inputs to T4/T5 are nonlinearly transformed ( Behnia et al . , 2014 ) , which also contributes to the biologically implemented non-multiplicative nonlinearityUnrestricted nonlinearity• The direction-selective neurons T4 receive inputs from more than two types of neurons ( Takemura et al . , 2013; Ammer et al . , 2015 ) • T4 receives inputs from both its major input channels at overlapping points in space ( Takemura et al . , 2013 ) Extra input nonlinearity• The direction-selective neuron T4 receives inputs from more than two discrete retinotopic locations ( Takemura et al . , 2013 )
The HRC is the dominant model of motion computation in flies and other insects . In this paper we describe several generalizations of the HRC , but it is helpful to first review this canonical model . The HRC comprises three stages of processing . First , two different temporal filters ( here , a low-pass filter ( f ( t ) ) and a high-pass filter ( g ( t ) ) ) are applied to each of two spatially filtered visual input streams ( Figure 1A , ‘Materials and methods’ ) . These four filtered signals are then paired and multiplied ( Figure 1A ) . Finally , the HRC takes the difference between the two multiplied signals to obtain a mirror anti-symmetric motion estimator ( Figure 1A ) . Because the HRC combines its two input channels via a multiplication operation , the average output of the HRC depends only on 2-point correlations in the visual stimulus . We thus refer to the HRC as a 2-point correlator , and we will return to this mathematical characterization of the HRC repeatedly throughout this work . 10 . 7554/eLife . 09123 . 004Figure 1 . The Hassenstein-Reichardt correlator ( HRC ) model is an incomplete description of Drosophila's motion estimator . ( A ) Diagram of the HRC model . ( B ) We assessed motion estimation performance across an ensemble of naturalistic motions , each of which consisted of a natural image ( van Hateren and van der Schaaf , 1998 ) and a velocity chosen from a normal distribution . ( C ) We quantified model accuracy by comparing the model response to the true velocity using the mean squared error . ( D ) We summarized the error with the correlation coefficient between the model output and the true velocity . ( E ) In previous work ( Clark et al . , 2014 ) , we used a panoramic display and spherical treadmill to measure the rotational responses of Drosophila to visual stimuli . ( F ) We presented flies with binary stimuli called gliders ( Hu and Victor , 2010 ) , which imposed specific 2-point and 3-point correlations ( Clark et al . , 2014 ) . ( G ) Flies turned in response to 3-point glider stimuli , but these responses cannot be predicted by the standard HRC . ( H ) Diagram of the converging 3-point correlator , which is designed to detect higher-order motion signals like those found in 3-point glider stimuli . ( I ) Adding the converging 3-point correlator to the HRC improved motion estimation performance with naturalistic inputs . We optimized weighting coefficients to minimize the mean squared error over the ensemble of naturalistic motions and used cross-validation to protect against over-fitting . ( J ) This model predicted that Drosophila would weakly turn in response to 3-point glider stimuli . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 004 No motion estimator is perfect for every stimulus , and this paper explores the hypothesis that evolution has tuned Drosophila's motion estimator for visual experiences that are likely to result from ordinary behavior in natural environments ( Appendix 1 ) . We assessed the accuracy of the HRC and other motion estimators by approximating naturalistic motion as the rigid translation of natural images ( Clark et al . , 2014 ) , with a velocity distribution that mimicked Drosophila's natural behavior ( Figure 1B , ‘Materials and methods’ ) ( Katsov and Clandinin , 2008 ) . We spatiotemporally filtered the input signals to simulate the responses of two neighboring photoreceptors ( ‘Materials and methods’ ) . We quantified the performance of each model as the mean squared error between the input velocity and model output . However , we report each model's accuracy as the correlation coefficient between its output and the true velocity ( Figure 1C ) , an intuitive metric that is equivalent to the mean squared error for correctly scaled model outputs ( ‘Materials and methods’ ) . In isolation , the local HRC was weakly correlated with the velocity of motion ( Figure 1D ) . Although the HRC's performance can be improved by averaging over space and time ( Dror et al . , 2001; Clark et al . , 2014 ) , this study explores how alternate nonlinear processing can improve motion estimation accuracy without sacrificing spatial or temporal resolution ( Clark et al . , 2014 ) . Researchers can probe a fly's motion estimate by measuring its behavioral optomotor turning response ( Hassenstein and Reichardt , 1956; Götz and Wenking , 1973; Buchner , 1976; Reichardt and Poggio , 1976 ) . We previously measured optomotor responses from flies walking on a spherical treadmill by recording their turning responses to various visual stimuli ( Figure 1E ) ( Clark et al . , 2014 ) . We emphasized binary stimuli called gliders ( Hu and Victor , 2010 ) ( Figure 1F ) , which enforce spatiotemporal correlations to interrogate the fly's motion estimation algorithm . For example , 2-point gliders contain only 2-point correlations ( first two stimuli , Figure 1F ) . Drosophila turned in response to these stimuli ( Clark et al . , 2014 ) ( black bars , left , Figure 1G ) , and the HRC correctly predicted that flies would respond to both positive and negative 2-point correlations ( gray bars , left , Figure 1G ) . On the other hand , 3-point gliders contain 3-point correlations without 2-point correlations ( last four stimuli , Figure 1F ) . These stimuli generated motion responses in flies ( black bars , right , Figure 1G ) that the HRC could not explain ( gray bars , right , Figure 1G ) . Thus , behavioral responses to glider stimuli show that the HRC is an incomplete description of fly motion estimation and provide a useful benchmark for evaluating alternate models . In this study , we tune our models to optimize motion estimation accuracy , rather than to fit the behavioral data , for two main reasons . First , we want to explore the hypothesis that Drosophila's glider responses follow from performance optimization within biologically plausible circuit architectures . Second , we seek models that will generalize well across visual stimuli , and the measured glider responses under-constrain possible motion estimation models . It's useful to illustrate our procedure with a simple example . The HRC does not account for 3-point glider responses because it is insensitive to 3-point correlations . Nevertheless , 3-point correlations are present in natural stimuli ( Clark et al . , 2014; Nitzany and Victor , 2014 ) , and their use might facilitate accurate motion estimation . We can explore this hypothesis by summing the HRC with a motion estimator designed to respond specifically to 3-point correlations . For instance , the mirror anti-symmetric ‘converging’ 3-point correlator multiplies one high-pass filtered signal with two low-pass filtered signals ( Figure 1H ) and mimics the converging structure present in certain glider stimuli ( last two stimuli , Figure 1F ) . We tune the model for motion estimation accuracy by choosing the weights of the HRC and the converging 3-point correlator to minimize the mean squared error ( ‘Materials and methods’ ) . The resulting model is more accurate than the HRC ( Figure 1I ) and it predicts that flies should respond to glider stimuli in the observed directions ( Figure 1J , ‘Materials and methods’ ) . Nevertheless , this simple model underestimates 3-point turning magnitudes ( Figure 1J ) , indicating a discrepancy between the fly's motion estimator and this performance-optimized model . In this study , we apply this same basic model building procedure to a series of increasingly general model architectures . There are four benefits to this approach . First , each model incorporates a type of computation that was neglected by earlier models . Thus , we can compare model accuracies to quantify how important various computations are for naturalistic motion estimation . Second , each model has a distinct biological interpretation in terms of Drosophila's motion estimation circuit ( Table 1 ) . This allows us to enumerate many directions for future experimental and computational research . Third , this set of models reveals several distinct principles of accurate naturalistic motion estimators , yet no single model illustrates every principle . Finally , by comparing the glider predictions of each model to behavioral data , we can gain insight into which principles underlie Drosophila's known glider responses . The HRC correlates pairs of photoreceptor signals ( Figure 1A ) . We previously assumed that each photoreceptor's response was generated from incoming contrast signals through linear spatiotemporal filtering . However , real photoreceptors are linear only over a limited range of inputs ( Laughlin , 1981; Juusola and Hardie , 2001 ) ( Table 1 ) . Our first model thus modifies the HRC by allowing the photoreceptor responses to become nonlinear ( Figure 2A ) . More specifically , we consider models in which a static nonlinearity transforms the filtered contrast signals before a standard HRC is applied to the two input streams ( Figure 2A , ‘Materials and methods’ ) . Since the nonlinearity occurs before the HRC , we refer to this model as the front-end nonlinearity model . By nonlinearly transforming the contrast signals , the front-end nonlinearity model is able to reshape natural sensory statistics . In particular , linear photoreceptor signals inherit complex non-Gaussian statistics from their natural inputs ( Figure 2B ) , but front-end nonlinearities ( Figure 2C ) can produce transformed signals with alternate statistics ( Figure 2D , ‘Materials and methods’ ) . Thus , optimal front-end nonlinearity models should reshape natural statistics into those that best suit the HRC . Previous studies have already demonstrated example front-end nonlinearity models that improve naturalistic motion processing by the HRC ( Dror et al . , 2001; Brinkworth and O'Carroll , 2009 ) . Here we provide new theoretical insight into these improvements and their consequences for glider responses . 10 . 7554/eLife . 09123 . 005Figure 2 . Front-end nonlinearities improved naturalistic motion estimation but did not reproduce the psychophysical results . ( A ) Diagram of the front-end nonlinearity model . The nonlinearity occurs after the spatiotemporal filtering of photoreceptors but before the temporal filtering of the HRC . ( B ) The distribution of contrast signals after photoreceptor filtering had a kurtosis of 9 . 6 . The kurtosis of unfiltered pixels in the image database was 7 . 8 . ( C ) Three different nonlinearities that transformed this input distribution into a Gaussian distribution , a uniform distribution , and a binary distribution . ( D ) After these transformations , the kurtosis of the contrast signal was reduced to 3 , 1 . 8 , and 1 , respectively . ( E ) Each front-end nonlinearity model improved the HRC's estimation accuracy , and uniform output signals worked best . ( F , G ) The front-end nonlinearity models reproduced the sign of the negative 2-point glider psychophysical responses but did not reproduce the pattern of psychophysical responses to 3-point gliders . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 00510 . 7554/eLife . 09123 . 006Figure 2—figure supplement 1 . Front-end nonlinearities modify the correlations present in natural scenes . ( A ) Example images with no front-end nonlinearity ( top ) , with an equalizing front-end nonlinearity ( middle ) , and with a binarizing front-end nonlinearity ( bottom ) . ( B ) The covariance between contrasts at 2 horizontally separated points is plotted as a function of distance between the points . The binary nonlinearity attenuated spatial correlations . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 006 Although the statistics of natural images are complicated , the mean squared error between the HRC's output and the velocity of motion depends only on a few statistical quantities . Since the HRC is a 2-point correlator , the mean velocity signal decoded by an HRC is determined by the second-order statistics of the image ensemble ( Dror et al . , 2001 ) . The variance of the motion signal comes from the square of a quadratic signal , and thus the noise statistics of the HRC depend on the fourth-order statistics of the image ensemble ( Appendix 2 ) . If the image ensemble is spatially uncorrelated , the situation simplifies further and the correlation between the estimated and true image velocity is determined entirely by the standardized fourth central moment of the input streams , a quantity known as kurtosis ( Appendix 3 ) . A larger kurtosis results in a larger error in the motion estimate . Note that some authors use ‘kurtosis’ to refer to the ‘excess kurtosis’ , which shifts kurtosis values such that the Gaussian distribution has zero excess kurtosis . This shift is not relevant for our purposes . Because large positive contrasts are relatively probable , naturalistic inputs are highly kurtotic ( kurtosis = 9 . 6 for the spatiotemporal filtering in our simulations ) and are thus expected to hinder HRC performance ( Figure 2B ) . The Gaussian , uniform , and symmetric Bernoulli distributions have much lower kurtosis values ( kurtosis = 3 . 0 , 1 . 8 , 1 . 0 , respectively , Figure 2D ) . In fact , the symmetric Bernoulli distribution has the lowest kurtosis of any probability distribution ( DeCarlo , 1997 ) . When we transformed the HRC's inputs to have these statistics ( ‘Materials and methods’ ) , we found that each nonlinearity substantially improved the accuracy of the HRC ( Figure 2E ) . The contrast equalizing nonlinearity , which produces uniform outputs , performed best and also plays a prominent role in efficient coding theory ( Laughlin , 1981 ) . It is interesting that contrast equalization improved the accuracy of the HRC more than binarization ( Figure 2E ) , even though it produced outputs with greater kurtosis . The reason for this is that natural images are spatially correlated , and the accuracy of the HRC over a general image ensemble depends on the ensemble's spatial correlation structure ( Appendix 2 ) . Binarization attenuated spatial correlations more strongly than contrast equalization over the natural image ensemble ( Figure 2—figure supplement 1 ) , and spatial correlations can enhance the performance of the HRC ( Appendices 4 , 5 ) . Designing a nonlinearity that optimally sculpts the correlation structure of natural images is not simple and goes beyond the scope of this study . Each front-end nonlinearity model is sensitive to a variety of higher-order correlations ( Appendix 6 ) . We thus tested whether accurate front-end nonlinearity models would predict Drosophila's glider response pattern . However , each front-end nonlinearity model performed poorly at this task ( Figure 2F , G ) . None of the three models predicted that Drosophila would invert its response to positive and negative 3-point gliders ( Figure 2G ) , even though they predicted that the 3-point glider responses would be nonzero . The simplest explanation for this observation is that the front-end nonlinearity models responded to fourth-order correlations that are common to the stimuli , rather than the third-order correlations that defined the glider stimuli and primarily drove the experimental response ( Clark et al . , 2014 ) . Mechanistically , this result follows from the fact that the nonlinearities that reduced kurtosis ( Figure 2C ) were not strongly asymmetric around zero contrast ( Appendix 6 ) . The binarizing front-end nonlinearity model also failed to predict that Drosophila would respond less to negative 2-point glider stimuli than positive 2-point glider stimuli ( Figure 2F ) . Since this effect was correctly predicted by the standard HRC ( Figure 1G ) , this observation shows that accurate front-end nonlinearity models can distort the processing of 2-point correlations . Although the front-end nonlinearity model did not explain the phenomenon of fly glider perception , future work should investigate whether its merits make it functionally relevant for motion processing in other contexts or species . Instead of a front-end nonlinearity , Drosophila could use an alternative non-Reichardtian motion estimation strategy that reflects natural sensory statistics , without necessarily requiring nonlinear preprocessing . Previous computational analyses show that motion estimation strategies that distinguish light and dark information can enhance motion processing with natural inputs ( Fitzgerald et al . , 2011; Clark et al . , 2014; Nitzany and Victor , 2014 ) , and recent experiments indicate that flies use separate channels to process the motion of light and dark edges ( Joesch et al . , 2010; Clark et al . , 2011; Behnia et al . , 2014; Clark et al . , 2014; Meier et al . , 2014; Strother et al . , 2014 ) ( Table 1 ) . Our next model explores the hypothesis that Drosophila segregates ON and OFF signals in order to facilitate naturalistic motion estimation ( Clark et al . , 2014 ) ( Figure 3A , ‘Materials and methods’ ) . There are four ways to pair the ON and OFF components of the two filtered signals that enter the HRC's multiplier . For example , one possibility is to pair the ON component of the low-pass filtered signals with the OFF component of the high-pass filtered signal . Since each pairing restricts the HRC's multiplier to a single quadrant of the Cartesian plane , we refer to these four signals as HRC-quadrants . If the quadrants are summed with equal weights , then this model is mathematically identical to the HRC ( Hassenstein and Reichardt , 1956; Clark et al . , 2011 ) . Unequal weighting coefficients enable the motion estimator to prioritize some quadrants over others , and here we select quadrant weightings that minimize the mean squared error between the model output and velocity ( Figure 3B , ‘Materials and methods’ ) . More generally , we refer to any model that linearly combines the four HRC-quadrants as a weighted 4-quadrant model . The precise manner in which the four HRC-quadrants might map onto circuitry remains unclear; we do not suggest there exists separate circuitry for each quadrant . For instance , studies have identified only two motion-processing channels in the Drosophila brain , which might suggest that the fly only uses a subset of the quadrants ( Eichner et al . , 2011; Joesch et al . , 2013; Maisak et al . , 2013 ) . On the other hand , each channel appears imperfectly selective for light vs dark signals ( Behnia et al . , 2014 ) , which in principle enables these two channels to access all four quadrants ( Table 1 ) . 10 . 7554/eLife . 09123 . 007Figure 3 . The weighted 4-quadrant model improved estimation performance and reproduced the directionality of psychophysical results . ( A ) Diagram of the weighted 4-quadrant model . Similar to ON/OFF processing in the visual system , the weighted 4-quadrant model splits the four differentially filtered signals into positive and negative components . As in the HRC , these component signals are paired , multiplied , and subtracted to produce four mirror anti-symmetric signals . We refer to these signals as HRC-quadrants . The model output is a weighted sum of the quadrant signals . We identify quadrants by whether they respond to the positive or negative components of each filtered signal and denote the four quadrants as ( + + ) , ( + − ) , ( − + ) , and ( − − ) . In this notation , the first index refers to the sign of the low-pass filtered signal ( emanating from f ( t ) ) , and the second refers to the high-pass filtered signal ( emanating from g ( t ) ) . ( B ) We measured the response of each quadrant to naturalistic motions and chose the quadrant weightings to minimize the mean squared error between the model output and the true velocity . ( C ) Comparison of the estimation performance of individual quadrants , multiple quadrants , and the HRC . The best two quadrants were ( − − ) and ( − + ) ; the best three also included ( + − ) . ( D ) The performance-optimized weighted 4-quadrant model reproduced the signs and approximate magnitudes of the psychophysical results . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 00710 . 7554/eLife . 09123 . 008Figure 3—figure supplement 1 . Separate ON and OFF processing improved motion estimation by supplementing the HRC with odd-ordered correlations . ( A ) By summing and subtracting the four quadrants ( top labels , e . g . , ‘++’ ) in four different patterns , we isolated the contributions of various correlation types ( side labels , e . g . , ‘odd’ ) to the weighted 4-quadrant model ( Appendix 7 ) . For example , the uniform sum of the four quadrants is the HRC , and we denote this quadrant combination as ‘even = 2’ ( top row of matrix ) . The other three rows of the matrix define quadrant combinations that are sensitive to two different classes of third and higher odd-ordered correlations ( ‘odd’ and ‘odd*’ rows ) and to fourth and higher even-ordered correlations ( ‘even >2’ row ) . The factor of 1/4 merely sets the magnitude of the quadrant contributions to match the formulas in Appendix 7 and is without conceptual importance . ( B ) The ‘even = 2’ correlation class worked best in isolation . Nevertheless , the ‘even = 2’ and ‘odd’ classes were highly synergistic ( their weighted sum is notated ‘best 2’ ) , and these classes together made the ‘odd*’ and the ‘even >2’ classes irrelevant . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 00810 . 7554/eLife . 09123 . 009Figure 3—figure supplement 2 . The weighted 4-quadrant model cannot reproduce the positive-negative parity asymmetry in the psychophysical data . In this numerical experiment , we tuned the coefficients of the weighted 4-quadrant model to optimize a fit to the psychophysical data . ( A ) The tuned model could reproduce the 2-point glider data well . ( B ) Although the tuned weighted 4-quadrant model could reproduce the signs of the 3-point glider data , it could not reproduce the differential amplitudes of the positive and negative parity responses . This demonstrates that the architecture of the weighted 4-quadrant model is too limited to reproduce the experimental response pattern . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 009 We began by examining how well individual quadrants predicted the velocity of motion . The four quadrants provided motion signals of strikingly different quality ( first four red bars , Figure 3C ) . The most accurate quadrant correlated negative low-pass filtered signals with negative high-pass filtered signals ( ( − − ) bar , Figure 3C ) . This isolated quadrant already outperformed the full HRC . The quadrant that correlated negative low-pass filtered signals with positive high-pass filtered signals also performed relatively well ( ( − + ) bar , Figure 3C ) , whereas the quadrants that involved positive low-pass filtered signals performed poorly ( ( + + ) and ( + − ) bars , Figure 3C ) . This shows that negative signals emanating from the low-pass filter better facilitate motion estimation , and the HRC's uniform weighting of all four quadrants is computationally detrimental . We next considered all subsets of two , three , or four quadrants . The best subsets for each number of predictors were nested , and the quadrants were incorporated in the order ( i ) ( − − ) ; ( ii ) ( − + ) ; ( iii ) ( + + ) ; ( iv ) ( + − ) . Although all four quadrants enhanced the accuracy of the weighted 4-quadrant model , the benefit of each added quadrant decreased with the number of quadrants ( Figure 3C ) . It is possible to reparameterize the weighted 4-quadrant model in a form that isolates the contributions of various higher-order correlations to the model's accuracy ( Appendix 7 ) . Interestingly , this parameterization showed that nearly all the accuracy of the weighted 4-quadrant model can be obtained by supplementing the HRC with a set of odd-ordered correlations that account for the asymmetry between positive and negative low-pass filtered signals ( Figure 3—figure supplement 1 , Appendix 8 ) . Principal component analysis ( PCA ) did not reveal this simple interpretation of the model's computation ( Appendix 9 ) . The performance-optimized weighted 4-quadrant model also offered an interesting interpretation of Drosophila's glider response pattern . First note that the model preserved the HRC's response pattern to 2-point glider stimuli ( compare left subpanels of Figure 3D and Figure 1G ) . More interestingly , the model predicted behavioral responses to 3-point glider stimuli that matched the experimentally observed turning directions , and even the response magnitudes were similar between the model and the data ( right , Figure 3D ) . Nevertheless , the model's predictions were imperfect . The primary qualitative discrepancy was that the model failed to predict that positive 3-point glider stimuli would generate smaller turning responses than negative 3-point glider stimuli . The simplest interpretation for this experimental result is that flies might incorporate both 3-point correlations and 4-point correlations into their motion estimation strategy . In particular , since the positive and negative 3-point glider stimuli have inverted 3-point correlations and matched 4-point correlations , third-order and fourth-order correlations would have the same sign for one parity and opposite signs for the other parity . This observation makes it easier to understand the glider predictions of the weighted 4-quadrant model . The optimized model does a good job accounting for the direction and approximate magnitude of the glider responses because it draws heavily on second-order and odd-order correlations , but it fails to predict the 3-point glider magnitude asymmetry because it finds little added utility in higher-order even correlations ( Figure 3—figure supplement 1 , Appendix 8 ) . This failure stems from architectural limitations in the weighted 4-quadrant model ( Figure 3—figure supplement 2 ) , so it is important to consider alternate model classes . The previous section suggested that the segregation of light and dark signals by Drosophila's motion estimation circuitry might enhance naturalistic motion estimation in a manner that also generates the observed glider responses . In this section , we introduce three hierarchical models to investigate other features of Drosophila's circuit that might have functional consequences for the processing of natural stimuli and gliders ( Table 1 ) . We refrain from modifying the temporal filtering of the motion estimator , and we focus on its nonlinear architecture . The first of these models recasts the HRC and the weighted 4-quadrant model in a more general architecture . This model is the class of mirror anti-symmetric models that apply a 2-dimensional nonlinearity to the low-pass filtered signal from one point in space and the high-pass filtered signal from a neighboring point in space ( Figure 4A ) . Since the observed glider responses indicate that flies use higher-order correlations of both even and odd order , we model this 2-dimensional nonlinearity as a fourth-order polynomial ( ‘Materials and methods’ ) . The HRC corresponds to the special case of this nonlinearity that multiplies the two inputs ( left , Figure 4B ) . To emphasize how the model class in Figure 4A generalizes the HRC , we refer to it as the non-multiplicative nonlinearity model . In comparison , the weighted 4-quadrant model corresponds to a different nonlinearity that separately scales a pure multiplication in each quadrant of the Cartesian plane . Compared to the HRC , the optimized forms of both the weighted 4-quadrant model and the non-multiplicative nonlinearity model substantially attenuated positive low-pass filtered signals ( middle and right , Figure 4B ) , though the non-multiplicative nonlinearity shows less attenuation . This model architecture provides enough flexibility to generate the glider response pattern ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 09123 . 010Figure 4 . Several biologically motivated generalizations of the motion estimator further improved estimation performance without sacrificing glider responses . See Table 1 for a description of the biological rationales behind these models . ( A ) The ‘non-multiplicative nonlinearity’ model substitutes a 2-dimensional nonlinearity for the pure multiplication of the HRC . Here , we approximated the nonlinearity with a fourth order polynomial . ( B ) Two-dimensional nonlinearities underlying the HRC , the weighted 4-quadrant model , and the non-multiplicative nonlinearity model . The latter models reflect optimized cases , in which the weighting coefficients maximized estimation performance with natural inputs . Iso-output lines are shown in each plot , and the horizontal and vertical limits are chosen to include 95% of the naturalistic input signals . ( C ) Another generalization , the ‘unrestricted nonlinearity’ model allows all 4 input signals to be combined nonlinearly . We approximate this 4-dimensional nonlinearity with a fourth-order polynomial . ( D ) A final generalization , the ‘extra input nonlinearity’ model , relaxes the restriction that the motion estimator only uses 2 spatial inputs . We approximate this 6-dimensional nonlinearity with a fourth-order polynomial . ( E ) Comparison of the estimation performance of these models to the HRC . We compare the extra input nonlinearity model to the average of two neighboring motion estimators . ( F , G ) The three models correctly predicted the directions of psychophysical responses . The pattern of 3-point responses differed somewhat across the models , and the extra input nonlinearity model was the first to predict a large asymmetry between positive and negative 3-point glider responses . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 01010 . 7554/eLife . 09123 . 011Figure 4—figure supplement 1 . The non-multiplicative nonlinearity model can be tuned to account for the positive-negative parity asymmetry in the psychophysical data . In this numerical experiment , we tuned the model nonlinearity to optimize a fit to the psychophysical data . ( A ) In this case , the tuned model could reproduce the 2-point glider data well . ( B ) This tuned model could also reproduce the differential amplitudes of the positive and negative parity responses . Thus , the non-multiplicative nonlinearity model repairs an architectural defect of the weighted 4-quadrant model ( Figure 3—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 01110 . 7554/eLife . 09123 . 012Figure 4—figure supplement 2 . The performance of the non-multiplicative nonlinearity model is plotted against the order of the fitted polynomial . With only zeroth or first-order terms , the model cannot predict motion . With second-order terms , it can perform slightly better than the HRC ( Appendix 10 ) . The biggest performance increase occurred when third-order terms were included , and the fourth-order terms also improved performance . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 012 The non-multiplicative nonlinearity model relaxes some restrictions of the 4-quadrant model . This is prudent because the exact nonlinear transformations implemented by neural circuits in the Drosophila brain remain poorly understood . For example , T4 and T5 are the first direction-selective neurons in the fly brain ( Maisak et al . , 2013 ) , but the mechanism by which they become direction-selective is not yet known . Furthermore , neurons upstream of T4 and T5 imperfectly segregate light and dark information ( Behnia et al . , 2014 ) and show overlap between the two motion pathways ( Silies et al . , 2013 ) , suggesting that ON/OFF segregation may not be crisply realized . We will discuss this model's estimation accuracy and glider performance in the next section . Drosophila's motion processing circuitry suggests two more generalizations of the non-multiplicative nonlinearity model . First , note that the non-multiplicative nonlinearity model inherits the HRC's assumption that each nonlinear unit only acts upon the low-pass filtered signal from one point in space and the high-pass filtered signal from the neighboring point ( Figure 4A ) . In contrast , the converging 3-point correlator ( Figure 1H ) shows that the accuracy of motion estimation can sometimes be enhanced by nonlinearly combining both low-pass filtered signals ( Figure 1I ) . Moreover , connectomic evidence conflicts with the non-multiplicative nonlinearity model's constraints , because each T4 cell receives synaptic connections from both the Mi1 cell and the Tm3 cells ( T4's two major input channels ) at overlapping points in space ( Takemura et al . , 2013 ) . The unrestricted nonlinearity model removes this restriction of the non-multiplicative nonlinearity model by allowing a 4-dimensional nonlinearity to act on all four filtered signals ( Figure 4C ) . Here , we again model this nonlinearity as a fourth-order polynomial ( ‘Materials and methods’ ) . The unrestricted nonlinearity allows the motion estimator to nonlinearly combine multiple temporal channels from the same point in space . Recent experiments indicate that the Mi1 and Tm3 cells alone are insufficient to account for the motion processing of the T4 channel ( Ammer et al . , 2015 ) . Future work might generalize the unrestricted nonlinearity model to include three or more temporal channels at each point in space . The models presented so far operate only on a pair of neighboring photoreceptors , and the final generalization incorporates a third point in space . Averaging EMDs over space improves the accuracy of whole-field motion estimation ( Dror et al . , 2001 ) , but Drosophila's neural circuitry suggests that it might adopt a more sophisticated strategy to combine signals across space . In particular , single T4 cells receive synaptic inputs from Mi1 cells and Tm3 cells from more than two retinotopic columns ( Takemura et al . , 2013 ) . This arrangement could allow the circuit to incorporate higher-order correlations that are distributed across three or more spatial input channels . To explore whether this possibility has computational significance , we generalized the unrestricted nonlinearity model to provide unrestricted access to six temporal channels distributed across three points in space ( Figure 4D ) . We refer to this model as the extra input nonlinearity model . We approximate its 6-dimensional nonlinearity as a fourth-order polynomial ( ‘Materials and methods’ ) . Having introduced the rationale behind the non-multiplicative , unrestricted , and extra input nonlinearity models , it is straightforward to examine their performance as motion estimators . First note that the polynomial non-multiplicative nonlinearity model was a better motion estimator ( Figure 4E ) than the weighted 4-quadrant model ( Figure 3C ) . This implies that some useful signatures of naturalistic motion are not made accessible by simply segregating ON and OFF motion signals . Interestingly , this performance improvement is largely due to 3-point correlations , and models that exclude fourth-order polynomial terms still outperform the weighted 4-quadrant model ( Figure 4—figure supplement 2 ) . Third-order correlations are only useful for motion estimation because of light–dark asymmetries in natural stimulus statistics ( Fitzgerald et al . , 2011; Clark et al . , 2014 ) , so this result implies that ON/OFF segregation provides an imperfect way to account for the complexity of light–dark asymmetries found in the natural world . The non-multiplicative nonlinearity model also made novel use of low-order correlations to improve its motion estimate ( Appendix 10 ) . The three models are hierarchical because the non-multiplicative nonlinearity model is a special case of the unrestricted nonlinearity model , which is itself a special case of the extra input nonlinearity model . Thus , we expect each model to perform at least as well as its predecessor , but it is possible that some circuit elaborations will not introduce useful computational cues . Nevertheless , we found that that the unrestricted nonlinearity model performed better than the non-multiplicative nonlinearity model , and the extra input model performed better than the average of two neighboring unrestricted nonlinearity models ( Figure 4E ) . Therefore , both models incorporated novel computational signatures with relevance for visual motion estimation . Although the relative improvements were fairly small , it's worth noting that the improvement from spatial averaging is also small , and it is possible that the fly brain builds an accurate motion estimator by combining a large number of weak predictors of motion . Each of these three generalized models predicted 2-point glider responses ( Figure 4F ) that closely resembled the standard HRC ( left , Figure 1G ) . Each model also correctly predicted the experimental turning directions to each of the 3-point glider stimuli ( Figure 4G ) . The magnitudes of the 3-point glider turning responses did not unambiguously favor any of the three hierarchical models ( Figure 4G ) or the weighted 4-quadrant model ( right , Figure 3D ) . Each model did better on some stimuli and worse on others . Nevertheless , the predicted glider responses did make several interesting points . First , the extra input nonlinearity model predicted a clear asymmetry between positive and negative 3-point gliders ( Figure 4G ) . This shows that some of the even-ordered correlations found in 3-point glider stimuli have relevance for naturalistic motion estimation . Second , the observation that each model provides qualitatively similar glider response patterns illustrates that animals could use multiple nonlinear mechanisms to access ethologically relevant higher-order correlations . Future experiments should directly assess the functional relevance of the different models in the hierarchy . Finally , the qualitative agreement between all of these predictions and the experimental data supports the general hypothesis that glider responses could reflect underlying nonlinear mechanisms that facilitate motion estimation in natural environments . In this paper , we sequentially introduced several models in order to isolate specific ideas about the relationships between Drosophila's behavior , its motion estimation circuit , and the statistical demands of accurate motion estimation in natural environments . The front-end nonlinearity model explored an interesting candidate principle for visual motion estimation , but it conflicted sharply with fly behavior ( Figure 2G ) and excluded the conceptual insights offered by other models . For example , the front-end nonlinearities we considered eliminated the asymmetry between light and dark contrasts ( Figure 2D ) , removing the need for separate ON and OFF processing . However , the remaining models embodied ideas that are complementary rather than exclusive , and these models should not be thought of as competitors . Instead , we will show here that the final , most general model incorporates the variety of conceptual points that were initially illustrated by specific models . The structure of the non-multiplicative nonlinearity models can be directly plotted ( Figure 4B ) , but is not easy to visualize the 6-dimensional nonlinearity that defines the extra input nonlinearity model . We therefore need an alternate technique to illustrate its computations . We proceed by leveraging three ideas . First , a wide variety of visual motion estimators can be expanded as an abstract series of multipoint correlators ( e . g . , see Poggio and Reichardt , 1980 , Fitzgerald et al . , 2011 , and Appendices 6 , 7 , 11 ) , and it is straightforward to pictorially represent a multipoint correlator ( e . g . , see Figure 1A , H , and more to come ) . In the extra input nonlinearity model , this expansion is immediate because we have already parameterized its 6-dimensional nonlinearity as a polynomial . Importantly , this expansion should be considered at the algorithmic level ( Marr and Poggio , 1976 ) , and we do not suggest that the wiring of brain circuits will reflect a large number of higher-order correlators . To the contrary , a large number of higher-order multipoint correlators may be implemented implicitly by high-dimensional nonlinearities suggested by Drosophila's visual circuitry . Second , we note that certain multipoint correlators can be recombined into a 2-dimensional non-multiplicative nonlinearity that facilitates easy comparisons with the HRC and weighted 4-quadrant models ( e . g . , see Figure 4B ) . Taken together , these two points mean that we can represent the performance-optimized extra input nonlinearity model in terms of non-multiplicative nonlinearity models and multipoint correlators , each of which are easy to represent graphically . This graphical representation could be unwieldy because of the shear number of higher-order correlators in the model . Thus the third and final point is that we need a way to identify a relatively small number of terms that substantially improve the accuracy of motion estimation and illustrate the conceptual content of the model . To achieve this , we used lasso regression ( Tibshirani , 1996 ) to identify models with fewer multipoint correlators that still enabled accurate motion estimation ( ‘Materials and methods’ ) . This analysis revealed that fewer than half of all multipoint correlators were needed to account for the full accuracy of the extra input nonlinearity model ( rightmost bars , Figure 5A ) . In fact , the accuracy of naturalistic motion estimation increased rapidly as the few correlators were sequentially added ( left bars , Figure 5A ) , and a model that used 16 out of the 209 possible predictors was already able to produce 74% of the gain offered by the full extra input nonlinearity model ( red bar , Figure 5A ) . 10 . 7554/eLife . 09123 . 013Figure 5 . Computational interpretation of the extra input nonlinearity model . ( A ) We used lasso regression to select subsets of predictors that might enable accurate estimation ( see ‘Materials and methods’ ) . With only 16 predictors , the model improved naturalistic performance over the HRC by 68% , and including fewer than half of the predictors improved it by the full 92% . The maximum number of predictors corresponds to the number of polynomial coefficients that were fit in the full model . ( B ) We visualized the 6-dimensional nonlinearity as the sum of several simpler computational modules . When only 16 predictors were used ( red bar in ( A ) ) , the model used four distinct types of computations . In particular , the model included nearest-neighbor and next-nearest-neighbor non-multiplicative nonlinearities ( top row ) . It also included a converging 3-point correlator from the two furthest photoreceptors and a 4-point correlator that combined three spatial inputs ( bottom row ) . ( C ) Venn diagram illustrating the hierarchical nesting of models used in this paper . All models in this paper contain sets of parameters that reproduce the HRC ( gray dot ) . The weighted 4-quadrant model is a subset of non-multiplicative nonlinearity models , which are themselves a subset of unrestricted nonlinearity models . The extra input nonlinearity encompasses all the models . When we approximated the nonlinearites with fourth order polynomials , we restricted them to a smaller portion of the model space . The 4-quadrant nonlinearities only overlapped with the fourth-order polynomial approximation at the HRC , because the weighted 4-quadrant model is infinite order when expanded as a polynomial ( see Appendix 7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 01310 . 7554/eLife . 09123 . 014Figure 5—figure supplement 1 . Structure of non-multiplicative nonlinearities in the extra input model of Figure 5B . ( A ) The nearest-neighbor non-multiplicative nonlinearity was made up of a standard HRC and a 3-point correlator in which the low-pass filtered input was squared before being multiplied by the adjacent receptor's high-pass filtered signal . ( B ) The next-nearest-neighbor non-multiplicative nonlinearity combined the analogous long-range terms . ( C ) Structure of the 2-dimensional nonlinearities , shown according to the same conventions as Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 09123 . 014 The leading 16 predictors compactly illustrated how the extra input nonlinearity model recapitulates the conceptual advances offered by the other models ( Figure 5B ) . Four of the predictors combine to implement a mirror-symmetric non-multiplicative nonlinearity model that acts on the first and second points in space ( first term , Figure 5B ) . The dominant contribution to the nonlinearity is the HRC's multiplier , but an additional third-order term breaks the symmetry between positive and negative low-pass filtered signals ( Figure 5—figure supplement 1 ) . Thus , the extra input nonlinearity model approximately correlates neighboring points in space , as the HRC would suggest , but it differentially weights positive and negative low-pass filtered signals , like the weighted 4-quadrant model . It also replicates the main insight from the non-multiplicative nonlinearity model: the best treatment of asymmetric light and dark information need not be as simple as pure ON/OFF segregation . The model used another eight predictors to construct two more non-multiplicative nonlinearity models , one that surveyed the second and third points in space and another that surveyed the first and third points ( first and second terms , Figure 5B , Figure 5—figure supplement 1 ) . These components make the previously highlighted conceptual points and add the observation that spatial averaging improves estimates . The final four predictors implemented two mirror anti-symmetric multipoint correlators ( third and fourth terms , Figure 5B ) . In particular , two predictors went towards implementing a converging 3-point correlator that spanned the first and third spatial points ( third term , Figure 5B ) . This estimator made the model's asymmetric treatment of light and dark signals more nuanced than permitted by the non-multiplicative nonlinearity model , and it also incorporated motion signals that combine multiple temporal signals from the same point in space . This latter point was the main conceptual motivation for the unrestricted nonlinearity model . Finally , the last two predictors implemented a 4-point correlator that combined temporal signals from three distinct points in space ( fourth term , Figure 5B ) . This component reinforces the conceptual motivation for the extra input nonlinearity model and gives a concrete example of a computationally relevant higher-order correlator that is distributed across three points in space . It's interesting that the leading fourth-order correlator spanned three spatial points , because the extra input nonlinearity model was the first performance-optimized model that generated a substantially asymmetric response to positive and negative 3-point gliders ( Figure 4G ) . This paper set out with the goal of exploring whether the statistical demands of naturalistic motion estimation could provide a useful lens for interpreting features of Drosophila's behavior and neural circuitry that push beyond the canonical HRC . Although we have considered several interesting classes of visual motion estimators , the space of possible motion estimators is much larger ( Figure 5C ) . For instance , these models have not explored the impact of temporal filter choice on naturalistic motion estimation . Nor have they assessed the possibility of more than two temporal filters , which is be suggested by anatomical ( Takemura et al . , 2013 ) and physiological ( Ammer et al . , 2015 ) experiments . More generally , the neural circuits contributing to Drosophila's motion estimator are still incompletely known , and the extent to which the fly brain's biological complexity reflects computational sophistication remains an open question . Theoretical considerations will be critical for resolving that question and pinpointing the most relevant principles underlying visual motion estimation .
Ongoing research is providing an increasingly detailed picture of the anatomy and physiology of the visual circuitry that implements motion processing in Drosophila . Through the combination of genetic silencing experiments , connectomic analysis , and functional recordings , researchers have identified many individual neurons in the fly brain that contribute to visual motion processing ( Silies et al . , 2014 ) . Although the HRC provided the initial theoretical impetus for these experiments , specific experimental outcomes have often been unanticipated . For instance , the fly brain contains multiple pathways that segregate different types of motion information ( Joesch et al . , 2010; Clark et al . , 2011; Silies et al . , 2013 ) ; its direction-selective neurons receive inputs from more than two neighboring points in visual space ( Takemura et al . , 2013 ) ; and the biological substrates for reverse-phi signals , which were fundamental to the formulation of the HRC , remain poorly understood ( Clark et al . , 2011; Tuthill et al . , 2011; Joesch et al . , 2013 ) . Theoretical work to illuminate the computational significance of these various discrepancies is critical for understanding Drosophila's motion estimator . The results presented in this paper provide a new theoretical perspective on these experimental results . While previous research has addressed how neural circuits could use four quadrants to carry out algebraic multiplication , here , the recurring theme of our models was that motion-processing circuits should treat light and dark signals differently for functional reasons . We first showed that visual systems could use ON and OFF processing channels that separately correlate light and dark signals to improve the accuracy of motion estimation ( Figure 3 ) . This model was inspired by the experimental observation that Drosophila's motion processing channels distinguish between light increments and decrements ( Joesch et al . , 2010; Clark et al . , 2011 ) , but this study is the first to explicitly demonstrate how such processing channels can improve the accuracy of motion estimation . Furthermore , our model shows that both the phi channels ( i . e . , the ( + + ) and ( − − ) and quadrants ) and the reverse-phi channels ( i . e . , the ( + − ) and ( − + ) quadrants ) can contribute productively to motion estimation in natural environments . Since many animals experience similar sensory statistics and ON and OFF visual processing channels are pervasive across visual systems ( Schiller , 1992; Westheimer , 2007 ) , these mechanisms might be very general . Ultimately , the performance gains from weighted quadrants were a consequence of statistical asymmetries between light and dark contrasts in natural images , and our models showed that neural circuits could perform even better if they made distinctions between light and dark signals that were subtler than simple ON/OFF segregation ( Figure 4 ) . Recent experimental evidence indicates that the fly's motion processing channels are imperfectly selective for ON vs OFF information ( Silies et al . , 2013; Behnia et al . , 2014; Strother et al . , 2014 ) , and it is important that future experiments characterize such subtleties in the computations performed by these circuits . Our most general model contained three spatial inputs and showed that spatial averaging of local motion detectors was suboptimal ( Figure 4E ) . Anatomy suggests that single T4 cells receive inputs from several different retinotopic columns , and also from multiple neuron types in a single retinotopic column ( Takemura et al . , 2013 ) . Our modeling suggests that these two forms of circuit heterogeneity could enhance motion estimation by facilitating computations that go beyond averaging to compute higher-order correlations that are distributed across multiple points in space ( Figure 5B ) . Overall , our results demonstrate how the subtleties of neural circuit nonlinearities can improve motion detection with naturalistic inputs . It therefore seems likely that some of the complexities of Drosophila's circuitry are critical to its performance under natural conditions . It is remarkable that our approximation of natural motion by the rigid translation of natural images revealed substantial utility for higher-order correlations in motion processing . Truly naturalistic motion would include spatial velocity gradients , occlusion , expansion , and contraction , yet the simplified naturalism we used to optimize our models already sufficed to account for many aspects of the fly's glider responses . This may be because the rotational optomotor response measured in the fly experiments is thought to be sensitive primarily to full-field rotations , which our naturalism emulates well . However , since other higher-order correlations may be associated with non-rigid translation ( Nitzany and Victor , 2014 ) , one might expect a different set of glider sensitivities to be optimal in the context of other motion-guided behaviors , such as looming responses ( Gabbiani et al . , 1999; Tammero and Dickinson , 2002; Card and Dickinson , 2008 ) . Since a common elementary motion detector might underlie many or all motion-guided behaviors , incorporating more complex optic flow patterns may even diminish discrepancies between our models and Drosophila's behavior . The approach of this study is also relevant to vertebrate vision , where researchers typically model motion estimation using the motion energy model ( Adelson and Bergen , 1985 ) . Like the HRC , the motion energy model only responds to 2-point correlations in the visual stimulus . Consequently , many of the theoretical considerations in this paper apply directly to the motion energy model . Furthermore , each of our computational models can be straightforwardly generalized to the architecture of the motion energy model . For example , one could incorporate non-multiplicative nonlinearities by replacing the squaring operation of the motion energy model with a more flexible nonlinearity . Nevertheless , the numerical benefits offered by each modification to the motion energy model might differ from those found for the HRC because the motion energy model and HRC use distinct spatial and temporal filtering . Such differences could in principle manifest themselves as a different pattern of predicted glider responses ( Hu and Victor , 2010; Clark et al . , 2014 ) , but comparative electrophysiology experiments in macaques and dragonflies currently suggest that similarities between primate and insect motion processing are abundant ( Nitzany et al . , 2014 ) . Our models make predictions that are testable with new experiments . Researchers hypothesize that the T4 and T5 neurons in the fly lobula nonlinearly combine visual inputs across space and time to become the first direction-selective neurons in Drosophila's visual system ( Maisak et al . , 2013 ) . In accordance with the HRC model , conventional wisdom says that these neurons will multiply their input channels . In contrast , we predict that T4 and T5 will combine their visual input streams with non-multiplicative nonlinearities that facilitate accurate motion estimation in natural sensory environments . It's crucial to note that subtle differences between biology's nonlinearity and a pure multiplication can correspond to substantial functional effects . In particular , the optimized nonlinearity that we found here ( Figure 4B ) is superficially similar a simple multiplication , yet its subtle distinctions manifest themselves by improving the local estimation accuracy of the HRC by an impressive margin ( Figure 2D ) . In this paper , we studied several simple models to most clearly illustrate the computational consequences of fundamental nonlinear circuit operations . Each of these operations individually provided a way for Drosophila to improve their motion estimation accuracy in natural environments , but they are not necessarily exclusive . For example , if a front-end nonlinearity does not fully remove the asymmetry between light and dark contrasts , then subsequent ON and OFF processing might further improve estimation accuracy . Similarly , non-multiplicative nonlinearities might enable an even better combination of ON and OFF signals for motion estimation . The general approach that we adopted here is to restrict the space of candidate models to those that have immediate biological relevance and to identify interesting models by optimizing the model's estimation accuracy over naturalistic stimuli . Future models should incorporate more biological details to better emulate the specifics of Drosophila's visual circuitry , which is rapidly being dissected through unprecedented anatomical , functional , and behavioral experiments ( Silies et al . , 2014 ) .
We simulated the linear responses of neighboring photoreceptors to naturalistic motion using methods similar to previous work ( Clark et al . , 2014 ) . We began with a database of natural images ( van Hateren and van der Schaaf , 1998 ) . We converted each natural image to a contrast scale , C ( x→ ) = ( I ( x→ ) −I0 ) /I0 , where C ( x→ ) is the contrast at the spatial point x→ , I ( x→ ) is its intensity , and I0 is the average intensity across the image . Since we only consider horizontal motion , we emulated the spatial blurring of Drosophila's photoreceptors in the vertical dimension by filtering across rows with a Gaussian kernel ( FWHM = 5 . 7° ) . We then took the central row of each filtered image to represent a one-dimension variant of the natural image , denoted c ( x ) . We applied reflective boundary conditions to generate images that covered 360° and down-sampled each resulting image to 1° pixels by averaging . Photoreceptor blurring from signals in the horizontal dimension depends on the velocity of motion . In particular , we model the response of the ith photoreceptor asVi ( t ) =∫dt′T ( t′ ) ∫dxM ( x−xi ) c ( x−v ( t−t′ ) ) , where T is a causal exponential kernel ( timescale = 10 ms ) , M is a Gaussian kernel ( FWHM = 5 . 7° ) , xi is the location of the ith photoreceptor , and ν is the velocity of motion . Each naturalistic motion comprised a randomly selected one-dimensional natural image , an offset to set the initial location of the photoreceptors , and a velocity drawn from a zero-mean normal distribution with a standard deviation of 90°/s . In this manner , we simulated the responses of three horizontally adjacent photoreceptors ( spaced by 5 . 1° ) to 5 × 105 naturalistic motions ( each with duration = 800 ms , time step = 5 ms ) . We then explicitly enforced left-right symmetry in the naturalistic ensemble by pairing each naturalistic motion with a new simulated motion , in which the natural image is reflected , the velocity is inverted , and the offset is chosen such that {V1 ( t ) , V2 ( t ) , V3 ( t ) } in the new naturalistic motion is exactly {V3 ( t ) , V2 ( t ) , V1 ( t ) } from its partner . The final symmetric ensemble thus consists of 106 naturalistic motions . The HRC applies two temporal filters to its photoreceptor inputs . We denote the kernels of the low-pass and high-pass filters as f and g , respectively , such that the output of a local HRC is R ( t ) = ( f*V1 ) ( t ) ( g*V2 ) ( t ) − ( g*V1 ) ( t ) ( f*V2 ) ( t ) , where * denotes convolution ( Figure 1A ) . We consider the HRC's velocity estimate for a given naturalistic motion as its value at the final time point of the simulation . We model the filter kernels asf ( t ) =te−t/τ , t≥0andg ( t ) =df ( t ) dt , where τ = 20 ms and g ( t ) is comparable to lamina monopolar cell responses ( Clark et al . , 2011; Behnia et al . , 2014 ) . We built the alternate motion estimators considered in this work from the same four filtered signals , { ( f∗V1 ) , ( g∗V1 ) , ( f∗V2 ) , ( g∗V2 ) } , always considering the estimator's output at the final time point as its velocity estimate . Thus , none of our models modified the spatial or temporal processing of the HRC , reflecting our emphasis on how nonlinear processing might be tuned for naturalistic motion estimation . The global output of an array of HRCs would be obtained by pooling signals across space . Here we focus on spatiotemporally local strategies for motion estimation and at most pool motion signals across two neighboring motion detectors . We evaluate motion estimators by the mean squared error between their output and the true velocity . To minimize the mean squared error of the HRC , we scale its output by r ( R ) σν/σR , where r ( R ) is the correlation coefficient between the HRC's output and the velocity of motion , σν is the standard deviation of the velocity distribution , and σR is the standard deviation of the HRC's output . Once the HRC is scaled in this manner , its mean squared error isϵ=σv2 ( 1− ( r ( R ) ) 2 ) . More generally , this equation rewrites the mean squared error of any optimally scaled motion estimator in terms of its correlation coefficient with the velocity . All motion estimators considered in this paper are optimally scaled , and we find the correlation coefficient to be more intuitive than the mean squared error . We thus always report the performance of each motion estimator in terms of the correlation coefficient between the true and estimated velocity . We fit the linear weighting parameters in the models of Figures 1I , 3A , 4A , C , D to maximize the estimation accuracy over a simulated ensemble of naturalistic motions . The formulas provided in subsequent sections of the ‘Materials and methods’ will cast each motion estimation scheme as a linear combination of a variety of motion predictors , ve=∑iwixi , where the xi are nonlinear combinations of { ( f∗V1 ) , ( g∗V1 ) , ( f∗V2 ) , ( g∗V2 ) , ( f∗V3 ) , ( g∗V3 ) } that depend on the model architecture , and the wi are associated weighting coefficients . We chose the weights to minimize the mean squared error between the true and predicted velocity , which is the standard scenario considered by ordinary least-squares regression . The same weights maximize the correlation coefficient between the true and predicted velocity , and we typically present model accuracies as correlation coefficients . We used twofold cross-validation to protect against over-fitting . In particular , we randomly divided the ensemble of naturalistic motions into a training set of 500 , 000 symmetrically paired examples and a testing set of the remaining 500 , 000 examples . We determined the weighting coefficients by minimizing the empirical error over the training set , and we reported accuracies over the test set . To estimate error bars for each model's accuracy , we computed twenty random divisions of the naturalistic motion ensemble and calculated the standard deviation of the estimation accuracy . We generated 25 random instantiations of each glider stimulus considered by our previous experimental work ( Figure 1F , duration = 3 s , update rate = 40 Hz , pixel size = 5° ) ( Hu and Victor , 2010; Clark et al . , 2014 ) . We evaluated the response of each model to these stimuli by averaging the outputs of 60 identical local motion estimators ( each separated by 5 . 1° ) over the last two seconds of visual stimulation . Glider predictions were equal and opposite for the left and right variants of the stimuli , so we pooled leftward and rightward stimuli in all figures ( Figure 1F shows the rightward variants ) . We scaled each model's output such that the average response to the positive 2-point glider was 1 . All figures associated with glider responses show the mean and standard error of each model's response across the 25 glider instantiations . The model in Figure 2A replaces the linear photoreceptor signals , V1 and V2 , with nonlinear photoreceptor signalsyi=h ( Vi ) , where h is some nonlinear function . Thus , the motion estimate from the front-end nonlinearity model isF= ( f*y1 ) ( g*y2 ) − ( g*y1 ) ( f*y2 ) . To implement the contrast equalizing nonlinearity , we replaced values of Vi ( t ) by their rank-order ( scaled and shifted to range between −1 and +1 ) . Note that all Vi ( t ) were sorted together ( i . e . , including all spatial points , temporal points , and simulated naturalistic motions ) . When multiple Vi ( t ) had the same value , they were given the same rank . To implement binarizing nonlinearities , we again sorted the Vi ( t ) and found the values corresponding to the threshold locations . For example , to calculate the binarizing nonlinearity with two steps ( Appendix 4 ) : ( i ) we found the Vi ( t ) values corresponding to the 25th and 75th percentiles; ( ii ) signals below the 25th percentile or above the 75th percentiles were assigned the value of −1; and ( iii ) signals between 25th and 75th percentiles were assigned the value of +1 . To implement the Gaussianizing nonlinearity , we again rank-ordered the Vi ( t ) ( scaled to range between 0 and 1 ) and applied the inverse Gaussian cumulative distribution function to these ranks . The HRC is the special case of this model where the front-end nonlinearity is linear . The weighted 4-quadrant model in Figure 3A separately correlates bright and dark signals . Mathematically , it isQ=∑a∈{+ , −}∑b∈{+ , −}wab ( Q ) Qab , where wab ( Q ) are adjustable weights that parameterize the model , Qab=[ ( f*V1 ) ]a[ ( g*V2 ) ]b−[ ( g*V1 ) ]b[ ( f*V2 ) ]a , [x]+ is equal to x when x is positive and zero otherwise , and [x]− is equal to x when x is negative and zero otherwise . The HRC is the special case of this model where w++ ( Q ) =w+− ( Q ) =w−+ ( Q ) =w−− ( Q ) . The non-multiplicative nonlinearity model in Figure 4A replaces the HRC's multiplication step with a more flexible two-dimension nonlinearity . In particular , it isN=η ( ( f*V1 ) , ( g*V2 ) ) −η ( ( f*V2 ) , ( g*V1 ) ) , where we approximate the nonlinearity , η , as a fourth-order polynomialη ( x , y ) =∑i=04∑j=04−iwij ( N ) xiyj , and wij ( N ) are adjustable weights that parameterize the model . We include terms up to fourth order in this model to ensure that it is flexible enough to describe the published glider response data . In particular: ( i ) the second-order terms accommodate responses to 2-point glider stimuli; ( ii ) the third-order terms accommodate parity-inverting responses to 3-point glider stimuli; and ( iii ) the fourth-order terms enable the model to respond with unequal magnitude to positive and negative parity 3-point glider stimuli ( Figure 4—figure supplement 1 ) . Thus , this model has 14 parameters . The HRC is the special case of this model where only w11 ( N ) is nonzero . Here we model the 4-dimensional nonlinearity in Figure 4C as a fourth-order polynomial of the four filtered signals in the HRC . In general , this motion estimator isS=∑i=04∑j=04−i∑k=04−i−j∑l=04−i−j−kwijkl ( S ) ( f*V1 ) i ( g*V1 ) j ( f*V2 ) k ( g*V2 ) l , where wijkl ( S ) are adjustable weights that parameterize the model , and we set w0000 ( S ) =0 because this term has no utility for naturalistic motion estimation . Thus , this model has 69 parameters . The HRC is the special case of this model where w1001 ( S ) = −w0110 ( S ) ≠0 , and all other parameters are zero . Here we model the 6-dimensional nonlinearity in Figure 4D as a fourth-order polynomial of the six filtered signals in two neighboring HRCs . In general , this motion estimator isE=∑i=04∑j=04−i∑k=04−i−j∑l=04−i−j−k∑m=04−i−j−k−l∑n=04−i−j−k−l−mwijklmn ( E ) ( f*V1 ) i ( g*V1 ) j× ( f*V2 ) k ( g*V2 ) l ( f*V3 ) m ( g*V3 ) n , where wijklmn ( E ) are adjustable weights that parameterize the model , and we set w000000 ( E ) =0 because this term has no utility for naturalistic motion estimation . Thus , this model has 209 parameters . The average of two neighboring HRCs is the special case of this model where w100100 ( E ) = −w011000 ( E ) =w001001 ( E ) =−w000110 ( E ) ≠0 , and all other parameters are zero . Lasso regression augments the squared error with an L1 penalty on nonzero weighting coefficients that favors sparse solutions ( Tibshirani , 1996 ) . We used lasso regression to identify subsets of predictors that might enable accurate motion estimation ( Figure 5A ) . Once we identified a predictor subset using lasso regression , we refit the nonzero model weights using ordinary least squares regression ( i . e . , without the weight penalty ) . | Many animals have evolved the ability to estimate the speed and direction of visual motion . They use these estimates to judge their own motion , so that they can navigate through an environment , and to judge how other animals are moving , which allows them to avoid predators or detect prey . In the 1950s , a physicist and a biologist used measurements of beetle behavior in response to visual stimuli to develop a model for how the brain estimates motion . The model became known as the Hassenstein-Reichardt correlator ( HRC ) . The HRC and related models accurately predict the behavioral and neural responses of insects and mammals to many types of motion stimuli . However , there are visual stimuli that generate motion percepts in fruit flies ( and humans ) that cannot be accounted for by the HRC . Are these differences between real brains and the HRC simply imperfections in visual circuits , whose neurons cannot perform idealized mathematical operations , or are these deviations intentional , somehow improving motion estimates ? In other words: are the observed deviations a bug or a feature of visual circuits ? To address this question , Fitzgerald and Clark evaluated how different models of motion detection performed when presented with natural scenes . Natural scenes are fundamentally different from most stimuli used in lab , since they contain a rich set of regularities that are not present in simple stimuli . Fitzgerald and Clark compared the ability of the HRC , along with new , more general models , to estimate the speed and direction at which images moved across a screen . This revealed that many models could out-perform the HRC by taking advantage of regularities in natural scenes . Those models that were tuned to perform well with natural scenes could also predict the paradoxical motion percepts that were not predicted by the HRC . This suggests that visual circuits may have evolved to perform well with natural inputs , and the paradoxical motion percepts represent a feature of the real circuit , rather than a bug . Models that performed well with natural inputs treated light and dark visual information differently . This different treatment of light and dark is a property of most visual systems , but not of the HRC or related models . In the future , these models of motion processing may help us understand how biological details of the fruit fly's visual circuits help it to estimate motion . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Nonlinear circuits for naturalistic visual motion estimation |
Metabolic specialization among major brain cell types is central to nervous system function and determined in large part by the cellular distribution of enzymes . Serine hydrolases are a diverse enzyme class that plays fundamental roles in CNS metabolism and signaling . Here , we perform an activity-based proteomic analysis of primary mouse neurons , astrocytes , and microglia to furnish a global portrait of the cellular anatomy of serine hydrolases in the brain . We uncover compelling evidence for the cellular compartmentalization of key chemical transmission pathways , including the functional segregation of endocannabinoid ( eCB ) biosynthetic enzymes diacylglycerol lipase-alpha ( DAGLα ) and –beta ( DAGLβ ) to neurons and microglia , respectively . Disruption of DAGLβ perturbed eCB-eicosanoid crosstalk specifically in microglia and suppressed neuroinflammatory events in vivo independently of broader effects on eCB content . Mapping the cellular distribution of metabolic enzymes thus identifies pathways for regulating specialized inflammatory responses in the brain while avoiding global alterations in CNS function .
The brain is a highly heterogeneous organ composed of diverse cell types with distinct structures , molecular compositions , and complementary functions ( Zhang et al . , 2014 ) . Major brain cell classes include neurons , which are the principal cells responsible for directing and transmitting information in the nervous system in the form of chemical and electrical signals , and glia , which are often referred to as support cells , but are now recognized to play critical roles in regulating neurotransmission , synaptic activity , and higher-order neurophysiological , behavioral , and disease processes ( Barres , 2008 , Gundersen et al . , 2015 ) . Understanding the biochemical pathways that furnish neurons and glia with different properties and functions requires knowledge of the content , activity , and regulation of proteins within these cells . Global gene expression studies have provided valuable insights into the molecular composition of different brain cell types ( Cahoy et al . , 2008 , Doyle et al . , 2008 , Beutner et al . , 2013 , Chiu et al . , 2013 , Zhang et al . , 2014 ) . In particular , a recent study using deep RNA sequencing generated a quantitative , high-resolution map of transcript abundances for major mouse brain cell types , including neurons , astrocytes , and microglia , among others ( Zhang et al . , 2014 ) . Thousands of cell type-enriched transcripts and splicing isoforms were identified , underscoring the unique molecular composition of different brain cell classes ( Zhang et al . , 2014 ) . How these transcriptional signatures relate to the protein content of brain cell types , however , remains mostly unknown . Several proteomic studies have been performed on isolated brain cells , in particular , neurons and astrocytes , using approaches that include two-dimensional gel electrophoresis and shotgun liquid chromatography-mass spectrometry ( LC-MS ) ( Yu et al . , 2004 , Yang et al . , 2005 , Glanzer et al . , 2007 , Liao et al . , 2008 , Bell-Temin et al . , 2012 ) . Many proteins , however , are regulated in their function by post-translational mechanisms ( Kobe and Kemp , 1999 ) that are difficult to discern with conventional expression-based proteomic methods . Activity-based protein profiling ( ABPP ) is a chemical proteomic method that uses active site-directed chemical probes to selectively target subsets of proteins in the proteome based on shared mechanistic and/or structural features ( Cravatt et al . , 2008 ) . Because ABPP probes modify protein targets based on conserved functional features , these reagents can detect , enrich , and identify many members of a class of proteins and illuminate changes in protein activity that are not reflected in transcript or protein abundance ( Jessani et al . , 2004 ) . ABPP has been used to map deregulated enzyme activities in a variety of physiological and pathological processes , including cancer ( Jessani et al . , 2004 , Nomura et al . , 2010 , Kohnz et al . , 2015 ) , metabolic disorders ( Dominguez et al . , 2014 ) , and infectious diseases ( Greenbaum et al . , 2002 , Bottcher and Sieber , 2008 , Heal and Tate , 2012 , Nasheri et al . , 2013 ) . Among the classes of proteins that have been addressed by ABPP , the serine hydrolases are a particularly large and diverse enzyme family that plays many key roles in the nervous system ( Simon and Cravatt , 2010 , Long and Cravatt , 2011 ) . Serine hydrolases regulate proteolysis at the synapse to modulate neuronal plasticity ( Melchor and Strickland , 2005 ) , the post-translational modification state of key brain signaling proteins ( Sontag et al . , 2007 , Siegel et al . , 2009 ) , and , perhaps most notably , the metabolism of a wide range of chemical messengers , including neurotransmitters ( e . g . , acetylcholine [Phillis , 2005] , neuropeptides ( e . g . , α-melanocyte-stimulating hormone [Wallingford et al . , 2009] ) , and lipid messengers ( e . g . , endocannabinoids; eCBs [Blankman and Cravatt , 2013] ) . Here we use ABPP combined with shotgun LC-MS ( Washburn et al . , 2001 ) to generate an in-depth portrait of serine hydrolase activities across three major mouse brain cell types—neurons , astrocytes , and microglia . We show that the output of this functional proteomic analysis correlates well with published RNA-seq data ( Zhang et al . , 2014 ) , although there are clear exceptions of enzymes where transcript and activity are un- and even anti-correlated . We also uncover strong evidence for the compartmentalization of functionally related serine hydrolases into distinct cell types , suggesting that their respective activities are anatomically segregated in the brain . A principal example was the selective enrichment of different sets of eCB biosynthetic and degradative enzymes in neurons versus microglia . We show that these activity patterns have functional relevance and mark pathways that regulate immunomodulatory eCB/eicosanoid signals without globally perturbing brain eCB content .
We globally assessed serine hydrolase activities in cultured mouse neurons , astrocytes , and microglia by treating proteomes from these cells with serine hydrolase-directed fluorophosphonate ( FP ) probes ( Figure 1A ) ( Simon and Cravatt , 2010 ) . Treatment with a rhodamine-coupled FP probe ( Patricelli et al . , 2001 ) followed by SDS-PAGE and in-gel fluorescence scanning revealed that the brain cell types express overlapping , but distinct sets of serine hydrolase activities ( Figure 1B ) . We then identified these serine hydrolases by treating brain cell proteomes with a biotin-coupled FP probe ( Liu et al . , 1999 ) , followed by enrichment with avidin chromatography , and multidimensional LC-MS ( MudPIT ) analysis ( Figure 1A ) ( Washburn et al . , 2001 , Jessani et al . , 2005 ) , which detected , in total , over 70 serine hydrolase activities across neuron , astrocyte , and microglia proteomes ( Figure 1—source data 1 ) . We assessed the relative amount of each serine hydrolase activity across brain cell types by the semi-quantitative method of spectral counting ( Jessani et al . , 2005 , Nahnsen et al . , 2013 ) . We found that the serine hydrolase activity profiles measured by ABPP-MudPIT showed excellent reproducibility across the four biological replicates performed for each cell type , with high correlations for replicates from the same cell type ( Pearson’s correlation , mean r = 0 . 91 ± 0 . 01 ) and much lower correlations for serine hydrolase activity profiles across differing cell types ( Pearson’s correlation , mean r = 0 . 62 ± 0 . 02 ) ( Figure 1—figure supplement 1A ) . Microglia and astrocytes displayed the highest pairwise correlation across the brain cell types examined , suggesting greater similarity of serine hydrolase metabolic activities among glia than between neurons and glia ( Figure 1—figure supplement 1A ) . 10 . 7554/eLife . 12345 . 003Figure 1 . Serine hydrolase activity profiles of mouse brain cell types . ( A ) Cartoon scheme of gel- and MS-based activity-based protein profiling ( ABPP ) methods used to measure serine hydrolase activities in primary mouse neurons , astrocytes , and microglia . For gel-based ABPP , a fluorophosphonate ( FP ) reactive group conjugated to a rhodamine reporter tag ( red oval; FP-Rh ) is used ( Patricelli et al . , 2001 ) . For MS-based ABPP ( ABPP-MudPIT ) , an FP reactive group conjugated to a biotin reporter tag ( purple diamond; FP-biotin ) is used ( Liu et al . , 1999 ) . ( B ) Gel-based ABPP of membrane proteomes from different brain cell types . ( C ) Hierarchically clustered heatmap of ABPP-MudPIT data ( left ) for serine hydrolases detected in neurons ( N ) , astrocytes ( A ) and microglia ( M ) . Data represent the mean spectral count values for each serine hydrolase ( from four independent experiments ) expressed as % of cell type with maximum number of spectral counts ( right heatmap shows the maximum spectral counts among cell types for each serine hydrolase ) . ( D ) Relationship between serine hydrolase activities , as measured by ABPP-MudPIT , and previously reported mRNA expression for these enzymes , as measured by RNA-Seq ( Zhang et al . , 2014 ) , in neurons , astrocytes , and microglia . Serine hydrolases showing ≥ three-fold enrichment in activity in a specific cell type as measured by ABPP-MudPIT are shown as filled colored circles and a Pearson’s correlation reported for the aggregate correlation between their ABPP and RNAseq profiles ( r = 0 . 54; p < 0 . 01 ) . ( E ) Examples of serine hydrolases where activity and mRNA expression measurements were uncorrelated ( ABHD16A , HTRA1 ) or anti-correlated ( MGLL ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 00310 . 7554/eLife . 12345 . 004Figure 1—source data 1 . Serine hydrolases identified in neuron , astrocyte , and microglia proteomes by ABPP-MudPIT . Peptide spectral counts ( SC ) of serine hydrolases in neuron , astrocyte and microglia proteomes . Average SC values of four individual ABPP-MudPIT experiments ± SEM are reported for proteins identified with a minimum of five SCs in at least one cell type . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 00410 . 7554/eLife . 12345 . 005Figure 1—figure supplement 1 . Serine hydrolase activity profiles of mouse brain cell types . ( A ) Heat map of Pearson’s correlation between spectral count values for serine hydrolases in biologically independent replicates of ABPP-MudPIT experiments performed on mouse primary neurons , astrocytes , and microglia . Each replicate consists of cells derived from 5-10 pooled brains . ( B ) Cell type enriched serine hydrolases displaying at least two-fold greater number of spectral counts in one cell type compared to the other two . Average spectral counts for neurons , astrocytes , and microglia , as well as fold-enrichment are reported from left to right in the table . ( C ) Relationship between fold-enrichment and average activity measurements for serine hydrolases in specific cell types . Fold-enrichment values were defined as spectral counts for a serine hydrolase in a given cell type divided by the average spectral count number for that serine hydrolase in all three cell types . Average activity measurements correspond to the mean spectral count for each serine hydrolase within a given cell type ( from four replicates ) . Enzymes that did not meet a two-fold enrichment cutoff for cell type-specific expression are shown in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 00510 . 7554/eLife . 12345 . 006Figure 1—figure supplement 2 . DAGLβ and ABHD12 activities are enriched in microglia . ( A ) Gel-based ABPP analysis of membrane proteomes from different brain cell types using a DAGL-directed probe ( HT01 ) ( Hsu et al . , 2012 ) shows enriched DAGLβ activity in microglia . ( B ) 2-AG hydrolytic activities of neuron , astrocyte , and microglia membrane proteomes derived from Abhd12+/+ and Abhd12–/– mice basally or following pre-treatment with the MGLL inhibitor KML29 ( 250 nM , 1 hr ) . While MGLL was found to be the major 2-AG hydrolase in all three brain cell types , the proportion of ABHD12-dependent 2-AG hydrolysis was greater in astrocytes and , in particular , microglia compared to neurons . Data represent average values ± SEM; N = 5 per cell type , genotype , and treatment . *p < 0 . 05 and **p < 0 . 01 for Abhd12–/– groups vs corresponding Abhd12+/+ groups . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 006 Unsupervised hierarchical clustering of the ABPP-MudPIT data highlighted distinct groups of serine hydrolases that were enriched in specific brain cell types ( Figure 1C; Figure 1—figure supplement 1B ) . These enrichment profiles included some enzymes with established neuronal expression , such as acetylcholine esterase ( ACHE ) ( Hicks et al . , 2011 ) and fatty acid amide hydrolase ( FAAH ) ( Egertova et al . , 1998 , Tsou et al . , 1998 ) , which catalyze the hydrolysis of the neurotransmitter acetylcholine and eCB anandamide , respectively ( Figure 1C; Figure 1—figure supplement 1B ) , as well as many less well characterized enzymes with strong preferential activity in astrocytes ( e . g . , ABHD2 , ABHD16A ) and microglia ( e . g . DAGLβ , AOAH ) ( Figure 1C; Figure 1—figure supplement 1B ) . Brain cell type-enriched serine hydrolases spanned a wide range of spectral count values ( Figure 1—figure supplement 1C ) , indicating that restricted cellular distribution was not correlated with the abundance of detected enzyme activities . Comparison of our ABPP data with recently reported mRNA expression results generated by RNA-seq ( Zhang et al . , 2014 ) revealed a good overall correlation across mouse brain cell types ( Pearson’s correlation r = 0 . 33 , ****p < 0 . 0001 ) , particularly for those serine hydrolases displaying prominent cell type-specific enrichment ( Pearson’s correlation r = 0 . 54 , **p < 0 . 01; Figure 1D ) . We also observed , however , a few notable exceptions of enzymes where transcript and activity were poorly correlated , including enzymes that showed i ) near-equivalent RNA-seq signals across all three brain cell types , but predominant activity signals in one cell type ( e . g . , ABHD16A ) ; ii ) strong RNA-seq and activity signals in distinct cell types ( e . g . , HTRA1 ) , and iii ) strong RNA-seq and activity signals that were anti-correlated across all three cell types ( e . g . MGLL ) ( Figure 1E ) . These data , taken together , demonstrate that ABPP furnishes an in-depth portrait of serine hydrolase activities across neurons , astrocytes , and microglia that complements and bolsters the information afforded by expression-based analytical platforms ( e . g . RNA-seq ) and illuminates candidate enzymes that contribute to the metabolic specialization of major brain cell types . Prominent among serine hydrolases displaying strong brain cell-enriched profiles were multiple biosynthetic and degradative enzymes of the eCB 2-arachidonoyl glycerol ( 2-AG ) ( Mechoulam et al . , 1995 , Sugiura et al . , 1995 ) . This arachidonic acid ( AA ) -derived retrograde lipid messenger broadly modulates synaptic function , neurophysiology , and behavior ( Fowler et al . , 2005 , Pacher et al . , 2006 , Kano et al . , 2009 ) by binding to G protein-coupled cannabinoid receptors CB1 ( CB1R ) and CB2 ( CB2R ) , which are also targets of the primary psychoactive component of marijuana Δ9-tetrahydrocannabinol ( Mechoulam and Hanus , 2000 ) . Enzymatic hydrolysis of 2-AG also serves as a major source of AA for the synthesis of brain eicosanoids that regulate neuroinflammatory processes ( Nomura et al . , 2011 , Chen et al . , 2012 , Piro et al . , 2012 ) . Previous studies have demonstrated that diacylglycerol lipase-α ( DAGLα ) ( Bisogno et al . , 2003 ) and monoacylglycerol lipase ( MGLL ) ( Karlsson et al . , 1997 ) are the principal 2-AG biosynthetic and degradative enzymes , respectively , in the brain ( Long et al . , 2009 , Chanda et al . , 2010 , Gao et al . , 2010 , Schlosburg et al . , 2010 , Tanimura et al . , 2010 ) . Our ABPP data revealed that both of these enzymes are preferentially expressed in neurons compared to astrocytes or microglia ( Figure 1C ) . On the other hand , microglia expressed high levels of the alternative 2-AG biosynthetic and degradative enzymes DAGLβ ( Bisogno et al . , 2003 ) and ABHD12 ( Blankman et al . , 2007 ) , respectively ( Figures 1C ) . Previous RNA-seq measurements also supported the enriched expression of DAGLβ and ABHD12 in microglia compared to neurons or astrocytes ( Zhang et al . , 2014 ) . We verified strong DAGLβ activity in microglia using the tailored activity-based probe HT-01 ( Hsu et al . , 2012 ) ( Figure 1—figure supplement 2A ) and confirmed that ABHD12 substantially contributed to 2-AG hydrolysis in microglia ( and astrocytes ) , but not neurons using cells isolated from Abhd12–/– mice ( Blankman et al . , 2013 ) ( Figure 1—figure supplement 2B ) . These data pointed to a potential anatomical demarcation of enzyme function within the eCB system where individual brain cell types would use distinct sets of enzymes to control 2-AG metabolism and signaling , including crosstalk with other lipid networks . We next set out to test this premise by evaluating the contributions of ABHD12 and DAGLβ to regulating 2-AG metabolism in microglia . ABHD12 exhibits 2-AG hydrolase activity in vitro ( Blankman et al . , 2007 , Navia-Paldanius et al . , 2012 ) , but Abhd12–/– mice have normal brain 2-AG content ( Blankman et al . , 2013 ) , and therefore the potential physiological relevance of ABHD12 as a regulator of 2-AG metabolism remains unclear . Having found that microglia from Abhd12–/– mice show ~40% reductions in 2-AG hydrolytic activity ( Figure 1—figure supplement 2B ) , we next examined 2-AG content in these cells . Abhd12–/– microglia displayed a modest , but significant increase in cellular 2-AG compared to Abhd12+/+ microglia ( Figure 2A ) ; in contrast , 2-AG content in Abhd12–/– neurons and astrocytes was unaltered and instead elevated in Mgll–/– neurons and astrocytes ( Figure 2A ) . Mgll–/– microglia also showed a strong elevation in cellular 2-AG that was much greater in magnitude compared to that observed in Abhd12–/– microglia ( Figure 2A ) . Notably , however , secreted 2-AG content was much higher from Abhd12–/– microglia and unaltered from Mgll–/– microglia compared to wild type microglia ( Figure 2B ) . These data suggested that ABHD12 primarily regulates extracellular pools of 2-AG in microglia ( Figure 2B ) , which is consistent with the enzyme’s luminal/extracellular orientation ( Blankman et al . , 2007 ) . Secreted 2-AG was unaltered in Abhd12–/– neurons and astrocytes ( but elevated in Mgll–/– neurons ) ( Figure 2B ) , indicating that ABHD12 specifically regulates extracellular 2-AG content in microglia . Consistent with this extracellular pool of 2-AG being competent for signaling , we observed tonically increased CBR-dependent ERK1/2 phosphorylation in Abhd12–/– microglia ( Figure 2C and D ) . 10 . 7554/eLife . 12345 . 007Figure 2 . ABHD12 modulates 2-AG content and CBR activation , but not eCB-eicosanoid crosstalk in microglia . ( A , B ) Intracellular ( A ) and secreted ( B ) 2-AG content in cultured wild type , Abhd12–/– and Mgll–/– neurons , astrocytes , and microglia . Data represent average values ± SEM; N = 4-6 per genotype . **p < 0 . 01 for Abhd12–/– or Mgll–/– cells vs corresponding wild-type cells . ( C , D ) Immunoblot ( C ) and quantification of band density ( D ) for Erk1/2 phosphorylation , which increases upon 2-AG-mediated activation of CBRs ( Walter et al . , 2003 ) , in wild-type and Abhd12–/– microglia . Abhd12–/– microglia display hyperphosphorylation of Erk1/2 ( lanes 1-4 ) , which can be partially mimicked in Abhd12+/+ cells by addition of exogenous 2-AG ( 1 μM , 15 min; lanes 9–12 ) and fully blocked by pre-treatment with CBR1 and CBR2 antagonists ( rimonabant and AM630 , respectively ) ( 5 µM each , 1 hr pre-treatment; lanes 5-8 and 13-16 ) . N = 4 per genotype and treatment . Immunoblot shows two representative replicate experiments for each group . Data represent average values ± SEM; N = 4 per genotype and treatment; *p < 0 . 05 and **p < 0 . 01 for vehicle-treated Abhd12–/– group and 2-AG-treated Abhd12+/+ group vs vehicle-treated Abhd12+/+ group; #p < 0 . 05 for CBR antagonist-treated Abhd12–/– group vs vehicle-treated Abhd12–/– group; &p < 0 . 05 and &&p < 0 . 01 for 2-AG-treated Abhd12–/– or 2-AG/CBR antagonist-treated Abhd12+/+ groups vs 2-AG-treated Abhd12+/+ group; ^p < 0 . 05 for 2-AG/CBR antagonist-treated Abhd12–/– group vs 2-AG-treated Abhd12–/– group . ( E , G ) 2-AG , AA , and PGE2/D2 content basally and following exposure to the pro-inflammatory agent LPS ( 100 ng/mL for 4 hr ) in Abhd12–/– and Abhd12+/+ microglia ( E ) or in wild-type microglia pre-treated with the MGLL inhibitor KML29 ( 250 nM , 3 hr prior to LPS; G ) . Data represent average values ± SEM; N = 5 per genotype and treatment . *p < 0 . 05 and **p < 0 . 01 for vehicle-treated Abhd12–/– microglia and LPS-treated Abhd12+/+ microglia groups ( E ) or KML29-treated wild-type microglia and LPS-treated wild type microglia groups ( G ) vs vehicle-treated Abhd12+/+ microglia or vehicle-treated wild-type microglia groups , respectively; #p < 0 . 05 and ##p < 0 . 01 for LPS-treated Abhd12–/– microglia ( E ) or KML29-treated , LPS-treated wild-type microglia ( G ) groups vs vehicle-treated Abhd12+/+ microglia or KML29-treated wild-type microglia groups , respectively; &&p < 0 . 01 for KML29-treated , LPS-treated wild-type microglia group vs LPS-treated wild-type microglia group . ( F , H ) Cytokine production basally and following exposure to LPS ( 100 ng/mL for 4 hr ) as measured by ELISA in Abhd12–/– and Abhd12+/+ microglia ( F ) or in wild-type microglia pre-treated with KML29 ( 250 nM , 3 hr prior to LPS; H ) . Data represent average values ± SEM; N = 5 per genotype and treatment; **p < 0 . 01 for LPS-treated Abhd12+/+ microglia ( F ) or LPS-treated wild-type microglia ( H ) groups vs vehicle-treated Abhd12+/+ microglia or vehicle-treated wild-type microglia groups , respectively; ##p < 0 . 01 for LPS-treated Abhd12–/– microglia ( F ) or KML29-treated , LPS-treated wild-type microglia ( G ) groups vs vehicle-treated Abhd12–/– or KML29-treated wild-type microglia groups , respectively; &&p < 0 . 01 for KML29-treated , LPS-treated wild-type microglia group vs LPS-treated wild-type microglia group . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 007 We next evaluated whether ABHD12-mediated 2-AG hydrolysis provided an AA source for prostaglandin production in microglia . We found that Abhd12+/+ and Abhd12–/– microglia exhibited similar concentrations of prostaglandins , either basally or after stimulation with bacteria-derived lipopolysaccharide ( LPS; 100 ng/mL , 4 hr ) ( Figure 2E ) and did not show differences in LPS-induced cytokine production ( Figure 2F ) . In contrast , treatment of microglia with the MGLL inhibitor KML29 ( Chang et al . , 2012 ) significantly lowered basal and LPS-induced prostaglandins ( Figure 2G ) , as well as LPS-induced cytokine production ( Figure 2H ) . These data are consistent with recent reports showing that MGLL-inactivated microglia are impaired in LPS-induced prostaglandin production and inflammatory responses ( Pihlaja et al . , 2015 , Viader et al . , 2015 ) . Our results thus indicate that ABHD12 is a major regulator of secreted 2-AG and CBR activation in microglia , while MGLL is the primary intracellular 2-AG hydrolase in these cells . Considering that the prostaglandin biosynthetic enzymes PTGS1 and PTGS2 are also localized intracellularly ( Soberman and Christmas , 2003 ) , it is perhaps not surprising that eCB-eicosanoid crosstalk was also found to be governed by MGLL rather than ABHD12 in microglia . We next examined the respective roles of DAGLα and DAGLβ in modulating eCB signaling pathways in brain cell types . Neurons and astrocytes from Dagla–/– mice displayed much lower 2-AG content compared to neurons and astrocytes from wild type or Daglb–/– mice , indicating that DAGLα is responsible for the bulk of 2-AG biosynthesis in these brain cell types ( Figure 3A and B ) . In contrast , microglia from Daglb–/– , but not Dagla–/– mice showed substantial reductions in 2-AG ( Figure 3C ) . A similar profile was observed for AA , which was lower in neurons and astrocytes from Dagla–/– mice and microglia from Daglb–/– mice ( Figure 3A–C ) , as well as for prostaglandins , with the exception that we did not detect prostaglandins in neurons ( Figure 3A–C ) . We also observed comparable lipid changes in wild-type microglia treated with two DAGL inhibitors ( KT109 and KT172 ) , but not with a structurally related inactive control compound ( KT195 ) ( Hsu et al . , 2012 ) ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 12345 . 008Figure 3 . DAGLβ is a principal 2-AG biosynthetic enzyme in microglia . ( A–C ) 2-AG , AA , and PGE2/D2 content in primary neurons ( A ) , astrocytes ( B ) , and microglia ( C ) derived from Dagla–/– , Daglb–/– and corresponding wild-type mice . Data represent average values ± SEM; N = 5–8 per genotype . **p < 0 . 01 for Dagla–/– or Daglb–/– groups vs corresponding wild-type groups . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 00810 . 7554/eLife . 12345 . 009Figure 3—figure supplement 1 . Pharmacological blockade of DAGLβ modulates eCB/eicosanoid metabolism in microglia . ( A–C ) 2-AG ( A ) , AA ( B ) , and PGE2/D2 ( C ) content in wild-type microglia treated with vehicle , the DAGLβ inhibitors KT109 ( 500 nM , 4 hr ) or KT172 ( 500 nM , 4 hr ) , or the inactive control compound ( KT195 , 500 nM , 4 hr ) . Data represent average values ± SEM; N = 5 per treatment . **p < 0 . 01 for inhibitor-treated groups vs corresponding vehicle-treated groups . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 009 We next assessed the lipid profiles of Daglb+/+ and –/– microglia following LPS stimulation ( 100 ng/mL , for 4 hr ) and found that DAGLβ inactivation resulted in the coordinated decreases in intracellular 2-AG , AA , and prostaglandins ( Figure 4A–C ) . No changes in these lipids were observed in Dagla–/– microglia . Daglb–/– microglia , but not Dagla–/– microglia also displayed marked reductions in LPS-induced inflammatory cytokine production compared to wild type microglia ( Figure 4D–J ) . Similar lipid and cytokine reductions were observed in LPS-stimulated wild type microglia that were pre-treated ( 3 hr ) with KT172 , but not KT195 ( Figure 4K–U ) . 10 . 7554/eLife . 12345 . 010Figure 4 . Genetic or pharmacologic inactivation of DAGLβ impairs LPS-induced eCB-eicosanoid crosstalk and cytokine production in microglia . ( A–C , F–H , K–M , Q–S ) 2-AG , AA , and PGE2/D2 content basally and following exposure to LPS ( 100 ng/mL for 4 hr ) in Daglb–/– ( A-C ) , Dagla–/– ( F–H ) and corresponding wild-type microglia , or in wild-type microglia treated with the DAGL inhibitor-treated KT172 ( 500 nM , 3 hr prior to LPS; K–M ) or an inactive control compound ( KT195 , 500 nM , 3 hr prior to LPS; Q–S ) . Data represent average values ± SEM; N = 5 per genotype and treatment . ( D , E , I , J , N , O , T , U ) Cytokine production basally and following exposure to LPS ( 100 ng/mL for 4 hr ) as measured by ELISA in Daglb–/– ( D , E ) , Dagla–/– ( I , J ) and corresponding wild-type microglia , or in wild-type microglia treated with KT172 ( 500 nM , 3 hr prior to LPS; N , O ) or a KT195 ( 500 nM , 3 hr prior to LPS; T , U ) . Data represent average values ± SEM; N = 5 per genotype and treatment . For A–U , **p < 0 . 01 for Daglb–/– or KT172-treated wild-type microglia or LPS-treated Daglb+/+ , Dagla+/+ , and wild-type microglia groups vs corresponding vehicle-treated wild-type microglia groups; #p < 0 . 05 and ##p < 0 . 01 for LPS-treated Dagl–/– or inhibitor ( KT172 or KT195 ) -treated groups vs corresponding vehicle-treated Dagl–/– or inhibitor ( KT172 or KT195 ) -treated groups; &&p < 0 . 01 for LPS-treated Daglb–/– or LPS-treated , KT172-treated groups vs corresponding LPS-treated Daglb+/+ or LPS-treated , vehicle-treated groups . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 010 These results , taken together , demonstrate that distinct DAGL enzymes are responsible for controlling eCB and eicosanoid metabolism in astrocytes and neurons ( DAGLα ) versus microglia ( DAGLβ ) , and that blocking DAGLβ genetically or pharmacologically attenuates LPS-induced inflammatory responses in microglia . Dagla–/– mice have been shown to display dramatic reductions in brain 2-AG ( ~80-90% ) , demonstrating that DAGLα is responsible for the bulk of 2-AG biosynthesis in the CNS ( Gao et al . , 2010 , Tanimura et al . , 2010 ) . The contribution of DAGLβ is less clear , as Daglb–/– mice have been reported to show either no changes ( Hsu et al . , 2012 , Tanimura et al . , 2012 ) or modest decreases ( Gao et al . , 2010 ) in brain 2-AG . Our data generally matched these previous studies , as we observed severely depleted brain 2-AG ( ~90% , Figure 5A ) in Dagla–/– mice , which was accompanied by a substantial increase in 1-stearoyl-2-arachidonoyl-sn-glycerol ( SAG ) ( Figure 5B ) , a main precursor of 2-AG ( Shonesy et al . , 2014 , Ogasawara et al . , 2015 ) , and corresponding reductions in brain AA ( Figure 5C ) and prostaglandins ( Figure 5D ) . In contrast , brain 2-AG , SAG , and AA content were unaltered in Daglb–/– mice ( Figure 5A–C ) , although these animals did exhibit a significant reduction in brain PGE2 ( Figure 5D ) , perhaps owing to the modulation of 2-AG and related lipids selectively in microglia , which , despite representing only 10% of all brain cells ( Aguzzi et al . , 2013 ) , display enriched expression of the prostaglandin synthase PTGS1 ( or cyclooxygenase-1 ( COX-1 ) ) ( Hoozemans et al . , 2001 , Font-Nieves et al . , 2012 , Zhang et al . , 2014 ) . 10 . 7554/eLife . 12345 . 011Figure 5 . DAGLβ modulates discrete brain 2-AG pools in vivo . ( A–D ) 2-AG ( A ) , SAG ( B ) , AA ( C ) , and PGE2 ( D ) content in brain tissue from Dagla–/– , Daglb–/– mice and their corresponding wild-type littermates . Data represent average values ± SEM; N = 6-8 mice per genotype . **p < 0 . 01 for Dagla–/– or Daglb–/– mice vs corresponding wild-type mice . ( E ) Gel-based ABPP analysis of brain membrane proteomes from Dagla–/– mice following treatment with the MGLL inhibitor MJN110 ( i . p . 10 mg/kg , 3 hr ) the DAGL inhibitor DO34 ( i . p . 100 mg/kg , 2 hr ) , or sequentially with both inhibitors ( DO34 , 2 hr; followed by MJN110 , 3 hr ) using a DAGL-directed probe ( HT-01 , left ) ( Hsu et al . , 2012 ) or the broad-spectrum serine hydrolase probe FP-Rh ( E , right ) ( Patricelli et al . , 2001 ) . Fluorescent gels are shown in grayscale and 2-AG metabolic enzymes are labeled . ( F ) Brain 2-AG content from Dagla–/– mice following treatment with the MGLL inhibitor MJN110 ( i . p . 10 mg/kg , 3 hr ) or sequentially with the DAGL inhibitor DO34 ( i . p . 100 mg/kg , 2 hr ) and MJN110 ( DO34 , 2 hr; followed by MJN110 , 3 hr ) . Data represent average values ± SEM; N = 5 per genotype and treatment . **p<0 . 01 for Dagla–/– groups vs vehicle-treated Dagla+/+ group; ##p < 0 . 01 for MJN110-treated for Dagla–/– group vs vehicle-treated Dagla–/– group; &&p < 0 . 01 for DO34 + MJN110-treated Dagla–/– group vs MJN100-treated Dagla–/– group . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 011 To more directly address whether a DAGLβ-dependent pool of 2-AG exists in the brain , we evaluated the inhibition of DAGLβ in the context of Dagla–/– mice . We first found that the pharmacological blockade of 2-AG hydrolysis with the MGLL inhibitor MJN110 ( Niphakis et al . , 2013 ) promoted a significant recovery of brain 2-AG content in Dagla–/– mice ( Figure 5E and F ) , consistent with a previous report ( Shonesy et al . , 2014 ) . We then tested inhibition of DAGLβ using a recently described CNS-active DAGL inhibitor DO34 ( Ogasawara et al . , 2015 ) and found that a high dose of this compound ( 100 mg/kg , i . p . ) near-completely blocked DAGLβ activity in brain ( Figure 5E ) and suppressed the MJN110-induced rescue of brain 2-AG in Dagla–/– mice ( Figure 5F ) . These data thus indicate that DAGLβ can contribute to bulk 2-AG content in the CNS , in particular , in the absence of DAGLα . MGLL blockade has been shown to lower prostaglandins and attenuate neuroinflammation in the brain ( Nomura et al . , 2011 , Chen et al . , 2012 , Piro et al . , 2012 ) . These previous findings , combined with the dramatic reductions in brain 2-AG , AA , and prostaglandins observed in Dagla–/–mice ( Figure 5A–D ) and the substantial suppression of inflammatory responses in Daglb–/–microglia ( Figure 4 ) , motivated us to test whether microglial activation , a common feature of diverse nervous system diseases ( Aguzzi et al . , 2013 ) , was regulated by DAGL enzymes in vivo . Wild type and DAGL-disrupted mice were treated with LPS ( 1 mg/kg , once per day for four days ) and then sacrificed and their lipid profiles and microglial activation state analyzed . This LPS treatment paradigm resulted in a significant increase in brain PGE2 , but other measured lipids ( 2-AG , SAG , AA , PGD2 ) were unaltered . As anticipated , Dagla–/– mice showed substantial reductions in brain 2-AG , AA , and prostaglandins , along with elevations in SAG , in both control and LPS-treated mice ( Figure 6A ) . Along with these lipids changes , Dagla–/– mice also displayed attenuated LPS-induced microglial activation in the brain ( Figure 6B and Figure 6—figure supplement 1 ) . In contrast , Daglb–/– mice challenged with LPS did not show changes in the measured brain lipids , including PGE2 , for which an LPS-stimulated increase appeared to override the basal reduction observed in Daglb–/– mice ( Figure 6C ) . Yet , LPS-treated Daglb–/– mice still exhibited reduced activation of brain microglia ( Figure 6D and Figure 6—figure supplement 1 ) . No obvious differences in basal cell morphology of microglia were observed between wild-type and Dagla–/– or Daglb–/– mice ( Figure 6 B , D , and Figure 6—figure supplement 1 ) . LPS-induced anapyrexia , a profound reduction in core body temperature mediated largely by central inflammatory processes ( Romanovsky , 2004 ) , which we have found to be sensitive to DAGLa disruption ( Ogasawara et al . , 2015 ) , was also greatly attenuated in Daglb–/– mice ( Figure 6E ) . These results indicate that disruption of either DAGLα or DAGLβ curbs neuroinflammatory responses in the brain , with blockade of the latter enzyme doing so while avoiding bulk changes in eCBs and eicosanoids . 10 . 7554/eLife . 12345 . 012Figure 6 . DAGL blockade attenuates microglial neuroinflammatory responses in vivo . ( A ) SAG , 2-AG , AA , PGE2 , PGD2 and AEA content in brain tissue from Dagla–/– mice and wild-type littermates basally and following exposure to LPS ( i . p . , 1 mg/kg once per day for 4 days ) . Data represent average values ± SEM; N = 5-6 per genotype and treatment . *p < 0 . 05 and **p < 0 . 01 for vehicle-treated Dagla–/– group or LPS-treated Dagla+/+ group vs vehicle-treated Dagla+/+ group; &&p < 0 . 01 LPS-treated Dagla–/– group vs LPS-treated Dagla+/+ group . ( B ) Representative pictures and quantification of microglial activation assessed by Iba-1 staining ( a microglia marker that becomes upregulated during inflammatory activation of these cells ) in hippocampal regions from Dagla–/– mice and wild-type littermates basally and following exposure to LPS ( i . p . , 1 mg/kg once per day for 4 days ) . Scale bar , 50 µm . Data represent average values ± SEM; N = 6 per genotype and treatment . **p < 0 . 01 for LPS-treated Dagla+/+ group vs vehicle-treated Dagla+/+ group; ##p < 0 . 01 for LPS-treated Dagla–/– group vs vehicle-treated Dagla–/– group; &&p < 0 . 01 for LPS-treated Dagla–/– group vs LPS-treated Dagla+/+ group . ( C ) SAG , 2-AG , AA , PGE2 , PGD2 and AEA content in brain tissue from Daglb–/– mice and wild-type littermates basally and following exposure to LPS ( i . p . , 1 mg/kg once per day for 4 days ) . Data represent average values ± SEM; N = 5-6 per genotype and treatment . *p < 0 . 05 for vehicle-treated Daglb–/– group or LPS-treated Daglb+/+ group vs vehicle-treated Daglb+/+ group; ##p < 0 . 01 for LPS-treated Daglb–/– group vs vehicle-treated Daglb–/– group . ( D ) Representative pictures and quantification of microglial activation assessed by Iba-1 staining ( a microglia marker that becomes upregulated during inflammatory activation of these cells ) in hippocampal regions from Daglb–/– mice and wild-type littermates basally and following exposure to LPS ( i . p . , 1 mg/kg once per day for 4 days ) . Scale bar , 50 µm . Data represent average values ± SEM; N = 6 per genotype and treatment . **p < 0 . 01 for LPS-treated Daglb+/+ group vs vehicle-treated Daglb+/+ group; ##p < 0 . 01 for LPS-treated Daglb–/– group vs vehicle-treated Daglb–/– group; &&p < 0 . 01 for LPS-treated Daglb–/– group vs LPS-treated Daglb+/+ group . ( E ) Time course of body temperature changes in Daglb+/+ and Daglb–/– mice following LPS ( 10 mg/kg , i . p . ) -induced anapyrexia . Data represent average values ± SEM; N = 6 mice per genotype and treatment . *p < 0 . 05 for Daglb+/+ + Veh vs Daglb+/+ + LPS groups; #p < 0 . 05 for Daglb–/– + Veh vs Daglb–/– + LPS groups; &p < 0 . 05 for Daglb+/+ + LPS vs Daglb–/– + LPS groups . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 01210 . 7554/eLife . 12345 . 013Figure 6—figure supplement 1 . DAGL inactivation attenuates LPS-induced microglial activation . Representative pictures of microglial activation assessed by Iba-1 staining ( a microglia marker that becomes upregulated during inflammatory activation of these cells ) in hippocampal regions from Dagla–/– , Daglb–/– and wild-type mice basally and following exposure to LPS ( i . p . , 1 mg/kg once per day for 4 days ) . Scale bar , 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 12345 . 013
Understanding the metabolic and signaling pathways expressed by major brain cell types should help to illuminate key features of nervous system physiology and disease ( Kreft et al . , 2012 , Kettenmann et al . , 2013 , Morrison et al . , 2013 ) . Gene expression profiling ( Cahoy et al . , 2008 , Doyle et al . , 2008 , Beutner et al . , 2013 , Chiu et al . , 2013 , Zhang et al . , 2014 ) and proteomic studies ( Yu et al . , 2004 , Yang et al . , 2005 , Glanzer et al . , 2007 , Liao et al . , 2008 , Bell-Temin et al . , 2012 , Sharma et al . , 2015 ) performed on isolated brain cells have provided valuable insights into the unique molecular makeup of different brain cell types , but the assessment of the functional state of proteins , and not just their abundance , is also needed to dissect the cellular distribution of signaling and metabolic pathways that regulate nervous system ( patho ) physiology . We have herein used ABPP to provide a global portrait of serine hydrolase activities in neurons , astrocytes , and microglia , and , through doing so , identified cell type-specific regulation of key chemical transmission pathways , in particular eCBs and eicosanoids , that exert differential impact on global versus inflammation-induced processes in the mouse brain . Previous studies have found that a substantial portion ( ~40% ) of the variance in protein content can be explained by differences in mRNA expression ( Schwanhausser et al . , 2011 ) . Consistent with this conclusion , we found that our activity-based proteomic data displayed good overall correlation with previously reported mRNA expression data obtained by deep RNA-sequencing ( Zhang et al . , 2014 ) . A few notable exceptions of enzymes where transcript and activity were un- or even anti-correlated were , however , identified , and it would be interesting , in future studies , to understand the basis for these differences . One anti-correlated protein – HTRA1 – for instance , is known to be regulated at the level of secretion from cells ( Skorko-Glonek et al . , 2013 ) , and it is possible that individual brain cell types differentially export this enzyme to create discordant measures of mRNA and cellular protein content . Other posttranslational events , such as phosphorylation , could regulate serine hydrolase activity and/or stability . Some of the discrepancies between our chemical proteomic results and previously reported mRNA expression datasets ( Zhang et al . , 2014 ) may also arise from differences in methodology , given that our studies used primary brain cell cultures instead of acutely isolated neurons and glia . Cells in culture are known to only partially model cells from an in vivo environment , and future ABPP studies using freshly isolated neurons and glia should help to further enrich our understanding of serine hydrolase activities in specific brain cell types . We were struck by the remarkable number of serine hydrolases that showed differential activity across neurons , astrocytes , and microglia ( Figure 1C ) . This observation supports the emerging recognition that neurons and glia are endowed with complementary enzymatic pathways to meet the overall metabolic needs of the nervous system ( Belanger et al . , 2011 ) . Among the most prominently compartmentalized serine hydrolases were those that regulate the eCB 2-AG . This lipid messenger , known to be biosynthesized by two different enzymes—DAGLα and DAGLβ ( Bisogno et al . , 2003 ) —and with the potential to be degraded by several enzymes—MGLL , ABHD6 , ABHD12 , and FAAH ( Blankman et al . , 2007 , Marrs et al . , 2011 ) — is an important regulator of inter-neuronal and neuron-glia interactions ( Navarrete and Araque , 2008 , 2010 , Martin et al . , 2015 , Viader et al . , 2015 ) . 2-AG signaling also impacts diverse neurophysiological processes , including pain , mood , and neuroinflammation ( Blankman and Cravatt , 2013 , Murataeva et al . , 2014 ) . Considering the widespread functions of 2-AG , the identification of mechanisms that allow for the selective modulation of specific aspects of this lipid transmitter’s action in the nervous system stands as an important objective . Toward this end , our work establishes that individual cell types in the brain control 2-AG metabolism and signaling by different enzymatic pathways . We found that microglia , in particular , are equipped with a distinct complement of 2-AG metabolic enzymes – ABHD12 and DAGLβ – the inactivation of which perturbs eCB signaling and crosstalk with eicosanoids specifically in this cell type . ABHD12 plays additional important roles in lipid metabolism in the brain , where the enzyme serves as a major lyso-PS hydrolase ( Blankman et al . , 2013 ) that is part of an immunomodulatory pathway ( Kamat et al . , 2015 ) linked to the human neurological disease PHARC ( Fiskerstrand et al . , 2010 ) . It will thus be important , in the future , to dissect in what physiological and disease contexts ABHD12 acts as a 2-AG versus lyso-PS hydrolase . Note also that , while our results indicate a role for ABHD12 in the modulation of microglia endocannabinoid signaling , our experiments do not discriminate the precise identity of the receptors involved . Whether CB1R , CB2R , or even other closely related receptors , mediate the observed ABHD12-regulated 2-AG effect on microglial ERK phosphorylation can be addressed in future studies using lower concentrations of selective CB1R and CB2R antagonists or microglia derived from knockout mice lacking these receptors . DAGLα has been shown to be the principal source for 2-AG production in the nervous system , and disruption of this enzyme impairs many forms of CB1R-dependent synaptic plasticity ( Gao et al . , 2010 , Tanimura et al . , 2010 ) . Our results extend these findings in an important way by demonstrating that DAGLα also regulates neuroinflammatory responses in the brain , in particular , LPS-induced microglial activation . The reductions in LPS-stimulated neuroinflammation observed in Dagla–/– mice may reflect , at least in part , crosstalk between eCB and eicosanoid pathways , since basal and LPS-induced prostaglandin synthesis were decreased in these animals . By also identifying DAGLβ as a principal 2-AG synthase in microglia , our work establishes a previously underappreciated and specialized role for this enzyme in 2-AG biosynthesis in the nervous system ( Reisenberg et al . , 2012 , Murataeva et al . , 2014 ) . Our results also do not rule out the presence of DAGLβ-regulated pools of 2-AG in other brain cell types , as might be supported by the significant reductions in this eCB observed in brain tissue from Dagla–/– mice treated with MGLL and DAGL inhibitors ( Figure 4F ) . Also potentially consistent with a role for DAGLβ in regulating bulk brain 2-AG in the absence of DAGLα , we have been unable to generate viable Dagla/b double-knockout mice . Nonetheless , most of the major forms of eCB-dependent synaptic plasticity have been shown to be regulated by DAGLα rather than DAGLβ ( Gao et al . , 2010 , Tanimura et al . , 2010 ) , underscoring the broad role that the former enzyme plays in eCB signaling throughout the nervous system . That Daglb deletion attenuates LPS-induced microglial activation in vivo without altering bulk eCB or eicosanoid content in the brain suggests modulation of restricted pools of 2-AG can impact neuroinflammatory processes while avoiding global effects on eCB signaling . This finding adds to a growing body of work implicating glial proteins and pathways as potential targets for nervous system disease that may have fewer neuron-related adverse side effects ( Barres , 2008 , Ilieva et al . , 2009 , Milligan and Watkins , 2009 , Sheridan , 2009 ) . The remarkable cell type-specific compartmentalization of 2-AG metabolic enzymes , and more broadly of serine hydrolases , may thus prove to be fertile ground for the identification of targets that can safely reverse or slow the course of diverse pathologies of the nervous system .
FP-rhodamine , FP-biotin , HT-01 , KML29 , MJN110 , KT172 , KT195 , and DO34 were synthesized in-house as previously described ( Patricelli et al . , 2001 , Chang et al . , 2012 , Hsu et al . , 2012 , Niphakis et al . , 2013 , Ogasawara et al . , 2015 ) . All deuterated lipid standards and substrates were purchased from Cayman Chemicals . Lipopolysaccharide from E . coli was purchased from Sigma ( 0111:B4 ) . The primary cell culture protocols used in this study were approved by the Scripps Research Institute Institutional Animal Care and Use Committee ( IACUC #09-0041-03 ) . Cortico-hippocampal neurons were prepared from embryonic day 18 mice from transgenic or wild-type mice as needed . Cortices/hippocampi were dissected , freed of meninges , and dissociated by incubation in Papain/DNase for 20 min at 37°C followed by trituration . Dissociated cortico-hippocampal neurons were then washed with DMEM media supplemented with 10% FBS and 2 mM glutamine , prior to seeding them onto poly-D-lysine coated 10 cm culture dishes in neurobasal medium containing 2% B27 supplement , 2 mM glutamine , and 5 μM 5-fluoro-2′-deoxyuridine at a density of 8 × 106 cells/dish . A third of the media was exchanged twice per week . Neurons were harvested for proteome isolation after 16 days in vitro in the presence of antimitotics , thus ensuring high neuronal enrichment as confirmed by western blot using neuron- , astrocyte- and microglia-specific markers ( Tuj1 , GFAP and Iba-1 , respectively; data not shown ) . Microglia were derived from mixed glial cultures prepared from postnatal day 2-3 mouse forebrains from transgenic or wild-type mice as needed . Briefly , forebrains were dissected , stripped of meninges , and digested in papain/DNAse ( 20 min at 37°C ) followed by 0 . 25% trypsin ( 15 min at 37°C ) and trituration . Dissociated cells were then cultured for 10 days in poly-D-lysine coated T75 tissue culture flasks in DMEM media supplemented with 10% FBS , 2 mM glutamine , and 5 ng/mL of granulocyte macrophage-colony stimulating factor . After establishment of the astrocyte monolayer , the flasks were shaken for 2 hr at 180 rpm to obtain the loosely attached microglia . Microglia were subsequently plated onto 10 cm dishes at a density of 2-3 × 106 cells/dish in Macrophage-SFM media ( Gibco ) supplemented with 1% FBS and 0 . 5 ng/mL of granulocyte macrophage-colony stimulating factor . The purity of these microglia cultures was >99% as determined by immunohistochemical quantification of the proportion of Iba-1 positive cells ( total cell number determined by DAPI nuclear staining ) in six different fields from two separate cultures . Cells were allowed to sit for at least 72 hr prior to harvesting them for proteome isolation . Following isolation of microglia , established mixed glial cultures were treated with 8 μM cytosine-arabinoside for 3–5 days to kill actively dividing cells ( e . g . microglia , fibroblast ) , and generate an astrocyte monolayer with >85% purity , as determined by immunohistochemical quantification of the proportion of GFAP positive cells ( total cell number determined by DAPI nuclear staining ) in six different fields from two separate cultures . These astrocytes were subsequently plated onto poly-D-lysine coated 10 cm dishes in DMEM media supplemented with 10% FBS and 2 mM glutamine , and allowed to become confluent . Upon reaching confluence , astrocytes were harvested for subsequent proteome isolation . All animal experiments were carried out in compliance with institutional animal protocols ( IACUC #09-0041-03 ) , and mice were housed on a normal 6AM/6PM light/dark phase with ad libitum access to water and food . All mice used in this study were generated by heterozygous matings . Mgll–/– ( Viader et al . , 2015 ) , Abhd12–/– ( Blankman et al . , 2013 ) , Dagla–/– ( Hsu et al . , 2012 ) , and Daglb–/– mice ( Hsu et al . , 2012 ) as well as their wild-type littermates were all in a homogeneous C57Bl/6 background . PCR genotyping of genomic tail DNA was performed using the following primers: Mgll 5’- cacctgtctttggagctc ccacc-3’ , 5’-cctttctttagggagagtccactgaatgtg-3’ , and 5’-ggcagcactgacaaatgtgtctgag-3’ ( 415-bp product in wild-type mice , 598-bp product in mice with the Mgll null allele ) ; Abhd12 5’- cagtgctggcctgtcagtcg-3’ , 5’-ggtgcccagtgaatggcc-3’ , and 5’-taaagcgcatgctccagactgcc-3’ ( 500-bp product in wild-type mice , 300-bp product in mice with the Abhd12 null allele ) ; Dagla 5’-tgagattggtatcaagacctttg-3’ , 5’-ccttgctcctgccgagaaagtatcc-3’ , and 5’- gaagaacaggtaaccaggaccat-3’ ( 300-bp product in wild-type mice , 600-bp product in mice with the Dagla null allele ) ; Daglb 5’-aaggaggcaaagacagcaaagtgc-3’ , 5’-tatcctaggtgcagacagattgtgc-3’ , and 5’-aaatggcgttacttaagctagcttgc-3’ ( 390-bp product in wild-type mice , 195-bp product in mice with the Daglb null allele ) . 8-week-old mice were anaesthetized with isofluorane , killed by cervical dislocation , and tissues harvested , immediately flash frozen in liquid nitrogen and kept frozen at −80°C until use . For preparation of proteomes , tissues were homogenized in cold PBS using a bullet blender ( Next Advance , Inc . ) as per the manufacturer’s instructions , or dounced homogenized in an isotonic buffer consisting of 20 mM Hepes , 2 mM DTT , 0 . 25 M sucrose , and 1 mM MgCl2 , pH 7 . 2 for improved DAGL visualization . Lysed proteomes were then subjected to a low-speed spin ( 1400 × g , 5 min ) to remove debris , and ultracentrifugation ( 100 , 000 × g , 45 min ) to separate membrane and cytosolic fractions . The supernatant was removed and saved as the soluble proteome , while the pellet was washed and resuspended in cold PBS by sonication or in isotonic resuspension solution ( 20 mM Hepes , 2 mM DTT ) by pipetting and saved as the membrane proteome . Total protein concentration for each proteome was determined using a Bio-Rad Dc Protein Assay kit , and proteomes were kept at -80°C until further use . Cultured cell proteomes were prepared in an analogous manner after scraping cells from culture dishes with 1 mL of cold PBS . 50 μL of 1 mg/mL cell or tissue membrane proteome were incubated with the broad-spectrum serine hydrolase probe FP-rhodamine ( 1 μM final concentration ) for 30 min at room temperature or with the DAGL-directed probe HT-01 ( 1 μM final concentration ) for 30 min at 37°C . When necessary , proteomes were pre-treated with inhibitors for 1 hr at 37°C prior to addition of ABPP probes . Probe labeling was terminated by quenching with 4x SDS/PAGE loading buffer , and labeled proteome samples were separated by SDS-PAGE ( 10% [wt/vol] acrylamide ) and visualized by in-gel fluorescence scanning using a flatbed fluorescence scanner ( Hitachi FMBio IIe ) . Gel fluorescence is shown in grayscale , and optical density of the signals was determined using Image-J . For the ABPP-MudPIT samples , whole cell proteomes ( 500 μg/mL in 500 μl of PBS ) were labeled with FP-biotin ( 10 μM ) for 2 hr at room temperature . After labeling , the proteomes were precipitated using cold MeOH ( 2 mL ) , CHCl3 ( 0 . 5 mL ) and PBS ( 1 mL ) . The resulting cloudy mixture was vortexed and then centrifuged ( 5000 × g , 15 min , 4°C ) . The organic and aqueous layers were aspirated leaving a protein disc which had formed between phases . The protein disc was washed with cold 1:1 MeOH:CHCl3 ( 3 × 1 mL ) while intact and then probe sonicated in cold 4:1 MeOH:CHCl3 ( 2 . 5 mL ) . Insoluble proteins were pelleted via centrifugation ( 5000 × g , 15 min , 4°C ) and the supernatant was removed . The remaining pellet was redissolved in 500 µL of 6 M urea in PBS , reduced using Tris ( 2-carboxyethyl ) phosphine ( TCEP , 10 mM ) for 30 min at 37°C , and then alkylated using iodoacetamide ( 40 mM ) for 30 min at 25°C in the dark . The biotinylated proteins were then enriched by addition of 10% ( w/v ) SDS ( 140 µL ) , followed by 5 . 5 mL of PBS and 100 µL of PBS pre-washed avidin-agarose beads ( 100 µL , 1 , Sigma-Aldrich ) and shaking at 25°C for 1 . 5 hr . Beads were pelleted by centrifugation ( 1000 × g , 2 min ) and sequentially washed with 0 . 25% SDS ( 3 × 10 mL ) , DPBS ( 3 × 10 mL ) and ddH2O ( 3 × 10 mL ) . The beads were transferred to a Protein LoBind tube ( Eppendorf ) and on-bead digestion was performed using sequence-grade trypsin ( 2 μg; Promega ) in 2 M urea in PBS with 2 mM CaCl2 for 12–14 hr at 37°C ( 200 μl ) . Peptides obtained from this procedure were acidified using formic acid ( 5% ) and stored at –80°C before analysis . MS analysis was performed using a LTQ following previously described protocols ( Jessani et al . , 2005 ) . Peptides were eluted using a five-step multidimensional LC/MS protocol in which increasing concentrations of ammonium acetate are injected followed by a gradient of increasing acetonitrile , as previously described ( Washburn et al . , 2001 ) . For all samples , data were collected in data-dependent acquisition mode over a range from 400–1800 m/z . Each full scan was followed by up to 7 fragmentation events . Dynamic exclusion was enabled ( repeat count of 1 , exclusion duration of 20 s ) for all experiments . The data were searched using the ProLuCID algorithm against a mouse reverse-concatenated , non-redundant ( gene-centric ) FASTA database that was assembled from the Uniprot database . ProLuCID searches allowed for variable oxidation of methionine ( +15 . 9949 m/z ) , static modification of cysteine residues ( +57 . 0215 m/z; iodoacetamide alkylation ) and accepted only half- or fully-tryptic peptides . The resulting peptide spectral matches were filtered using DTASelect ( version 2 . 0 . 47 ) , and only half-tryptic or fully tryptic peptides were accepted for identification . Peptides were restricted to a specified false positive rate of <1% . Only proteins detected in at least two out of four replicates for any one cell type preparation , with ≥ 5 spectral counts were used for downstream analysis . Hierarchical clustering analysis shown in Figure 1C was performed using CIMminer . In vitro 2-AG hydrolysis was determined by liquid chromatography-mass spectrometry ( LC-MS ) monitoring of the generation of AA product . Briefly , 10 μg of primary cell proteomes from Abhd12+/+ and Abhd12–/– mice ( N = 5 per genotype ) were treated with vehicle or 250 nM KML29 prior to incubation with 100 μM 2-AG in PBS ( 100 μL final volume ) for 10 min . Reactions were quenched by addition of 300 μL of 2:1 ( vol/vol ) CHCl3:MeOH doped with 0 . 5 nmol of AA-d8 lipid standard , vortexed to mix , and spun at 2000 rpm × 5 min to separate phases . The bottom organic phase was extracted and 20 μL were injected onto an Agilent 6460 triple quadrupole ( QQQ ) MS . Chromatography was performed on a 50 × 4 . 60 mm 5-μm Gemini C18 column ( Phenomenex ) coupled to a guard column ( Gemini; C18; 4 × 3 . 0 mm; Phenomenex SecurityGuard cartridge ) . The LC method consisted of 0 . 5 mL/min of 100% buffer A [95:5 ( vol/vol ) H2O:MeOH plus 0 . 1% ( vol/vol ) ammonium hydroxide] for 1 . 5 min , 0 . 5 mL/min linear gradient to 100% buffer B [65:35:5 ( vol/vol ) iPrOH:MeOH:H2O plus 0 . 1% ( vol/vol ) ammonium hydroxide] over 5 min , 0 . 5 mL/min 100% buffer B for 3 min , and equilibration with 0 . 5 mL/min 100% buffer A for 1 min ( 10 . 5 min total run time ) . MS analysis was performed in negative scanning mode with an electrospray ionization ( ESI ) source using the precursor to product ion transition and collision energies for AA ( 303 , 303 , 0 , negative ) and AA-d8 ( 311 , 311 , 0 , negative ) . Dwell times for each lipid were set to 100 ms , and the following MS parameters were used: capillary voltage = 3 . 5 kV , drying gas temperature = 350°C , drying gas flow rate = 9 L/min , and nebulizer pressure = 50 psi , sheath gas temperature = 375°C , and sheath gas flow rate = 12 L/min . AA release was quantified by measuring the area under the peak in comparison with the AA-d8 internal standard and correcting for non-enzymatically formed AA present in heat inactivated ( 10 min at 90°C ) control reactions , and relative 2-AG hydrolytic activity for conditional Abhd12–/– proteomes was calculated by comparing to activity of wild-type proteomes . Metabolite levels were quantified by multiple reaction monitoring ( MRM ) of each lipid species using and Agilent 6460 QQQ instrument . For cultured cells , lipids were extracted from each cell type by scraping cells from dish in 1 . 1 mL of cold PBS , of which 100 μL were saved to measure protein concentration for normalization , prior to re-scraping with 1 mL of cold MeOH . The 2 mL of cell lysate were then transferred to 2 mL of CHCl3 with 20 μL of formic acid and the following lipid standards: 1 nmol of 2-AG-d5 , 0 . 5 nmol of AEA-d4 , 1 nmol of AA-d8 , and 0 . 5 nmol of PGE2-d9 . After centrifugation at 2000 rpm for 5 min , the organic bottom fraction was carefully collected and dried under a nitrogen stream . Lipids were resolubilized in 140 μL of 1:1 ( v/v ) CHCl3:MeOH , 20 μL of which were injected for mass spectrometry analysis . When necessary , 5 mL of media were also extracted in an analogous manner by combining with 5 mL of MeOH and 10 mL of CHCl3 . For LPS studies , lipids were similarly extracted from 2-3 × 106 microglia pre-treated with inhibitors as needed for 3 hr prior to addition of 100 ng/mL of LPS for 4 hr . For brains , 2 month-old Dagla–/– , Daglb–/– , or corresponding wild-type littermates were injected with LPS in PBS ( 1 mg/kg , once a day for 4 days , i . p . ) or with PBS vehicle alone . On day 5 , mice were sacrificed by cervical dislocation , and their brains rapidly removed and immediately flash frozen in liquid nitrogen . One brain hemisphere was then weighted and dounce homogenized in 8 ml of cold 2:1:1 ( v/v/v ) CHCl3:MeOH:PBS doped with the following lipid standards: 1 nmol of 2-AG-d5 , 0 . 5 nmol of AEA-d4 , 1 nmol of AA-d8 , 1 nmol of SAG-d8 , and 0 . 5 nmol of PGE2-d9 . Note that brains were not allowed to thaw prior to contact with organic solvents . Brain homogenates were vortexed and centrifuged at 2000 rpm for 5 min . The organic bottom fraction was carefully collected , and the remaining solution was re-extracted by adding another 2 mL of CHCl3 with 20 μL of formic . The organic fractions from both extractions were combined , and dried under a nitrogen stream . Lipids were resolubilized in 200 μL of 1:1 ( v/v ) CHCl3:MeOH , 10 μL of which were injected for mass spectrometry analysis . LC separation of lipid metabolites was performed on a 50 × 4 . 60 mm 5 μm Gemini C18 column ( Phenomenex ) coupled to a guard column ( Gemini; C18; 4 × 3 . 0 mm; Phenomenex SecurityGuard cartridge ) . Mobile phase A consisted of 95:5 ( vol/vol ) H2O:MeOH and mobile phase B consisted of 60:35:5 ( vol/vol ) iPrOH:MeOH:H2O , with 0 . 1% formic acid or ammonium hydroxide added to both mobile phases to assist in ion formation in positive and negative ionization modes , respectively . For the targeted detection of MAGs , NAEs , and DAGs in positive mode , each run started at 0 . 1 mL/min of 100% A . At 5 min , the solvent was immediately changed to 60% B with a flow rate of 0 . 4 mL/min and increased linearly to 100% B over 15 min . 100% B was allowed to flow at 0 . 5 mL/min for an additional 8 min prior to equilibrating for 3 min with 100% A at 0 . 5 mL/min . For targeted detection of eicosanoids and fatty acids in negative mode , each run started at 0 . 1 mL/min of 100% A . At 3 min , the flow rate was increased to 0 . 4 mL/min with a linear increase of solvent B to 100% over 17 min . 100% B was allowed to flow at 0 . 5 mL/min for 7 min prior to equilibrating for 3 min with 100% A at 0 . 5 mL/min . MS analysis was performed in either positive or negative scanning mode with an electrospray ionization ( ESI ) source , and the following parameters were used to measure the indicated metabolites by SRM ( precursor ion , product ion , collision energy in V , polarity ) : 2-AG ( 379 , 287 , 8 , positive ) ; 2-AG-d5 ( 384 , 287 , 8 , positive ) ; AEA ( 348 , 62 , 11 , positive ) ; AEA-d4 ( 352 , 66 , 11 , positive ) ; SAG ( 662 . 5 , 341 , 40 , positive ) ; SAG-d8 ( 670 . 5 , 341 , 40 , positive ) ; AA ( 303 , 303 , 0 , negative ) ; AA-d8 ( 311 , 311 , 0 , negative ) ; PGE2/D2 ( 351 , 271 , 4 , negative ) ; PGE2/D2-d9 ( 360 . 5 , 280 , 4 , negative ) . Dwell times for each lipid were set to 100 ms , and the following MS parameters were used: capillary voltage = 3 . 5 kV , drying gas temperature = 350°C , drying gas flow rate = 9 L/min , and nebulizer pressure= 50 psi , sheath gas temperature = 375°C , and sheath gas flow rate = 12 L/min . Metabolite species were quantified by measuring the area under the peak in comparison with the appropriate unnatural internal standard and normalizing for wet tissue weight or protein concentration for cultured cells . Cytokines were measured using a DuoSet ELISA kit ( R&D systems ) as per the manufacturer's instructions . For cultured cells , 100 μL of media were used . For western blotting , samples ( 30 μg protein ) were separated by SDS-PAGE [10% ( wt/vol ) acrylamide] then transferred to a nitrocellulose membrane . The membrane was then blocked in 5% milk in 0 . 5% TBS-Tween and incubated overnight with P-Erk1/2 ( 1:1000 , rabbit , Cell Signaling ) and Erk 1/2 ( 1:1000 , rabbit , cell signaling ) primary antibodies . Following incubation with IRdye680 secondary antibody ( 1:5000 , Licor Biosciences ) , membranes were visualized with a Licor Odyssey CLx near-infrared imager . Mice were deeply anesthetized using isofluorane and perfused with PBS followed by 4% ( wt/vol ) paraformaldehyde . Brains were then carefully dissected , postfixed in 4% paraformaldehyde overnight , cryoprotected in 30% ( wt/vol ) sucrose , and rapidly frozen on dry ice . Free floating coronal sections 40 μm in thickness were cut on a Leica CM1850 cryostat . Frozen , free-floating sections were blocked with Bloxall ( Vector labs , 10 min ) and 0 . 2% Triton-X100 and 3% goat serum in PBS ( blocking solution , 1 hr ) prior to O/N incubation at 4°C with a rabbit anti-ionized calcium-binding adaptor molecule ( Iba ) 1 primary antibody ( 1:500 dilution in blocking solution; Wako ) . Sections were then incubated with secondary antibody ( anti-rabbit biotin; Vector Laboratories; 1:300 dilution of 1 . 5 mg/mL stock ) in 0 . 5% ( wt/vol ) BSA in 0 . 1 M PB for 1 hr at room temperature and developed with ABC Elite Vectastain ( Vector Laboratories ) . After staining , sections were mounted in Permount and imaged using a Leica SCN400 whole slide scanner . Images were processed using ImageJ and Photoshop using global adjustments in brightness and contrast . Iba1-positive microglia were quantified in matching sections from at least 5 mice per genotype using ImageJ by first converting images to be counted to 8 bit greyscale and applying global adjustments in brightness and contrast in the same way to the whole set of images being analyzed . The area for all cells being quantified was then highlighted using ImageJ’s Threshold macro , setting the same range of pixel intensities for the entire set of images being processed . Total microglia area in the resulting binary image was determined using ImageJ’s Particle Analyzer macro . Area for all particles larger than 10 pixels was included in the quantification . Core body temperature was measured by radiotelemetry as previously described ( Sanchez-Alavez et al . , 2015 ) . Briefly , mice were anesthetized with isofluorane ( induction 3–5% , maintenance 1–1 . 5% ) and surgically implanted with radiotelemetry devices ( TA-F20 , Data Sciences , St . Paul , MN ) into the peritoneal cavity for core body temperature ( CBT ) and locomotor activity ( LA ) evaluation . Following surgical implantation and appropriate wound closure , the animals were allowed to recover for 2 weeks and then subjected to telemetry recordings . Mice were individually housed in a Plexiglas cage in a room maintained at 25 ± 0 . 5°C . The cages were positioned onto the receiver plates ( RPC-1; Data Sciences ) and radio signal reporting CBT and LA information from the implanted transmitter was recorded continuously with a fully automated data acquisition system ( Dataquest ART , Data Sciences , St . Paul , MN ) . Access to food and water was ad libitum and the light:dark cycle was of 12 hr:12 hr . Anapyrexia was induced by intraperitoneal injection of Bacterial lipopolysaccharides ( LPS ) ( 0127:B8 , Sigma , St . Louis , MO ) at a dose of 10 mg/kg in saline . Data are shown as the mean ± SEM . A Student’s t test ( unpaired , two-tailed ) was used to determine differences between two groups . Data that included more than two groups were analyzed by one-way ANOVA , with post hoc Sidak’s multiple comparisons test . Correlations were determined by two-tailed Perason’s correlation . All statistical analyses were conducted using Excel or GraphPad Prism version 6 , and a p value <0 . 05 was considered significant throughout . No statistical methods were used to predetermine sample sizes , which were comparable to those in previous publications ( Viader et al . , 2015 ) and based on previous knowledge of the variability associated with the different experiments and the expected differences . | The brain is made up of many types of cells . These include the neurons that transmit messages throughout the nervous system , and microglia , which act as the first line of the brain’s immune defense . The activity of both neurons and microglia can be influenced by molecules called endocannabinoids that bind to proteins on the cells’ surface . For example , endocannabinoids affect how a neuron responds to messages sent to it from a neighbouring neuron , and help microglia to regulate the inflammation of brain tissue . Enzymes called serine hydrolases play important roles in several different signaling processes in the brain , including those involving endocannabinoids . Viader et al . have now studied the activities of these enzymes – including two called DAGLα and DAGLβ – in the mouse brain using a technique called activity-based protein profiling . This revealed that DAGLα plays an important role in controlling how neurons respond to endocannabinoids , while DAGLβ performs the equivalent role in microglia . When Viader et al . shut down DAGLβ activity , this only affected endocannabinoid signaling in microglia . This also had the effect of reducing inflammation in the brain , without affecting how endocannabinoids signal in neurons . These results suggest that inhibitors of DAGLβ could offer a way to suppress inflammation in the brain , which may contribute to neuropsychiatric and neurodegenerative diseases , while preserving the normal pathways that neurons use to communicate with one another . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"neuroscience"
] | 2016 | A chemical proteomic atlas of brain serine hydrolases identifies cell type-specific pathways regulating neuroinflammation |
The precise assembly of inner ear hair cell stereocilia into rows of increasing height is critical for mechanotransduction and the sense of hearing . Yet , how the lengths of actin-based stereocilia are regulated remains poorly understood . Mutations of the molecular motor myosin 15 stunt stereocilia growth and cause deafness . We found that hair cells express two isoforms of myosin 15 that differ by inclusion of an 133-kDa N-terminal domain , and that these isoforms can selectively traffic to different stereocilia rows . Using an isoform-specific knockout mouse , we show that hair cells expressing only the small isoform remarkably develop normal stereocilia bundles . However , a critical subset of stereocilia with active mechanotransducer channels subsequently retracts . The larger isoform with the 133-kDa N-terminal domain traffics to these specialized stereocilia and prevents disassembly of their actin core . Our results show that myosin 15 isoforms can navigate between functionally distinct classes of stereocilia , and are independently required to assemble and then maintain the intricate hair bundle architecture .
The inner ear detects sound using mechanosensitive hair bundles that project from the apical surface of cochlear hair cells ( reviewed in Schwander et al . , 2010 ) . Each hair bundle is composed of actin-based stereocilia that are arranged into rows of increasing height to create a staircase-like architecture; a feature evolutionarily conserved in vertebrates ( Manley , 2000 ) . Extracellular tip-link filaments connect the tip of each stereocilium to the lateral shaft of its taller neighbor ( Pickles et al . , 1984 ) . Tensioning of these links during hair bundle deflection initiates mechanoelectrical transduction ( MET ) by gating mechanotransducer ion channels that are located at the tips of shorter rows of stereocilia ( Beurg et al . , 2009 ) . Formation of the mature staircase architecture involves a complex program of differential elongation and thickening of the individual stereocilia actin cores ( Tilney et al . , 1988 ) . How this process is developmentally specified with sub-nanometer tolerances and subsequently maintained throughout adult life is poorly understood . Unconventional myosin 15 ( encoded by Myo15 ) is an actin-based molecular motor and a key regulator of hair bundle development . In mice with mutations of the myosin 15 motor or tail domain , shaker 2 ( Myo15sh2/sh2 ) or shaker 2-J ( Myo15sh2-J/sh2-J ) , respectively , hair bundles are short and this results in profound hearing loss and vestibular dysfunction ( Probst et al . , 1998; Anderson et al . , 2000 ) . Mutations in the human ortholog MYO15A similarly cause non-syndromic autosomal recessive deafness , DFNB3 ( Friedman et al . , 1995; Wang et al . , 1998 ) . Myosin 15 localizes to the tips of stereocilia ( Belyantseva et al . , 2003; Rzadzinska et al . , 2004; Belyantseva et al . , 2005 ) , a site of barbed-end actin filament growth and turnover ( Schneider et al . , 2002; Zhang et al . , 2012; Drummond et al . , 2015; Narayanan et al . , 2015 ) . Myosin 15 is required for stereocilia elongation and traffics molecules to the stereocilia tips , including whirlin , a cytoskeletal scaffolding protein ( Mburu et al . , 2003; Belyantseva et al . , 2005; Delprat et al . , 2005 ) , and epidermal growth factor receptor pathway substrate 8 ( Eps8 ) which has actin binding , bundling and barbed-end capping activity ( Disanza et al . , 2004; Manor et al . , 2011 ) . The loss of either whirlin or Eps8 recapitulates the short hair bundle phenotype and deafness of Myo15sh2/sh2 mice ( Mburu et al . , 2003; Belyantseva et al . , 2005; Manor et al . , 2011; Zampini et al . , 2011 ) , consistent with these proteins forming a complex with myosin 15 to promote stereocilia growth . Alternative splicing creates two major protein isoforms from the 66 exon Myo15 gene ( Liang et al . , 1999 ) . Isoform 2 transcripts skip exon 2 and use a translation start codon in exon 3 to encode a 262 kDa protein including the motor ATPase domain and C-terminal MyTH4 , SH3 and FERM moieties ( Figure 1A ) . Isoform 1 transcripts include exon 2 that contains an alternate translation start codon and adds a 133-kDa N-terminal extension in frame with the motor domain and tail ( Figure 1A ) . Both isoform transcripts are detected in inner ear cDNAs ( Belyantseva et al . , 2003 ) and are expressed by hair cells ( Liang et al . , 1999; Anderson et al . , 2000; Caberlotto et al . , 2011 ) . Overexpression of isoform 2 can induce stereocilia elongation in Myo15sh2/sh2 cochleae in vitro ( Belyantseva et al . , 2005 ) , but the function of isoform 1 remains unknown . However , given that mutations in exon 2 are associated with DFNB3 deafness in humans , it strongly suggests that isoform 1 also has a critical role in the auditory system ( Nal et al . , 2007; Cengiz et al . , 2010; Bashir et al . , 2012; Fattahi et al . , 2012 ) . 10 . 7554/eLife . 08627 . 003Figure 1 . A mutation targeting isoform 1 causes deafness in Myo15ΔN/ΔN mice . ( A ) Two protein isoforms are generated from alternatively spliced transcripts of Myo15 . Transcripts incorporating exon 2 encode isoform 1 ( 396 kDa; Genbank: NM_010862 . 2 ) , while exclusion of this exon produces isoform 2 ( 262 kDa; Genbank: NM_182698 . 2 ) . Both isoforms have identical motor and tail domains , including a PDZ ligand , SH3 , MyTH4 and FERM moieties . The mutant Myo15 alleles used in this study are shown along with antibody epitopes . ( B ) Auditory brainstem response ( ABR ) thresholds at 4 , 20 and 48 kHz for Myo15+/+ , Myo15+/ΔN and Myo15ΔN/ΔN mice at 6 weeks of age . Data are mean ± SD ( n = 3–6 animals per group ) . ( C ) ABR thresholds at 20 kHz , measured from 2 , 4 and 6 weeks old Myo15+/+ , Myo15+/ΔN and Myo15ΔN/ΔN mice . Data are mean ± SD ( n = 3–6 animals per group ) . ( D ) Distortion product otoacoustic emission ( DPOAE ) levels ( 2F1-F2 ) at 12 kHz in Myo15+/+ and Myo15ΔN/ΔN mice at 6 weeks . Data are mean ± SD ( n = 3–4 animals per group ) . ( E–G ) Relative expression of Myo15 isoforms in wild-type cochleae measured with RT-qPCR at ages indicated . Probes target exon junction 13–14 , detecting the motor domain common to both isoforms 1 and 2 ( E ) ; exon junction 2–3 , detecting isoform 1 ( F ) ; exon junction 1–3 , detecting isoform 2 ( G ) . Relative expression 2^ ( -∆∆CT ) for each Myo15 transcript was normalized first to the housekeeping gene ( Tbp ) and then to the respective isoform expression at P0 . The total expression of both isoforms remains stable ( E ) , however there is a transition from isoform 2 to isoform 1 , which becomes the dominant mRNA species by P21 ( see Figure 1—figure supplement 1D ) . Data are mean ± SD ( n = 3–5 biological replicates per condition ) . Asterisks indicate significance: n . s . , p > 0 . 05; ** , p < 0 . 01; **** , p < 0 . 0001 ( ANOVA with Tukey's multiple comparison test ) . ( H ) Identical qPCR probes were used to assay Myo15 expression in Myo15+/ΔN and Myo15ΔN/ΔN cochleae at P0 . Relative expression 2^ ( -∆∆CT ) values were normalized to Tbp and then to expression in heterozygous Myo15+/∆N samples . Data are mean ± SD ( n = 3–4 biological replicates per condition ) . n . s . , p > 0 . 05 ( t-test of independent variables ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 00310 . 7554/eLife . 08627 . 004Figure 1—figure supplement 1 . Generation of a mouse model for human p . E1105X DFNB3 deafness . ( A ) Genomic locus of the wild-type mouse Myo15 allele and targeting strategy to introduce the p . E1086X mutation into exon 2 by homologous recombination . Southern blot probes ( black bars ) and loxP sites ( yellow triangles ) are shown . A single loxP site remains after excision of the PGK-Neo cassette by cre recombinase . Restriction sites unique to the recombined allele are highlighted in red . Homozygous mutants of both sexes were fertile and obtained in normal Mendelian ratios expected for an autosomal recessive allele ( data not shown ) . ( B ) Validation of homologous recombination by Southern blot analysis . ES cell genomic DNA was digested with Asp718 ( left panel ) or NsiI ( right panel ) and hybridized with DNA probes to the 3′ arm ( left panel ) or 5′ arm ( right panel ) . The restriction fragment sizes expected for a correctly recombined locus are shown ( arrows ) . ( C ) PCR confirmation of germ-line transmission from mouse tail genomic DNA . Introduction of the p . E1086X mutation creates a new MseI endonuclease restriction site that is used for genotyping . Amplicons were digested with MseI and restriction fragments analyzed by gel electrophoresis . The expected restriction fragments for different genotypes are shown ( arrows ) . ( D ) Quantitative PCR ( qPCR ) analysis of Myo15 isoform expression in mouse cochleae . The qPCR data for isoforms 1 and 2 ( previously presented in Figure 1 ) , were reanalyzed as fold changes ( 2^-ΔCT ) relative to the housekeeping gene Tbp . Note: the large errors bars at day 10 are due to the use of semi-logarithmic axes . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 004 In this study , we show that both isoforms of myosin 15 are expressed in auditory hair cells at different developmental stages , and that they traffic to distinct sub-cellular locations within the stereocilia hair bundle . To understand their individual functions , we engineered a mouse model carrying a nonsense mutation in exon 2 that ablates isoform 1 , leaving isoform 2 intact . We found that hair bundles depend critically upon two phases of myosin 15 activity throughout their lifetime; isoform 2 orchestrates development of the staircase architecture , while a postnatal transition to isoform 1 is required to maintain the shorter , mechanosensitive stereocilia rows .
To selectively disrupt myosin 15 isoform 1 without altering the coding sequence of isoform 2 , we used homologous recombination in mouse embryonic stem ( ES ) cells to knock-in a p . E1086X nonsense mutation into exon 2 ( Figure 1A and Figure 1—figure supplement 1 ) , mimicking the p . E1105X DFNB3 allele that causes hearing loss in humans ( Nal et al . , 2007 ) . Because isoform 2 transcripts skip exon 2 , we hypothesized that the p . E1086X mutation ( referred to as Myo15ΔN ) would specifically disrupt isoform 1 . Auditory brainstem response ( ABR ) testing was used to measure the hearing thresholds of 6 week old Myo15ΔN/ΔN mice and their littermates at 4 , 20 and 48 kHz ( Figure 1B ) . Myo15ΔN/ΔN mice were profoundly deaf at all frequencies tested ( Figure 1B ) . However , around the onset of hearing at 2 weeks , Myo15ΔN/ΔN mice did respond to loud sounds of 75 dB of sound pressure level ( dB SPL ) at 20 kHz , the most sensitive frequency range of mouse hearing ( Figure 1C ) . However , by 4 and 6 weeks of age ABR thresholds at 20 kHz exceeded 100 dB SPL in Myo15ΔN/ΔN mice , indicating a rapid progression to profound deafness ( Figure 1C ) . In control Myo15+/+ and Myo15+/ΔN littermates , the average thresholds measured at 20 kHz were between 27 and 40 dB SPL and did not change significantly with age ( Figure 1C ) . Distortion product otoacoustic emissions ( DPOAEs ) were collected to evaluate active cochlear amplification by outer hair cells ( OHCs ) . There was a complete absence of DPOAEs in Myo15ΔN/ΔN mice at 2 weeks ( data not shown ) and 6 weeks of age ( Figure 1D ) , where in contrast Myo15+/+ littermates had normal DPOAEs at 2 weeks ( data not shown ) and 6 weeks ( Figure 1D ) . We conclude that cochlear amplification is disrupted in Myo15ΔN/ΔN mice , and that this contributes to the profound deafness evident by 4 weeks of age . Myo15sh2/sh2 and Myo15sh2-J/sh2-J mice have profound congenital deafness and vestibular dysfunction typified by persistent head bobbing and circling behaviors ( Probst et al . , 1998; Anderson et al . , 2000 ) . The residual hearing function in 2 week old Myo15ΔN/ΔN mice hinted that the hearing impairment caused by the disruption of exon 2 had a different underlying pathophysiology to these previously reported Myo15 mouse models . This interpretation was supported by functional studies of the vestibular system . Myo15ΔN/ΔN mice lacked circling behavior and performed normally in swimming tests ( data not shown ) . Thus , Myo15ΔN/ΔN mice differ from the Myo15sh2/sh2 and Myo15sh2-J/sh2-J models in the onset and severity of deafness and also by the absence of an overt vestibular pathology . We hypothesized that the varying severity of sensory pathology evident in Myo15 mutant mice was due to isoforms 1 and 2 being differentially targeted . The Myo15sh2 and Myo15sh2-J alleles contain a mutation in the motor domain ( p . C1779Y ) or a genomic deletion of exons encoding the C-terminal FERM domain and PDZ ligand , respectively ( Probst et al . , 1998; Anderson et al . , 2000 ) ; both are predicted to ablate isoforms 1 and 2 ( Figure 1A ) . In contrast , the p . E1086X mutation in exon 2 is expected to selectively target isoform 1 , but preserve isoform 2 ( Figure 1A ) . To investigate the expression of Myo15 isoforms during cochlea development and maturation , we used quantitative PCR ( qPCR ) to measure their relative abundance in total RNA at P0 through P21 . Total RNA was used as Myo15 is primarily expressed in hair cells ( Anderson et al . , 2000; Caberlotto et al . , 2011; Shen et al . , 2015 ) . Probe sets were designed to detect mRNA splicing between exons 2 and 3 ( representing isoform 1 ) , exons 1 and 3 ( isoform 2 ) , and of exons 13 and 14 within the ATPase motor domain , which independently reported the total pool of isoform 1 and 2 transcripts . The total pool of Myo15 transcripts remained stable between P0 and P21 ( Figure 1E ) , but this concealed an underlying switch in exon 2 splicing . Transcripts encoding isoform 1 became progressively more abundant , increasing ∼nine-fold from P0 through to P21 ( Figure 1F and Figure 1—figure supplement 1D ) . Conversely , the amount of isoform 2 sharply decreased ∼21-fold between P0 and P6 and remained stable thereafter ( Figure 1G and Figure 1—figure supplement 1D ) . We used the same qPCR assays to examine the effect of the Myo15∆N allele upon isoform-specific transcript levels . There was no statistically significant difference in the quantity of isoform 1 , isoform 2 or the total pool of transcripts in control Myo15+/ΔN vs Myo15∆N/∆N cochleae at P0 ( Figure 1H ) or at P7 ( data not shown ) . These data show that the p . E1086X mutation did not induce nonsense mediated mRNA decay or change the relative abundance of isoform transcripts . We conclude that isoform 2 mRNA was predominantly expressed during early neonatal development , but that alternative splicing progressively shifts to favor isoform 1 in postnatal cochleae . To investigate the underlying pathophysiology causing deafness in Myo15ΔN/ΔN mice , we examined cochlear hair cells using scanning electron microscopy ( SEM ) . Remarkably , unlike the short stereocilia bundles of Myo15sh2/sh2 and Myo15sh2-J/sh2-J hair cells ( Probst et al . , 1998; Anderson et al . , 2000 ) , we found that stereocilia bundles of Myo15ΔN/ΔN inner hair cells ( IHCs ) had the characteristic staircase architecture and were indistinguishable from normal hearing Myo15+/ΔN littermate controls at P4 ( Figure 2A ) . We quantified the architecture of Myo15ΔN/ΔN and control Myo15+/ΔN IHC bundles at P4 and found no differences in either the distribution of heights ( Figure 7—figure supplement 2 ) , or the aspect ratio of the second row mechanosensitive stereocilia tips ( Figure 7—figure supplement 3 ) . Tip-links were present in both Myo15ΔN/ΔN and Myo15+/ΔN IHCs ( Figure 2A , lower panels ) . The normal morphology of Myo15ΔN/ΔN IHC hair bundles at P4 was striking compared to those of age-matched Myo15sh2/sh2 mice that were abnormally short and had immature , omnidirectional links ( Figure 2A ) . Hair bundles of Myo15sh2/sh2 IHCs also had supernumerary stereocilia rows , which were not observed in Myo15ΔN/ΔN IHCs ( Figure 2A ) . Similar to IHCs , Myo15ΔN/ΔN OHCs at P4 also developed a normal hair bundle morphology that was clearly distinct from age-matched Myo15sh2/sh2 OHCs ( Figure 2B ) . Although the hair bundles of IHCs and OHCs initially developed normally at P4 , they subsequently degenerated ( Figure 6 ) , and this likely contributes to the deafness observed in Myo15ΔN/ΔN mice at P14 and older ( Figure 1C ) . 10 . 7554/eLife . 08627 . 005Figure 2 . Myo15ΔN/ΔN hair cells initially develop normal stereocilia bundles . ( A , B ) Scanning electron microscopy ( SEM ) of inner ( A ) and outer ( B ) hair cells from Myo15ΔN/ΔN , normal hearing Myo15+/ΔN littermates and Myo15sh2/sh2 cochleae at P4 . Highlighted regions of IHC bundles ( white boxes ) are shown at higher magnification below . Myo15ΔN/ΔN stereocilia bundles develop the characteristic staircase architecture ( see Figure 7—figure supplement 2 ) that is strikingly absent from age-matched Myo15sh2/sh2 hair cells . Scale bars are 1 µm ( A , upper row and B ) and 250 nm ( A , lower row ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 005 We conclude that the p . E1086X ( ΔN ) mutation in exon 2 causes deafness through a mechanism that is fundamentally different to the previously reported Myo15 alleles that interfere with stereocilia elongation . The initial formation of normal stereocilia in p . E1086X mutants suggests that isoform 2 is sufficient for normal hair bundle development , consistent with our finding that the dominant Myo15 transcript in developing P0 hair cells encodes isoform 2 . Conversely , the postnatal switch to isoform 1 expression points to a distinct function for myosin 15 that is also essential for normal hearing . Myosin 15 has been detected at the tips of all stereocilia rows using antibodies raised against common epitopes present in both isoforms 1 and 2 ( Belyantseva et al . , 2003 ) . To determine the localization of isoform 1 specifically , we developed antibodies PB888 and PB886 to an epitope uniquely encoded by exon 2 ( Figure 1A ) . Labeling of P14 IHC bundles with PB888 ( Figure 3A ) and PB886 ( data not shown ) revealed that isoform 1 was concentrated at the tips of the shorter second and third row stereocilia . Isoform 1 was inconsistently detected at the tips of the tallest row , with ( data not shown ) , or without antigen retrieval ( Figure 3—figure supplement 1F ) , indicating that the asymmetric distribution in stereocilia rows was unlikely due to epitope masking . We used transmission immuno-gold electron microscopy ( immuno-TEM ) to study the localization of isoform 1 at higher spatial resolution . Gold particles were concentrated within the prolate tips of wild-type shorter stereocilia of P16 IHCs ( Figure 3C , D ) and were infrequently observed along the stereocilia core ( Figure 3C ) . Consistent with this , isoform 1 was not detected at the upper tip-link insertion site on the tallest row in mechanically splayed IHC bundles ( Figure 3F ) . In OHCs at P14 , isoform 1 was similarly detected at the tips of shorter stereocilia ( Figure 3B , E ) but additionally at the tips of the tallest row ( Figure 3B , B′ ) . We conclude that isoform 1 localizes to the tip density of shorter row stereocilia in close proximity to the site of MET ( Beurg et al . , 2009 ) . 10 . 7554/eLife . 08627 . 008Figure 3 . Isoform 1 targets to the tips of shorter mechanotransducing stereocilia . ( A ) PB888 antibody ( green ) detects isoform 1 at the tips of shorter row stereocilia of IHCs from normal hearing Myo15+/∆N mice . ( B ) Isoform 1 is present at the tips of all stereocilia in OHCs of normal hearing Myo15+/∆N mice at P14 . An oblique view confirms the presence of isoform 1 on the tallest row ( B′ ) . ( C–E ) TEM micrographs of immuno-gold labeled PB886 in ultrathin stereocilia sections . Isoform 1 is localized in proximity to the stereocilia tip density . Labeling was infrequently observed along the stereocilia core . ( F ) PB888 does not localize to the upper tip-link insertion point on the tallest IHC stereocilia row in intentionally splayed bundles . ( G , H ) Loss of reactivity in Myo15∆N/∆N IHCs and OHCs confirms the loss of isoform 1 protein from the hair bundle and the specificity of PB888 labeling . ( I ) PB888 does not label the stereocilia tips in Myo15sh2/sh2 hair cells which have the p . C1779Y motor domain mutation . Scale bars are 5 µm ( A , B , B′ , F , G–I ) , 500 nm ( C ) , 100 nm ( D , E ) . Immunofluorescence samples are counter-stained with rhodamine phalloidin ( red ) to reveal the stereocilia actin cytoskeleton . See also Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 00810 . 7554/eLife . 08627 . 009Figure 3—figure supplement 1 . Isoform 1 localization during cochlear development . ( A–D ) PB888 antibody labeling ( green ) of isoform 1 in cochlear whole mounts from wild-type Myo15+/+ ( A , C ) and isoform 1-null Myo15∆N/∆N ( B , D ) cochleae at P1 ( A , B ) and P7 ( C , D ) . ( E ) PB888 antibody ( green ) in adult wild-type Myo15+/+ IHCs and OHCs at P28 . The same OHCs are presented at different focal planes to show the tips of each stereocilia row . ( F ) PB888 antibody ( green ) in normal Myo15+/∆N IHCs at P14 . Two focal planes of the same IHCs are shown . Note the presence of isolated punctae ( arrows ) on the tallest row . Rhodamine phalloidin ( magenta ) labels actin filaments in all panels . Scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 009 Given the developmental regulation of Myo15 splicing , we examined the localization of isoform 1 in cochleae at different postnatal ages . Similar localization patterns of isoform 1 were observed not only at P14 ( Figure 3A , B ) but also from P7 through to P28 in both IHCs and OHCs ( Figure 3—figure supplement 1C , E ) , indicating that this was the distribution of isoform 1 in mature hair cells . Only trace PB888 labeling was detected in stereocilia of hair cells at P1 ( Figure 3—figure supplement 1A ) , consistent with the relatively low abundance of isoform 1 transcripts detected at this age by qPCR ( Figure 1F and Figure 1—figure supplement 1D ) . The apparent absence of isoform 1 in young hair bundles provides an explanation for why mutations of exon 2 do not affect stereocilia elongation or establishment of the hair bundle architecture . To test the specificity of immuno-labeling for isoform 1 , PB888 was examined in mutant Myo15ΔN/ΔN cochleae at P1 , P7 and P14 ( Figure 3G , H , Figure 3—figure supplement 1B , D ) and at P14 for PB886 ( data not shown ) . We observed no stereocilia labeling with either PB888 or PB886 , confirming antibody specificity and also that the Myo15ΔN allele resulted in the loss of isoform 1 from the hair bundle . Since the premature stop codon ( p . E1086X ) did not trigger nonsense-mediated mRNA decay in Myo15ΔN/ΔN cochleae ( Figure 1H ) , we hypothesized that a truncated isoform 1 could not localize to stereocilia without the motor and tail domains . Consistent with this , PB888 labeling was absent from stereocilia of Myo15sh2/sh2 hair cells ( Figure 3I ) , indicating that the actin-binding motor domain is critical to actively localize isoform 1 within stereocilia . Having discovered that isoform 1 was restricted to shorter row IHC stereocilia , we hypothesized that previous reports of pan-specific myosin 15 immunolabeling on the tallest row must represent isoform 2 ( Belyantseva et al . , 2003; Rzadzinska et al . , 2004 ) . There is no unique epitope to generate isoform 2 specific antibodies ( Figure 1A ) . Instead , we labeled Myo15ΔN/ΔN cochleae with an antibody raised against a common epitope present in both isoforms 1 and 2 ( PB48 , Figure 1A ) , reasoning that a pan-specific myosin 15 antibody should recognize only isoform 2 in isoform 1-null hair cells . In IHCs from normal controls at P7 and P14 , strong PB48 labeling was detected at the tips of all stereocilia rows ( Figure 4A , B ) , consistent with previous reports ( Belyantseva et al . , 2003 ) . PB48 still strongly localized to tips of the tallest stereocilia row of littermate Myo15ΔN/ΔN IHCs , however there was a selective loss from the shorter ( second and third ) stereocilia rows at both P7 ( Figure 4A ) and P14 ( Figure 4B ) . Because isoform 1 is absent from Myo15ΔN/ΔN IHC bundles ( Figure 3G ) , we infer that PB48 was detecting isoform 2 on the tallest row . We quantified the intensity of PB48 signal on the shorter , second row IHC stereocilia relative to the tallest row . In IHCs from normal controls at P7 , the intensity of PB48 labeling on the second row was 54 ± 22% ( mean ± SD ) of the first row signal ( Figure 4C ) . In Myo15ΔN/ΔN IHCs this values was significantly reduced to 9 . 7 ± 5 . 9% of the tallest row intensity ( Figure 4C ) . This indicates that the majority of myosin 15 in the second row is isoform 1 and that isoform 2 is the minor species at this location . These observations , taken together with PB888 data , support the conclusion that myosin 15 isoforms are sorted into different stereocilia rows as early as P7 in IHCs . 10 . 7554/eLife . 08627 . 010Figure 4 . Isoform 2 traffics predominantly to the tallest stereocilia row and is sufficient to target Eps8 and Whirlin . ( A , B ) PB48 antibody ( green ) was raised to an epitope common to isoforms 1 and 2 ( Figure 1A ) and labels all stereocilia rows in wild-type Myo15+/+ and Myo15+/∆N hair cells at P7 ( A ) and P14 ( B ) . In isoform 1-null Myo15∆N/∆N hair cells , PB48 is predominantly detected on the tallest stereocilia row at P7 ( A ) and at P14 ( B ) , identifying isoform 2 at these locations . ( C ) Quantification of PB48 fluorescence on the shorter second stereocilia row of IHCs at P7 normalized to the first ( tallest ) row . Data points represent individual stereocilia from wild-type ( blue , Myo15+/+ and Myo15+/∆N combined , n = 125 stereocilia , 3 animals ) or Myo15∆N/∆N ( red , n = 112 stereocilia , n = 4 animals ) IHCs at P7 , overlaid with mean ± SD . Asterisks indicate significance: * , p < 0 . 0001 ( t-test of independent variables ) . ( D , E ) Whirlin antibody localizes primarily to the tallest stereocilia row of control Myo15+/∆N hair cells at P7 ( D ) and at P14 ( E ) . The localization of whirlin remains unchanged in isoform 1-null Myo15∆N/∆N hair cells at P7 ( D ) or P14 ( E ) . ( F ) Eps8 antibody localizes primarily to the tallest stereocilia row of control Myo15+/∆N and isoform 1-null Myo15∆N/∆N IHCs at P14 . All samples are co-labeled with rhodamine phalloidin ( red ) . Scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 01010 . 7554/eLife . 08627 . 011Figure 4—figure supplement 1 . Isoform 2 is sufficient to traffic Eps8 and Whirlin within the hair bundle . ( A , B ) PB48 labeling ( green ) of isoforms 1 and 2 in whole mount Myo15+/+ ( A ) and isoform 1-null Myo15∆N/∆N ( B ) hair cells at P1 . ( C ) PB48 labeling ( green ) of P14 Myo15+/∆N ( top panel ) and Myo15∆N/∆N OHCs ( lower panel ) . ( D ) Whirlin antibody ( HL5136 ) labeling ( green ) of Myo15+/∆N ( top panel ) and Myo15∆N/∆N ( lower panel F ) OHCs at P14 . ( E and F ) Eps8 antibody labeling ( green ) in Myo15+/∆N ( upper panels ) and Myo15∆N/∆N ( lower panels ) hair cells at P7 ( E ) and P14 ( F ) . Only OHCs are shown at P14 ( F ) . ( G ) Alternate view of stereocilia bundles in Myo15∆N/∆N IHCs labeled for Eps8 at P7 . The Eps8 antibody signal on the second row was boosted ( remapped ) to reveal weak labeling at the tips . Actin filaments are labeled in all samples with rhodamine phalloidin ( red ) . Scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 011 At P1 , a strong PB48 signal was observed at the stereocilia tips of both IHCs and OHCs of isoform 1-null Myo15ΔN/ΔN mice that was comparable with the PB48 signal in control mice of the same age ( Figure 4—figure supplement 1A , B ) . This is consistent with the early expression of isoform 2 detected by qPCR ( Figure 1G ) and its sufficient role in normal development of the hair bundle ( Figure 2A , B ) . Later in development , PB48 labeling persisted at the tallest row of Myo15ΔN/ΔN IHC stereocilia at P7 and P14 ( Figure 4A , B ) but was progressively diminished in Myo15ΔN/ΔN OHC stereocilia at P7 and P14 ( Figure 4A and Figure 4—figure supplement 1C ) , as compared to the labeling in control OHCs from littermates ( Figure 4A , Figure 4—figure supplement 1C ) . These observations suggest that the overall decline of isoform 2 mRNA detected by qPCR ( Figure 1G ) may originate from a change in OHC expression . In summary , our data show a developmental transition from isoform 2 to isoform 1 by the time the mature hair bundle architecture is almost fully developed ( i . e . by P6 ) . Furthermore , the independent segregation of myosin 15 isoforms within the hair bundle provides a mechanism for why the Myo15ΔN and Myo15sh2 alleles cause strikingly different hair bundle phenotypes . Whilst both isoforms are absent from Myo15sh2/sh2 hair bundles , isoform 2 is still present in stereocilia of Myo15ΔN/ΔN hair cells and is sufficient to drive hair bundle elongation . It follows that isoform 1 has a function critical for hearing that is unrelated to stereocilia bundle development . Myosin 15 regulates hair bundle development by transporting a molecular complex containing Eps8 and whirlin ( Belyantseva et al . , 2005; Manor et al . , 2011 ) to sites of actin polymerization at the stereocilia tip ( Schneider et al . , 2002; Drummond et al . , 2015 ) . Since stereocilia elongated normally in isoform 1-null Myo15ΔN/ΔN hair cells , we investigated whether isoform 2 was sufficient to traffic whirlin and Eps8 within the developing hair bundle . In normal Myo15+/ΔN cochleae , whirlin was concentrated at the tips of the tallest row of IHC and OHC stereocilia at both P7 and P14 ( Figure 4D , E and Figure 4—figure supplement 1D ) . Weaker signal ( relative to the tallest row ) was also detected on the shorter stereocilia rows of IHCs at both ages ( Figure 4D , E ) . When examined in isoform 1-null Myo15ΔN/ΔN cochleae at both P7 and P14 , the localization of whirlin in IHCs and OHCs was indistinguishable from their Myo15+/ΔN littermates ( Figure 4D , E and Figure 4—figure supplement 1D ) . Similarly , Eps8 was detected at the tips of the tallest stereocilia row of control Myo15+/ΔN IHCs at both P7 and P14 ( Figure 4F , Figure 4—figure supplement 1E ) . Weaker Eps8 labeling was also detected on the second stereocilia row of control IHCs at P7 ( Figure 4—figure supplement 1E ) , and on the tallest row of control OHC stereocilia at both P7 and P14 ( Figure 4—figure supplement 1E , F ) . In isoform 1-null Myo15ΔN/ΔN cochleae , Eps8 labeling in IHCs and OHCs appeared unchanged from the wild-type littermates ( Figure 4F , Figure 4—figure supplement 1E , F , G ) . We conclude that isoform 2 by itself is sufficient to establish the wild-type distribution of Eps8 and whirlin in IHCs and OHCs . Whirlin and Eps8 bind to domains in the C-terminal tail of myosin 15 that are common to both isoforms 1 and 2 ( Belyantseva et al . , 2005; Delprat et al . , 2005; Manor et al . , 2011 ) . It is striking therefore that the localization of whirlin and Eps8 do not depend upon isoform 1 , despite the abundance of isoform 1 in hair bundles from P7 onwards . The apparent selectivity for isoform 2 explains how Eps8 and whirlin are primarily trafficked to the tallest stereocilia row and why the ablation of isoform 1 does not interfere with stereocilia development in Myo15ΔN/ΔN cochleae . To test whether isoform 1 is a critical component of the transduction machinery , MET currents were measured from Myo15ΔN/ΔN hair cells , which have normal staircase morphology and correctly oriented tip-links ( Figure 2 and Figure 5A , B ) . Whole cell MET current responses of young postnatal IHCs and OHCs ( P3-4 + 3–5 days in vitro ) were evoked by graded stereocilia deflections using a rigid probe ( Figure 5C , D ) . Maximal MET current amplitudes were not statistically different between mutant Myo15ΔN/ΔN ( n = 12 ) and control Myo15+/ΔN IHCs ( n = 10 ) ( Figure 5E ) . However , the responses to small bundle deflections ( 150–300 nm ) were significantly larger in mutant Myo15ΔN/ΔN IHCs , indicating an increased deflection sensitivity of the transduction apparatus in the absence of isoform 1 ( Figure 5E ) . This sensitivity depends on the mechanical stiffness of a theoretical ‘gating spring’ that is connected to the MET channel ( Howard and Hudspeth , 1987 ) . A stiffer gating spring would transmit the maximal opening force to the MET channel at a smaller bundle deflection , resulting in earlier saturation of the current-displacement relationship . This was indeed observed in Myo15ΔN/ΔN IHCs ( Figure 5E ) . Our data show that isoform 1 is not essential for MET responses in IHCs but may contribute to the stiffness of this gating spring . We did not observe a similar increase of MET sensitivity in OHCs ( Myo15ΔN/ΔN , n = 9; Myo15+/ΔN , n = 7 ) , perhaps due to the MET current degradation already present in Myo15ΔN/ΔN OHCs compared to Myo15+/ΔN controls ( Figure 5H ) . 10 . 7554/eLife . 08627 . 012Figure 5 . Isoform 1 is not required for MET but influences the deflection sensitivity of IHCs . ( A , B ) SEM images of IHC ( A ) and OHC ( B ) stereocilia bundles in Myo15+/ΔN ( left panels ) and Myo15ΔN/ΔN ( right panels ) hair cells . Higher magnification of the second row IHC stereocilia tips are shown ( inset ) . ( C , D ) Whole cell current responses ( top traces ) evoked by graded deflections of the stereocilia bundles ( bottom traces ) in IHCs ( C ) and OHCs ( D ) in Myo15+/ΔN ( left ) and Myo15ΔN/ΔN ( right ) hair cells . ( E , H ) Relationship between the peak MET current and stereocilia bundle displacement in IHCs ( E ) and OHCs ( H ) from Myo15+/ΔN ( open circles ) and Myo15ΔN/ΔN ( closed circles ) cochleae . ( F , I ) Time constants of MET adaptation in IHCs ( F ) and OHCs ( I ) for Myo15ΔN/ΔN and Myo15+/ΔN . Time constants were determined from a single exponential fit of MET responses evoked by the small bundle deflections of ∼150 nm ( see black traces in C , D ) . ( G , J ) Percent changes of the MET current 10 ms after a stimulation step ( extent of adaptation ) as a function of stimulus intensity in IHCs ( G ) and OHCs ( J ) . The same MET records contribute to all averaged data . Data are mean ± SE . Asterisks indicate statistical significance: * , p < 0 . 01; ** , p < 0 . 001; *** , p < 0 . 0001 ( t-test of independent variables ) . Holding potential was −90 mV . Age of the cells: P3-4 + 3–5 days in vitro . SEM images were obtained from cultured samples used for MET recordings . Number of cells: n = 10 ( IHCs , Myo15+/ΔN ) , n = 12 ( IHCs , Myo15ΔN/ΔN ) , n = 7 ( OHCs , Myo15+/ΔN ) , n = 9 ( OHCs , Myo15ΔN/ΔN ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 012 MET responses in both IHCs and OHCs of Myo15ΔN/ΔN mice exhibited prominent adaptation , that is , a fast decay of the MET current following stereocilia deflection ( Figure 5C , D ) . The time constant of adaptation was not affected in either IHCs or OHCs of Myo15ΔN/ΔN mice compared to Myo15+/ΔN controls ( Figure 5F , I ) . The extent of adaptation , represented as the percentage of the MET current decay during a step-like bundle deflection , was identical in Myo15ΔN/ΔN and Myo15+/ΔN IHCs ( Figure 5G ) , and larger in Myo15ΔN/ΔN OHCs as compared to Myo15+/ΔN OHCs ( Figure 5J ) . We conclude that isoform 1 is not essential for assembling the MET machinery in early postnatal development , but can contribute to the overall stiffness of the MET apparatus . Stereocilia and MET currents developed normally in Myo15ΔN/ΔN hair cells , however mutant mice still had severe hearing loss by P14 . We investigated whether isoform 1 was essential for maintenance of the hair bundle architecture . Stereocilia ultrastructure was examined in Myo15ΔN/ΔN and control Myo15+/ΔN cochlear hair cells using SEM . At P4 through P10 , the gross morphology of Myo15ΔN/ΔN hair cells was similar to control Myo15+/ΔN littermates ( Figure 6A , B , top row ) , consistent with isoform 2 being sufficient for stereocilia development . However , in older Myo15ΔN/ΔN cochleae from P32 onwards , the mechanotransducing stereocilia in the second row were evidently reduced in height compared with normal hearing Myo15+/ΔN controls . Furthermore , the third stereocilia row was almost completely resorbed by P32 in Myo15ΔN/ΔN hair cells . Remarkably , this degeneration was specific to the shorter rows that harbor active MET channels ( Beurg et al . , 2009 ) , and it did not affect the tallest rows , at least up until P50 . The same degenerative phenotype was observed in Myo15ΔN/sh2-J compound heterozygotes , which produce normal isoform 2 from the Myo15ΔN allele and a mutant isoform 1 lacking the tail domain from the Myo15sh2-J allele ( Figure 6—figure supplement 1 ) . These data indicate that a single genetic copy of isoform 2 is sufficient for the production of a normal hair bundle ( Table 1 ) and show that isoform 1 must be full-length to maintain shorter row stereocilia . 10 . 7554/eLife . 08627 . 006Figure 6 . Degeneration of mechanotransducing shorter row stereocilia in isoform 1-null hair cells . ( A ) SEM micrographs of IHC stereocilia bundles of normal Myo15+/ΔN ( left ) and Myo15ΔN/ΔN mutant ( right ) mice at different stages of postnatal development . Arrows point to examples of almost completely resorbed stereocilia . Note that the tallest stereocilia row does not thin , or shorten . ( B ) SEM micrographs of OHC stereocilia bundles show a similar degeneration pattern to IHCs . Shorter row stereocilia are retracted ( arrows ) but the tallest stereocilia row remains unaffected . All cells were located approximately at the middle of the cochlea . Scale bars are 1 µm ( A ) and 0 . 5 µm ( B ) . See also Figure 6—figure supplement 1 and Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 00610 . 7554/eLife . 08627 . 007Figure 6—figure supplement 1 . Myo15sh2-J does not complement Myo15ΔN . ( A ) Scanning electron micrographs of stereocilia hair bundles from wild-type Myo15+/+ and Myo15sh2-J/ΔN compound heterozygotes at P28 . The shorter stereocilia rows in compound heterozygotes undergo retraction ( arrows ) and resorption ( arrowheads ) similar to Myo15ΔN/ΔN hair cells ( see Figure 6 ) . Scale bars are 1 µm ( IHCs ) and 0 . 5 µm ( OHCs ) . ( B ) ABR thresholds measured at 4 , 20 and 48 kHz in Myo15+/+ and Myo15sh2-J/ΔN mice at P21 and P70 . Compound heterozygotes have profound hearing impairment at both ages tested . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 00710 . 7554/eLife . 08627 . 013Table 1 . Hair bundle phenotypes resulting from the combination of Myo15 allelesDOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 013Myo15 genotypeFunctional isoforms generated by:Allele AAllele BAllele AAllele BHair bundle phenotype++1 and 21 and 2Normal+∆N1 and 22Normal *+sh21 and 2-Normal †+sh2-J1 and 2-Normal ‡∆N∆N22Normal staircase , short rows degenerate *sh2-J∆N-2Normal staircase , short rows degenerate *sh2sh2--Short staircase , additional stereocilia rows †sh2-Jsh2-J--Short staircase , additional stereocilia rows ‡sh2sh2-J--Short staircase , additional stereocilia rows ‡*Data reported in this study . †Probst et al . ( 1998 ) . ‡Anderson et al . ( 2000 ) . In wild-type hair cells both Myo15 alleles can independently generate mRNA and encode protein for isoforms 1 and 2 . The Myo15sh2-J and Myo15sh2 alleles disrupt production of functional isoform 1 and isoform 2 , whilst the Myo15∆N allele reported in this study disrupts isoform 1 , but leaves isoform 2 functionally intact . Comparing different combinations of Myo15 alleles reveals a clear genotype–phenotype correlation . Mice deficient for both isoform 1 and 2 ( Myo15sh2/sh2 , Myo15sh2-J/sh2-J , Myo15sh2/sh2-J ) have short hair bundles with additional stereocilia rows . In the presence of at least one Myo15 allele competent to generate isoform 2 ( Myo15∆N/∆N , Myo15∆N/sh2-J ) , the stereocilia bundle develops the normal staircase architecture , but shorter stereocilia rows degenerate postnatally . At least one functional copy of isoform 1 , in addition to isoform 2 is required for normal hair bundle development and its long-term maintenance . We performed a more detailed survey of stereocilia ultrastructure at P4–P8 to understand how the degenerative process initiated in young postnatal hair cells . In the second row of Myo15ΔN/ΔN IHCs at P8 , the shape of the normally prolate tips were frequently exaggerated and over-elongated ( Figure 7B ) , compared to Myo15+/+ IHCs ( Figure 7A ) . We observed these over-elongated tips in Myo15ΔN/ΔN IHCs at P8 , but not at P4 ( Figure 7—figure supplement 3 ) , suggesting this abnormality was not a developmental defect , but rather the abnormal maintenance of stereocilia in isoform 1-deficient hair cells . The over-elongated tips of Myo15ΔN/ΔN IHCs were filled with actin filaments ( Figure 7B ) , indicating that actin polymerization was dysregulated at this location . The tips of stereocilia actin filaments are normally capped in an electron dense material that likely contains components of the actin polymerization machinery ( Tilney et al . , 1983 ) . In agreement with a previous report ( Rzadzinska et al . , 2004 ) , we confirmed that the tip density was absent from the tips of young postnatal Myo15sh2/sh2 stereocilia ( Figure 7—figure supplement 1E–I ) , supporting the role of this structure in stereocilia elongation and actin polymerization . In contrast to Myo15sh2/sh2 stereocilia , a prominent tip density was detected at the tips of Myo15ΔN/ΔN stereocilia at P7 , similar to littermate controls ( Figure 7A , B and Figure 7—figure supplement 1A–D ) . These data are consistent with isoform 2 being necessary to form the tip density complex and drive stereocilia elongation . Whilst isoform 1 is not required for initial tip density formation , its incorporation into the postnatal structure appears to regulate actin dynamics , specifically on the shorter stereocilia rows that have active MET channels . 10 . 7554/eLife . 08627 . 014Figure 7 . Isoform 1 maintains the diameter of mature mechanosensitive stereocilia and regulates the actin cytoskeleton at their tips . ( A–D ) SEM images of stereocilia bundles from Myo15+/+ and Myo15ΔN/ΔN IHCs at P8 ( A , B ) and P11 ( C , D ) . The inset ( top right ) is a higher magnification image of the second row stereocilia tips . The insets ( lower right ) are TEM images of either longitudinal ( A , B ) or axial ( C , D ) cross-sections of the stereocilia core . Note the electron-dense material at the tips of Myo15ΔN/ΔN stereocilia ( B ) and similar density of actin filaments in the thinning stereocilia of second row Myo15ΔN/ΔN IHCs ( D ) . ( E ) Distribution of diameters of second row stereocilia in Myo15+/+ ( upper histogram ) and Myo15ΔN/ΔN ( lower histogram ) IHCs at P8 ( left ) , and at P11 ( right ) . The diameters were measured from SEM images and are likely underestimated by ∼30% due to the uniform tissue shrinkage during critical point drying . Number of IHCs: n = 9 ( P8 , Myo15+/ΔN ) , n = 9 ( P8 , Myo15ΔN/ΔN ) , n = 8 ( P11 , Myo15+/ΔN ) , n = 8 ( P11 , Myo15ΔN/ΔN ) . ( F ) Average diameter of stereocilia in the first and second rows of Myo15+/+ ( white bars ) and Myo15ΔN/ΔN ( black bars ) IHC bundles at P8 and P11 ( data from panel E ) . Data are shown as mean ± SE . Asterisks indicate statistical significance: ** , p < 0 . 001; *** , p < 0 . 0001 ( t-test of independent variables ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 01410 . 7554/eLife . 08627 . 015Figure 7—figure supplement 1 . Isoform 2 is necessary for tip density formation . ( A–D ) TEM images of IHCs ( A ) and OHCs ( B–D ) stereocilia bundles in P7 Myo15ΔN/ΔN and control Myo15+/ΔN littermates . An electron-dense material ( arrows ) is present at the tips of stereocilia of all rows in both isoform 1-null Myo15ΔN/ΔN ( A–C ) and Myo15+/ΔN ( D ) littermates . ( E–I ) TEM images of OHCs ( E , F , I ) and IHCs ( G , H ) stereocilia bundles in P8 Myo15sh2/sh2 and P7 Myo15+/sh2 mice . The electron-dense material at the tips of stereocilia is absent from Myo15sh2/sh2 stereocilia that lack isoforms 1 and 2 ( arrows ) , compared with Myo15+/sh2 . Note: the smaller insertional tip density ( arrowheads ) can be still observed in P8 Myo15sh2/sh2 hair cells ( F , H ) . Scale bars are 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 01510 . 7554/eLife . 08627 . 016Figure 7—figure supplement 2 . Analysis of stereocilia staircase architecture in Myo15ΔN/ΔN hair cells . ( A–C ) Tip to tip measurements of IHC stereocilia heights from SEM preparations of mutant Myo15ΔN/ΔN and normal hearing Myo15+/ΔN littermates at P4 , P6 and P8 . ( A ) Distance from the tip of the tallest stereocilia ( first ) row to the tip of the corresponding second row stereocilium . ( B ) Distance from the tip of the second stereocilia row to the third row . ( C ) Height of third row stereocilia . Each data point corresponds to an individual stereocilium with mean ± SD overlaid ( number of hair cells sampled is shown in brackets ) . Asterisks indicate statistical significance: *** , p < 0 . 0001 ( ANOVA with Tukey's multiple comparisons test ) . SEM images were obtained from cultured samples used for MET recordings ( see also Figure 5 ) . The same images were analyzed in Figure 7—figure supplements 2 , 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 01610 . 7554/eLife . 08627 . 017Figure 7—figure supplement 3 . Analysis of stereocilia tip morphology in Myo15ΔN/ΔN hair cells . ( A ) Quantification of stereocilia tip shape from SEM preparations of mutant Myo15ΔN/ΔN and Myo15+/ΔN IHCs at P4 , P6 and P8 . An ellipse was fitted to the tip of mechanotransducing stereocilia ( second row ) and used to calculate an aspect ratio ( major axis divided by minor axis ) . An aspect ratio of 1 indicates a perfect circle , whilst aspect ratios >1 indicate elongation . Note the presence of aspect ratios >2 in mutant Myo15ΔN/ΔN hair cells at P8 ( a split scale is used for the ordinate ) . Data points correspond to individual stereocilia with mean ± SD overlaid ( number of hair cells sampled in brackets ) . Asterisks indicate statistical significance: *** , p < 0 . 0001 ( ANOVA with Tukey's multiple comparisons test ) . SEM images were obtained from cultured samples used for MET recordings ( see also Figure 5 ) . The same images were analyzed in Figure 7—figure supplements 2 , 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 08627 . 017 In further support of this idea , ablation of isoform 1 resulted in deterioration of the staircase architecture of the IHC bundle from P6 onwards ( Figure 7—figure supplement 2 ) . In addition , we observed a clear reduction in second row stereocilia widths measured between P8 to P11 in Myo15ΔN/ΔN IHCs ( Figure 7E ) . At P8 , the distribution of diameters in the second stereocilia row of Myo15ΔN/ΔN IHCs was comparable to controls ( Figure 7E ) , although there was a small , statistically significant reduction in the average diameter ( Figure 7F ) . By P11 , the average diameter of second row stereocilia in Myo15ΔN/ΔN hair cells was significantly reduced from 256 ± 2 nm ( mean ± SEM ) to 189 ± 4 nm and included a distinct population of ‘super-thin’ stereocilia with diameters of less than 150 nm ( Figure 7E , F ) . In contrast , the diameter of the tallest stereocilia row did not decrease between P8 and P11 in IHCs ( Figure 7F ) . Stereocilia thinning was specific to Myo15ΔN/ΔN IHCs as shorter row stereocilia in OHCs of the same mutant mice retracted without any noticeable changes in diameter ( Figure 6B , bottom row ) . TEM cross-sections taken towards the base of the stereocilia core revealed a similar density of actin filaments in normal and thinned stereocilia from Myo15+/+ and Myo15ΔN/ΔN IHCs at P28 ( Figure 7C , D ) . We infer that stereocilia thinning in Myo15ΔN/ΔN IHCs was caused by the loss of core actin filaments , rather than by an increased density of actin filament packing . We conclude that isoform 1 is required for the postnatal maintenance of the actin cytoskeleton , specifically in shorter row mechanotransducing stereocilia .
We demonstrate that mutations in Myo15 cause deafness through two fundamentally different mechanisms: ( i ) a failure to initially develop mechanosensory hair bundles , or ( ii ) a failure to maintain them once assembled . The involvement of myosin 15 isoform 2 in stereocilia development was previously established ( Belyantseva et al . , 2005 ) , however our finding that isoform 1 is required to maintain mature stereocilia was unexpected . These contrasting functions for myosin 15 are underlined by their different expression profiles during development . Furthermore , the isoforms have distinct spatial localizations within the hair bundle . A growing number of proteins exhibit asymmetrical distributions within different stereocilia rows of the hair bundle ( Peng et al . , 2011 ) , yet how these gradients are formed remains unclear . Our study shows that myosin 15 isoforms can selectively traffic to different stereocilia rows and establish the distribution of other proteins ( e . g . , whirlin and Eps8 ) critical for stereocilia development . What molecular signposts influence the trafficking of myosin 15 isoforms to specific stereocilia rows ? Because isoform 2 has no unique sequence when compared with isoform 1 , the unique N-terminal extension of isoform 1 must be responsible for regulating its ultimate destination within the hair bundle . Several targeting mechanisms are conceivable . Under a selective retention model , both isoforms would uniformly enter all stereocilia , but be captured at the tips of the tallest or shorter rows by isoform-specific interactions . Alternatively , a selective entry model would specifically restrict isoforms from entering at the base of each stereocilium . A physical ‘gate keeper’ structure restricts cytosolic access to primary cilia ( Dishinger et al . , 2010; Hu et al . , 2010 ) , but an equivalent apparatus has yet to be identified in stereocilia or actin-rich protrusions like microvilli . Rather than being restricted by a physical barrier , the two myosin 15 isoforms may exhibit different motility on stereocilia actin filaments . The shaker 2 ( p . C1779Y ) mutation within the actin-binding motor domain prevents both isoforms from accumulating in stereocilia , and indicates that the interaction with actin is likely critical in the trafficking process . Actin-binding proteins can direct myosin motility within cells ( Brawley and Rock , 2009 ) by enhancing , or inhibiting the acto-myosin binding interface to create permissive or restrictive actin tracks . Although row-specific variations in actin topology have not yet been identified ( Tilney et al . , 1983 ) , stereocilia and rootlets contain a plethora of actin-binding proteins that may provide guidance cues to spatially regulate the activity of myosin 15 isoforms ( Drummond et al . , 2012; Shin et al . , 2013; Kitajiri et al . , 2010 ) . The recent purification of myosin 15 will allow its motility to be tested on different types of cross-linked actin filaments , and to examine how the N-terminal extension might modulate this activity ( Bird et al . , 2014; Hartman et al . , 2011 ) . We found that Eps8 and whirlin do not require isoform 1 to traffic to their normal locations within the hair bundle . The sole dependence upon isoform 2 is intriguing since both myosin 15 isoforms share identical Eps8 and whirlin binding sites in the tail domain ( Liang et al . , 1999; Belyantseva et al . , 2005; Manor et al . , 2011 ) . The molecular basis for this selectivity is unknown , although a distinct possibility is that the N-terminal extension may interact with the tail domain of myosin 15 to regulate binding to cargo proteins . This type of intra-molecular regulation is common in other myosin classes ( Sellers and Knight , 2007 ) . We postulate that selective binding of cargo protein to isoform 1 , or 2 , forms the basis of a targeted protein trafficking system within mechanosensory hair bundles . Binding partners of the N-terminal extension are now of great interest , since isoform 1 may be responsible for trafficking proteins to the vicinity of the MET machinery on the shorter stereocilia rows . The spatial segregation of myosin 15 isoforms within the hair bundle has implications for existing models of stereocilia length regulation , which remain controversial . In one model , isoform 2 is required for elongation but does not set the absolute length ( Belyantseva et al . , 2003 ) . This model is supported by the fact that exogenous expression of isoform 2 does not cause additional elongation of developing stereocilia in vitro ( Belyantseva et al . , 2003 ) . An alternate model posits that the quantity of myosin 15 at each stereocilia tip determines its length , and that graded increases on each row specify the overall staircase architecture ( Rzadzinska et al . , 2004; Manor and Kachar , 2008 ) . We found that although isoform 1-null Myo15∆N/∆N hair cells have a reduced quantity of myosin 15 on their shorter stereocilia rows ( Figure 4C ) , no gross change in the initial development of the staircase architecture was observed ( Figure 7—figure supplement 2 ) . These data argue against the hypothesis that myosin 15 ‘dose’ controls the length of individual stereocilia . Isoform 1 is dispensable for the normal elongation of stereocilia in young postnatal hair cells; by inference , we conclude that isoform 2 is sufficient to drive stereocilia development . There is growing experimental evidence that myosin 15 regulates the actin cytoskeleton in developing stereocilia: ( i ) Myo15sh2/sh2 hair cells have short stereocilia where the actin core fails to elongate ( Probst et al . , 1998 ) ; ( ii ) exogenous expression of isoform 2 in Myo15sh2/sh2 hair cells rescues elongation of the stereocilia core , which entails actin polymerization ( Belyantseva et al . , 2005 ) ; ( iii ) expression of isoform 2 in COS-7 cells induces filopodia formation , again entailing actin polymerization ( Belyantseva et al . , 2003 ) ; ( iv ) our present study shows that the loss of isoform 1 results in progressive disassembly of mature mechanotransducing stereocilia . Considering the exceptional stability of the stereocilia actin core ( Zhang et al . , 2012; Drummond et al . , 2015; Narayanan et al . , 2015 ) , this implies changes to how the actin cytoskeleton is regulated at the stereocilia tips of isoform 1-null hair cells . Actin filaments at the stereocilia tip are embedded within an electron-dense plaque , which is proposed to cap the filament barbed ends and regulate polymerization ( Tilney et al . , 1983; DeRosier and Tilney , 2000; Rzadzinska et al . , 2004 ) . The tip density was prominent in all rows of Myo15ΔN/ΔN stereocilia ( Figure 7—figure supplement 1 ) , indicating that isoform 1 was not necessary for formation of this critical structure . However , once stereocilia are assembled , the postnatal incorporation of isoform 1 into the tip density , along with its cargoes , may repurpose the actin polymerization machinery from driving elongation towards the fine-tuning of local actin dynamics; which are normally restricted to the tips of mature stereocilia ( Zhang et al . , 2012; Drummond et al . , 2015; Narayanan et al . , 2015 ) . Accordingly , we note that the onset of stereocilia abnormalities in the second row ( Figure 7—figure supplement 3 ) and changes to the staircase architecture ( Figure 7—figure supplement 2 ) occur around the same time as the appearance of isoform 1 within normal hair bundles ( i . e . by P7 ) . The selective degeneration of only mechanotransducing stereocilia in Myo15∆N/∆N hair cells may result from the previously hypothesized link between MET currents and actin polymerization ( Tilney et al . , 1988 ) . This is consistent with the tallest stereocilia rows not degenerating in Myo15∆N/∆N hair cells , since MET currents are not detected on the tallest rows ( Beurg et al . , 2009 ) . A role for mechano-transduction has been further evidenced in hair cells of mutant Ush1gflox/flox mice , which lose tip-links and MET currents prior to degeneration of the shorter row stereocilia ( Caberlotto et al . , 2011 ) . However , the loss of MET currents can occur alone without any apparent effects on stereocilia morphology ( Kawashima et al . , 2011 ) . Furthermore , the absence of MET currents was unlikely the cause of stereocilia disassembly in Myo15∆N/∆N hair cells , since we found grossly normal transduction in these cells . We argue that membrane tension is instead a critical factor controlling actin polymerization at the tips of mature mechanosensitive stereocilia ( Diz-Munoz et al . , 2013; Barr-Gillespie , 2015 ) and that isoform 1 may contribute to a tension-sensing mechanism . The selective degeneration of shorter mechanotransducing stereocilia in Myo15∆N/∆N hair cells is reminiscent of the phenotype described in Ush1g , Tmhs , Eps8L2 , Xirp2 , Fscn2 , Dstn , Wdr1 and Pls1 mouse mutants ( Caberlotto et al . , 2011; Xiong et al . , 2012; Furness et al . , 2013; Perrin et al . , 2013; Francis et al . , 2015; Narayanan et al . , 2015; Scheffer et al . , 2015; Taylor et al . , 2015 ) . Our observations allude to isoform 1 being part of a larger macromolecular complex , or signaling pathway , that maintains shorter row stereocilia with active MET channels . How might isoform 1 regulate the stereocilia actin cytoskeleton in response to tension ? One possibility is that isoform 1 traffics components of a tension-sensing mechanism directly to the MET machinery . The 133-kDa N-terminal extension , unique to isoform 1 , contains dense clusters of poly-proline helices ( data not shown ) that are ligands for SH3 , WW and Enabled/VASP homology ( EVH1 ) domains commonly found in actin-regulatory proteins . The N-terminal extension may thus interact with a broad range of proteins capable of directly orchestrating cytoskeletal dynamics . The unusual physical properties of the N-terminal extension raise another distinct possibility . The N-terminal extension is predicted as intrinsically disordered due to its high-proline content ( 16 . 8% ) and low sequence complexity ( http://dis . embl . de; data not shown ) . Intrinsically disordered domains can explore multiple structural conformations ( Dunker et al . , 2002 ) , and in some cases act as entropic spring elements , such as the PEVK domains within the giant muscle protein , titin ( Linke et al . , 1998; Watanabe et al . , 2002 ) . We found that hair bundle deflection sensitivity was increased in isoform 1-null Myo15∆N/∆N IHCs , consistent with the alteration of a spring element within the MET machinery . In a further parallel with the N-terminal extension , the reversible unfolding of titin's PEVK domains during muscle extension is proposed to expose buried SH3 ligands and enable force-dependent biochemical signaling ( Ma et al . , 2006 ) . If the N-terminus were similarly mechano-sensitive , this may result in tension-dependent interactions with SH3 , WW and EVH1 domains , and retention of isoform 1 at sites of mechanical stress , e . g . in the vicinity of the lower tip-link insertion point on stereocilia . A mechano-sensitive N-terminal extension may also explain the localization of isoform 1 to the tallest rows of OHC stereocilia , since the tips likely experience tension from their attachment to the tectorial membrane during sound-induced deflections . Understanding the properties of isoform 1 and the N-terminal extension are now key to deciphering how the actin cytoskeleton is regulated in adult stereocilia , and to ultimately understand how these mechanosensory organelles are continually maintained throughout life .
To model the c . 3313 G > T ( p . E1105X ) allele of MYO15A ( Genbank: NM_016239 . 3 ) that causes DFNB3 deafness in humans ( Nal et al . , 2007 ) , we made the equivalent amino acid change c . 3256GAG > TAA ( p . E1086X ) in Myo15 ( Genbank: NM_010862 . 2 ) using homologous recombination in mouse ES cells ( Figure 1—figure supplement 1A ) . Homologous recombination arms encompassing the genomic sequences of Myo15 exon 2 ( 5′ arm ) , and exons 3 through exon 7 ( 3′ arm ) , were amplified from 129X1/SvJ genomic DNA using polymerase chain reaction ( PCR ) , and ligated into a pflox plasmid ( Chui et al . , 1997 ) previously modified to remove the HSVtk cassette . The c . 3256GAG > TAA mutation was introduced using the QuickChange II XL Site-Directed Mutagenesis Kit ( Agilent , Santa Clara , CA ) . This created a new MseI site that was used for genotyping . An additional Asp718 site was engineered adjacent to the loxP site in intron 2 to aid Southern analysis . The targeting vector was fully sequenced on both strands . Correctly recombined ES cells were screened by Southern blot analysis ( Figure 1—figure supplement 1B ) and injected into C57BL/6J blastocysts . Positive F1 progeny were crossed with EIIA–Cre mice ( Lakso et al . , 1996 ) to remove the neomycin-resistance cassette . Except for a single residual loxP site and Asp718 site in non-conserved regions of intron 2 , the resulting B6 . 129P2-Myo15 ( E1086X/E1086X ) mouse strain ( referred to as Myo15ΔN/ΔN ) had a structurally intact Myo15 genomic locus . Mouse tail biopsies were genotyped by PCR ( primers 5′-CCACAGTCTGAGGACCGAGT-3′ and 5′ GGTCTTGGTCTGGATGCTCT-3′ ) . The resulting amplicon was analyzed by MseI restriction endonuclease digestion; the Myo15+ allele generates 445 bp and 30 bp restriction fragments , whereas the Myo15ΔN allele generates 324 bp , 121 bp and 30 bp products ( Figure 1—figure supplement 1C ) . Shaker 2 ( Myo15sh2/sh2 ) , shaker 2J ( Myo15sh2-J/sh2-J ) , EIIa-Cre and C57BL/6J mice were obtained from the Jackson Laboratory ( Bar Harbor , ME , USA ) . All animal procedures were approved by the Animal Use and Care Committees ( ACUC ) at the University of Michigan ( #PRO00004639 , #PRO00005913 , #PRO00005128 ) , the University of Kentucky ( #903M2005 ) and the NIDCD/NIH ( #1263-12 ) . ABRs and DPOAEs were recorded and analyzed as described previously ( Karolyi et al . , 2007 ) . Mice were tested at 4 , 20 and 48 kHz for ABRs ( n = 3–6 mice per condition ) and 12 , 24 , and 48 kHz for DPOAEs ( n = 3–4 mice per condition ) . The otic capsule and vestibular labyrinth were dissected in RNAlater ( Life Technologies , Frederick , MD ) and total RNA isolated from cochlear tissues using RNAqueous-4PCR ( Life Technologies ) . First strand cDNA was generated using random-primed SuperScript III First-Strand Synthesis System for RT-PCR ( Life Technologies ) . TaqMan assay IDs ( Life Technologies ) #Mm00465026_m1 ( Myo15 exon 13–14 ) , #Mm04205306_m1 ( Myo15 exon 2–3 ) and #AJMSGCH ( Myo15 exon 1–3 ) were run on a Real-Time PCR System ( ABI7500 , Life Technologies ) . Transcripts for TATA-binding protein ( Tbp; #Mm00446971_m1 ) were used as the housekeeping reference gene . All reactions were performed in triplicate and averaged . A minimum of 3 independent biological replicates ( from different animals ) was performed for each age group and/or genotype . The difference in cycle threshold between Myo15 and Tbp was calculated for each sample ( ΔCT ) , and then normalized ( ΔΔCT ) to either P0 ( Figure 1E–G ) , or Myo15+/ΔN ( Figure 1H ) samples . Relative expression was calculated as 2^ ( -ΔΔCT ) . PB886 and PB888 antisera were generated by immunizing New Zealand White rabbits ( Covance , Denver , PA ) with a KLH-coupled peptide [H]-CKKFLGQHHDPGPGQLTKSAD-[NH2] ( Princeton Biomolecules Corp , NJ ) . Antisera were peptide affinity purified before use . Immunofluorescence protocols were performed as described ( Belyantseva et al . , 2003 ) . Briefly , mouse cochleae were fixed with 4% paraformaldehyde in PBS for 2 hr at room temperature . Dissected cochleae were permeabilized in 0 . 5% Triton X-100 in PBS for 30 min . An optional 30 min incubation in 0 . 1M sodium citrate ( pH 6 ) at 60C was included for antigen retrieval . Samples were blocked in 5% NGS/2% BSA in PBS and incubated in primary antibodies at 4C: PB888/886 , Eps8 ( #610143 , BD Biosciences , San Jose , CA ) , PB48 ( Liang et al . , 1999 ) or HL5136 ( Belyantseva et al . , 2005 ) . Primary antibodies were detected using Alexa Fluor 488 conjugated secondary antibodies ( Life Technologies ) . Samples were labeled with rhodamine-phalloidin ( Life Technologies ) and mounted in Prolong Gold ( Life Technologies ) , before imaging with a LSM780 confocal microscope ( Zeiss , Thornwood , NY ) and a 63x oil objective ( Plan-Apochromat , 1 . 4 N . A . ) . Confocal z-stacks of P7 IHCs labeled with PB48 were analyzed using ImageJ ( http://imagej . nih . gov ) . For each hair cell , ∼10 circular regions of interest ( ROI; 500 nm diameter ) were placed at the tips of the tallest stereocilia ( first row ) and additionally ∼10 more ( 400 nm diameter ) at the tips of the second row . The z-stack position was adjusted for each individual ROI to obtain the maximum fluorescence signal . Integrated fluorescence values were divided by the corresponding ROI area to yield an integrated fluorescence density for each stereocilia tip . To allow comparisons between stereocilia from different samples , fluorescence densities of individual ROIs ( from the first and second row ) from a single hair cell were normalized to the mean fluorescence density of the first row measured from the same hair cell . This yielded a relative intensity ( RI ) index , that expresses the fluorescence density at each stereocilia tip as a ratio of the fluorescence intensities observed ( on average ) at the tips of the tallest stereocilia row . Implicit in this analysis is the assumption that fluorescence values on the tallest stereocilia row remain constant between different genotypes , and can thus act as an internal calibration standard . Although PB48 fluorescence intensities at the tallest stereocilia row were subjectively comparable between Myo15+/+ and Myo15∆N/∆N IHCs , we cannot exclude the possibility of a systematic difference . We stress that the RI index does not allow for comparisons of absolute fluorescence between different genotypes , but only of relative changes . All statistical analyses ( t-test and ANOVA ) were performed using Prism v6 . 0 ( GraphPad , La Jolla , CA ) with two-tailed p-values reported . Cochleae were perfused with 2 . 5% glutaraldehyde ( Electron Microscopy Sciences , Hatfield , PA ) in 0 . 1M sodium cacodylate ( pH 7 . 4 ) with 2 mM CaCl2 for 1–2 hr at room temperature , and then micro-dissected in distilled water . Specimens were dehydrated in a graded ethanol series , critical-point dried in liquid CO2 , mounted on stubs using double-stick carbon tape and sputter-coated with 4–5 nm of platinum ( Q150T , Quorum Technologies , United Kingdom ) . Samples were examined using a field-emission SEM ( S-4300 or S-4800 , Hitachi , Japan ) . The stereocilia staircase architecture was quantified using a previously published approach ( Xiong et al . , 2012 ) . Briefly , direct tip-to-tip measurements ( in nm ) were made from the tip of the tallest stereocilia row ( first row ) to the tip of the second row , and from the tip of the second row to the tip of the third row . Additionally , the height of the third stereocilia row was measured relative to its insertion into the apical surface of the hair cell . Hair bundles were imaged as perpendicular to the stereocilia axis as possible to minimize projection errors . To quantify the shape of the second row stereocilia tips in IHCs , an outline was traced around the edge of each stereocilium , extending up to 300 nm down from the most distal point of the tip . An ellipse was fit to this region of interest using ImageJ ( http://imagej . nih . gov/ij/ ) and the aspect ratio calculated as the ellipse major axis divided by the ellipse minor axis . Both types of morphological analyses were made blinded to genotype . Cochleae were fixed in 2 . 5% glutaraldehyde and 2% paraformaldehyde in 0 . 1 M cacodylate buffer overnight at 4C , and then micro-dissected in PBS , followed by 3 × 20 min washes in 0 . 1 M cacodylate buffer ( pH 7 . 2 ) . Samples were post-fixed for 30 min in 1% osmium tetroxide , washed again in cacodylate buffer , dehydrated in series of 25% , 35% and 50% ethanol , then stained with 1% uranyl acetate in 50% ethanol and dehydrated further in 75% and 100% ethanol . Samples were embedded in Epon ( PolyBed 812 , Polysciences Inc , Warrington , PA ) and 60–80 nm thin sections cut using an Ultracut UCT ( Leica , Buffalo Grove , IL ) , collected on copper grids , post-stained with uranyl acetate and lead citrate , and imaged using a JEM 1010 electron microscope ( JEOL , Japan ) . Post-embedding immunogold labeling was performed as described ( Petralia and Wenthold , 1999 ) with minor modifications . Cochleae were perfused with 4% paraformaldehyde and 0 . 25% glutaraldehyde in PBS , cryoprotected in 30% sucrose , freeze-substituted and embedded in Lowicryl HM-20 resin ( Electron Microscopy Sciences ) . Ultrathin sections were cut using an Ultracut UCT ( Leica ) , collected on nickel grids and treated with 0 . 1% sodium borohydride with 50 mM glycine in TBST , incubated in 10% NGS in TBST , followed by overnight incubation at 4C in primary antibody ( PB886 ) diluted in 1% NGS/TBST . After washing in TBST , grids were blocked in 1% NGS/TBST for 10 min followed by incubation using 1:20 dilution of goat F ( ab ) 2 anti-rabbit IgG conjugated to 10 nm gold particles ( Ted Pella Inc . , Redding , CA ) in 1% NGS and 0 . 5% polyethylene glycol ( 20 , 000 MW ) in TBST for 1 hr at room temperature . Finally , sections were stained with 1% uranyl acetate and 0 . 3% lead citrate and examined using a JSM-1010 TEM microscope ( JEOL ) . Organ of Corti explants were dissected at P3-P4 and cultured in glass bottomed WillCo Wells ( Chemglass , Vineland , NJ ) for 3–5 days in DMEM ( Life Technologies ) supplemented with 7% fetal bovine serum ( Atlanta Biologicals , Flowery Branch , GA ) and 10 mg/l ampicillin ( EMD Millipore , Billerica , MA ) at 37°C ( 5% CO2 ) as previously described ( Stepanyan and Frolenkov , 2009 ) . Experiments were performed at room temperature in Leibovitz L-15 ( Life Technologies ) containing the following inorganic salts ( in mM ) : NaCl ( 137 ) , KCl ( 5 . 4 ) , CaCl2 ( 1 . 26 ) , MgCl2 ( 1 . 0 ) , Na2HPO4 ( 1 . 0 ) , KH2PO4 ( 0 . 44 ) , MgSO4 ( 0 . 81 ) . Hair cells were observed with a TE2000 inverted microscope ( Nikon , Melville , NY ) using an 100x oil-immersion objective and differential interference contrast optics . To access the basolateral plasma membrane of the hair cells , the outermost cells were removed by gentle suction with a ∼5 µm micropipette . Smaller pipettes for whole-cell patch-clamp recordings were filled with intracellular solution containing ( in mM ) : CsCl ( 140 ) , MgCl2 ( 2 . 5 ) , Na2ATP ( 2 . 5 ) , EGTA ( 1 . 0 ) , HEPES ( 5 . 0 ) . The pipette resistance was typically 4–6 MΩ when measured in the bath . Patch clamp recordings were performed with an AxoPatch 200B amplifier ( Molecular Devices , Sunnyvale , CA ) . Series resistance was compensated ( up to 80% , lag 7–10 µs ) . After compensation , the time constant of the recording system was in the range of 35–70 µs . Hair cells were held at −60 mV except for short periods of MET recordings , when the holding potential was temporarily changed to −90 mV . All recorded hair cells were approximately at the middle turn of the cochlea . Hair bundles were deflected using a stiff glass probe that was fire-polished to a diameter of ∼5–7 µm , matching the shape of the hair bundle . The protruding part of the probe was 2–3 mm long , which prevented lateral resonances . The probe was moved by a fast piezo actuator ( PA 8/14 SG , Piezosystem Jena , Hopedale , MA ) , custom-modified for a faster response with a time constant of 26–28 μs . The built-in strain gauge sensor of this actuator provided a direct reading of the probe's axial displacement . The angle between the axis of the probe movement and the bottom surface of a dish was kept at ∼30° . | Sound is detected by the cochlea , a coiled structure encapsulated within the inner ear of humans and other mammals . Inside this organ , intricate arrays of sensory hair cells are stimulated by sound to generate neural signals that are transmitted to the brain . The ‘hairs’ that give hair cells their name are actually structures called stereocilia that act like antennas to detect sound waves . Damage to these delicate mechanical sensors , through genetic mutations or loud noise , are a significant cause of hearing loss in humans . A protein called myosin 15 is a molecular motor needed for stereocilia to develop and grow to their normal height . Mutations of this protein cause hereditary deafness in humans . Hair cells produce two versions of myosin 15 , which are identical except for one version having an extra region called the N-terminal extension . Using genetic engineering , Fang et al . created mutant mice that only produce the smaller version of myosin 15 that lack the N-terminal extension . These mutant mice helped reveal that when hair cells are young , they mostly produce the smaller version of myosin 15 , and this is sufficient for stereocilia to grow normally . Once hair cells mature however , they switch to producing the larger version of myosin 15 that contains the N-terminal extension . In the mutant mice that lacked the larger version of myosin 15 , stereocilia ultimately deteriorate , leaving the hair cells unable to detect sound . Myosin 15 was previously known to help stereocilia grow , but Fang et al . now show that this protein is also required to maintain stereocilia throughout life . The next challenge is to understand how the N-terminal extension enables myosin 15 to preserve the structure of adult stereocilia , and to investigate whether this activity might be stimulated to prevent hearing loss . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2015 | The 133-kDa N-terminal domain enables myosin 15 to maintain mechanotransducing stereocilia and is essential for hearing |
In the human brain , a default mode or task-negative network shows reduced activity during many cognitive tasks and is often associated with internally-directed processes , such as mind wandering and thoughts about the self . In contrast to this task-negative pattern , we show increased activity during a large and demanding switch in task set . Furthermore , we employ multivoxel pattern analysis and find that regions of interest within default mode network are encoding task-relevant information during task performance . Activity in this network may be driven by major revisions of cognitive context , whether internally or externally focused .
Functional magnetic resonance imaging ( fMRI ) has repeatedly demonstrated that cognitive tasks of many kinds decrease activity in a large-scale cortical network , variously termed the task-negative or default mode network ( DMN ) ( Shulman et al . , 1997; Raichle et al . , 2001; Andrews-Hanna et al . , 2010 ) . The DMN consistently shows reduced activity during task performance compared to rest ( Raichle and Snyder , 2007 ) and often reduced activity for harder compared to easier task versions ( Gilbert et al . , 2012 ) . In contrast to this general pattern , increased activity has been reported in a cluster of mental states involving thinking about the self , one's own perspective , or the perspective of others ( Buckner and Carroll , 2007 ) . Examples include recollecting previous experiences ( Vincent et al . , 2006 ) or imagining future ones ( Addis et al . , 2007 ) , mind-wandering ( Mason et al . , 2007 ) , and theory of mind tasks ( Young et al . , 2010 ) . The DMN has thus been linked to a number of high-level cognitive processes , such as self-referential processing ( Gusnard et al . , 2001 ) and imaginary scene construction ( Hassabis and Maguire , 2007 ) . Recently , Andrews-Hanna et al . ( 2010 ) have argued that the DMN separates into three sub-networks . Using graph theoretical analytic approaches to resting-state fRMI data , Andrews-Hanna et al . identified a core sub-network comprising bilateral posterior cingulate cortex ( PCC ) and anterior medial prefrontal cortex ( AMPFC ) a medial temporal lobe ( MTL ) sub-network made up of ventromedial prefrontal cortex ( VMPFC ) , bilateral hippocampal formation ( HF ) , parahippocampus ( PHC ) , retrosplenial cortex ( Rsp ) , and posterior inferior parietal lobule ( pIPL ) and a dorsomedial prefrontal cortex ( DMPFC ) sub-network which includes the DMPFC , bilateral temporal parietal junction ( TPJ ) , lateral temporal cortex ( LTC ) , and the temporal Pole ( TempP ) . Andrews-Hanna et al . argue for a degree of functional segregation between these sub-networks , with the MTL sub-network especially linked to construction of mental scenes based on memory , and the DMPFC network more involved in mentalising . These segregations , however , are likely to be relative , with many experiments , for example , linking all three DMN sub-networks to conscious recollection ( Vilberg and Rugg , 2012 ) . Here , we consider a simple conceptualisation of DMN function and apply it within the Andrews-Hanna framework . Arguably , imagination , mind-wandering , and taking another's perspective have in common a substantial change from the current cognitive context . Similarly , conscious recollection is typically conceived as reactivation of a previously-experienced episode , with components linked into a complex surrounding , context . Substantial shifts of context may be common in everyday activity , for example , a shift from cooking dinner to giving directions to guests over the phone , but less common in the constrained setting of typical laboratory tasks . For example , in a recent review of neuroimaging studies of task switching ( Kim et al . , 2011 ) , tasks that they argue required a contextual switch involved either a change in simple attended features or binary categorization rules using a fixed , small set of possible stimuli . Irrespective of specific high-level processes involved , we reasoned that the DMN may be involved in any large switch of cognitive context—an operation presumably calling for relaxation of many aspects of a current attentional focus , with concomitant activation of representations and processes relevant to the new context . To test this hypothesis , we used a novel experimental paradigm that required participants to switch between similar and dissimilar tasks , within a relatively large set of six tasks ( Figure 1 ) . The six tasks were each associated with a different rule , as determined by the colour border surrounding the task stimulus . The tasks were split into three groups defined by stimulus category , with two possible tasks per stimulus type . A no-switch trial occurred when participants had to apply the same rule that was applied on the previous trial . A similar-task-switch trial—resembling switches in typical neuroimaging studies—occurred when participants had to apply the other rule from the same category as the previous trial . A dissimilar-task-switch occurred when participants had to apply a rule from a different category compared to the previous trial . 10 . 7554/eLife . 06481 . 003Figure 1 . Task description . The experiment required participants to learn six tasks prior to scanning . ( A ) The six tasks were each associated with a different rule , as determined by the colour border . The tasks were split into three groups defined by stimulus category , with two possible tasks per stimulus type . ( B ) Experimental design . Within each run , trials using the six tasks occurred in random order . A no-switch trial occurred when participants had to apply the same task that was applied on the previous trial . A similar-task-switch trial occurred when participants had to apply the other task from the same category as the previous trial . A dissimilar-task-switch occurred when participants had to apply a task from a different category compared to the previous trial . DOI: http://dx . doi . org/10 . 7554/eLife . 06481 . 003 Contrary to the common concept of a task-negative system , we predicted increased DMN recruitment for the most difficult condition of switching between dissimilar tasks . We found this to be the case and furthermore , that the activity increase was selectively found in the Core and MTL sub-networks . In addition , we provide evidence that all sub-networks of the DMN represented task-related information during task performance .
Accuracy on all tasks was high ( median accuracy for all tasks >95% , inter quartile range <6% ) . As predicted , response time ( RT ) was significantly longer when switching between dissimilar tasks ( 2043 ms ) compared both to trials when no task switch occurred ( 1670 ms; t17 = 8 . 6 , p < 0 . 001 ) , and to trials with switches between similar tasks ( 1746 ms; t17 = 8 . 1 , p < 0 . 001 ) . Switches between similar tasks also produced significantly longer RTs compared to no-switch trials ( t17 = 2 . 8 , p = 0 . 006 ) . Preprocessing steps for fMRI data included realignment of the raw echo-planar images ( EPI ) , slice-time correction , coregistration of the EPI images with the structural image , normalisation to the Montreal Neurological Institute ( MNI ) template brain , smoothing with an 8 mm full-width at half-maximum Gaussian kernel and filtering with a high-pass filter ( see ‘Materials and methods’ ) . Univariate analysis of fMRI data was used to compare dissimilar-task-switch with no-switch trials through the standard general linear model ( GLM ) approach . A regressor was constructed for each switch type with events modelled from stimulus onset until response and convolved with the haemodynamic response function . The resulting beta values for each switch type were compared and thresholded at p < 0 . 05 , correcting for the false discovery rate . We identified widespread activation predominantly in regions of in the DMN ( Figure 2A ) , with peaks found in bilateral HF , PHC , Rsp , PCC , AMPFC , and left pIPL ( Table 1 ) . It is worth noting that all of these regions fall within either the Core or MTL sub-networks . In comparison , no regions from the DMPFC sub-network showed significant activation at the whole-brain level . A contrast of similar-task-switch against no-switch trials revealed no significant activation across the whole brain . 10 . 7554/eLife . 06481 . 004Figure 2 . Activation of the default mode network ( DMN ) for dissimilar task switches . Region labels and regions of interest ( ROIs ) are color-coded according to the sub-network to which they belong: yellow for Core , green for medial temporal lobe ( MTL ) , blue for dorsomedial prefrontal cortex ( DMPFC ) . ( A ) Whole brain rendering in axial slices: the numbers above each slice indicate z-coordinate of that slice . The contrast of dissimilar-task-switch > no-switch ( T = 3 . 23 , p < 0 . 05 , FDR corrected ) shows activations in regions previously identified as the DMN . ( B ) Locations of DMN ROIs distinguished by Andrews-Hanna et al . ( C ) Change in activation of similar-task-switch ( darker colours ) and dissimilar-task-switch ( lighter colours ) relative to no-switch trials in the DMN ROIs . APMFC: anterior medial prefrontal cortex , PCC: posterior cingulate cortex , pIPL: posterior inferior parietal lobe , Rsp: retrosplenial cortex , PHC: parahippocampal cortex , HF: hippocampal formation , VMPFC: ventromedial prefrontal cortex , TPJ: temporoparietal junction , LTC: lateral temporal cortex , TempP: temporal pole , DMPFC: dorsomedial prefrontal cortex . * indicates p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06481 . 00410 . 7554/eLife . 06481 . 005Table 1 . Peak coordinates of DMN regions that showed significantly greater activation for dissimilar-task-switch over no-switchDOI: http://dx . doi . org/10 . 7554/eLife . 06481 . 005ROIhemispherexyzt-statisticHFleft−30−36−63 . 64right33−36−93 . 73PHCleft−21−4294 . 87right30−3966 . 81Rspleft−9−48123 . 97right9−51123 . 89PCCleft−12−54245 . 12right12−51245 . 16AMPFCleft−951−63 . 24right948−33 . 72pIPLleft−39−75333 . 58Coordinates are in MNI space . HF = hippocampal formation , PHC = parahippocampus , Rsp = retrosplenial cortex , PCC = posterior cingulate cortex , AMPFC = anterior medial prefrontal cortex , pIPL = posterior inferior parietal lobe . To examine changes in activation from the perspective of the DMN sub-networks , we used individual DMN regions of interest ( ROIs ) previously defined ( Buckner et al . , 2009; Andrews-Hanna et al . , 2010 ) . The mean beta value was extracted from each ROI following each switch type . Planned , paired two-tailed t-tests revealed significant increase in activity during dissimilar-task-switch compared to no-switch in core ( bilateral AMPFC , PCC ) and MTL ( Rsp , PHC ) sub-networks , with a tendency to de-activation in the DMPFC sub-network ( significant in right TPJ ) ( Figure 2C ) . Again , no ROIs revealed a significant difference between the similar-task-switch trials and no-switch trials . Two-way repeated measures ANOVAs were performed separately for each sub-network , with factors of ROI ( Core: 4 , MTL: 9 , DMPFC: 7 ) and switch type ( no-switch , dissimilar-task-switch ) . Main effects of task-switching were found for the Core ( F ( 1 , 17 ) = 16 . 7 , p = 0 . 001 ) and MTL ( F ( 1 , 17 ) = 6 . 1 , p = 0 . 03 ) sub-networks , showing increased activity for dissimilar switches . In contrast , the DMPFC sub-network showed a marginally significant de-activation ( F ( 1 , 17 ) = 4 . 1 , p = 0 . 06 ) . Corresponding ANOVAs were performed to test for the difference between similar-task-switch and no-switch , but these revealed no main effect in any sub-network . To investigate differences at the sub-network level , beta values were averaged across the ROIs within each sub-network each of the three trial types , and a two-way repeated measures ANOVA ( factors of sub-network and switch type ) was performed on the mean beta values . This analysis revealed a main effect of sub-network ( F ( 2 , 34 ) = 18 . 9 , p < 0 . 001 ) and an interaction of switch type and sub-network ( F ( 4 , 68 ) = 17 . 8 , p < 0 . 001 ) . These data therefore show a dissociation within the DMN: While the DMPFC sub-network displayed the characteristic pattern of reduced activity during executive control , switching between dissimilar tasks showed an opposite pattern of increased activity in Core and MTL sub-networks . In an exploratory analysis , we looked at the univariate activation associated with dissimilar switches between specific categories compared to no-switch trials , in the three DMN sub-networks , for example , from a semantic task to a lexical task , compared with repetition of the lexical task . This was performed for all six between category switch types . Figure 3 shows that all types of switch showed a relative increase in Core sub-network activation and decrease in DMFPC sub-network activation . The MTL sub-network shows increases for 4 of the 6 switch types and a marginal decrease when switching from the perceptual task . Especially for the Core and DMPFC sub-networks , these data suggest little variation in the pattern of results across different task types . 10 . 7554/eLife . 06481 . 006Figure 3 . Activation associated with each between category switch . An exploratory analysis looking at the activation/deactivation associated with switching between each of the three task categories . Core and MTL sub-networks predominantly show increased activation following a dissimilar task switch across switch types , whereas DMPFC shows a relative decrease in activation . Abbreviations: sem = semantic category , per = perceptual , lex = lexical . * denotes p < 0 . 05 from a paired , two-tailed t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 06481 . 006 For the multivariate analysis , the same preprocessing pipeline was followed with the omission of the smoothing step . We reasoned that if the DMN was involved in switching between tasks , then the differences between tasks might be represented within the network . To test this hypothesis , we performed a multivariate pattern analysis on the same ROIs . For each ROI , classifiers were trained to discriminate between the voxel-wise pattern of activation for each task pair ( 6 tasks , therefore 15 task pairs ) and these classifiers were subsequently tested on independent data using leave-one-run-out cross-validation ( see ‘Materials and methods’ ) . The matrices in Figure 4A show the classification accuracy ( CA ) for each task pair in each ROI , averaged across participants . The strongest decoding of task was found in bilateral HF , pIPL and the PCC , while bilateral TPJ , Rsp , AMPFC , and DMPFC on the midline showed weaker but still significant decoding . 10 . 7554/eLife . 06481 . 007Figure 4 . Classification accuracy ( CA ) within the DMN sub-network ROIs . ( A ) Classification accuracies between different task pairs in all DMN ROIs . Large correlation matrix used as example is the same as left HF . The colour of the circle in the key matches with the colour borders used to indicate each task in Figure 1A: Red for ‘bigger than a shoebox ? ’ , blue for ‘living ? ’ , brown for ‘same shape ? ’ , pink for ‘same height ? ’ , green for ‘does adding A make a word ? ’ , purple for ‘does adding I make a word’ ? Matrices show the classification accuracy of decoding each task pair; values below the diagonal show classification accuracy for all task pairs , while non-grey values above the diagonal show only decoding that survived the threshold for statistical significance . The colour borders indicate the sub-network that the ROIs belong to: core ( yellow ) , MTL ( green ) , and DMPFC ( blue ) . ROIs on the left side of each box are from the left hemisphere , those on the right are from the right hemisphere . ( B ) All three sub-networks demonstrated above-chance classification accuracy when decoding dissimilar tasks , while only the DMPFC sub-network demonstrated significant decoding of similar task pairs . Error bars indicate standard error . * indicates p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06481 . 007 Decoding of task between dissimilar task pairs is likely driven by many differences in task features . In contrast , differences between similar tasks will be predominantly driven by the internal representation of the specific decision rule . To quantify the extent to which rule and other features were driving the CA scores , CAs for ROIs within the Core , MTL and DMPFC sub-networks were averaged separately for similar task pairs and dissimilar task pairs . This analysis ( Figure 4B ) revealed significant decoding of dissimilar task pairs in all DMN sub-networks , and weaker but significant decoding of similar task pairs in the DMPFC sub-network . Recently concern has arisen that differences in RT may be driving differences in CA ( Todd et al . , 2013; for contrary arguments see; Woolgar et al . , 2014 ) . We performed a regression analysis of CA against absolute difference in RT in each of the three sub-networks separately . First , we extracted the CA associated with each task pair in each ROI in each subject . We then calculated the mean CA across the component ROIs of the Core , MTL , and DMPFC sub-networks in each individual , producing a 3-dimensional matrix of CA values for 3 sub-networks × 18 subjects × 15 task pairs . A similar matrix was produced for absolute RT differences . We segregated the dissimilar–task pairs and similar–task pairs and conducted a Spearman's correlational analysis of RT against CA for each task-pair type within each sub-network . The results showed strong discrimination of dissimilar task pairs , and weak discrimination of similar task pairs , irrespective of RT difference ( Figure 5A ) . 10 . 7554/eLife . 06481 . 008Figure 5 . The influence of response time ( RT ) on classification accuracy . ( A ) Correlation between classification accuracy and RT difference in the three DMN sub-networks . Each point represents data for a single task pair in a single subject , with mean CA across ROIs of the named sub-network plotted against absolute RT difference . The darker shades in each graph are taken from the similar task pairs , while the lighter shades are taken from the dissimilar task pairs . ( B ) Beta estimates for the association of CA and RT in each subject for similar and dissimilar task pairs of the three DMN sub-networks . In each graph subjects' beta estimates are sorted is in ascending order . Top row , a–c , displays beta estimates for similar task pairs in the Core , MTL and DMPFC , respectively . Bottom row , d–f , shows beta estimates for dissimilar task pairs in the Core , MTL and DMPFC , respectively . The p-value from a 2-tailed , 1-way t-test of each graph's beta values is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 06481 . 008 A second analysis considered data from each subject separately , with a separate regression analysis for similar and dissimilar task pairs in each ROI . A general linear model was constructed with a regressor for the absolute RT difference for either the similar-task-pairs or dissimilar-task-pairs , which was fit to the corresponding CA data . This produced a beta estimate for similar and dissimilar-task-pairs in each ROI , in each subject . Beta estimates were subsequently averaged across the component ROIs of each sub-network . Figure 5B shows the mean beta estimate for the similar and dissimilar task pairs in each subject for the Core , MTL , and DMPFC sub-networks . Within each graph , data from the 18 subjects are sorted in ascending order . Overall , the graphs suggest that RT does not systematically predict CA across participants , especially for the dissimilar task pairs ( bottom row ) for which CA was highest . Together these analyses demonstrate that simple RT differences were not strongly driving the classification accuracies .
In conclusion , we propose that the DMN may be recruited whenever large changes of cognitive context are required . This may apply in complex cases of self-referential processing , mind wandering etc , but also in relatively simple acts of cognitive or executive control . The DMN , widely seen as a ‘task-negative’ network , may respond positively to any task which demands a switch from one broad context to another .
18 right-handed participants ( 10 females ) aged between 18 and 40 were recruited from the Medical Research Council Cognition and Brain Sciences Unit subject panel . Of 21 original subjects scanned , three had to be removed for excessive head-movements ( over 10 mm translation and/or 6° rotation ) . No participant had a history of neurological or psychiatric illness . Participants were reimbursed for their time . Ethics approval was given by the Cambridge Psychology Research Ethics Committee . All three tasks were created using the Psychophysics Toolbox for MATLAB ( Brainard , 1997 ) . Within the scanner , the stimulus display was projected onto a mirror mounted to a 32-channel head-coil . Participants were required to learn six different tasks ( Figure 1A ) . Each task was associated with a different rule , with the appropriate rule determined by the colour border in which the task stimulus appeared . The six tasks rules are shown in Figure 1A . Prior to scanning , participants practised the task until they had completed at least 20 trials with an accuracy exceeding 80% . Importantly , the six tasks were grouped into three categories of two tasks each , where the stimuli within a category could be relevant to either rule within that category , but not to rules of other categories . Furthermore , all categories included trials which required a positive answer for both rules , for one rule but not the other , or for neither rule; therefore subjects needed to remember and apply the correct rule on all trials . All questions were framed in a true/false format , so that arbitrary response mappings for each rule did not have to be learned in addition to the rules themselves . Each trial began with the simultaneous appearance of the colour border ( visual angle = 7 . 9° ) and the task stimulus ( Figure 1B ) . Participants were requested to respond as quickly as possible with a true or false answer ( right thumb button press = true , left thumb button press = false ) . The border and stimulus remained on screen until the subject had made their response . A low tone was played to participants if they made an incorrect response . There was a jittered interval between the response to one trial and the onset of the next . Interval jittering followed an exponential distribution between 1 s and 11 s , with a mean of 4 . 1 s . Participants learned the tasks prior to scanning . An event-related design was adopted , with 73 trials per run . Each run had at least 12 trials of each of the six task types . Task switch type was also balanced within a run: 24 no-switch trials , 24 similar-task-switches , and 24 dissimilar-task-switches . Post scanning , when questioned , no participants reported having any sense of what task to expect on a given trial . Scans were acquired with a 3T Siemens Trim Trio scanner . 32 3-mm slices ( 0 . 75 mm interslice gap ) in axial orientation gave an in-plane resolution of 3 × 3 mm and were acquired using a TR of 2 s . T2*-weighted EPI capturing blood oxygen level dependent contrast was employed with a flip angle of 78° . For both experiments , the first eight images were discarded to avoid T1 equilibration effects . For the univariate analysis , images were preprocessed and analysed with SPM5 ( Wellcome Department of Cognitive Neurology ) . In the first preprocessing step , data were checked for obvious artefacts , and all images were realigned to the first image . Next we performed slice time correction and coregistration of the structural with the functional EPI images . Finally , data were normalized to the standard MNI template , smoothed with an 8 mm full-width at half-maximum Gaussian kernel and subjected to a high-pass filter with cut-off at 128 s . Fixed-effects analyses were performed on each individual's data using a general linear model . In the first univariate analysis investigating switching related activity three regressor functions were used ( no-switch trials , similar-task-switch trials , and dissimilar-task-switch trials ) . Each regressor was modelled as a rectangular function from the onset of each stimulus to the moment of response and convolved with the canonical hemodynamic response function . Beta weight images were contrasted for the conditions dissimilar-task-switch > no-switch and similar-task-switch > no-switch . Contrasts were further examined by random-effects analysis . Activation maps ( threshold 0 . 05 , FDR-corrected ) were visualised using MRIcroGL ( Rorden et al . , 2007 ) . For ROI analysis , mean contrast values within each ROI were extracted for each subject , using the MarsBaR SPM toolbox ( Brett et al . , 2002 ) . ROIs were spherical , with 8 mm radius , based around peak coordinates ( Figure 2B ) taken from Andrews-Hanna et al . ( 2010 ) . For the exploratory analysis into the activation associated with switches between specific categories a separate GLM was constructed . 36 regressors were used ( one regressor for each possible switch type ) and modelled as before . The resulting beta estimates were then processed using the same ROI analysis method as before with the same ROIs . Multivoxel pattern analysis was performed using the Decoding Toolbox ( Christophel et al . , 2012; Görgen et al . , 2012 ) . Preprocessing of the data was the same as for the univariate whole-brain analysis , except for the omission of the smoothing step . Again , a fixed effects analysis was performed on each participant's data using a general linear model . For this GLM , each task was modelled as a separate regressor , constructed as a rectangular function from the onset of each stimulus to the moment of response and convolved with the canonical hemodynamic response function . The same ROIs as previously ( Figure 2B ) were used . Prior to pattern analysis , beta values were Z-scored across tasks within each voxel of the ROI . This step was intended to reduce any impact of task differences in overall ROI activity . Pattern discrimination between tasks was then estimated using pairwise classification , that is , only 1 of the 15 possible task pairs was decoded at a time . A support vector machine ( LIBSVM ) ( Fan et al . , 2005 ) was used to train and classify data from three of the four runs , with the remaining run used to test the classifier . Test and training runs were always kept separate and each run was used to test the classifier once , that is , fourfold cross-validation . The CA for a given ROI was averaged across test-train splits , yielding a single CA for each ROI , in each individual , for each task pair . | The default mode network is a network in the brain that is often active when we think about ourselves , reminiscence about the past or just let our minds wander . However , this network—which involves many different regions of the brain—usually becomes inactive when we focus on a specific cognitive task . Now Crittenden et al . have used a technique called functional MRI to show that the default mode network can become active again if we switch from one task to another . Functional MRI works by measuring the blood flow in the brain: regions of the brain that are active have more blood flow than regions that are not active . Crittenden et al . studied the brains of human subjects as they performed a series of different tasks . These experiments showed that the activity of the default mode network does not change when the subject is focused on a single task . This is also true for when the subject switches between two similar tasks . However , when the subject switches between two very different tasks , the network becomes significantly more active . Moreover , the patterns of activity in the network seem to reflect the nature of the tasks . The work of Crittenden et al . strongly suggests that in order to successfully switch between two different tasks , the brain needs to engage the default mode network and allow the mind to wander . Future studies will involve exploring how different the two tasks need to be in order to activate the default mode network , and studying how brain damage within the network may impair patients ability to switch between different tasks . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Recruitment of the default mode network during a demanding act of executive control |
Human natural killer ( NK ) cells are defined as CD56+CD3− . Despite its ubiquitous expression on human NK cells the role of CD56 ( NCAM ) in human NK cell cytotoxic function has not been defined . In non-immune cells , NCAM can induce signaling , mediate adhesion , and promote exocytosis through interactions with focal adhesion kinase ( FAK ) . Here we demonstrate that deletion of CD56 on the NK92 cell line leads to impaired cytotoxic function . CD56-knockout ( KO ) cells fail to polarize during immunological synapse ( IS ) formation and have severely impaired exocytosis of lytic granules . Phosphorylation of the FAK family member Pyk2 at tyrosine 402 is decreased in NK92 CD56-KO cells , demonstrating a functional link between CD56 and signaling in human NK cells . Cytotoxicity , lytic granule exocytosis , and the phosphorylation of Pyk2 are rescued by the reintroduction of CD56 . These data highlight a novel functional role for CD56 in stimulating exocytosis and promoting cytotoxicity in human NK cells .
Natural killer ( NK ) cells are innate immune effectors that play an important role in the clearance of virally infected and tumorigenic cells and modulation of immune responses . Human NK cells represent approximately 10% of circulating lymphocytes and can be defined within this population as CD56+CD3− cells . CD56dim cells are the predominant subset in peripheral blood while the minority of the circulating NK cell population are CD56bright cells . CD56bright and CD56dim NK cells have unique expression of cell surface receptors , transcription factors and intracellular effector molecules that contribute to their distinct phenotypic and functional capacities ( Caligiuri , 2008; Freud and Caligiuri , 2006 ) . Despite its conserved expression on human NK cells and its use as a phenotypic identifier , the role of CD56 in immune function is poorly defined . CD56 is the cluster of differentiation nomenclature for neural cell adhesion molecule ( NCAM ) . In addition to being abundantly expressed in cells of neuronal origin , NCAM is also expressed in other tissues including the heart , kidney , skeletal muscles , liver , and on hematopoietic-derived cells including dendritic cells and human natural killer ( NK ) and NKT cells ( Galuska et al . , 2017; Lanier et al . , 1989; Roothans et al . , 2013 ) . NCAM is a member of the immunoglobulin superfamily of cell adhesion molecules and has multiple isoforms due to RNA splicing ( Cunningham et al . , 1987; Shapiro et al . , 2007; Williams and Barclay , 1988 ) . The most abundant isoforms are the glycophosphatidylinositol ( GPI ) -anchored 120 kDa form ( NCAM120 ) and two isoforms that contain transmembrane endodomains , NCAM140 and NCAM180 ( Cunningham et al . , 1987; Kleene and Schachner , 2004 ) . The extracellular portion of NCAM is highly conserved between the isoforms and contains 5 Ig-domains and two fibronectin-III domains that mediate homophilic ( NCAM-NCAM ) and heterophilic interactions between cells and the extracellular matrix ( Soroka et al . , 2003; Walmod et al . , 2004 ) . NCAM−/− mice have impaired learning and memory and behavioral disorders that are attributed to impaired nervous system development and post-differentiation maintenance of signaling and plasticity ( Brandewiede et al . , 2014; Cremer et al . , 1994 ) . As such , NCAM plays an important role in non-immune cellular differentiation and function . In addition to its role as an adhesion molecule , NCAM signaling in neuronal cells is important for neurite outgrowth and synaptic plasticity . NCAM140 is constitutively associated with the membrane-associated Src-family tyrosine kinase Fyn in axonal growth cones ( Beggs et al . , 1997 ) . Signaling through NCAM140 by agonist antibodies induces the recruitment of the non-receptor tyrosine kinase focal adhesion kinase ( FAK ) in neuronal cells , leading to a rise in intracellular Ca++ and neurite outgrowth ( Beggs et al . , 1997; Ditlevsen et al . , 2008; Schmid et al . , 1999 ) . In addition to this signaling pathway , NCAM activates fibroblast growth factor receptor −1 ( FGFR1 ) following trans-homophilic binding or binding of soluble extracellular NCAM and activates a non-canonical signaling pathway via phospholipase Cγ ( PLCγ ) that activates Erk1/2 in a Src-kinase dependent manner ( Francavilla et al . , 2009 ) . NCAM is unique amongst other glycoproteins in that its fifth Ig-domain contains 2 N-linked glycosylation sites that can be highly polysialylated ( Cunningham et al . , 1987; Kleene and Schachner , 2004 ) . PSA-NCAM is particularly important for synaptic plasticity and is highly expressed in embryonic development , with decreasing and restricted expression found in adults . The phenotype of olfactory bulb precursor cell migration in NCAM-deficient mice can be recapitulated by enzymatic removal of PSA , demonstrating the importance of PSA in NCAM signaling ( Ono et al . , 1994 ) . The lack of expression of NCAM or a known NCAM homologue on murine NK cells precludes the use of mouse models to test the in vivo requirement for NCAM or PSA-NCAM on NK cells . However , while murine NK cells are not modified by PSA , a subset of murine myeloid precursors in bone marrow are PSAhigh , and mice deficient for the ST8Sia IV sialyltransferase exhibit a sustained and severe contact hypersensitivity response , suggesting a role for PSA in the functional immune response ( Drake et al . , 2008 ) . Further investigation demonstrated that C-C motif chemokine receptor 7 ( CCR7 ) is a carrier of PSA on dendritic cells . Binding of PSA on CCR7 to the C-C motif chemokine ligand 21 ( CCL21 ) relieves CCL21 autoinhibition , leading to DC migration and helping confer specificity for response to chemokine ( Kiermaier et al . , 2016 ) . Together these studies underscore the significance of this unusual protein modification in both neuronal cells and certain immune contexts , however the role of PSA addition to NCAM on human NK cells has not been defined . NK cell cytotoxic function is exerted through the directed secretion of perforin- and granzyme-containing lytic granules following formation of an immunological synapse ( IS ) that includes adhesion , effector , and termination stages ( Orange , 2008 ) . Cell adhesion is mediated by integrins , particularly LFA-1 , that play a critical role in IS formation and actin polymerization . Prior to the polarization of the microtubule organizing center ( MTOC ) to the IS , lytic granules converge to the MTOC in a dynein-dependent minus-ended directed manner that is independent of actin polymerization and microtubule dynamics; as such , ligation of LFA-1 or CD28 alone is sufficient to induce convergence ( James et al . , 2013; Mentlik et al . , 2010 ) . LFA-1-mediated outside-in signaling leads to F-actin polymerization and reorganization , phosphatidylinositol 4 , 5-bisphosphate ( PtdIns ( 4 , 5 ) P2 ) generation , and the activation of protein tyrosine kinases , including Src family kinases ( Mace et al . , 2009; Steblyanko et al . , 2015 ) . Activating signaling and sustained F-actin remodeling promotes the polarization of the MTOC , with converged lytic granules , to the lytic IS ( Chen et al . , 2007; Krzewski et al . , 2008; Lagrue et al . , 2013 ) . Ultimately , lytic granules transverse a dynamic yet pervasive cortical actin network to access the plasma membrane , where they undergo SNARE-mediated exocytosis at the IS ( Brown et al . , 2011; Carisey et al . , 2018; Krzewski and Coligan , 2012; Rak et al . , 2011 ) . Previous studies have demonstrated that CD56 can promote human NK cell cytotoxic function against some CD56+ target cells ( Jarahian et al . , 2007; Nitta et al . , 1989; Taouk et al . , 2019; Valgardsdottir et al . , 2014 ) , although in some cases lysis is not enhanced by target cell CD56 expression ( Lanier et al . , 1991 ) . As these studies focused on homotypic CD56-mediated interactions , the significance of CD56 binding to heterotypic ligands is unclear yet is likely relevant . In neuronal cells , CD56 binds FGFR1 in cis and in trans , and CD56 on NK cells binding to FGFR1 on T cells leads to IL-2 production ( Kos and Chin , 2002 ) . CD56 also mediates direct recognition of A . fumigatus leading to fungal-induced NK cell production of MIP-1α , MIP-1β and RANTES ( Ziegler et al . , 2017 ) . This interaction is marked by accumulation of CD56 at the interface between the NK cell and A . fumigatus and is actin-dependent ( Ziegler et al . , 2017 ) . While CD56 has been implicated in NK cell development , migration , and cytotoxicity ( Nitta et al . , 1989; Taouk et al . , 2019; Lanier et al . , 1991; Chen et al . , 2018; Mace et al . , 2016 ) , the signaling pathways that regulate its function in immune cells have not been described . Given signaling downstream of CD56 that is mediated by FAK in neuronal cells , one potential link between CD56 and IS formation is the closely related non-receptor tyrosine kinase 2 ( Pyk2 ) , which is highly expressed in NK cells ( Gismondi et al . , 1997 ) . FAK and Pyk2 , with its expression more restricted to cells of hematopoietic origin , play critical roles in cell adhesion , cell migration and actin remodeling . Stimulation or engagement through multiple receptors , including T cell receptors , integrins and G protein coupled receptors , leads to Pyk2 phosphorylation and activation . As has been reported for Fyn-dependent activation of FAK in neuronal cells , tyrosine 402 ( Y402 ) of Pyk2 is a substrate for Fyn-dependent signaling downstream of TCR ligation ( Qian et al . , 1997 ) . In addition , Pyk2 clustering leads to rapid autophosphorylation on Y402 by trans-acting intermolecular interactions ( Eide et al . , 1995; Park et al . , 2004 ) . Phosphorylation on Pyk2 Y402 , which is equivalent to Y397 of FAK , enables binding and activation of SH2 domain-containing proteins , including Src kinases , and downstream activation of multiple signaling pathways that mediate cell adhesion and migration ( Parsons , 2003 ) . In NK cells , Pyk2 is phosphorylated downstream of integrin β2 ligation as part of an ILK-PINCH-PARVIN signaling cascade that leads to activation of Cdc42 , which can control microtubule dependent polarity through CLIP-170 and actin remodeling through WASp and the Arp2/3 complex ( Zhang et al . , 2014 ) . Pyk2 colocalizes with the MTOC in the uropod of migrating NK cells , however following activation it is translocated to the IS and is required for MTOC polarization in IL-2 activated primary NK cells ( Sancho et al . , 2000 ) . Expression of dominant negative Pyk2 disrupts cytotoxicity in this system , and its interactions with β1 integrin , paxillin , and other protein tyrosine kinases suggests that Pyk2 plays a role as a scaffolding protein that helps orchestrate NK cell cytotoxicity ( Gismondi et al . , 1997; Zhang et al . , 2014; Sancho et al . , 2000 ) . Here , we describe a requirement for CD56 in human NK cell function and show that deletion of CD56 in two human NK cell lines leads to impaired secretion and accompanying lytic dysfunction . Furthermore , we identify Pyk2 as a critical signaling intermediate downstream of CD56 . These data demonstrate a direct role for CD56 in the NK cell-mediated lysis of CD56-negative target cells and describe a novel activation pathway for cytotoxicity that is unique to human NK cells .
We previously used CRISPR-Cas9 to generate stable CD56-knockout ( KO ) NK92 cell lines and define a requirement for CD56 in human NK cell migration ( Mace et al . , 2016 ) . To extend our findings to a second NK cell line , we generated YTS CD56-KO cell lines using the same approach and CRISPR guides . CD56-negative YTS cells were isolated by FACS and the absence of CD56 protein was confirmed in both YTS and NK92 CD56-KO cell lines by Western blot analysis and flow cytometry ( Figure 1A , B ) . While transcripts for all three commonly expressed NCAM isoforms ( NCAM140 , NCAM180 , NCAM120 ) can be detected in human NK cells , NCAM140 has been previously reported to be the only isoform expressed ( Lanier et al . , 1989; Lanier et al . , 1991 ) . The extracellular domain of NCAM can be also post-translationally modified by the addition of polysialic acid ( PSA ) , which affects the molecular weight of CD56 when detected by Western blotting . We noted that Western blot analyses of NK92 and YTS cell lines suggested that CD56 is highly polysialated , particularly in the NK92 cell line , leading to a range of apparent molecular weights including a band corresponding to the 140 kDa isoform ( Figure 1A ) . In contrast , the YTS cell line expressed primarily the 140 kDa isoform with less polysialation , whereas primary NK cells expressed higher molecular weight isoforms and , as previously reported , also expressed polysialated CD56 ( Figure 1A; Moebius et al . , 2007 ) . To determine the contribution of polysialation to the molecular weight of CD56 on YTS and NK92 cell lines , we performed enzymatic treatment of cell lysates with PNGase F to cleave polysialic acid followed by Western blotting with anti-CD56 antibody . These data showed that removal of PSA reduced the variability of CD56 molecular weights and suggested that the 140 kDa or 120 kDa isoform were predominantly expressed in NK cell lines ( Figure 1C ) . Similar treatment followed by immunoblotting with a PSA-NCAM-specific antibody demonstrated that PSA-NCAM was not detectable following PNGase F treatment ( Figure 1—figure supplement 1 ) , thus confirming that the treatment was highly effective in PSA removal . Despite the reduction in molecular weight variability generated by enzyme treatment , it was difficult to define whether the most predominant band in the cell lines was the 120 kDa or 140 kDa isoform . To determine whether NCAM120 , which is GPI-anchored , was expressed on the cell surface of NK cell lines , we treated cells with phosphoinositide phospholipase C ( PI-PLC ) and quantified expression of NCAM on the cell surface by flow cytometry . CD56 expression on WT NK92 and YTS cells was resistant to PI-PLC cleavage while CD55 , a GPI-anchored protein , present on Raji and Jurkat cells , was cleaved ( Figure 1D ) . Together , these data strongly suggest that the isoform expressed on the surface of NK92 and YTS cells is not GPI-anchored NCAM120 , but is primarily NCAM140 as previously reported ( Kleene and Schachner , 2004; Lanier et al . , 1991 ) . Previous reports on the role of CD56 in NK cell lytic function have focused on homotypic interactions between CD56 on NK cells and target cells and have led to conflicting conclusions about the role that CD56 plays in cytotoxicity ( Nitta et al . , 1989; Taouk et al . , 2019; Valgardsdottir et al . , 2014; Lanier et al . , 1991 ) . To test the cytotoxic function of wild-type ( WT ) and CD56-knockout ( KO ) cell lines against CD56-negative target cells , we performed 51Cr-release cytotoxicity assays . NK92 or YTS cells were used as effectors against K562 target cells , which are susceptible to NK92-mediated lysis , or 721 . 221 , which are susceptible to lysis by both cell lines ( Gwalani and Orange , 2018 ) . Deletion of CD56 in the NK92 cell line severely abrogated its cytolytic function against susceptible K562 and 721 . 221 target cell lines ( Figure 2A ) . However , YTS CD56-KO cells exerted normal lytic function against 721 . 221 target cells when compared to their WT counterparts ( Figure 2B ) . To test a second susceptible target with YTS cells , we used KT86 cells , which are K562 cells that express CD86 and thus are susceptible to YTS-mediated lysis ( Banerjee et al . , 2007 ) . The use of KT86 targets did not reveal a requirement for CD56 in their lysis by YTS cells despite the failure of NK92 CD56-KO cells to lyse K562 targets ( Figure 2B ) . These data demonstrated the conserved defect in NK92 CD56-KO cytolytic function , whereas YTS cells were not significantly affected by loss of CD56 in their killing of multiple target cell lines in a 4 hr 51Cr assay . NK cell cytotoxic function includes serial killing of target cells , with target cell lysis followed by NK cell detachment and re-engagement with subsequent targets ( Anft et al . , 2020; Bhat and Watzl , 2007; Jenkins et al . , 2015; Vanherberghen et al . , 2013 ) . Given the demonstrated defect in NK cell migration in CD56-KO NK92 cells ( Mace et al . , 2016 ) we sought to measure whether the defect in NK cell killing could be attributed to impaired serial killing . The average time to first kill of a target cell by an NK92 or YTS cell ranges from 35 to 45 min ( Gwalani and Orange , 2018 ) , therefore to test the efficacy of early cytotoxic function we performed 51Cr cytotoxicity assays with 1 hr incubations with target cells . As seen in 4 hr assays , the lytic function of NK92 CD56-KO cells was severely impaired following 1 hr of incubation with target cells , suggesting that the decrease seen at 4 hr is not due to defects in serial killing . In addition , 1 hr 51Cr-release assays with YTS CD56-KO cells revealed a decreased specific lysis compared to WT YTS ( Figure 2C ) . While the killing decrease in the YTS CD56-KO cells at 1 hr was not as substantive as observed in NK92 CD56-KO cells , it was statistically significant at all effector to target ratios and consistent between multiple experiments . Therefore , while the impairment in CD56-deficient NK92 cells was more profound , there was a conserved phenotype between the two human NK cell lines tested . To determine whether the observed defect in cytotoxicity was due to dysregulated expression of molecules associated with NK cell lytic function , multiparametric flow cytometry was performed using panels designed to measure expression of receptors required for NK cell development , adhesion , activation , and inhibition ( Mahapatra et al . , 2017 ) . Small increases in the percent positive cells and mean fluorescence intensity of CD2 , CD11a , CD18 , CD45 , CD94 , and NKG2A were observed in NK92 CD56-KO cells ( Figure 2—figure supplement 1 ) . However , these slight differences may be attributed to better ligand accessibility by antibodies , as deleting CD56 removes the long chains of negatively charged polysialic acid attached to CD56 ( Nicoll et al . , 2003 ) . No significant differences were observed in the frequency of NK92 CD56-KO cells that were positive for expression of granzyme A or B and for IFNγ at rest or after activation with PMA/ionomycin relative to the parental NK92 cell line ( ( Figure 2—figure supplement 1 ) . Similarly , we did not observe differences in surface receptors or effector molecules at rest or after activation in YTS CD56-KO cells when compared to wild-type YTS ( ( Figure 2—figure supplement 1 ) . Therefore , we concluded that deletion of CD56 did not lead to dysregulated expression of several common NK cell receptors and , importantly , the capacity to produce granzymes A and B and interferon gamma was retained in CD56-KO cells . To confirm that the defect in NK92 cytotoxicity was specifically conferred by CD56 deletion , NK92 CD56-KO cells were retrovirally transduced with the NCAM140 isoform fused with an mApple fluorescent reporter to re-introduce CD56 expression . CD56 protein expression and expression of polysialic acid were confirmed in the reconstituted cells by flow cytometry ( ( Figure 2—figure supplement 2 ) . Importantly , re-expression of CD56 restored the lytic function of NK92 CD56-KO cells against both 721 . 221 and K562 targets ( Figure 2D ) , demonstrating that the impairment we observed was specific to the deletion of CD56 and that re-expression of NCAM140 was sufficient to rescue this defect . We sought to further define the requirements for the intracellular and extracellular domains of CD56 in NK cell cytotoxicity . The 140 kDa isoform of CD56 ( NCAM ) expresses 5 Ig-like domains and 2 FNIII repeats , a transmembrane domain and an intracellular domain . We generated CD56 chimeric constructs and retrovirally transduced these into NK92 CD56-KO cells . ΔICD lacks the intracellular domain ( ICD ) of CD56 but contains the extracellular domain ( ECD ) and transmembrane domain ( TM ) ; ΔECD lacks the CD56 ECD but includes the transmembrane and intracellular domains . Stable NK92 CD56-KO cell lines were generated and confirmed to be mApple positive and the ΔICD cells were verified to be CD56 positive ( Figure 2E ) . 51Cr-release assays were performed to determine the requirements for CD56 domains in cytotoxicity in NK92 cells . Sole expression of either the ECD or ICD was not sufficient to rescue cytotoxicity ( Figure 2F ) , whereas full-length CD56 restored cytotoxicity as previously demonstrated ( Figure 2D , F ) . Collectively , these data indicate that the full-length NCAM140 protein is required to promote human NK cell cytotoxicity and suggests a role for intracellular interactions that mediate this process . Both NK92 and YTS cell lines kill target cells via perforin- and granzyme-dependent exocytosis . In addition , YTS in particular are potent producers of IFNγ upon stimulation ( Gunesch et al . , 2019 ) . To further assess the effect of CD56 deletion on NK cell function , two methods of detection were used; a colorimetric benzyloxycarbonyl-L-lysine thiobenzyl ester ( BLT ) esterase assay for detection of granzyme A function as a readout for granule exocytosis and detection of surface-exposed lysosomal-associated membrane protein-1 ( LAMP-1 or CD107a ) by flow cytometry after plate-bound activation or coculture with target cells ( Alter et al . , 2004; Betts and Koup , 2004; Suhrbier et al . , 1991 ) . We activated WT , CD56-KO , and CD56 rescued CD56-KO NK92 cells for 90 min with immobilized anti-CD18 and -NKp30 antibodies and measured BLT esterase activity ( Rak et al . , 2011 ) . As suggested by the impaired lytic function demonstrated by 51Cr assays , NK92 CD56-KO cells had significantly decreased release of BLT esterase compared to WT NK92 , demonstrating that degranulation is impaired in the CD56-KO cells . Re-expression of full-length CD56 in NK92 CD56-KO cells restored esterase activity comparable to that of WT NK92 ( Figure 3A ) . To further support our observation that CD56-KO NK92 cells did not undergo lytic granule exocytosis , we measured CD107a exposure following activation by plate-bound antibodies or target cells . Following plate-bound stimulation , NK92 CD56-KO had a significantly reduced percentage of cells positive for CD107a compared to WT NK92 , and those cells that did express CD107a did so with a lower mean fluorescent intensity , supporting our data showing decreased secretion in the absence of CD56 ( Figure 3B ) . As with the BLT esterase assay , re-expression of CD56 in the knockout cells restored their capacity to degranulate ( Figure 3B ) . Furthermore , the same effect was observed when NK92 cells were co-incubated with target cells ( Figure 3C ) . Taken together , these data demonstrated that the deletion of CD56 in NK92 cells leads to a defect in lytic granule exocytosis in response to target cell activation or activating receptor cross-linking . We failed to detect BLT esterase secreted from YTS cell lines , likely due to their very low expression of granzyme A ( Gunesch et al . , 2019 ) . Furthermore , YTS cells undergo fewer degranulation events on a per cell basis when killing target cells than NK92 ( Gwalani and Orange , 2018 ) . Measurement of exposed CD107a on YTS cell lines ( WT and CD56-KO ) after plate-bound activation proved difficult and inconclusive after coculture with 721 . 221 target cells ( data not shown ) . However , YTS cells robustly produce IFNγ in response to activation ( Gunesch et al . , 2019 ) . Co-incubation of WT YTS with 721 . 221 target cells led to secretion of IFNγ detectable by ELISA , whereas this secretion was significantly decreased in the CD56-KO YTS cells ( Figure 3D ) . We confirmed robust IFNγ production in YTS CD56-KO cells in response to stimulation with PMA/ionomycin by intracellular flow cytometry ( ( Figure 2—figure supplement 1 ) , confirming that this decrease in IFNy levels from YTS cells conjugated with 721 . 221 is due to a defect in secretion , rather than production . Therefore , despite their intact target cell killing in 4 hr 51Cr assays , YTS cells had significant impairment in secretion in response to contact-dependent activation when cytokine production was measured . In summary , deletion of CD56 in the NK92 and YTS cell lines impairs their secretion in response to activation , including degranulation in NK92 CD56-KO cells , demonstrating a critical role for CD56 in NK function that is independent from CD56 homotypic interactions . The differential functional response to CD56 deletion of these cell lines supports their differing properties and speaks to relevant differences in their biology ( Gunesch et al . , 2019 ) ; here we have chosen to focus on the mechanism by which CD56 is mediating lytic function in the NK92 cell line . NK cell lytic function is exerted through the formation of an immunological synapse , which serves to focus directed secretion and mitigate bystander killing . The steps leading to IS formation can broadly be defined by adhesion/activation , polarization , secretion , and termination stages ( Orange , 2008 ) . Initial activation through ITAM-based activating receptors , integrins , or cytokine stimulation leads to lytic granule convergence to the MTOC , a step that precedes actin remodeling and the re-orientation of the MTOC to the synapse and ultimately lytic granule exocytosis and target cell death . Having shown that the secretion stage was impacted by CD56 deletion , we sought to further define how activation and polarization were affected in NK92 CD56-KO NK cell lines . Fixed cell confocal microscopy was performed to visualize granule convergence , MTOC polarization , and actin accumulation at the cell-cell interface and each of these parameters was quantified as previously described ( Hsu et al . , 2017; Sanborn et al . , 2010; Banerjee and Orange , 2010 ) . Both WT and CD56-KO NK92 cells were found in conjugates with K562 target cells , suggesting that their ability to adhere to targets was not impaired ( Figure 4A ) . While actin remodeling was consistently decreased in CD56-KO cells relative to WT NK92 , this effect was not significant ( Figure 4B ) . In contrast , the MTOC was not polarized towards the immunological synapse in CD56-KO NK92 cells as it was in NK92 cells , suggesting decreased activation . This observation was quantified by significantly greater MTOC-synapse distance ( Figure 4C ) . However , convergence of lytic granules to the MTOC was not affected by deletion of CD56 ( Figure 4D ) . Given the importance of cell adhesion in IS formation , we sought to further confirm that impaired conjugate formation between CD56-KO NK92 cells and their targets was not underlying their observed cytotoxic defect . We performed flow cytometry-based conjugate assays in which effector and target cells were differentially labeled , co-incubated , and then fixed prior to flow cytometric analyses . These analyses demonstrated that there was not an impairment in conjugate formation and that the frequency of conjugates formed with CD56-KO cells was , in fact , higher at 30- and 60 min timepoints but not significantly different at 10- and 120 min timepoints ( Figure 4E ) . Therefore , loss of CD56 expression in the NK92 cell line , which results in impaired cytotoxic function , is accompanied by impaired formation of the immunological synapse reflected by reduced MTOC polarization towards target cells despite intact conjugate formation . The introduction of the CD56-domain specific constructs into the NK92 CD56-KO cells illustrated the observation that both the extracellular and intracellular domain of CD56 are required to recover cytotoxicity . This suggests that the intracellular domain of CD56 may be interacting with intracellular molecules within NK92 cells to enable cytotoxic function . Pyk2 is closely related to FAK and a predicted interacting partner of CD56 through binding to Fyn ( Sancho et al . , 2000; Szklarczyk et al . , 2015 ) . In non-immune cells , engagement of NCAM140 ( CD56 ) recruits and activates FAK , which in turn activates the MAPK signaling pathway that is important for neurite outgrowth and cell survival ( Beggs et al . , 1997; Schmid et al . , 1999; Ditlevsen et al . , 2003 ) . Additionally , Pyk2 is an important regulator of NK cell cytotoxicity by stimulating lytic granule polarization and target cell conjugation ( Zhang et al . , 2014; Sancho et al . , 2000; Gismondi et al . , 2000 ) . To determine the effect of CD56 deletion on Pyk2 phosphorylation WT , CD56-KO , and CD56-KO reconstituted NK92 cells were activated by plate-bound anti-NKp30 and -CD18 . Cells were dissociated and phosphorylation of Pyk2 ( Y402 ) was measured by intracellular flow cytometry . We found that phosphorylation of Pyk2 Y402 was significantly decreased in the NK92 CD56-KO cells , a phenotype that was recovered by the reconstitution of CD56 ( Figure 5A ) . We hypothesized that the differential impact upon NK cell function on YTS and NK92 cell lines following CD56 deletion may be related to differential usage of Pyk2-dependent signaling pathways . To test this hypothesis , we analyzed Pyk2 phosphorylation in YTS and NK92 WT and CD56-KO cells using flow cytometry as described above . While we noted a similar trend in YTS cells , notably decreased Pyk2 phosphorylation in CD56-KO cells relative to WT , this effect was not as profound as that seen in NK92 cells and was not statistically significant ( Figure 5B ) . In addition , the magnitude of Pyk2 phosphorylation in WT YTS cells was lower than that of NK92 cells based upon normalized fluorescence intensities . The lower relative fluorescence intensity of Pyk2 Y402 phosphorylation in YTS cells and greater relative impairment in NK92 CD56-KO cells suggests a greater dependence on Pyk2 phosphorylation in NK92 cells and a potential mechanism whereby CD56 deletion has a greater effect on Pyk2-mediated cytotoxic function in NK92 than YTS cells . To further investigate this hypothesis , we performed 51Cr assays in the presence of PF431396 , a Pyk2/FAK inhibitor that inhibits Pyk2 autophosphorylation and blocks its kinase function ( Buckbinder et al . , 2007 ) . Consistent with the differential phosphorylation of CD56 in YTS and NK92 cells , the presence of Pyk2 inhibitor significantly decreased the cytotoxic function of WT NK92 cells , whereas both WT and CD56-KO YTS cells were only minimally affected ( Figure 5C , D ) . These data demonstrate that , in the absence of CD56 expression in NK92 cells , Pyk2 phosphorylation is decreased and that inhibition of Pyk2 phosphorylation in NK92 cells impairs cellular cytotoxicity . In YTS cells , a reduced dependence on Pyk2 for target cell lysis is reflected by retained lytic function both in the presence of Pyk2 inhibitor and the absence of CD56-mediated function . We further sought to determine whether Pyk2 localization at the immune synapse was affected by loss of CD56 . Fixed cell confocal microscopy was performed by acquiring 3D volumes of WT or CD56-KO NK92 cells conjugated to K562 target cells and immunostained for pPyk2 Y402 , perforin and actin . As suggested by intracellular flow cytometric analysis of Pyk2 phosphorylation , we detected higher fluorescence intensity of pPyk2 Y402 in WT NK92 relative to CD56-KO NK92 ( Figure 5E ) . Quantification of conjugates confirmed greater pPyk2 accumulation at the IS of WT NK92 cells when compared to CD56-KO NK92 cells ( Figure 5F ) . In contrast , we measured no difference in the localization or intensity of total Pyk2 between WT and CD56-KO cell lines , which primarily localized to the MTOC as previously reported ( Sancho et al . , 2000 ) ( ( Figure 5—figure supplement 1 ) . These measurements were made difficult by the significant phosphorylation of Pyk2 detected in the target cells , however the effect of CD56 deletion on Pyk2 phosphorylation is underscored by our functional experiments in a target cell-free system ( Figure 5A ) . Both Pyk2 and CD56 are localized to the uropod of migrating human NK cells ( Mace et al . , 2016; Sancho et al . , 2000 ) . We noted that CD56 in NK92 cells conjugated to targets remained partially localized to the uropod , however a fraction of CD56 localized to the IS , where it strongly co-localized with actin and pPyk2 Y402 ( Figure 5E ) . This was observed both early and late in immune synapse formation ( data not shown ) , suggesting that the effect we observed was not due to the kinetics of CD56 re-distribution . These data support our functional studies demonstrating that loss of CD56 in NK92 cells impairs immune synapse function through deregulated Pyk2 activation , while additionally illustrating the recruitment of CD56 to the IS in human NK cells . We sought to determine the effect of CD56 deletion on primary NK cell function . Using CRISPR-Cas9 , we deleted CD56 in bulk peripheral blood NK cells or CD56bright NK cells activated with IL-15 . CD56 expression was reduced by 9-fold when cells were incubated with low dose IL-15 but was still detectable on the surface of NK cells ( data not shown ) . Because CD56 protein appeared stable , we utilized higher doses of IL-15 to induce NK cell proliferation and CD56 dilution . Using this approach , the deletion of CD56 was highly efficient following delivery and expansion with IL-15 ( Figure 6A ) ; notably , NK cell cytotoxic function was intact ( Figure 6B ) . NK92 cells are phenotypically and genotypically more aligned with the CD56bright NK cell subset , whereas YTS cells are more similar to the CD56dim subset ( Gunesch et al . , 2019 ) . Therefore , we reasoned that the effect of CD56 deletion may be greater on the CD56bright subset , especially given its significantly higher expression of CD56 . As the CD56dim subset represents the majority of NK cells found within peripheral blood and has significantly decreased expression of CD56 relative to the CD56bright subset , we isolated CD56bright NK cells and repeated the deletion of CD56 and measurement of cytotoxic function following IL-15 expansion . While CD56bright cells do not have significant lytic function when freshly isolated , brief cytokine stimulation with IL-15 is sufficient to confer substantive cytotoxic function on this subset ( Wagner et al . , 2017 ) . Following IL-15 expansion , both mock transfected and CD56-KO CD56bright NK cells showed robust cytolytic function against K562 target cells ( Figure 6C ) . Therefore , despite the requirement for CD56 in NK92-mediated cytolytic function , loss of CD56 in CD56bright or CD56dim NK cells does not lead to a defect in primary NK cell cytotoxicity . We hypothesized that the expansion of primary cells in IL-15 during the process of cell editing may be overcoming the requirement for Pyk2 in primary NK cells , particularly given previous reports of IL-15 signaling altering primary NK cell dependence on Pyk2 ( Lee et al . , 2010 ) . Therefore , we expanded primary NK cells for 15 days in IL-15 and tested the effect of Pyk2 inhibition on cytotoxic function . While Pyk2 inhibition moderately decreased lytic function against K562 target cells following IL-15 activation , the effect was greater in cells that had not been stimulated with IL-15 , suggesting that long-term culture in IL-15 can reduce dependency of primary NK cells on Pyk2 Y402 phosphorylation for lytic function ( Figure 6E ) . We further hypothesized that if long-term stimulation in IL-15 could overcome Pyk2 dependency in primary cells , it may be possible to rescue the defect in CD56-KO NK92 cells with IL-15 stimulation . We incubated NK92 WT and CD56-KO cells in IL-15 for 5 days to mimic the conditions used to generate CD56-deficient NK cells . While this stimulation increased the lytic function of WT NK92 cells , activation by IL-15 failed to rescue the cytolytic deficiency in NK92 CD56-KO cells ( Figure 6F ) . However , NK92 cells stimulated with IL-15 also remained sensitive to Pyk2 inhibition , suggesting that the mechanism used to overcome Pyk2 dependence in primary cells in response to IL-15 was not functional in NK92 cells . Despite intact cytolytic function in CD56-deficient primary NK cells , we sought to define the localization of CD56 in primary NK cell conjugates and we performed fixed cell confocal imaging of freshly isolated ex vivo NK cells conjugated to K562 target cells . As with NK cell lines , we found that CD56 was localized to the uropod of migrating cells as previously described ( Sancho et al . , 2000 ) . However , we also frequently found redistribution of a pool of CD56 to the immune synapse ( Figure 7A ) . This was quantified by measurement of the mean fluorescence intensity of CD56 , which demonstrated greater CD56 intensity at the immune synapse than in non-synaptic regions of the NK cell ( Figure 7B ) . Furthermore , CD56 and pPyk2 Y402 were spatially co-localized at the immune synapse , and intensity of pPyk2 was similarly greater at the synapse than in non-synaptic regions as previously described ( Figure 7B; Sancho et al . , 2000 ) . Therefore , as with NK cell lines , CD56 and Pyk2 colocalize in primary NK cells and are recruited to the immunological synapse .
While CD56 is the prototypical identifier of human NK cells in peripheral blood , its function has been poorly defined . Early studies of its role in cytotoxicity largely focused on CD56 homotypic interactions and found that cytotoxicity of IL-2 expanded primary NK cells was diminished against CD56+ target cells in the presence of anti-CD56 blocking antibodies ( Nitta et al . , 1989 ) . In contrast , Lanier et al . found no significant difference in NK cell-mediated lysis of CD56– KG1a and CD56+ KG1a target cells ( Lanier et al . , 1991 ) . In both cases , CD56 functionality was tested with IL-2 expanded primary NK cells using antibody to inhibit CD56 homotypic interactions , through the manipulation of CD56 expression on target cells , or a combination of both . CD56 was also implicated in alloantigen-specific recognition by human NK cells , and the cytotoxic activity of the NS2 cell line against its LCL stimulator cell line was diminished in the presence of monoclonal antibodies specific for CD56 ( Suzuki et al . , 1991 ) . Conversely , the expression of NCAM on some target cell lines inhibits NK cell-mediated cytotoxicity ( Jarahian et al . , 2007 ) . Here , using CD56-deficient NK92 and YTS human NK cell lines , we demonstrate a requirement for CD56 in killing of CD56-negative targets by human NK cell lines and define a mechanism for the requirement for CD56 in human NK cell cytotoxicity . Our previous study defined a role for CD56 in human NK cell migration , and we had previously modeled this requirement using the NK92 cell line ( Mace et al . , 2016 ) . Here we extended our studies to the use of YTS cells as a second human NK cell line . Despite a significant and consistent effect of CD56 deletion on the cytotoxic function of NK92 cells , this effect was not conserved between the two human NK cell lines that we tested . The profound deficiency in lytic function in the NK92 cell line was observed even when effector cells were incubated with targets for just one hour in a 51Cr release assay , a timescale which likely only permits a single target cell killing to occur ( Gwalani and Orange , 2018 ) . As this incubation time led to significant killing of target cells by WT NK92 cells , these data suggest that the observed decrease in killing by the NK92 CD56-KO cells is not due to an inability to mediate serial killing against multiple targets despite the previously demonstrated requirement for CD56 in NK92 cell migration ( Mace et al . , 2016; Bhat and Watzl , 2007 ) . Furthermore , our subsequent recapitulation of cytotoxicity using antibody-coated glass to activate NK cells uncoupled the defect that we describe in exocytosis from one of cell migration . For YTS cells , killing was only impaired at the 1 hr time point , signifying an initial delay in their ability to kill which could be rescued at later time points . The milder effect on cytotoxicity in the YTS cell line when compared to the NK92 cell line suggests fundamental differences in the requirements of the two cell lines for cytolytic function , and it remains possible that the effect on YTS cell cytotoxicity could be due to impaired cell migration affecting serial killing of targets at later time points given the requirement for CD56 in human NK cell migration ( Mace et al . , 2016 ) . However , impaired IFNγ production by YTS CD56-KO cells following contact-dependent co-incubation with 721 . 221 target cells further underscores underlying differences in the requirement for CD56 in these cell lines and also demonstrates a functional relevance for CD56 in the YTS cell line . Previously demonstrated phenotypic and genotypic differences between the cell lines ( Gunesch et al . , 2019 ) supports this hypothesis and suggests that further investigation is required to fully delineate the relative role of CD56 in these contexts . Here , we have chosen to focus on the mechanism underlying the requirement for CD56 in the cytotoxic function of NK92 cells . Our investigations into the mechanism of CD56 function in human NK cells were informed by extensive investigations into the functional role of NCAM in neural cells , where NCAM-mediated signaling plays a critical role in axonal growth , survival and proliferation ( Ditlevsen et al . , 2008 ) . While there are multiple pathways by which this can occur that include homotypic and heterotypic NCAM interactions , NCAM signaling includes the formation of complexes that signal through Ras-MAPK-ERK pathways via interactions with p59fyn and FAK . Specifically , NCAM140 constitutively interacts with p59fyn , whereas FAK is recruited upon cross-linking of the NCAM extracellular domain ( Beggs et al . , 1997 ) . Recruitment of FAK leads to its phosphorylation and kinase function , and ultimately growth cone migration . The demonstrated requirement for Src family kinases , including Fyn , in NK cell cytotoxicity suggested that a common pathway could be functioning in NK cells ( Dong et al . , 2012 ) . Furthermore , while FAK is not known to be expressed in human NK cells , Pyk2 is expressed and is required for polarization of the MTOC during IS formation and resultant cytotoxic function ( Sancho et al . , 2000 ) . Here , we define a requirement for CD56 in phosphorylation of Pyk2 on Y402 , the primary autophosphorylation site that serves as a docking site for the SH2 domain of Src family kinases ( Eide et al . , 1995 ) . This requirement was most significant in the NK92 cell line , which we found had a significantly greater amount of Pyk2 Y402 phosphorylation both at baseline and upon activation . The functional implication of this finding was the significantly greater impairment in NK92 cytotoxicity following pre-treatment of both cell lines with a Pyk2 inhibitor . Despite showing that Pyk2 is required for NCAM-mediated function , it is still unclear precisely the mechanism by which this signaling occurs . We demonstrate that loss of CD56 leads to abolishment of lytic granule exocytosis , as surface CD107a was decreased on NK92 CD56-KO cells compared WT NK92 after activation with either plate-bound antibodies or with target cells . In addition , the maximum activity of granzyme A in the supernatant , an indirect readout of exocytosis , was decreased for NK92 CD56-KO cells . Given that the exocytosis defects we observed in the NK92 cell line were demonstrated using antibody cross-linking , and thus independently of target cell activation , it is unlikely that CD56 is binding directly to unknown target cell ligands in this context . Furthermore , our use of CD56-negative target cells defines independence from homotypic NCAM interactions . NCAM can also bind both in cis and in trans to FGFR , and interactions between CD56 on NK cells and FGFR1 on T cells is sufficient to induce T cell IL-2 production ( Kos and Chin , 2002 ) . Despite this indication that productive interactions occur within immune cells upon CD56-FGFR1 interactions , we have repeatedly failed to detect FGFR1 on human NK cells or the target cells used in this study ( data not shown ) , suggesting that FGFR1 is not the mechanism by which NK cell function is mediated by CD56 . Given the defect in exocytosis in CD56-deficient NK92 cells , we sought to define earlier steps in NK cell cytotoxicity and found that actin accumulation at the synapse was decreased in CD56-KO cells , although not significantly , and that MTOC polarization was impaired . While these features could be attributed to impaired integrin-mediated adhesion , conjugation of effectors to targets was not impaired in CD56-KO cells . Given the previously described requirement for Pyk2 in MTOC polarization during NK cell cytotoxicity ( Zhang et al . , 2014; Sancho et al . , 2000 ) , these data further suggest that the requirement for CD56 is in Pyk2-mediated function during cytotoxicity . This was further defined by impaired Pyk2 phosphorylation and recruitment to the IS in NK92 CD56-KO cells . Taken together , we propose a mechanism by which CD56 helps to localize or retain Pyk2 at sites of potential activation , where it could be available for subsequent autophosphorylation and activation of downstream signaling pathways required for polarization . Such a model has been proposed as the mechanism by which NCAM signals through FAK ( Beggs et al . , 1997 ) . In addition , the well-defined requirement for activating microcluster formation in activation at the immune synapse is mediated by adhesion receptors , and integrin engagement promotes microcluster formation , mobility and signaling at the NK cell IS ( Steblyanko et al . , 2015 ) . Inhibition of Pyk2 phosphorylation abrogates this effect , underscoring the importance of Pyk2 in amplifying activating signaling that is promoted by integrin engagement ( Steblyanko et al . , 2015 ) . Here , we show another example of how an adhesion molecule may be playing a similar role; whether this is through direct binding of Pyk2 to CD56 through Fyn , or through indirect mechanisms , such as regulation of integrin spatial localization , remains to be defined . There are other mechanisms that could be at play , including the previously described role for NCAM in regulating lipid raft inclusion and exclusion of signaling receptors ( Niethammer et al . , 2002 ) . Finally , given the significant degree of polysialation of CD56 on NK cells , particularly NK92 cells , the steric hindrance and negative charge of polysialated CD56 may be modulating membrane dynamics by physically influencing the organization and activity of integrins and activating receptors ( Paszek et al . , 2014 ) . While the mechanisms of the regulation of this signaling remain to be elucidated , our data speak to CD56 being an important player in the orchestration of adhesion and cellular signaling at the NK cell membrane . Delineating the role that CD56 is playing in directly mediating signaling , as opposed to spatially modulating other players such as integrins and activating receptors , represents an exciting direction for NK cell biology . While demonstrating a requirement for CD56 in cytotoxicity mediated by NK92 cells , we found that primary NK cells mediated target cell killing when CD56 was deleted from either total NK cells or specifically from the CD56bright subset . Given the challenges inherent in manipulating gene expression in primary cells , we were reliant upon the use of cytokines , in this case IL-15 , to promote NK cell survival and expansion following gene editing . It remains unclear as to whether this expansion bypasses a requirement for CD56 function in primary NK cells , although we do demonstrate that IL-15 stimulation of ex vivo NK cells significantly reduces reliance on Pyk2 Y402 phosphorylation for NK cell lytic function based upon the effect of a Pyk2 inhibitor . Despite the apparent reduced dependence on Pyk2 phosphorylation in IL-15 activated primary cells and the robust phosphorylation of Pyk2 Y402 in NK92 cells , activation of the NK92 cell line with IL-15 failed to rescue the defect in cytotoxicity in NK92 CD56-KO cells . This , combined with the observation that Pyk2 inhibition does not abrogate NK92 lytic function to the same extent as CD56 deficiency , suggests that there may be other mechanisms by which CD56 is exerting function in human NK cells . Ultimately , we show that despite the requirement for CD56 function in cell lines , ex vivo human NK cells are independent of this requirement , a finding that has relevance for clinical applications of NK cell-mediated adoptive therapies . Consistent with previous reports ( Lanier et al . , 1991 ) , our data demonstrate that the transmembrane-containing 140 kDa NCAM ( CD56 ) isoform is primarily expressed on human NK cells . Western blot analyses demonstrated a molecular weight range of ~130–160 kDa , which was significantly decreased following treatment of NK cell lines with PNGase F to remove polysialic acid . While these experiments , which led to the generation of a band that appeared close to 120 kDa , suggested that NCAM-120 could be expressed intracellularly ( Drake et al . , 2008 ) , treatment with PI-PLC , which we demonstrated cleaved GPI-anchored proteins , did not affect the detection of CD56 on the cell surface . The single band that we detected was also consistent with the predicted weight of the NCAM140 isoform containing 858 amino acids ( Lanier et al . , 1989 ) . Furthermore , expression of the extracellular domain ( ΔICD ) or the intracellular domain ( ΔECD ) of CD56 failed to restore cytotoxic function to the NK92 CD56-KO , whereas expression of full-length 140 kDa NCAM was sufficient . These experiments also concur with previous evaluations of transcript level expression that also describe the predominant isoform in human PBMCs as the 140 kDa isoform ( Lanier et al . , 1989 ) . While other reports have suggested that NCAM120 is the primarily expressed isoform on freshly isolated NK cells based upon qPCR ( Van Acker et al . , 2019 ) , the use of Western blotting in our study allows us to directly assess protein expression . The degree of polysialation in human NK cells is notable , however , particularly in NK92 cells and primary NK cells , and this high level of polysialation in human NK cells suggests that polysialic acid itself may be playing a role in CD56 function through its steric properties . In summary , here we describe a novel , functional role for CD56 , the canonical identifier of human NK cells . We build upon previous studies describing a role for CD56 in NK cell cytotoxicity and those that define its role in cell migration and NK cell maturation to elucidate , for the first time , its role in the lysis of CD56-negative targets . In doing so we describe an unexpected role for CD56 in the function of commonly used NK cell lines , including NK92 cells , but demonstrate that primary NK cells can execute cytotoxicity independently of CD56 . Finally , we link the cytotoxicity defect in CD56-KO NK92 cells to impaired signaling through the non-receptor tyrosine kinase Pyk2 , and thus define novel roles for CD56 and Pyk2 in human NK cell cytotoxic function .
For imaging experiments and cytotoxicity assays where indicated , primary human NK cells were isolated from healthy donors by venipuncture followed by NK cell enrichment using RosetteSep ( Stemcell Technologies ) and separation by Ficoll-Paque density gradient . Cells were resuspended in complete R10 media then co-cultured with K562 target cells and prepared for fixed cell confocal imaging as described below . Primary cells for CRISPR-Cas9 deletion experiments were obtained from leukopaks obtained from anonymous healthy platelet donors and enriched using RosetteSep and Ficoll-Paque Plus according to manufacturer’s instructions . Where indicated , primary NK cells were cultured in complete R10 media containing 3 ng/ml IL-15 . Primary NK cells were obtained in accordance with the Declaration of Helsinki with the written and informed consent of all participants under the guidance of the Institutional Review Boards of Baylor College of Medicine and Columbia University . NK cell lines and target cell lines were maintained at an approximate concentration of 1–5 × 105 cells/ml and cultured at 37°C 5% CO2 . YTS , Jurkat , Raji , 721 . 221 , K562 and KT86 cells ( Banerjee et al . , 2007 ) were a gift from Dr . J . Orange ( Columbia University ) and NK92 , Phoenix and Jurkat cells were acquired from ATCC . RPMI supplemented with 10% fetal bovine serum ( FBS ) plus essential nutrients was used to culture YTS , Jurkat , Raji , 721 . 221 , K562 and KT86 cells . NK92 cell lines were maintained in Myelocult media ( Stemcell Technologies ) supplemented with 200 U/ml of IL-2 ( Roche ) and 50 U/ml of penicillin/streptomycin ( Gibco ) . All cell lines were confirmed to be mycoplasma negative every 3 months and cell line identify was routinely confirmed by flow cytometry assessing expression of known surface markers . WT NK92 and YTS cells were validated by their expression of CD56 and absence of CD3 expression , combined with expression of CD28 ( YTS only ) or CD2 ( NK92 only ) . CD56-KO cells were similarly validated but excluding consideration of CD56 expression . Jurkat cells were confirmed to be CD55+ and Raji cells were confirmed to be CD20+ by flow cytometry . NCAM1 ( CD56 ) was deleted in YTS using CRISPR gene editing as previously described for our generation of the NK92 cell line ( Mace et al . , 2016 ) . In brief , a U6gRNA-Cas9-2A-GFP vector ( Sigma-Aldrich ) containing a NCAM1-specific gRNA sequence was incorporated into the NK cell lines through nucleofection . 106 cells per condition were nucleofected with 4 µg of DNA using the Amaxa Nucleofector II ( Lonza , Kit R; program R-024 ) . Nucleofected cells were allowed to recover for 24–48 hr before sorting for GFP positivity and lack of CD56 expression . NCAM reporter plasmids were generated by Epoch Life Sciences Inc and were made by subcloning NCAM1 ( NCBI reference sequence: NM_000615 . 6 , transcript variant 1; Origene ) into BamHI and SalI digested pBABE-puro-mApple retroviral plasmid . For construction of chimeric constructs , inserts were assembled by PCR and cloned into pBABE-puro-mApple between BamHI and SalI using ligation independent cloning by Epoch Life Sciences Inc . Re-expression of CD56 or expression of chimeric constructs in NK92 CD56-KO cell lines was performed by retroviral transduction . Phoenix-AMPHO cells ( ATCC ) were transfected with 4 . 5 µg of plasmid DNA using Fugene 6 Transfection Reagent ( Promega ) . Supernatant containing viral particles was collected and concentrated using PEG-IT ( System Biosciences ) followed by centrifugation at 2600 rpm . NK92 CD56-KO cells were transduced with viral particles in the presence of MAX Enhancer and TransDux ( System Biosciences ) . Within 48 hours cells were sorted for CD56 and mApple and cells were expanded and maintained under antibiotic selection with routine phenotypic validation . Primary human NK cells were incubated after enrichment for 16 hr at 37°C in 3 ng/mL IL-15 ( Miltenyi ) in RPMI media supplemented with 10% Human AB serum ( Sigma ) . Cells were washed with PBS , twice , and suspended at 2 × 107 cells/ mL in EP Buffer ( MaxCyte ) . Using OC-100 processing assemblies and the MaxCyte GT ( Maxcyte ) , cells were electroporated in the presence of Cas9 mRNA ( TriLink ) and CD56 guide RNA ( CGCUGAUCUCCCCCUGGCU; IDT ) using the WUSTL-3 setting and transferred to a 12-well plate and allowed to rest for 10 min at 37°C ( Cooper et al . , 2018 ) . Pre-warmed media containing 3 ng/mL IL-15 was added to the cells which were then cultured in IL-15 containing media as indicated . Cells were sorted based on CD56 expression on BD FACS Aria II to >90% purity . Flow-based killing assays were performed for 1 hr using CFSE-labeled K562 as previously described ( Leong et al . , 2014 ) . NK cell effector cells were co-cultured with target cells that had been pre-incubated with 100 µCi 51Cr for 1 or 4 hr in 96-well round-bottomed plates at 37°C 5% CO2 . 1% IGEPAL ( v/v ) ( Sigma-Aldrich ) was used to lyse maximal release control wells and plates were centrifuged . Supernatant was transferred to a LUMA plate ( Perkin Elmer ) and dried overnight . Plates were read with a TopCount NXT and % specific lysis was calculated as follows: ( sample – average spontaneous release ) / ( average total release – average spontaneous release ) x 100 . For experiments done with Pyk2 inhibition , effectors were pre-incubated for 10 min with 5 µM PF431396 ( Tocris ) then assays were performed in the presence of 5 µM PF431396 or equivalent volume of DMSO as a vehicle control . 96-well flat-bottomed plates were pre-coated overnight at 4°C with 5 µg/ml anti-NKp30 ( clone P30-15 , Biolegend ) and -CD18 ( clone IB4 ) or with mouse IgG1aκ as an isotype control . Wells were blocked with phenol red-free RPMI complete medium then washed three times with PBS . 105 WT or CD56-KO NK92 cells were added per well . Plates containing cells were briefly centrifuged then incubated for 90 min at 37°C 5% CO2 . Following incubation 1% v/v IGEPAL Sigma-Aldrich ) was added to maximum release wells and plates were centrifuged at 1000 rpm . Supernatant was transferred and substrate solution containing PBS , HEPES ( Gibco ) , N-α-Cbz-L-lysine thiobenzyl ester hydrochloride ( BLT; Sigma-Aldrich ) and 5 , 5’-Dithiobis ( 2-nitrobenzoic acid ) ( DTMB; Sigma-Aldrich ) was added . The plate was incubated at 37°C 5% CO2 for 30–60 min and luminescence was read at 415 nm using the BioTek Synergy H4 Hybrid plate reader . % maximum activity was calculated as: ( sample absorbance – average background ) / ( average total release – average background ) x 100% . NK92 and YTS cell lines were assessed for surface CD107a expression as a marker for degranulation after activation . 2 × 105 WT or CD56-KO NK92 and YTS cell lines were added to 12-well flat-bottomed plates that were pre-coated overnight at 4°C with 5 µg/ml anti-NKp30 ( clone P30-15 , Biolegend ) and -CD18 ( clone IB4 ) or 5% bovine serum albumin as a negative control . Alternatively , YTS cells were cocultured with 105 721 . 221 target cells at a 2:1 effector to target ratio in 5 ml polystyrene round-bottom tubes . Each condition was run in triplicate . Anti-CD107a antibody ( eBioscience , clone H4A3 ) was added at the onset of incubation for co-culture experiments . Cells added to the pre-coated antibody plates were spun briefly at 200 rpm before incubation for 90 min at 37°C 5% CO2 . Cells activated by plate-bound antibodies were mechanically dislodged and transferred to 5 ml polystyrene round-bottom tubes and incubated for an additional 25 min with anti-CD107a . Cells were fixed using 300–400 µl of 2% paraformaldehyde ( Electron Microscopy Sciences ) , then CD107a surface expression was measured using a modified LSR Fortessa ( BD Biosciences ) with the CD107a+ gate defined by unstained negative control cells . Data were analyzed using FlowJo 10 ( BD Biosciences ) . For experiments performed with target cell activation the average background of media alone was subtracted from samples . For detection of phospho-Pyk2 , cells were activated on plates as above and then fixed and permeabilized with BD Cytofix/Cytoperm ( BD Biosciences ) followed by detection with pPyk2 Y402 ( Abcam , ab4800 , 1:50 ) and secondary detection with goat anti-rabbit IgG Alexa Fluor 488 ( Invitrogen , 1:100 ) . Mean fluorescence intensity was adjusted to relative fluorescent intensity by normalizing to the WT condition . FACS analysis of WT or CD56-KO NK92 and YTS cell lines was performed using modified NK cell panels designed to assess the expression of adhesion , inhibitory , and developmental ligands and intracellular molecules ( Mahapatra et al . , 2017 ) . For panels evaluating response to activation 10 ng/ml phorbol 12-myristate 13-acetate ( PMA , Sigma-Aldrich ) and 1 µg/ml of ionomycin were used to stimulate cells for 4 hr at 37°C 5% CO2 . 10 µg/ml of Brefeldin A ( Sigma-Aldrich ) was added at the onset of incubation to stimulated and unstimulated controls to prevent protein transport of newly synthesized proteins in response to cellular activation . After incubation , activated cells were stained for surface ligands for 25 min followed by permeabilization and fixation using BD Cytofix/Cytoperm ( BD Biosciences ) . Antibody staining was performed in the dark and at room temperature; cells were then stained for intracellular markers for 1 hr before being washed and resuspended in PBS 1% paraformaldehyde ( Sigma-Aldrich ) . Cells in panels without PMA/ionomycin stimulation were also resuspended in PBS 1% paraformaldehyde after being incubated with antibodies for 25 min . Detection of CD56 on cell lines was performed using clone HCD56 ( 1:100 , Biolegend ) and polysialated NCAM ( PSA-NCAM ) was detected by clone 12F8 ( BD Biosciences ) followed by anti-rat goat IgG FITC ( Invitrogen ) . 104 events per sample were acquired using a modified LSR Fortessa ( BD Biosciences ) . Fluorescence minus one controls ( FMO ) were used to define positive and negative populations . Prism 6 . 0 ( GraphPad Software ) was used to graph the percent positive and the mean fluorescence intensity calculated using FlowJo 10 ( BD Biosciences ) software . For measurement of cell conjugates , NK cells and targets were pre-incubated with eFluor 670 ( Thermo ) and Cell Tracker Green ( Thermo ) respectively , then 105 cells were co-incubated at a 1:1 effector:target ratio in 5 ml polystyrene FACS tubes for 0 , 10 , 30 , 60 or 120 min . At the indicated timepoints 2% paraformaldehyde was added to the tubes to fix conjugates then all samples were analyzed on a BD Fortessa flow cytometer . The frequency of cells in conjugates was calculated using FlowJo 10 ( BD Biosciences ) and data were plotted and tested for statistical significance using Prism 8 . 0 ( GraphPad Software ) . The mean of each triplicate condition was calculated , and these means were compared by rank sum ( Mann-Whitney ) test for four independent technical replicates . WT YTS and CD56-KO cells were incubated with 721 . 221 target cells at a 2:1 ratio at 37°C 5% CO2 in round-bottomed 96-well plates for 22 hr in triplicate . Plates were centrifuged and the supernatant collected . Human IFNγ was detected by ELISA ( Abcam ) following the manufacturer’s procedure . Absorbance was read at 415 nm on a BioTek Synergy H4 Hybrid plate reader . IFNγ concentrations were calculated using a standard curve following subtraction of background ( media only ) from all conditions . Results were graphed and statistical analyses were performed using Prism 6 . 0 ( GraphPad software ) . NK92 and YTS NK cells , or Raji and Jurkat cells as a positive control were incubated with 1 U/ml of phosphatidylinositol-specific phospholipase C ( PI-PLC; Invitrogen ) at 4°C for 30 min in cold PBS . Cells were then washed twice with cold PBS and transferred to polystyrene tubes . Cells were immunostained for 25 min at room temperature with anti-CD56 ( BV421; clone HCD56 , Biolegend ) or anti-CD55 ( PE; clone JS11 , Biolegend ) . Cells were washed once with PBS and then resuspended in 2% paraformaldehyde and analyzed by flow cytometry . Prism ( GraphPad Software ) was used to graph the percent positive and the mean fluorescence intensity calculated using FlowJo 10 ( BD Biosciences ) software . Cell lysates from 5 to 10 × 106 cells were generated using CHAPS Cell Extract Buffer ( Cell Signaling Technology ) supplemented with 1X Halt protease inhibitor cocktail ( Thermo Fisher Scientific ) . Samples were incubated at 95°C with NuPAGE sample reducing agent ( Thermo Fisher Scientific ) and 4X NuPAGE LDS sample buffer ( diluted to 1X ) for 10 min . 2–4 × 105 cell equivalents per well were loaded into a NuPAGE 4–12% Bis-Tris density gradient gel ( Thermo Fisher Scientific ) and ran at a constant 150V for 80 min . Separated proteins were transferred onto nitrocellulose membranes using a Mini Gel Tank/Mini Blot Module ( Life Technologies ) at a constant 0 . 2A for 105 min . The nitrocellulose membranes were then blocked with 5% nonfat milk in PBS 0 . 05% Tween-20 for 90–120 min at 4°C . Nitrocellulose membranes were incubated overnight at 4°C with primary antibodies in 5% ( w/v ) BSA in PBS 0 . 05% Tween 20 at the following dilutions: 1:1000 anti-CD56 ( mouse monoclonal , clone 123C3; Cell Signaling Technology ) and 1:4000 anti-actin ( rabbit polyclonal; Sigma-Aldrich ) as a loading control . Detection of polysialated NCAM ( PSA-NCAM ) was by clone 2-2B ( Millipore , 1:1000 ) . Membranes were washed with 0 . 5M NaCl in PBS 0 . 05% Tween 20 . Primary antibodies were probed with either IRDye 680RD goat anti-mouse IgG or IRDye 800CW goat anti-rabbit IgG ( Li-COR Biosciences , 1:10 , 000 ) secondary antibodies for 1 hr at room temperature . Nitrocellulose membranes were imaged using the Odyssey CLx imaging system ( Li-COR Biosciences ) . Image Studio Lite software was used for densitometry and analysis of Western blot images . For fixed cell imaging , WT and CD56-KO NK92 cells were co-cultured with K562 target cells at a 2:1 effector to target ratio in complete R10 medium . Cells were incubated for 20 min at 37°C 5% CO2 then were transferred to poly-L-lysine coated #1 . 5 coverslips for an additional 25 min . Following incubation , cells were fixed and permeabilized with CytoFix/CytoPerm ( BD Biosciences ) at room temperature for 15 min and were washed twice with 50–100 µl of PBS . Conjugate immunostaining was performed with biotinylated monoclonal mouse anti-tubulin ( Invitrogen ) and Brilliant Violet 421-conjugated streptavidin ( Invitrogen ) ; Alexa Fluor 488-conjugated mouse anti-perforin ( clone dG9 ) ; phalloidin Alexa Fluor 568; pPyk2 Y402 ( Abcam , ab4800 ) and goat anti-rabbit IgG Alexa Fluor 488 ( Invitrogen ) . Coverslips were mounted to slides using ProLong Gold antifade reagent ( ThermoFisher Scientific ) . For detection of CD56 in conjugates , NK cells were pre-incubated with HCD56 Alexa Fluor 647 ( Biolegend ) prior to conjugate formation . For experiments done with Pyk2 inhibition , effectors were pre-incubated for 10 min with 5 µM PF431396 ( Tocris ) then assays were performed in the presence of 5 µM PF431396 or equivalent volume of DMSO as a vehicle control . Images were acquired through a 63 × 1 . 40 NA objective on a Zeiss AxioObserver Z1 microscope stand equipped with a Yokogawa W1 spinning disk . Illumination was by solid state laser and detection by Prime 95B sCMOS camera . Data were acquired in 3i Slidebook software and exported as TIFF files for further analysis . Fiji ( Schindelin et al . , 2012 ) , Volocity ( Perkin Elmer ) or Imaris ( Bitplane ) were used to process and analyze confocal image sequences . MTOC polarization to the synapse was determined using the line measurement tool after denoting the highest point of fluorescence intensity of α-tubulin as the MTOC . For actin accumulation at the synapse , the stamp tool was used to measure the fluorescence intensity of actin at the synapse and distal region of both the effector and target cells ( Banerjee and Orange , 2010 ) . The background and initial synapse intensity ( AU ) of actin was calculated as the area ( µm2 ) x the mean of fluorescence intensity . Total intensity of actin at the synapse was calculated as follows: Total synapse intensity = synapse intensity – ( effector background intensity + target background intensity ) . Lytic granule convergence was calculated by measuring the distance from individual granules to the MTOC as defined as the brightest point of α-tubulin intensity ( Hsu et al . , 2017 ) . For the accumulation of pPyk2 Y402 and CD56 at the synapse of primary NK cells , masks of synaptic vs . non-synaptic actin were generated following auto thresholding of intensity in the actin channel . Intensity of pPyk2 Y402 or CD56 staining was measured in Fiji following auto thresholding , with the ‘limit to threshold’ box checked . Intensity of respective channels of interest at the synapse or the non-synaptic cortical region were plotted for individual cells . All microscopy experiments were repeated by multiple individuals and in some cases masking of experimental conditions was performed . Prism 6 . 0 or 8 . 0 ( GraphPad software ) was used for statistical analyses . Unless otherwise stated , analyses used either a Student t-test with Welch’s correction or a one-way ANOVA with multiple comparisons . Data with non-normal distribution were tested with a Mann-Whitney ( rank sum ) test . Statistical significance was denoted when differences in measurements produced a p-value of <0 . 05 . Sample size computations were not performed but experiments with cell lines were replicated at least three times on different days using different passages of cell lines to generate technical replicates of at least three for each experiment . Experiments with ex vivo NK cells were performed with at least three biological and technical replicates . | The immune system deploys different cell types to take out cancer cells . True to their name , one type of immune cell known as natural killer cells kills tumor target cells by releasing toxic proteins that kill the harmful cells . In humans , these immune cells are defined , among other things , by the presence of a protein called CD56 on their cell surface . This protein ( which is also known as NCAM ) is thought to help cells to stick to their surroundings and control their movements . However , it was not clear whether CD56 also plays a role in the destructive abilities of natural killer cells . Gunesch et al . have now looked to see what would happen if natural killer cells lacked CD56 on their surface . The experiments included deleting the gene for CD56 from two kinds of human natural killer cell that are commonly grown in the laboratory ( called NK92 and YTS ) . In both cases , the cells lacking CD56 killed fewer cancer cells than the unedited natural killer cells . The NK92 cells were much more affected by the loss of CD56 than the YTS cells , and after Gunesch et al . compared the two kinds of cell they identified another protein called Pyk2 as the potential reason behind the difference . The Pyk2 protein is known to help a natural killer cell latch onto target cancer cells and release its toxic proteins . To do this , Pyk2 must first be activated with phosphate groups via a process known as phosphorylation . Gunesch et al . showed that Pyk2 protein in unedited NK92 cells was more highly phosphorylated than those of the YTS cells , and that Pyk2 activation by phosphorylation was greatly decreased in NK92 cells when the gene for CD56 was deleted . Together these and other results suggest that CD56 on natural killer cells helps to promote Pyk2 to activate the cells’ cancer-killing abilities through Pyk2 phosphorylation , especially in NK92 cells . These findings open up new lines of investigation into the relationship between sticky surface proteins and the activation of immune cells . They may also have important implications for the use of the immune system to treat cancer via immunotherapy . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2020 | CD56 regulates human NK cell cytotoxicity through Pyk2 |
Temperate bacteriophages are viruses that can incorporate their genomes into their bacterial hosts , existing there as prophages that refrain from killing the host cell until induced . Prophages are largely quiescent , but they can alter host phenotype through factors encoded in their genomes ( often virulence factors ) or by disrupting host genes as a result of integration . Here we describe another mechanism by which a prophage can modulate host phenotype . We show that a temperate phage that integrates in Escherichia coli reprograms host regulation of an anaerobic respiratory system , thereby inhibiting a bet hedging strategy . The phage exerts this effect by upregulating a host-encoded signal transduction protein through transcription initiated from a phage-encoded promoter . We further show that this phenomenon occurs not only in a laboratory strain of E . coli , but also in a natural isolate that contains a prophage at this site .
Bacteria and the phages that infect them have a generally antagonistic relationship , with evolution arming each side to defeat the other . Sometimes , though , a bacterium and a temperate phage can form an uneasy truce through lysogeny , wherein the integrated prophage confers some beneficial attribute to its host cell that provides a fitness advantage; after all , unless the host cell dies on the phage’s own terms , the phage dies too . Prophage alteration of host phenotype , known as lysogenic conversion ( Lederberg , 1955 ) , can benefit the host by conferring abilities to produce toxins , resist antibiotics , increase virulence , and repel further phage infections ( for recent reviews , see Argov et al . , 2017; Bondy-Denomy and Davidson , 2014; Davies et al . , 2016; Fortier and Sekulovic , 2013; Harrison and Brockhurst , 2017; Howard-Varona et al . , 2017; Obeng et al . , 2016; and Touchon et al . , 2017 ) . Oftentimes these traits are encoded within prophage genetic elements called morons , which contain genes that are regulated by their own promoters and are not involved in the phage lytic cycle ( Hendrix et al . , 2000; Juhala et al . , 2000 ) . Although lysogenic conversion has been under study since its first description nearly 100 years ago ( Frobisher and Brown , 1927 ) , there are likely entire classes of phage-encoded proteins that impact host fitness in as-yet-undescribed ways , as most phage genes have unknown function and no homology to any genes with known function . Phages can alter their hosts’ behavior in more subtle or indirect ways than , say , carrying a moron that enables toxin production; indeed , in most cases the effects of lysogeny on host physiology are unknown . One study found that deleting all of the cryptic prophages in Escherichia coli BW25113 increased the strain’s susceptibility to exogenous stresses and decreased its growth rate through mechanisms yet to be understood ( Wang et al . , 2010 ) . Other studies have shown that host gene expression can be regulated by phage-encoded transcription factors , as in the case of the cI repressor of phage λ that is expressed during lysogeny . This protein prevents expression of the λ lytic genes but was also discovered to act directly at the promoter of the metabolic gene pckA ( phosphoenolpyruvate carboxykinase ) , repressing its expression and producing a slow growth phenotype in some conditions ( Chen et al . , 2005 ) . Prophages can also alter host gene expression by means of the position in the host genome where they integrate ( Bondy-Denomy and Davidson , 2014; McShan and Ferretti , 2007 ) . For instance , the Φ13 phage of Staphylococcus aureus integrates into the 5’ end of the hlb gene , disabling β-toxin expression ( Coleman et al . , 1991 ) . Disruption of host genes by prophage integration can be reversed by prophage excision , and in some bacteria prophages act as switches that regulate host gene expression through controlled excision from interrupted genes , a phenomenon called active lysogeny ( Feiner et al . , 2015 ) . In Listeria monocytogenes , for example , the Φ10403S prophage integrates in and disrupts a gene that is required for efficient escape of the mammalian phagosome ( Rabinovich et al . , 2012 ) . The prophage excises during infection , restoring gene function , but its bacterial lysis genes remain repressed . The excised phage later reintegrates back into the same gene without killing its host . In some cases , the prophages involved in active lysogeny have lost the genes required for production of virions but still act as key regulators of cellular processes such as differentiation ( Feiner et al . , 2015 ) . We became interested in a particular temperate phage that infects E . coli—HK022 ( Dhillon and Dhillon , 1972 ) —because its integration site lies precisely between the genes torT and torS ( Yagil et al . , 1989 ) . These genes produce a periplasmic binding protein and a sensor kinase , respectively , that together detect trimethylamine oxide ( TMAO ) in the periplasm and transduce this signal to the cytoplasm to phosphorylate the response regulator TorR . Phosphorylated TorR then activates transcription of the torCAD operon , which encodes TMAO reductase ( see Figure 1A ) . This pathway enables E . coli to use TMAO as a respiratory electron acceptor . TMAO is widespread in the environment ( Gibb and Hatton , 2004; Hatton and Gibb , 1999 ) and is particularly abundant in the tissues of many marine organisms ( Eisert et al . , 2005; Seibel and Walsh , 2002; Yancey et al . , 1982 ) . Animals can ingest significant amounts of this compound from seafood-rich diets ( Eisert et al . , 2005; Zhang et al . , 1999 ) . In addition , humans and other mammals synthesize TMAO from trimethylamine ( TMA ) that is liberated from dietary precursors by the gut microbiota ( Fennema et al . , 2016; Zhang et al . , 1999 ) . Circulating TMAO accumulates in urine and is excreted ( Hai et al . , 2015; Velasquez et al . , 2016 ) . TMAO respiration allows E . coli to grow anaerobically , but it occurs even when oxygen is available ( Ansaldi et al . , 2007 ) . This is surprising because of anaerobic respiration’s relatively poor energy yield compared to aerobic respiration . We recently showed that aerobic expression of torCAD occurs with high cell-to-cell variability ( Roggiani and Goulian , 2015 ) , which can benefit the population by serving as a metabolic bet-hedging strategy in the face of a rapid decrease in oxygen availability ( Carey et al . , 2018 ) . Highly variable torCAD expression is regulated by oxygen and is mediated by torS and torT , the genes that flank the HK022 integration site ( Carey et al . , 2018; Roggiani and Goulian , 2015 ) . The torS and torT genes are divergently transcribed but share a repressor binding site for the transcription factor IscR ( Carey et al . , 2018 ) . Under aerobic conditions , IscR repression at this site leads to exceptionally low abundance of TorS and TorT protein and noisy transcription of the torCAD operon ( Carey et al . , 2018; Li et al . , 2014; Taniguchi et al . , 2010 ) . Curiously , the HK022 integration site separates the torS coding sequence from the IscR binding site that regulates its transcription . In this work , we show that the HK022 prophage reprograms the regulation of torCAD transcription in E . coli by disrupting the native torS promoter and introducing a phage-encoded promoter that drives torS transcription . By hijacking the regulation of torS transcription , HK022 reconfigures how cells respond to the presence of oxygen—in uninfected cells , oxygen regulates cell-to-cell variability in torCAD transcription without changing the population mean expression level ( Roggiani and Goulian , 2015 ) ; in infected cells , oxygen regulates the mean torCAD expression level and not cell-to-cell variability . Consequently , the HK022 prophage disables the bet-hedging strategy that aids cells during rapid oxygen depletion ( Carey et al . , 2018 ) . We further show that this phenomenon is not unique to HK022 lysogeny in a laboratory strain of E . coli , since the E . coli isolate NRG 857C , which naturally has a different prophage integrated at the HK022 integration site , shows similar behavior . The mechanism uncovered here , whereby phage cis-acting factors replace those of the host at a particular locus , may be a general mechanism used by temperate phages to alter their hosts’ behavior .
The integration site for bacteriophage HK022 is in the short intergenic region between the divergently transcribed genes torS and torT and separates the torS open reading frame from the IscR binding site that negatively regulates torS transcription ( Figure 1A ) . We suspected that the presence of a prophage at this integration site would disrupt the regulation of torS transcription and , ultimately , torCAD transcription , which depends on TorS and TorT ( Figure 1A ) . To investigate the impact of HK022 lysogeny on torCAD transcription , we constructed an HK022 lysogen in an E . coli K-12 strain carrying a fluorescent protein reporter of torCAD transcription . We grew this lysogenized reporter strain aerobically in the presence of TMAO and measured torCAD transcription in single cells by fluorescence microscopy . Transcription of torCAD was undetectable in the lysogen but was observed in the non-lysogen control strain ( Figure 1B ) . The simplest explanation for the loss of aerobic torCAD transcription in the lysogen is that the presence of the prophage destroys the torS promoter , as cells without TorS cannot phosphorylate TorR and activate torCAD transcription ( Figure 1A ) ( Jourlin et al . , 1996 ) . Unexpectedly , however , when we measured torCAD transcription in cells grown anaerobically in the presence of TMAO , we observed no difference between the lysogen and the non-lysogen ( Figure 1B ) . These results indicate that torS is still transcribed in the lysogen and that the above explanation is incorrect . We previously showed that high cell-to-cell variability in aerobic torCAD expression can function as a bet-hedging strategy that helps a population tolerate a rapid transition to anaerobiosis ( Carey et al . , 2018 ) . Only cells with a recent history of high torCAD expression are able to continue growth after oxygen depletion when TMAO is present and no other respiratory electron acceptors or fermentative substrates are available . Because the HK022 lysogen does not express torCAD aerobically , we suspected that it would be unable to employ this bet-hedging strategy and would therefore be unable to grow through an aerobic-to-anaerobic transition under the conditions described above . We tested this hypothesis by growing aerobic liquid cultures of the HK022 lysogen and non-lysogen in media containing TMAO and the non-fermentable carbon source glycerol , combining the cultures , and then transferring to an anaerobic agarose pad , which we used to observe the fates of single cells by time-lapse microscopy . Both strains contained the same fluorescent protein reporter of torCAD transcription . Cellular fluorescence was used as a measure of recent torCAD transcription and was correlated with cell growth after the transition to anaerobiosis , as in Carey et al . ( 2018 ) . To differentiate the lysogen from the non-lysogen , each strain was engineered to express a second fluorescent protein constitutively . Both strains carried deletion mutations of the HK022 receptor gene ( fhuA ) to prevent any infection of the non-lysogen by phage particles produced by spontaneous prophage induction in the lysogen . The results of this experiment , shown in Figure 1C , D , indicate that only the non-lysogen contains a subpopulation of cells that can grow substantially after oxygen depletion and that this subpopulation has high torCAD expression at the time of transition . From this we conclude that the HK022 prophage deactivates TMAO-dependent bet hedging on rapid oxygen depletion . Our finding that the HK022 lysogen expresses torCAD in the absence of oxygen indicates that the prophage does not simply eradicate the torS promoter . To investigate the effect of the prophage on torS transcription , we measured β-galactosidase activity produced from an operon fusion of lacZ to torS in the HK022 lysogen and non-lysogen , both with and without oxygen . We found that torS expression was substantially elevated in the HK022 lysogen ( Figure 2A ) . When we performed analogous experiments to measure torT transcription , we found no difference between the two strains ( Figure 2B ) . These results suggest that the HK022 prophage shuts off aerobic torCAD transcription not by disrupting torS transcription but rather by increasing torS transcription while leaving torT transcription unchanged . TorS molecules that are not bound to TorT are unable to detect TMAO and are in a state that dephosphorylates TorR ( Figure 1A ) . Therefore , cells with a large excess of TorS over TorT would strongly favor TorR dephosphorylation and not express torCAD ( Ansaldi et al . , 2001; Carey et al . , 2018; Roggiani and Goulian , 2015 ) . We note that the HK022 lysogen also shows elevated torS transcription in the absence of oxygen , and yet torCAD is still expressed in these conditions . This suggests that anaerobic TorT levels are sufficiently high for any additional TorS not to have much impact on TorR phosphorylation and torCAD expression . If the model described above is correct , then it should be possible to compensate for elevated TorS levels in an HK022 lysogen and restore aerobic torCAD expression by increasing expression of TorT . To test this , we introduced a plasmid containing torT under control of a weakened trc promoter into the lysogen carrying the fluorescent torCAD transcriptional reporter and quantified torCAD expression ( Figure 2C ) . The result of this experiment agrees with our prediction that the lysogen carrying the torT overexpression plasmid is able to express torCAD in the presence of oxygen . We next wanted to probe the mechanism by which the HK022 prophage increases torS transcription . We hypothesized that the phage encodes some cis-acting element ( s ) near its attP site that , upon integration , affect torS transcription . A simple explanation would be an outward-reading promoter that produces an mRNA transcript originating from within the prophage and reading through torS . To test this explanation , we inserted a synthetic terminator construct—an Ω element ( Prentki and Krisch , 1984 ) —at the boundary between bacterial and prophage sequence ( attLHK022 ) ( Figure 3A ) . We measured torS transcription in strains containing this terminator and found that torS expression was very low during both aerobic and anaerobic growth ( Figure 3B ) . This result strongly suggests that the mechanism by which the HK022 prophage activates torS expression is the production of torS transcripts that originate from within the prophage and are driven by a phage-encoded promoter . The annotated HK022 gene nearest the torS-proximal phage/host junction encodes the viral integrase ( int ) ; there are 73 bp between the int stop codon and the junction with the E . coli chromosome . In HK022 ( as in phage λ ) , expression of int is repressed during lysogeny ( Yagil et al . , 1989 ) , but it is conceivable that transcription of torS could be coupled with leaky expression of int . To determine whether the HK022 lysogen encodes a separate torS promoter , we performed in vitro transcription using DNA sequence upstream of the torS start codon . Approximately 200 bp of upstream sequence from the lysogen was cloned into a plasmid , and an analogous plasmid was constructed using upstream sequence from the non-lysogen . In vitro transcription from both plasmids produced transcripts , and the transcripts were different lengths when produced from lysogen sequence than when produced from non-lysogen sequence ( Figure 3C ) . Interestingly , transcripts of two distinct lengths were produced from the non-lysogen sequence , suggesting that there are two torS promoters when the prophage is absent . To confirm that the transcripts produced by in vitro transcription were truly torS transcripts and to map the transcription start sites associated with each of them , we performed primer extension assays ( Figure 3D , E ) . The transcript produced by the HK022 lysogen sequence mapped to a single transcription start site located within the prophage ( position indicated in Figure 3A ) . The transcripts produced by the non-lysogen mapped to one transcription start site on the torS-proximal side of attBHK022 and one start site on the torS-distal side of attBHK022 ( Figure 3F ) . The transcription start site for the shorter transcript ( hereafter TSS2 ) is identical to a computationally predicted start site ( Huerta and Collado-Vides , 2003 ) , while the longer transcript ( hereafter TSS1 ) has not previously been predicted or reported . TSS2 is so close to attBHK022 that its promoter must be at least partially ablated after prophage integration; this likely explains why no TSS2 transcripts are observed in the lysogen . TSS1 transcripts , on the other hand , are likely absent in the lysogen because TSS1 lies on the far side of the attBHK022 site from the torS coding sequence; in the lysogen , the promoter and coding sequence are separated by the entire HK022 genome . It appears , then , that the only torS transcripts made in the HK022 lysogen are produced by a phage-encoded torS promoter and that the int promoter is not required for prophage-regulated torS transcription . After identifying the transcription start sites , we realized that TSS2 is within three base pairs of a predicted translation start site for torS and that transcripts produced from TSS2 would not have room for a ribosome binding site . This translation start site uses a GTG start codon , indicated in Figure 3A , F ( UniProt Consortium , 2019 ) . A second translation start site for torS , downstream of the GTG start codon and employing an ATG start codon , has also been inferred ( Figure 3A , F ) ( Jourlin et al . , 1996 ) . Neither of these putative start codons is associated with a canonical Shine-Dalgarno sequence , suggesting that translation initiation is inefficient . To determine if either serves as a bona fide start codon , we constructed lacZ translational fusions to each and measured β-galactosidase activity . Only the lacZ fusion to the downstream ATG produced β-galactosidase activity ( Figure 3—figure supplement 1 ) . Accordingly , we have indicated the torS coding sequence as beginning with the ATG codon in Figure 3A , F . The above results reveal that the increased torS expression caused by the HK022 prophage restricts torCAD expression to anaerobic conditions in E . coli K-12 strain MG1655 , resulting in the loss of bet hedging . This prompted us to investigate the prevalence of prophages integrated at the HK022 integration site in wild E . coli strains . We searched for the torS and torT genes by BLAST ( Boratyn et al . , 2013 ) against all complete E . coli genome sequences available through NCBI ( https://www . ncbi . nlm . nih . gov/genome/microbes/ ) at the time of this analysis and calculated the torS-torT intergenic distance for each strain . For all strains with large insertions between torS and torT ( relative to E . coli MG1655 ) , we used the PHASTER web server ( Arndt et al . , 2016 ) to identify prophages . Roughly 5% of sequenced E . coli genomes carried prophages integrated immediately upstream of torS , and prophage-containing strains were not restricted to closely related E . coli phylogenetic groups ( Supplementary file 1 ) . PHASTER indicated that the prophage integrases were all more similar to HK022 integrase than to any other phage integrase , and a multiple sequence alignment of the genomic region from torS to int revealed that , for every prophage , the attL site was in the same position relative to torS as attLHK022 ( Supplementary file 2 ) . Sequence conservation was high upstream of TSSHK022 , suggesting conservation of the associated promoter ( despite there being no clearly identifiable −10 or −35 sequences ) . In most of these strains the torS-torT intergenic distance was roughly the same size as the HK022 genome , which is 40 , 751 bp long ( Juhala et al . , 2000 ) , although several of the strains appeared to have large genomic rearrangements relative to MG1655 in this region . Even in the strains with rearrangements , however , a prophage was integrated immediately upstream of torS at attLHK022 . We previously showed that torCAD expression in various wild E . coli strains lacking a prophage between torS and torT follows a similar pattern to what is seen in MG1655 ( Roggiani and Goulian , 2015 ) . As prophage integration at attBHK022 appears to be widespread in wild E . coli strains , we wondered whether torCAD expression in prophage-containing strains would resemble torCAD expression in HK022-infected MG1655 . We introduced the fluorescent reporter of torCAD expression into one such strain , the Crohn’s disease-associated strain NRG 857C ( Eaves-Pyles et al . , 2008; Nash et al . , 2010 ) . This strain belongs to phylogenetic group B2 ( Supplementary file 1 ) and is thus only distantly related to the laboratory strain MG1655 ( which belongs to phylogenetic group A ) . The torCAD promoter sequences are identical between NRG 857C and MG1655 , enabling us to use the same PtorCAD-yfp reporter construct that we used in MG1655-derived strains to assess torCAD transcription in NRG 857C . We integrated the transcriptional reporter into the NRG 857C chromosome by conjugation with an MG1655-derived Hfr donor strain , as NRG 857C is immune to genetic manipulation by P1 transduction . We used genetic markers in the donor and recipient strains to confirm that the tor and isc loci of NRG 857C were not replaced upon introduction of the transcriptional reporter ( see Materials and methods ) . When we measured torCAD transcription in NRG 857C during aerobic and anaerobic growth by fluorescence microscopy ( Figure 4 ) , we found that the pattern of expression was much more similar to MG1655 HK022+ than to MG1655 without the prophage ( Figure 1B ) . This suggests that in at least some wild E . coli strains there is prophage-mediated regulation of torCAD expression that is mechanistically similar to the HK022-mediated regulation seen in MG1655 .
In this work , we have shown that bacteriophage HK022 reconfigures the regulation of TMAO reductase expression in E . coli . Although other cases have been described wherein a prophage alters the expression of host metabolic genes , we are unaware of other instances in which a prophage so dramatically modifies its host’s response to the presence of a metabolite . By restricting torCAD expression to anaerobic conditions , HK022 converts oxygen-dependent regulation of the variance in torCAD expression ( seen in non-lysogens ) into oxygen-dependent regulation of mean torCAD expression ( Figure 5 ) . HK022 reconfigures torCAD regulation by increasing expression of the regulatory protein TorS ( Figure 5 ) . The phage appears to achieve this by replacing the native torS promoters with a promoter located within the prophage . We mapped the transcription start site associated with this promoter to within the prophage and were able to abolish torS transcription initiated at this site by inserting a transcriptional terminator into the junction between the prophage and the E . coli chromosome ( Figure 3 ) . These results indicate that an outward-reading transcript originates from within the prophage and reads through torS . We can only speculate on why HK022 shuts off aerobic torCAD expression and , consequently , the bet hedging associated with it . We have argued previously that there must be a fitness cost to the expression of torCAD , or else its expression would not be regulated ( Carey et al . , 2018 ) . The HK022 prophage may prevent aerobic torCAD transcription to alleviate a fitness cost and thereby increase the rate of its own replication . If the primary function of aerobic torCAD expression is bet hedging on rapid oxygen depletion , shutting down aerobic expression could be a useful strategy if phages like HK022 primarily lysogenize E . coli in environments where TMAO is present but rapid oxygen loss is unlikely to occur . One can conceive of such niches existing within habitats enriched in TMAO , such as the mammalian urinary tract , animal latrines , or the marine environment ( especially in association with marine animals ) . The HK022 integration site is occupied by a prophage in roughly 5% of fully sequenced E . coli strains , and these prophages are found in E . coli of diverse origins and phylogenetic groups . This suggests that HK022 is not an oddity in its integration between torS and torT . All of the prophages we identified occupying this site have , like HK022 , an outward-reading integrase gene as the final identifiable gene before torS , and all share the general genomic architecture of phages belonging to the lambda supercluster ( Grose and Casjens , 2014 ) . ( The one partial exception is the prophage from strain STEC299 , which maintains the HK022-like integrase oriented towards torS but appears to be more closely allied with the GF-2 phage supercluster ( Casjens and Grose , 2016 ) than the lambda supercluster ) . We investigated oxygen-dependent torCAD expression in one of these prophage-containing strains , the Crohn’s disease-associated strain NRG 857C , and found that the torCAD expression pattern in this strain is similar to the pattern we observe in MG1655 containing the HK022 prophage . In contrast , a previous study found that two other E . coli isolates , Nissle 1917 and HS , which do not have prophages integrated at attBHK022 , have torCAD expression patterns that are like that of wild-type ( uninfected ) MG1655 ( Roggiani and Goulian , 2015 ) . These observations suggest that it may be a general capability of the phages that integrate between torS and torT to alter host regulation of torCAD expression . However , despite the similarity of the torCAD expression pattern in NRG 857C and HK022-infected MG1655 and the high conservation of the region around TSSHK022 in prophage-carrying wild strains , we would not necessarily expect torCAD expression in all such strains to have the same behavior: there is considerable variation in overall sequence and gene content among the prophages that occupy the attBHK022 site , the host genomes they inhabit , and the habitats from which they were isolated ( see Supplementary file 1 ) . Studying the diversity and distribution of prophage-mediated torCAD expression could provide insight into the evolutionary advantages for a phage to reconfigure the control of TMAO respiration . Prophage-mediated effects on host physiology remain largely enigmatic , and knowledge is mostly restricted to cases where the effects are readily apparent ( as when prophage morons confer an observable phenotype such as toxin production; Hendrix et al . , 2000 ) . Cases where a prophage directly alters host metabolism have been described infrequently and generally with little mechanistic detail . To our knowledge , the phenomenon described in this study , where a prophage rewires the regulation of a metabolic pathway by modulating the expression of a signaling gene , has not been reported before and may exemplify a general class of mechanisms phages use to control host behavior . Phage infections certainly play a significant role in bacterial community dynamics , and much of our knowledge about the effects of phage infection is centered on lytic infection , horizontal gene transfer , and bacterial pathogenesis . A greater appreciation of the subtler effects of phage infection on host phenotype is a likely platform for developing enhanced understanding of the structure and behavior of microbial communities .
Media and growth conditions were as described in Carey et al . ( 2018 ) except that minimal A glucose medium was supplemented with 0 . 1% casamino acids and 10 mM TMAO for all experiments . Antibiotics were added to media at the following concentrations unless otherwise indicated: streptomycin , 250 μg/mL; ampicillin , 50 μg/mL; kanamycin , 25 μg/mL; and spectinomycin 20 μg/mL . Lists of all strains and plasmids used in this study are provided in Supplementary file 3 and Supplementary file 4 , respectively . HK022 was a generous gift from M . E . Gottesman ( Columbia University ) . P1vir transductions were performed as in Miller ( 1992 ) to create strains JNC173 ( JW0146 × DFE12 ) and JNC174 ( JW0146 × MMR65 ) . HK022 lysogens were generated using a method adapted from protocols for making λ lysogens ( Silhavy et al . , 1984 ) and for making mycobacteriaphage lysogens ( Sarkis and Hatfull , 1998 ) . Briefly , the strain to be lysogenized was grown to saturation in LB and harvested by centrifugation . Cells were resuspended at 2× concentration in 10 mM MgSO4 , and 100 μL of the suspension was added to 3 mL molten LB top agar at 45°C . The top agar was mixed , layered onto an LB agar plate prewarmed to 42°C , and allowed to solidify . HK022 lysate ( 50 μL ) was spotted onto the top agar and allowed to dry , and the plate was incubated at 37°C overnight . On the following day , an LB plate was spread with 100 μL HK022 lysate and allowed to dry . Selection for lysogens was carried out by streaking from the turbid zone of lysis formed on the top agar plate onto the HK022-spread LB plate and incubating at 37°C overnight . HK022-resistant colonies were patched onto LB agar , and the same colonies were tested for lysogeny by patching onto a top agar lawn containing an HK022-sensitive strain ( MG1655 ) . After overnight incubation at 37°C , candidate lysogens that produced a zone of lysis around the area of the patch ( from spontaneous phage release ) were nonselectively purified by streaking for single colonies from the LB plate patches and incubating at 37°C overnight . The entire patch test procedure was then repeated using the purified colonies . Candidate lysogen colonies that still produced a zone of lysis around the patch after purification were tested for the presence of the HK022 prophage by PCR . Strains produced by this method were JNC151 ( HK022 lysogen of MG1655 ) , DFE12 ( HK022 lysogen of MMR8 ) , JNC168 ( HK022 lysogen of JNC163 ) , and JNC169 ( HK022 lysogen of JNC166 ) . These strains were assayed for tandem polylysogeny by PCR essentially as in Powell et al . ( 1994 ) using primers HK022-P1 ( 5’-GGAATCAATGCCTGAGTG-3’ ) , HK022-P2 ( 5’-GCTGATACACTACAGCAATG-3’ ) , HK022-P3 ( 5’-GACAGGAGCTTGTTGACTAA-3’ ) , and HK022-P4 ( 5’-GGCATCAACAGCACATTC-3’ ) . All appeared to be tandem polylysogens ( denoted ( HK022 ) n in the Key Resources Table and Supplementary file 3 , following the convention of King et al . , 2000 ) , although the possible presence of contaminating virion DNA in the PCR template could not be ruled out . The Ω element strain JNC175 was constructed by recombineering ( Datsenko and Wanner , 2000 ) . The Ω element was amplified by PCR from pJB31 using primers LRpJB31_JNC169U1 ( 5’-CAGAGTCTTCGGGTCAGGGTTAAATTCACGGTCGGTGCACTTTAGGTGAAATCCCGAATGTGCAGTTAAC-3’ ) and LRpJB31_JNC169L1 ( 5’-TACTTACATTAATTTACTGATAATTAAAGAGATTTTAAATATACAACTTAGGCGCTGAAAGAAACCGCAA-3’ ) . The PCR product was digested with DpnI and purified before electroporation into JNC169 carrying helper plasmid pKD46 . Cultures were spread on LB agar plates containing streptomycin ( 20 μg/mL ) , spectinomycin ( 20 μg/mL ) , and kanamycin ( 25 μg/mL ) to select for integration of the Ω element and maintenance of the torS-lacZ fusion . The strain was cured of pKD46 , and correct integration of the Ω element into attLHK022 was verified by sequencing . The attLHK022::Ω construct was transduced into a clean JNC169 background to create JNC175 . To construct strains containing chromosomally encoded PtorS-lacZ translational fusions , a DNA fragment encompassing -152 to +28 bp relative to the torS ATG start codon ( as determined in this study ) was PCR amplified from the MG1655 chromosome and cloned into the XhoI and BamHI sites of pPK7035 upstream of lacZ’ , creating pPK12792 . pPK12792 served as a template for site directed mutagenesis in which bases downstream of predicted torS start codons ( GTG or ATG ) through the native lacZ start codon were deleted to create plasmids harboring the translational fusion constructs kan-PtorS- ( GTG ) lacZ’ ( pPK13169 ) and kan-PtorS- ( ATG ) lacZ’ ( pPK13171 ) . pPK13169 and pPK13171 were used as templates for PCR amplification of the translational fusion constructs using primers with homology to the native Plac region . The amplicons were electroporated into PK12556 , and kanamycin resistance was used to select for integrants . The kan-PtorS- ( GTG ) lacZ and kan-PtorS- ( ATG ) lacZ constructs were then moved into PK4854 using P1vir transduction to create PK13196 and PK13199 , respectively . For in vitro transcription and primer extension assays , promoter regions were cloned into the XhoI and BamHI sites of pPK7179 . For native PtorS , the aforementioned DNA fragment encompassing -152 to +28 bp relative to the torS ATG start codon was used , generating pPK12669 . To identify the HK022-derived promoter driving torS expression , a DNA fragment encompassing -231 to +28 bp relative to the torS start codon was PCR amplified from the chromosome of JNC151 and cloned into pPK7179 , generating pPK13256 . The PtorCAD-yfp reporter was introduced into NRG 857C by conjugation . As E . coli K-12 and E . coli NRG 857C are not closely related , synteny of their chromosomes was confirmed by genomic alignment using Mauve ( Darling et al . , 2004; Darling et al . , 2010 ) before proceeding . The PtorCAD-yfp reporter was first moved from strain MMR129 into the Hfr strain SASX41B by P1vir transduction , creating DFE33 . DFE33 retains the hemA41 allele of SASX41B and is therefore a δ-aminolevulinic acid auxotroph . DFE33 was mated with NRG 857C by superimposed patching of one colony of each strain onto an LB agar plate supplemented with 25 μg/mL δ-aminolevulinic acid . The plate was incubated overnight at 37°C , and on the following day bacteria growing in the patch area were streaked for single colonies onto LB agar supplemented with kanamycin and lacking δ-aminolevulinic acid . This selective media permitted growth of cells that had received the PtorCAD-yfp reporter , which is linked to a kanamycin resistance gene , but had not received the hemA41 allele: as the hemA locus is proximal to the tor locus in the direction of conjugative transfer , growth without δ-aminolevulinic acid indicated retention of the NRG 857C tor genes . Colonies were purified nonselectively on LB agar , and the resulting strain was named DFE34 . The metB1 allele of DFE33 , which confers methionine auxotrophy , was also verified not to have been transferred to DFE34 by confirming that DFE34 could grow on minimal glucose medium without amino acid supplementation . The metB locus is proximal to the iscR locus ( and the hemA and tor loci ) in the direction of conjugative transfer , so growth on minimal medium without amino acid supplementation indicated that DFE34 retained the NRG 857C iscR allele . Based on the genetic markers analyzed , the maximum amount of NRG 857C genomic sequence that could have been replaced by K-12 sequence during the construction of DFE34 is 1 . 1 Mbp . Microscopy was performed as described in Carey et al . ( 2018 ) except that cultures were grown to OD600 = 0 . 1–0 . 4 before being put on ice . Cultures were chilled on ice for 30 min at the time of streptomycin addition and then aerated on a roller drum at 37°C for 2 hr before being held at 4°C overnight . Imaging was performed the next day with no additional aeration beforehand . Figures 1 and 4 were generated using the R package ggridges ( R Development Core Team , 2018; Wickham , 2009; Wilke , 2018 ) . The density curves were generated using a Gaussian kernel function with the bandwidth selected by applying Silverman’s rule of thumb ( Silverman , 1986 ) to the entire data set . Aerobic-to-anaerobic transition microscopy was performed as described in Carey et al . ( 2018 ) except that no ΔtorC control strain was included . β-Galactosidase assays were performed as in Carey et al . ( 2018 ) except that cultures were grown to OD600 = 0 . 1–0 . 5 before harvesting . For the β-galactosidase assays using the PtorS-lacZ translational fusion strains PK13196 and PK13199 , chloramphenicol was added to the cultures at a final concentration of 20 μg/mL before placing on ice . Following purification of pPK12669 and pPK13256 with a HiSpeed Plasmid Maxi Kit ( Qiagen ) , 2 nM supercoiled plasmid was incubated with 5 μCi of [α-32P]UTP , 50 μM unlabeled UTP , and 500 μM final concentrations each of ATP , CTP , and GTP for 5 min at 37°C in 40 mM Tris ( pH 7 . 9 ) , 30 mM KCl , 100 μg/mL bovine serum albumin , 1 mM dithiothreitol , and 10 mM MgCl2 . E . coli σ70 RNA polymerase holoenzyme ( 50 nM ) was added , and each reaction ( 20 μl total volume ) was terminated after 10 min by adding 10 μL 95% ( vol/vol ) formamide , 20 mM EDTA , 0 . 05% ( wt/vol ) bromophenol blue , and 0 . 05% ( wt/vol ) xylene cyanol FF . After the mixture was heated to 90°C for 30 s , 5 μl was loaded onto an 8% polyacrylamide-7 M urea gel ( 0 . 5× TBE ) and run at 1400 V for 3 hr . The gel was then dried and exposed to a PhosphorImager screen . RNA was synthesized using the same protocol as for the in vitro transcription assays , with the exception that UTP was unlabeled . After phenol extraction and ethanol precipitation , 5 μg RNA was hybridized with a 32P-labeled primer ( ‘native torS’ or ‘HK022/torS’ ) by heating at 95°C for 5 min followed by slow cooling for 1 hr . Primer extension with the MMLV Reverse Transcriptase 1st-Strand cDNA Synthesis Kit ( Lucigen ) was carried out according to the manufacturer’s instructions . Sequencing reactions using the same primer from the primer extension assays were performed using the Sequenase Version 2 . 0 DNA Sequencing Kit ( USB ) . Phylogenetic group assignments of the prophage-carrying strains listed in Supplementary file 1 were made as described in Clermont et al . ( 2013 ) using the ClermonTyping web tool ( Beghain et al . , 2018 ) and strain sequences available from NCBI . Isolation source was identified from information in the NCBI sequence entry or linked BioSample entry ( Barrett et al . , 2012 ) . The most similar phage was identified using PHASTER ( Arndt et al . , 2016 ) and is the fully sequenced phage with the highest overall protein sequence similarity to the query prophage . Prophage completeness was assessed using PHASTER . Multiple sequence alignment was performed using MUSCLE ( Edgar , 2004 ) as implemented in SnapGene ( SnapGene , 2019 ) . | Animals and plants can all fall prey to viruses – and so can bacteria . The viruses that infect bacteria are called bacteriophages ( or phages for short ) , and they are found everywhere bacteria live and probably outnumber bacteria by at least ten to one . While some phages quickly kill every bacterial cell they infect , others enter a dormant state by inserting their DNA into the DNA of their host cell . Here they lie in wait for a signal that reactivates them , triggering the production of more phages and the death of the host cell . While the phage lies dormant its DNA may harm the host by interfering with nearby bacterial genes , or it may actually provide new genes that benefit the host . In most cases the effects of dormant phages are unknown . A bacterium known as Escherichia coli is commonly found in the intestines of humans and other mammals . It can use a nutrient called trimethylamine oxide ( TMAO ) to help it survive rapid decreases in oxygen levels that can occur in its environment . When a phage called HK022 infects E . coli , the phage enters a dormant state by inserting its DNA between two genes that are critical for E . coli to use TMAO . However , it is not clear what effect , if any , HK022 has on E . coli’s behavior . To address this question , Carey et al . used genetic approaches to study E . coli cells carrying dormant HK022 phages . The experiments showed that the bacteria lost the ability to use TMAO to survive rapid decreases in oxygen because the dormant phages switched on one of the neighboring E . coli genes . Unexpectedly , the phage achieved this by neatly replacing the gene’s own promoter – the stretch of DNA that contains information about when the gene should be switched on , and how strongly – with a substitute promoter carried in the phage’s DNA . This substitute promoter is stronger than the normal version – meaning that the gene is more active than it should be . Phages are key players in every natural population of microbes and are therefore entwined in the health of humans and the environment . The findings of Carey et al . show a new mechanism through which phages modify their hosts . In the future it may be possible to develop this mechanism into a tool to manipulate bacteria in complex environments like infection sites , for example by introducing phages that block the mechanisms that allow bacteria to tolerate antibiotics . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | Phage integration alters the respiratory strategy of its host |
The emergence of drug resistance can defeat the successful treatment of pathogens that display high mutation rates , as exemplified by RNA viruses . Here we detail a new paradigm in which a single compound directed against a ‘dominant drug target’ suppresses the emergence of naturally occurring drug-resistant variants in mice and cultured cells . All new drug-resistant viruses arise during intracellular replication and initially express their phenotypes in the presence of drug-susceptible genomes . For the targets of most anti-viral compounds , the presence of these drug-susceptible viral genomes does not prevent the selection of drug resistance . Here we show that , for an inhibitor of the function of oligomeric capsid proteins of poliovirus , the expression of drug-susceptible genomes causes chimeric oligomers to form , thus rendering the drug-susceptible genomes dominant . The use of dominant drug targets should suppress drug resistance whenever multiple genomes arise in the same cell and express products in a common milieu .
Drug-resistant variants of positive-strand RNA viruses are generated due to error-prone replication of parental , drug-susceptible genomes . Even if individual host cells are infected with only a single virus each , error rates of 1 × 10−4 per nucleotide or higher ensure that each cell will experience a mixed infection as variant progeny genomes arise within it ( Holland et al . , 1989 ) . The new variants act as templates for both continued genome synthesis and translation of new proteins . What happens in the face of a new selective pressure , such as the addition of an antiviral drug ? In most cases , antiviral treatments prevent the growth of drug-susceptible viral genomes , but any drug-resistant variants in the same cell continue to amplify according to their fitness ( Figure 1A , top ) . 10 . 7554/eLife . 03830 . 003Figure 1 . Emergence of drug-resistant variants in non-dominant and dominant drug targets in mice . ( A ) Representation of drug-resistant viral amplification when a non-dominant drug target is inhibited ( top ) but its inhibition by drug-susceptible variants when a dominant drug target , such as an oligomeric structure , is inhibited ( bottom ) . See text for discussion . ( B–K ) Tnfr1+/+ ( B , C ) or Tnfr1−/− ( d-k ) PVR-expressing mice were infected with 1 × 107 PFU Mahoney type 1 poliovirus by intramuscular inoculation and treated with 76 mg/kg/day guanidine ( B–E , green symbols ) or 10 mg/kg/day V-073 ( 1-[ ( 2-chloro-4-methoxyphenoxy ) methyl]-4-[ ( 2 , 6-dichlorophenoxy ) methyl]benzene ) ( F–K , orange symbols ) . Inoculated muscles were harvested at times indicated . Total viral yields ( PFU/mg tissue ) in muscle samples of mice treated with guanidine ( B , D ) , V-073 ( F , H , J ) or vehicles ( black ) are shown . Fold changes in the frequency of drug-resistant variants in mice treated with guanidine ( C , E ) or V-073 ( G , I , K ) are shown . Fold change calculated by dividing sample value by mean value of control mice . p-values: 0 . 02 ( B ) , 0 . 003 ( C ) , 0 . 0002 ( D ) 0 . 02 ( E ) , and 0 . 01 ( F , H , J combined ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03830 . 003 In contrast , the dominant drug target approach presented here exploits the potential interactions between the products of drug-resistant and drug-susceptible genomes within the same cell that render the drug-susceptible phenotype dominant ( Crowder and Kirkegaard , 2005 ) . The most intuitive scenario for such dominance involves highly oligomeric assemblages such as viral capsids , because drug-resistant subunits will assemble with drug-susceptible ones , and the chimeric structure is likely to be inhibited by the drug ( Figure 1A , bottom ) . This situation is akin to a dominant–negative interaction and examples of such dominance by vector-mediated expression of mutant proteins are abundant . Expression of a non-functional version of the oligomeric Tat protein of HIV drastically reduces viral yield ( Meredith et al . , 2009 ) . Similarly , co-expression of non-functional NS5A protein of hepatitis C virus inhibits replication of genomes that encode functional NS5A ( Graziani and Paonessa , 2004 ) . In poliovirus , viral genomes with mutations in several proteins , including capsid proteins and the RNA-dependent RNA polymerase , interfere with the amplification of co-transfected wild-type viral genomes ( Crowder and Kirkegaard , 2005 ) . The idea of dominant drug targeting differs from conventional dominant–negative interaction experiments in two ways . First , the dominant products are encoded by the wild-type , drug-susceptible genome that initiated the infection as well as its drug-susceptible progeny . The second unique feature is that the mixed infection of drug-susceptible and drug-resistant viruses does not result from co-infection or co-transfection . Instead , the intracellular diversity arises from the error-prone replication of the viruses themselves , even if the infection in any individual cell is initiated by a single viral genome . Here , we show that V-073 , an inhibitor of viral capsid function that is currently in development for use in the poliovirus eradication campaign ( Buontempo et al . , 1997; Collett et al . , 2008; Oberste et al . , 2009 ) , fulfills the expectations of dominant drug targeting , dramatically suppressing the growth of drug resistant viruses in murine infections and in cell culture via the formation of mixed assemblages .
The active sites of enzymes are often considered to be promising drug targets , due to a wealth of biochemical analyses and thousands of precedents . For poliovirus , one such target is 2C protein , a membrane-associated NTPase that is required for viral genome replication . The NTPase activity of protein 2C is inhibited by low concentrations of guanidine in cell culture and in solution ( Pfister and Wimmer , 1999 ) . We investigated the effectiveness of guanidine in inhibiting viral growth in mice and whether treatment with guanidine caused the selection of compound-resistant viruses in a mouse model of poliovirus infection ( Ren and Racaniello , 1992; Crotty et al . , 2002 ) . Transgenic mice that express the human poliovirus receptor ( Mendelsohn et al . , 1989 ) under the control of the murine actin promoter ( Crotty et al . , 2002 ) were bred to be homozygously deleted for TNF receptor 1 in order to delay poliomyelitis in infected mice ( See results below ) . Subcutaneous delivery of guanidine began immediately after infection and muscle tissue was harvested at 4 or 5 days post infection . Virus stocks were made from tissue and titered . Guanidine treatment caused significant reductions in total viral yield in Tnfr1+/+ PVR+/+ mice after 4 days of viral infection ( Figure 1B ) and in Tnfr1−/− PVR+/+ mice after 5 days of infection ( Figure 1D ) . These viral stocks were then titered in the presence of guanidine . Due to the heterogeneous nature of the viral population , even samples from untreated mice contain a measureable background frequency of guanidine-resistant variants . These variants naturally occur at a frequency of approximately two in one thousand . The frequency of guanidine resistance in each experimental sample is presented as the fold change from this average value . Large and significant increases in the frequency of guanidine resistance were observed in all treated mice ( Figure 1C , E ) . Therefore , newly arising guanidine-resistant virus could be readily selected during guanidine treatment in mice infected with poliovirus . This selection for drug resistance mimics the treatment-dependent emergence of drug resistance observed in patients infected with rapidly evolving viruses . Targeting the active sites of monomeric enzymes is likely to select for drug resistance . The dominant drug target hypothesis predicts that the poliovirus capsid , a highly oligomeric structure , is very likely to be a dominant drug target ( Crowder and Kirkegaard , 2005 ) . For several picornaviruses , the function of viral capsid can be inhibited by ‘WIN’ compounds , which have sixty binding sites on each virion ( Diana et al . , 1985; Smith et al . , 1986 ) , These binding sites are hydrophobic pockets arrayed about each of the twelve fivefold axes ( Diana et al . , 1985; Smith et al . , 1986 ) . Drug binding reduces capsid flexibility thereby preventing the conformational changes required for cell entry and capsid uncoating ( Lewis et al . , 1998; Dove and Racaniello , 2000 ) . A chemically related drug , V-073 , inhibits all three poliovirus serotypes and is currently in clinical development ( Buontempo et al . , 1997; Collett et al . , 2008; Oberste et al . , 2009 ) . To test the ability of V-073 to inhibit poliovirus growth and to select for resistant virus in mice , infected Tnfr1−/− PVR+/+ mice were treated with V-073 by oral gavage as described previously ( Buontempo et al . , 1997 ) . As with guanidine-treated mice , the drug regimen began immediately following infection and muscle samples were collected and titered in the absence and presence of the drug . V-073 treatment caused a significant decrease in viral titer after 4 , 5 and 7 days of infection ( Figure 1F , H , J ) . However , no increase in the frequency of V-073 resistance was observed at any time point ( Figure 1G , I , K ) . Therefore , the selection of V-073 resistance , unlike selection for guanidine-resistant variants , did not occur during murine infection . V-073 capsid inhibition provides the first in vivo example of a potential dominant drug target interaction . It was possible that a very high fitness cost accounted for the failure of V-073-resistant variants to emerge during drug treatment . To test the fitness of guanidine- and V-073-resistant viruses directly , we isolated such variants from the infected muscle tissue of mice inoculated with wild-type virus . The presence of drug-resistant viruses in the absence of any selective pressure results from the natural variation in the viral pools . Both guanidine- and V-073-resistant variants naturally occurred at a frequency of approximately one in ten thousand in samples from untreated mice . Pooled variants of both types were selected and amplified in cell culture . Drug-resistant variants grown in cell culture were propagated at low infectious doses , so that each cell was infected with no more than one virus . When these stocks were used to infect mice ( Figure 2A ) , the guanidine-resistant viruses were viable but displayed significantly reduced fitness compared to wild-type virus . On the other hand , only a slight and statistically insignificant difference in fitness between wild-type virus and the V-073-resistant virus stock was observed . Thus , guanidine-resistant viruses emerged readily under selection pressure even though their fitness is low . V-073-resistant variants did not emerge in spite of showing very little growth defect . Therefore , we conclude that fitness costs do not account for the lack of outgrowth of V-073-resistant virus . Furthermore , the fact that guanidine-resistant variants are readily selected during treatment despite their fitness cost suggest that , at least in this example , relying on high fitness cost to prevent the emergence of resistance is insufficient; additional factors must be at play to prevent the emergence of resistance from a single drug regimen . 10 . 7554/eLife . 03830 . 004Figure 2 . Differences in frequencies of guanidine-resistant and V-073-resistant viruses do not result from differences in viral fitness or murine health . ( A , B ) Total viral yield ( PFU/mg ) of mouse-adapted wild-type and guanidine-resistant virus ( p = 0 . 0006 ) or wild-type and V-073-resistant virus in muscles of Tnfr1−/− PVR+/+ mice 4 days post infection . ( C ) Survival of Tnfr1+/+ and Tnfr1−/− mice after infection . Tnfr1+/+ , n = 10; Tnfr1−/− mice , n = 15 . p = <0 . 0001 . ( D ) Viral yield ( PFU/mg ) in muscles of Tnfr1−/− mice compared to Tnfr1+/+ mice 4 days post infection . p = 0 . 035 . ( E , F ) Survival ( E ) and frequency of drug resistance ( F ) of cPVR mice in the presence and absence of treatment with 76 mg/kg/day guanidine . For both groups in survival analysis , n = 11 . Triangles in ( E ) indicate individual mice that are analyzed in ( F ) . ( G , H ) Survival ( G ) and frequency of drug resistance ( H ) of cPVR mice in the presence and absence of treatment with 3 mg/kg/day V-073 . For survival analysis , untreated mice , n = 9; treated mice n = 8 . Triangles in ( G ) indicate individual mice that are analyzed in ( H ) . Mouse tissue was harvested immediately upon the first observation of paralysis ( E–H ) . All mice were inoculated with 1 × 106 PFU virus . DOI: http://dx . doi . org/10 . 7554/eLife . 03830 . 004 To test another example of potential selection pressure for drug-resistant viruses , we monitored the viral populations present at the time that paralytic symptoms developed . Mice that express the human poliovirus receptor ( Crotty et al . , 2002 ) and retain normal TNF signaling developed paralysis beginning at 3 days post-infection ( Figure 2C ) . This is in contrast to Tnfr1−/− PVR+/+ mice , which displayed reduced pathogenesis ( Figure 2C ) . To test whether drug resistance had broken through in mice that developed poliomyelitis , muscle tissue from infected mice was harvested as soon as the animal developed signs of paralysis . When testing drug resistance in Tnfr1+/+ mice , the yield of total virus after 4 days of infection was higher than in Tnfr1−/− mice ( Figure 2D ) , perhaps providing more viral diversity on which selection pressure could act during drug treatment . Treatment of poliovirus-infected transgenic mice with guanidine throughout a course of infection did not affect pathogenesis ( Figure 2E ) . However , the compound certainly had some effect: in mice that developed poliomyelitis , large and significant increases in the frequency of guanidine-resistant virus were observed ( Figure 2F ) . On the other hand , when the few mice that developed paralysis during V-073-treatment were analyzed , no selection for V-073-resistant virus was seen ( Figure 2G ) . We conclude that , during viral growth under either extended periods of selection or conditions that led to disease , viruses resistant to guanidine were readily selected in mice while viruses resistant to V-073 were not . These data argue that treatment-dependent emergence of resistance is not an inevitable outcome of antiviral monotherapy . In this example , an increase in resistance was never observed despite the fact that drug-resistant variants with relatively high fitness arose throughout the course of infection . We propose that the poliovirus capsid constitutes the first confirmed case of a dominant drug target . To test directly whether drug-susceptible viruses were dominant over drug-resistant viruses in the cases of guanidine , V-073 and a proteinase inhibitor , rupintrivir , we used deliberate co-infection in cell culture . If drug-susceptible variants dominantly interfere with the growth of drug-resistant variants , the yield of the resistant variant should decrease when co-infected with drug-susceptible virus . For guanidine ( Figure 3A ) , a known variant , 2C-N179G ( Pincus and Wimmer , 1986 ) , was reconstructed and shown to confer complete resistance to 0 . 5 mM guanidine ( Figure 3B ) . To mimic the mix of genomes in which a new drug-resistant variant would first arise , HeLa cells were co-infected with 2C-N179G virus and increasing amounts of wild-type , drug-susceptible virus in the presence of guanidine . The multiplicity of infection ( MOI ) of the drug-resistant 2C-N179G variant was maintained at 10 plaque-forming units ( PFU ) /cell while that of the drug-susceptible virus ranged from 0 to 100 PFU/cell . As shown in Figure 3C , the yield of guanidine-resistant virus in the presence of the drug was the same whether susceptible viruses were present or not . Even when the guanidine-resistant variant was outnumbered 100:1 , its growth was unimpeded by the presence of wild-type , drug-susceptible virus ( Figure 3D ) . Therefore , targeting the viral NTPase with guanidine allows the growth of drug-resistant genomes even in the presence of an excess of drug-susceptible ones . This is consistent with the results in mice , in which guanidine-resistant variants were readily selected . 10 . 7554/eLife . 03830 . 005Figure 3 . Co-infection experiments to test genetic dominance of drug-susceptible genomes . Differential growth of defined viral variants resistant to compounds inhibitory to three different viral targets in mixed infections . ( A ) Chemical structure of guanidine , which inhibits the NTPase activity of poliovirus 2C protein . ( B ) A previously characterized guanidine-resistant poliovirus variant ( 2C-N179G ) was genetically reconstructed . The yield of 2C-N179G virus at ( C ) MOI of 10 PFU/cell and ( D ) MOI of 0 . 1 PFU/cell in the presence of guanidine and increasing amounts of wild-type , drug-susceptible virus in the same cells was determined . ( E ) Chemical structure of V-073 , which inhibits the function of the icosahedral poliovirus capsid ( F ) A V-073-resistant poliovirus variant ( VP3-A24V ) was identified and genetically reconstructed in the Mahoney type 1 viral background . The yield of VP3-A24V virus in the ( G ) presence and ( H ) absence of V-073 and increasing amounts of wild-type , drug-susceptible virus in the same cells was determined . ( I ) V-073-resistant variant VP3-A24V and ( J ) VP1-I94F in the epidemic Hispaniola strain of poliovirus were identified and reconstructed . The yields of ( K ) VP3-A24V and ( L ) VP1-I94F Hispaniola type 1 poliovirus in the presence of V-073 and increasing amounts of wild-type , drug-susceptible Hispaniola virus in the same cells were determined . ( M ) Chemical structure of rupintrivir ( trans- ( 4S , 2R , 5S , 3S ) -4-{2-4- ( 4-fluorobenzyl ) -6-methyl-5’-[ ( 5-methylisoxazole ) -3-carbonylamino]-4-oxoheptanoylamino}-5- ( 2-oxopyrrolidin-3-yl ) pent-2-enoic acid ethyl ester ) , which targets the active site of poliovirus 3C proteinase . ( N ) A rupintrivir-resistant poliovirus variant ( 3C-G128Q ) was identified and genetically reconstructed . The yield of 3C-G128Q virus at ( O ) an MOI of 10 PFU/cell and at ( P ) 0 . 1 PFU/cell in the presence of rupintrivir and increasing amounts of wild-type , drug-susceptible virus in the same cells was determined . *p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 03830 . 005 To test whether viruses susceptible to V-073 ( Figure 3E ) were genetically dominant in cell culture , we selected a V-073-resistant variant of Mahoney type 1 poliovirus by repeated passage at low multiplicities of infection so that any pre-existing genome that was V-073-resistant could be amplified . Such variants were readily isolated and the responsible mutation , VP3-A24V ( Figure 3F ) was identified . Unlike with guanidine , co-infection of wild-type , drug-susceptible virus with V-073 resistant virus resulted in a dominant inhibition in drug-resistant virus growth ( Figure 3G ) . As in mice , this did not result from the decreased fitness of V-073-resistant virus ( Figure 3H ) . Even at only a fivefold excess of V-073-susceptible virus , the yield of drug-resistant virus was reduced by 98% . A new drug-resistant variant created within a cell is likely to be greeted by an excess of drug-susceptible genomes much greater than this 5:1 ratio , making genetic dominance of the drug-susceptible virus a viable mechanism for the suppression of V-073 resistance in the mouse model . To determine whether selection for V-073-resistant virus was also suppressed in a viral strain isolated from a poliomyelitis patient ( Kew et al . , 2002 ) , we tested whether the Hispaniola strain of vaccine-derived neurovirulent poliovirus , known to be susceptible to V-073 ( Oberste et al . , 2009 ) , was dominant over two different V-073-resistant variants . One such variant harbored the VP3-A24V mutation mentioned above and the second contained an I194F mutation in another capsid protein , VP1 ( Kouiavskaia et al . , 2011; Liu et al . , 2012 ) . Each of these mutations alters residues that line the pocket known to bind V-073 , confers drug resistance and allows growth comparable to wild-type virus in the absence of drug ( Figure 3I , J ) . Like the laboratory-passaged Mahoney type 1 strain ( Figure 3G ) , these V-073-resistant Hispaniola-derived variants were profoundly inhibited by the presence of the parental strain ( Figure 3K , L ) . Therefore , drug-susceptible virus was genetically dominant over drug-resistant variants in both lab-passaged and disease-causing strains , suggesting that V-073 is a general dominant drug target inhibitor . Viral proteinases are often effective antiviral drug targets ( Malcolm , 1995; Patick and Potts , 1998 ) , so it was of interest to determine whether inhibition of the major proteinase of poliovirus , termed 3C , was a dominant drug target . Rupintrivir ( Figure 3M ) is an active site inhibitor of enterovirus 3C proteinases ( Patick et al . , 1999 , 2005; Wang et al . , 2011 ) , and is , in cell culture , the most effective inhibitor of poliovirus growth tested here ( Figure 3N ) . Although no rupintrivir-resistant viruses have been reported previously , we could select such a variant during multiple viral passages in the presence of the drug . The phenotype of the rupintrivir-resistant variant ( Figure 3N ) was caused by a single amino acid change , G128Q , in the 3C coding region . As was the case with guanidine , the yield of rupintrivir-resistant virus grown in the presence of rupintrivir was relatively unperturbed by the presence of susceptible virus ( Figure 3O ) , up to ratios of 100:1 of susceptible to resistant virus ( Figure 3P ) . Thus , viral proteinase 3C , like viral NTPase 2C , is a non-dominant drug target . We hypothesized that the lack of infectivity of virions made in V-073-treated cells was due to the formation of chimeric capsids ( Figure 1A , bottom ) . To test explicitly whether capsid proteins encoded by V-073-susceptible and V-073-resistant genomes assemble into mixed capsid structures during co-infection ( Figure 4A ) , which could be distinguished from separate , homogeneous structures ( Figure 4B ) , we constructed viruses whose capsid proteins could be independently identified by antibodies . To this end , a FLAG epitope was engineered into the VP3-A24V virus and an HA epitope was engineered into wild-type , drug-susceptible virus ( Figure 4C ) . Following a strategy employed in foot and mouth disease virus ( Seago et al . , 2012 ) , the tags were inserted into a flexible , antibody-accessible loop in the assembled capsid structure . These constructs enabled immunoprecipitation of intact , viable virions that could be assessed for their drug susceptibility by soaking the virus in the capsid inhibitor before titering . In single-cycle infections , the FLAG-tagged virus retained its V-073-resistant phenotype , and the HA-tagged virus its drug susceptibility ( Figure 4D ) . Anti-FLAG immunoprecipitation of lysates obtained from these single infections retrieved , as expected , FLAG-tagged drug-resistant but not HA-tagged drug-susceptible virus ( Figure 4E ) or proteins ( Figure 4F , lanes 1 and 2 , respectively ) . When the two lysates were pooled , only FLAG-tagged , V-073-resistant virus was recovered ( Figure 4E , F , lane 3 ) , demonstrating that the immunoprecipitation was specific for capsids harboring proteins from the V-073-resistant genome . When the drug-resistant , FLAG-tagged virus and the drug-susceptible HA-tagged virus were co-infected into cells and the anti-FLAG immunoprecipitation was performed , the immunoprecipitated virions displayed reduced infectivity when soaked in the capsid inhibitor prior to titering ( Figure 4E ) . Furthermore , the anti-FLAG immunoprecipitate contained both FLAG- and HA-tagged VP1 proteins ( Figure 4F , lane 4 ) . Therefore , chimeric virions with reduced drug-resistance were formed in cells co-infected with V-073 susceptible and V-073-resistant viruses , even at the 1:1 ratio tested here . The formation of mixed capsids with reduced drug resistance is the most plausible explanation for the observed genetic dominance of V-073-susceptible variants and the blunted selection for drug-resistance during growth in mice . 10 . 7554/eLife . 03830 . 006Figure 4 . V-073-susceptible and V-073-resistant virus form mixed capsids during co-infection . ( A , B ) Depiction of ‘chimeric’ ( A ) and ‘pooled’ ( B ) viral preparations . Red subunits represent V-073-resistant , FLAG-tagged capsids , and grey subunits represent V-073-susceptible , HA-tagged capsids . Anti-FLAG antibodies are represented in pink . ( C ) Construction of FLAG- and HA-tagged virus . Sequence of FLAG and HA peptides inserted into antigenic site 1 in the poliovirus genome , replacing amino acids 96-101 and 97-102 of VP1 , respectively . ( D ) Viral yield ( PFU/ml ) of the wild-type , FLAG-tagged and HA-tagged poliovirus variants in the absence or presence of V-073 over a single infectious cycle at an MOI of 10 . Data represent means ± s . t . d . of two replicates . ( E ) Fraction of capsids with V-073-resistant phenotype in anti-FLAG immunoprecipitates as labeled . HA-tagged sample was below the limit of detection . Data represent means ± s . t . d . of two replicates . ( F ) Immunoblots of VP1 proteins collected by anti-FLAG immunoprecipitation of cell lysates prepared from infections indicated . ‘Pooled’ sample contains lysates from independently grown FLAG- and HA-tagged viruses , whereas ‘co-infected’ sample was obtained from cells infected with both FLAG- and HA-tagged variants . All variants were infected at an MOI of 10 PFU/cell . DOI: http://dx . doi . org/10 . 7554/eLife . 03830 . 006 For a viral product to be a dominant drug target , drug-susceptible virus must inhibit the growth of drug-resistant virus . Relying on this property and taking advantage of the intrinsic genetic diversity in ‘wild-type’ virus stocks , we used guanidine and V-073 to develop a simple MOI-based screen for compounds that inhibit dominant drug targets . Similar to the heterogeneity observed in virus stocks from mouse muscles , any stock of lab-passaged ‘wild-type’ virus contains pre-existing drug-resistant variants . As depicted in Figure 5A , when a ‘wild-type’ virus stock infects cells at low MOI , any pre-existing drug-resistant variant in the stock will infect a cell in isolation ( left panel ) and have the opportunity to express its phenotype . Thus , it can be selected in the presence of drug , regardless of the drug target type . However , at high MOI , drug-resistant variants will occupy cells in concert with drug-susceptible variants ( Figure 5A , middle ) . In cases of dominant drug targets , drug-susceptible viruses are genetically dominant and the inhibitor will prevent the outgrowth of drug-resistant variants . Consistent with this logic , when HeLa cells were infected at an MOI of 1 infectious virus per cell in the presence of either guanidine or V-073 , the yields of guanidine-resistant and V-073-resistant viruses were similar ( Figure 5B ) . At higher MOIs , the yield of guanidine-resistant progeny increased , as would be expected because more virus was present in the inoculum and nothing interfered with the outgrowth of the drug-resistant variants ( Figure 5C , green ) . However , the yield of V-073-resistant progeny decreased at high MOIs ( Figure 5B and Figure 5C , orange ) . To test whether the inhibition in growth of V-073-resistant virus was an intracellular event , as predicted if the formation of mixed capsids suppressed the drug-resistant phenotype , the virus stock was soaked in V-073 before the infections at different MOIs were performed ( Figure 5A , right panel ) . When entry of the drug-susceptible viruses was thus inhibited , a dose-dependent increase in the yield of V-073-resistant virus was observed even at high MOIs ( Figure 5C , dotted orange ) . The virus stock used was deliberately prepared from cells infected at low MOI to ensure that drug-resistant genomes were packaged in drug-resistant capsids . Thus , failure to select V-073-resistant genomes at high MOIs could be overcome by preventing the entry of the drug-susceptible genomes . 10 . 7554/eLife . 03830 . 007Figure 5 . Assaying the dominance of drug-susceptible genomes via the MOI-dependent detection of drug-resistance . ( A ) Cartoon of proposed infection at low MOI ( left ) and high MOI without V-073 pretreatment ( middle ) or with V-073 pretreatment ( right ) . Grey represents drug-susceptible variants , red represents drug-resistant variant and drug is depicted as yellow triangles . ( B ) Drug-resistant virus plaques following infections with wild-type Hispaniola strain at increasing MOIs in the presence of guanidine or V-073 . ( C ) Drug-resistant viral yield ( PFU/ml ) in presence of guanidine ( green line ) or V-073 without pretreatment of viral stock ( solid orange line ) or with pretreatment ( dotted orange line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03830 . 007 This MOI-dependent emergence of drug resistance provides a rapid screen to identify inhibitors of dominant drug targets . Its utilization requires only the identification of a compound that inhibits viral growth , the ability to infect at MOIs above 1 virus/cell in cell culture , and the ability to titer drug-resistant viruses after infection at various MOIs . This assay does not require knowledge of the drug target , although such information will become even more interesting if it is determined that it is a dominant drug target .
In this study we have shown that a single inhibitor of a positive-strand RNA virus , capsid poison V-073 , can both inhibit viral growth and suppress the emergence of newly arising drug-resistant variants during infection of mice and of cultured cells . Deliberate mixed infections performed in cell culture showed that the formation of chimeric capsids , which contained both drug-inhibitable and drug-resistant proteins , correlated with the suppression of drug resistance . These observations support the hypothesis that targeting the viral constituents of an oligomeric assemblage , such as a viral capsid , can lead to the genetic dominance of drug-susceptible genomes . The magnitude of the suppression of V-073 resistance in murine infection was higher than expected given an assumption that all viruses in the inoculum contributed to the eventual infection . There are two potential sources of V-073-resistant viruses in infected mice . The first is the quaisispecies present in any inoculum . At an initial infectious dose of 107 viruses per mouse , there should have been as many as 104 pre-existing drug-resistant variants in the inoculum ( Holland et al . , 1989 ) . Why were they unable to emerge from the infection ? Given the effectiveness of the innate immune response , there is often a minimum effective dose of inoculated microorganism . For poliovirus in this mouse model , more than 104 viruses are required to achieve any detectable viral growth in muscle tissue ( data not shown ) . This threshold predicts correspondingly fewer pre-existing drug-resistant variants in the effective inoculum . The second source of V-073-resistant viruses is the quasispecies addressed in the present study and depicted in Figure 1A: the newly arising variants formed within each individual cell by error-prone replication . Therefore , we conclude that V-073 was effective at suppressing resistance even as a monotherapy because V-073-susceptible variants suppress the growth of V-073-resistant variants as they arise intracellularly throughout the course of infection . In the mouse experiments , the compounds were administered within hours of infection . Such rapid treatment is unlikely to occur in patients , except in cases where drugs are delivered prophylactically . However , a delay in treatment will not alter the degree to which dominant drug target inhibitors suppress the selection of newly arising resistant viral genomes from drug-susceptible parents . Late treatment may be directed at a population that contains 0 . 2% drug-resistant genomes , but many of these will have been trans-encapsidated by the other viral variants present in their cells of origin . Thus , for a capsid inhibitor , trans-encapsidation of newly arising drug-resistant genomes by the intracellular excess of drug-susceptible capsids should provide an additional mechanism to reduce the reservoir of drug-resistant genomes packaged in drug-resistant capsids . The use of dominant drug targets is expected to reduce the number of antiviral compounds necessary for effective treatment of individuals and populations . In the case of V-073 , the only poliovirus antiviral currently in human trials , there has been hesitation to use it as a monotherapy ( Collett et al . , 2008; De Palma et al . , 2008 ) . Our data suggest that V-073 may be sufficient for the effective treatment of poliovirus infection . How can a dominant drug target be identified experimentally ? The most direct way for a drug-susceptible virus to be genetically dominant over a drug-resistant one is the formation of mixed oligomers , as exemplified here . For poliovirus , co-transfection experiments identified known oligomers such as the viral capsid and the RNA-dependent RNA polymerase as viral products whose fitness could be suppressed by the presence of defective subunits ( Crowder and Kirkegaard , 2005 ) . Other candidates also emerged , however , whose mechanisms are less straightforward and remain under investigation: intramolecularly cleaving protease 2A and an RNA structure required for protein priming were also identified as potential dominant drug targets ( Crowder and Kirkegaard , 2005 ) . Here , the MOI-dependent of the outgrowth of drug-resistant variants in cell culture is suggested to be a rapid screen for whether an antiviral compound inhibits a dominant drug target or not . How can dominant drug targets be identified a priori ? All dominant drug targets should share a unique feature: binding of the drug to its target generates a product that is toxic to the drug-resistant variant . Three factors are required to accomplish this effect . 1 ) The drug-bound product must interact with proteins made by the drug-resistant genome; 2 ) the interaction between drug-bound products and resistant proteins must inhibit an essential step in the infectious cycle of the virus; and 3 ) the ratio of drug-susceptible to drug-resistant products synthesized in the cell must be high enough to poison the function of the resistant variant . For viral structural proteins , criterion ( 1 ) is met through the direct binding of drug-inhibited and drug-resistant subunits within the capsid or core complex . However , a dominant drug target does not have to be an oligomeric protein itself . For example , poliovirus 2A protease has been illuminated as a potential dominant drug target , because when it fails to make an obligatorily intramolecular cleavage in the viral polyprotein , the uncleaved product is toxic ( Crowder and Kirkegaard , 2005 ) . Criterion ( 1 ) dictates that the toxic product generated by drug binding retains the ability to interact with other proteins . This suggests one reason for a current lack of dominant drug target inhibitors . Small-molecule screens are often designed to detect loss of viral product function , rather than the deranged function required for the drug-bound , susceptible viral product to join and poison drug-resistant complexes . V-073 , for example , hyper-stabilizes viral capsids . Thus , the drug-inhibited subunits can still insinuate themselves into capsid complexes . A theoretical capsid inhibitor that prevented assembly by binding at an intersubunit interface should display the characteristics of a non-dominant drug target inhibitor , because the drug-susceptible subunits would not have the opportunity to poison the drug-resistant subunits . To fulfill criterion ( 2 ) , we must identify stages of the viral infectious cycle that rely on the cooperation of proteins donated by multiple genomes . Capsid and cores are not the only such structures . Indeed , trans-assembly of highly oligomeric complexes is a hallmark of RNA viral infection , exemplified by the polymerase and polymerase-containing oligomers of positive-strand RNA viruses ( Crowder and Kirkegaard , 2005; Marcotte et al . , 2007; Kemball et al . , 2010 ) and the large nucleoprotein-RNA complexes that form during negative-strand RNA replication ( Reviewed in reference Morin et al . ( 2013 ) ) . For drug-susceptible subunits to inhibit chimeric complexes , they must be abundant enough to dictate the overall phenotype of the structure , fulfilling criterion ( 3 ) . Again using the poliovirus capsid as an example , we can calculate the approximate number of susceptible subunits required to render the entire virion drug-susceptible . Assuming free mixing of drug-susceptible and drug-resistant capsids and relative abundances that reflect the ratio of their genomes , the data argue that about twenty subunits , or one third , of the capsid proteins must be drug-susceptible to poison the complex . For the abundance of drug-susceptible and drug-resistant proteins to mirror the representation of the RNA that encodes them , both genomes must both continue to amplify or must both be inhibited . This suggests another reason that inhibitors of structural proteins , which are predicted to be dominant drug targets , are often overlooked . V-073 , for example , would not have been detected as a viable inhibitor in a screen that used a decrease in genome amplification , protein abundance or innate immune response activation during a single cycle of infection as a proxy for viral inhibition . Similarly , replicons , which cannot perform second rounds of infection , are useful tools for some kinds of screens but not for seeking inhibitors of many dominant drug targets . The paradigm of targeting inhibitors to structures whose function can be inhibited by non-functional subunits has applications to other viruses , especially those with very high error rates due to the repeated amplification of RNA . Thus , capsids , cores and other oligomeric assemblages of rhinovirus , enterovirus 71 , hepatitis C and dengue virus are likely to be dominant drug targets . This strategy may not apply to retroviruses , because the step with the least fidelity , reverse transcription , gives rise to a low copy number DNA intermediate . However , bona fide DNA viruses , present at a high copy number per cell , should also be amenable to this dominant drug targeting strategy . For other error-prone , intracellular genetic amplification events , such as the growth of bacterial and eukaryotic intracellular pathogens and the gene duplication associated with oncogenesis , the choice of oligomeric drug targets may also suppress the outgrowth of drug-resistant variants immediately after their inception . Deliberate targeting of oligomeric structures should reduce the numbers of inhibitors needed for effective treatment , but will require revision of decision-making processes during drug development .
Poliovirus strains were either Mahoney type 1 poliovirus or the vaccine-derived type 1 Hispaniola strain ( Kew et al . , 2002 ) . Solid guanidine hydrochloride from Sigma–Aldrich ( St . Louis , MO ) was dissolved to a concentration of 0 . 5 mM in water . Rupintrivir ( trans- ( 4S , 2R , 5S , 3S ) -4-{2-4- ( 4-fluorobenzyl ) -6-methyl-5'-[ ( 5-methylisoxazole ) -3-carbonylamino]-4-oxoheptanoylamino}-5- ( 2-oxopyrrolidin-3-yl ) pent-2-enoic acid ethyl ester ) from International Laboratory ( San Bruno , CA ) was dissolved to a concentration of 0 . 1 mM in DMSO . V-073 ( 1-[ ( 2-chloro-4-methoxyphenoxy ) methyl]-4-[ ( 2 , 6-dichlorophenoxy ) methyl]benzene ) from ViroDefense ( Rockville , MD ) was dissolved to a concentration of 24 mM in DMSO . For all infections , HeLa cell monolayers were infected with poliovirus for 5 hr at 37°C , except for infections with the VP1-I194F Hispaniola variant which were carried out at 34°C to compensate for its slight temperature sensitivity . For growth of single variants in the presence and absence of drug , the MOI was 10 PFU/cell . For co-infections and the MOI-dependent selection experiments , the various MOIs are indicated on the figures and were performed in the presence of the drug of interest . Following the infectious cycle , intracellular virus was harvested by washing cells with PBS+ ( PBS supplemented with 0 . 1 mg/ml MgCl2 and CaCl2 ) , resuspending in 0 . 5 ml PBS+ and subjecting stocks to three freeze–thaw cycles before clearing the cell debris by low-speed centrifugation . Viral titer was determined by plaque assay as described previously ( Kirkegaard , 1990 ) in the absence or presence of the drug as indicated . Drug concentrations were 0 . 5 mM guanidine , 0 . 1 μM rupintrivir and 24–47 μM V-073 . For stocks incubated with V-073 prior to infection , the concentration was 0 . 47 μM . Co-infections were performed with the Mahoney strain except where indicated . The MOI-dependent selection was performed with the Hispaniola strain . The guanidine-resistant variant , 2C-N179G has been previously published ( Pincus and Wimmer , 1986 ) . The rupintrivir-resistant variant was identified as follows: first , three rounds of plaque purification were performed with rupintrivir in the overlay . Independent plaque isolates were used to infect HeLa cell monolayers at MOIs of approximately 0 . 25 PFU/cell for 10 hr with rupintrivir in the medium . From this stock , a purified plaque was isolated and RNA was extracted and the sequence was determined . The 3C-G128Q mutation was identified and cloned into the T7pGempolio vector using the splicing by overlapping extension PCR method ( Lefebvre et al . , 1995 ) , changing codon 128 in the 3C coding region from GGT to GAA . Viral RNA was made by linearizing 10 μg of the plasmid with EcoRI , phenol:chloroform extracting the DNA , and transcribing from the linearized DNA using the MEGAscript kit ( Applied Biosystems , Foster City , CA ) . Viral RNA was transfected into HeLa cells with Lipofectamine 2000 ( Invitrogen , Carlsbad , CA ) using the manufacturer's protocol and agar overlays containing rupintrivir were added 4 hr post transfection to confirm the rupintrivir-resistant phenotype . Stocks of the virus were made in the presence of rupintrivir . V-073-resistant mutations in the Hispaniola strain have been published ( Kouiavskaia et al . , 2011; Liu et al . , 2012 ) . The V-073-resistant Mahoney type1 mutant was identified by three rounds of plaque purification with V-073 in the overlay . Viral RNA was extracted and reverse-transcribed as described above and the amino acid change was identified by sequencing with primer 5′-GTGGATTACCTCCTTGGAAATG-3′ . Once the VP3-A24V mutation was identified it was introduced in isolation into the T7pGEM polio plasmid . Transfection , stock preparation , and confirmation of the compound-resistant phenotype were performed as described above for 3C-G128Q; the change in codon 24 of the VP3 coding region was from GCG to GTA . The FLAG tag was inserted into the VP3-A24V T7pGempolio vector using the splicing by overlapping extension ( SOE ) PCR method ( Lefebvre et al . , 1995 ) . The forward primer was: 5′-ACCATTATGACCGTGGATAACCCAGACTACAAAGATGATGATGATAAGCTATTTGCAGTGTGGAAGA-3′ and the reverse primer was: 5′-TCTTCCACACTGCAAATAGCTTATCATCATCATCTTTGTAGTCTGGGTTATCCACGGTCATAATGGT-3′ . The HA tag was inserted into the wild type T7pGempolio vector also by the SOE method using the forward primer: 5′-ATTATGACCGTGGATAACCCAGCTTATCCTTATGATGTCCCTGATTATGCTAAGCTATTTGCAGTGTGGAAGATC-3′ and the reverse primer: 5′- ATCTTCCACACTGCAAATAGCTTAGCATAATCAGGGACATCATAAGGATAAGCTGGGTTATCCACGGTCATAAT-3′ . RNA was in vitro transcribed , transfected and virus stocks were made from individual plaque isolates as described above . HeLa cells were infected with either FLAG-tagged or HA-tagged virus at an MOI of 10 PFU/cell or they were co-infected with both viruses , each at an MOI of 10 PFU/cell . For the control ‘pooled’ sample , both viruses were grown separately and then pooled for the immunoprecipitation . Virus was grown for 7 hr at 37°C . Cells were washed once with TBS ( 50 mM Tris-Cl , 150 mM NaCl , pH 7 . 4 ) and then harvested in TBS . Stocks were made by freeze-thawing cells three times and cell debris was cleared by centrifuging at 2500 g for 5 min . Capsids were immunoprecipitated using the Anti-FLAG M2 Affinity Gel ( Sigma–Aldrich ) according to the immunoprecipitation protocol . Beads were incubated with virus stock for 2 . 5 hr at room temperature followed by three 15 min washes with TBS before releasing the capsid from the antibody by competitive elution with the 3× FLAG peptide ( Sigma Aldrich ) . Samples were prepared for SDS-PAGE gel by boiling for 5 min in protein sample buffer ( final concentration: 62 . 5 mM Tris , 2% SDS , 10% glycerol , 100 mM DTT , 0 . 0005% bromophenol blue ) and analyzed on a 12% SDS-PAGE gel . Proteins were visualized by immunoblot . For the anti-FLAG immunoblot , a monoclonal anti-FLAG M2 antibody was used ( Sigma–Aldrich ) at a 1:1000 dilution in 2% BSA in TBST ( TBS with 0 . 05% Tween ) . For anti-HA immunoblot , the THE HA Tag Antibody ( GenScript , Piscataway , NJ ) was used according to the manufacturer's protocol . Previously described transgenic cPVR mice ( Crotty et al . , 2002 ) and Tnfr1−/− PVR+/+ mice ( described below ) were infected with Mahoney type 1 virus by intramuscular inoculation . Mice were inoculated with 1 × 106 PFU for all experiments except the extended drug-treatments in which Tnfr1−/− PVR+/+ mice were inoculated with 1 × 107 PFU . For guanidine treatments , female mice were injected subcutaneously with 0 . 5 ml of 100 mM guanidine in PBS twice daily . Control mice were injected with PBS only . V-073 was delivered to mice by oral gavage in 100 μl corn oil . Control mice were given DMSO in corn oil . To titer virus in muscle samples taken , calf muscles were harvested , homogenized in 1 ml PBS+ , subjected to one round of freeze-thaw , cleared by low-speed centrifugation and titered as described above . V-073-resistant virus was titered by infecting between three and twenty plates at an MOI of 0 . 02 PFU/cell and adding 24 μM V-073 in the overlay; low amounts of virus were used to minimize toxicity from wild-type virus . As many plates as needed were infected in order to observe at least ten V-073-resistant plaques . If fewer than ten plaques were observed , the observed value was assumed to be k and the Poisson distribution was used to determine the highest possible λ value with a 5% likelihood ( Poisson , 1837 ) . Tnfr1−/− mice that express the human poliovirus receptor were derived by crossing cPVR mice with Tnfr1tm1Imx/J mice ( JAXS , Bar Harbor , ME ) and breeding both the transgenic PVR and the Tnfr1−/− allele to homozygosity . Stocks of mouse-adapted wild-type and drug-resistant virus were made as follows: Using viral stocks obtained from mouse muscle tissues harvested 4 days post infection , plaque assays were performed by adding overlays containing V-073 , guanidine or no drug . Eight to ten large plaques were picked for each condition , pooled and amplified for two cycles in the presence of V-073 or guanidine or in the absence of drug , depending on the conditions under which the plaques were originally grown . Tnfr1−/− PVR+/+ mice were infected with 1 × 106 PFU of these mouse-adapted wild-type , V-073-resistant or guanidine-resistant virus stocks . 4 days post infection , muscles were harvested and made into viral stocks for titering ( as outlined above ) . For co-infection experiments , a two-tailed equal variance Student's t test was used to determine p values . For co-infections , n = 3 for all viral yield data and each point was compared to the yield of drug-resistant virus in the absence of drug-susceptible virus . For total viral yield in mouse muscles , a Mann–Whitney t test was used . For V-073-treated Tnfr1−/− samples , all days were combined for analysis by converting values to a fraction of the control mean ( i . e . titer of sample/control average ) . For the frequency of drug resistance , a two-tailed equal variance Student's t test was used to determine p values . | Treating a viral infection with a drug sometimes has an unwanted side effect—the virus quickly becomes resistant to the drug . Viruses whose genetic information is encoded in molecules of RNA mutate faster than DNA viruses and are particularly good at developing resistance to drugs . This is because the process of copying the RNA is prone to errors , and by chance some of these errors , or mutations , may allow the virus to resist the drug's effects . Treating viral infections with most drugs destroys the viruses that are susceptible to the drug and inadvertently ‘selects’ for viruses that are resistant to the drug's effects . These drug-resistant viruses are harder to treat and often require physicians to switch between different drugs . Sometimes these new drug-resistant viruses spread and these new infections cannot be treated with drugs that would have worked in the past . So far , the best strategy to prevent drug-resistant viruses from growing in patients is to use multiple drugs , such as the life-saving treatments for HIV infection . However , for many viral infections—such as those that cause the common cold , dengue fever , Ebola , and polio—no drugs are yet available to treat infected people . Moreover , there are concerns that , if a new drug is used on its own , the viruses will quickly develop resistance to the drug and render it ineffective . Tanner et al . now show that an antiviral drug that interferes with the formation of the outer layer ( or capsid ) of the poliovirus inhibits the emergence of drug resistance . The drug , called V-073 , is currently being tested as a treatment for poliovirus and will be useful in the worldwide eradication effort . Tanner et al . show that treating poliovirus-infected mice with V-073 does not select for drug-resistant strains of the virus—and provide evidence that this occurs because the drug targets an assemblage of proteins . The poliovirus capsid is assembled from a mix of proteins from different naturally occurring strains of the virus within the infected cell . A new strain of virus is always ‘born’ into a cell that is already infected by other viruses , which could be thought of as its parents , cousins and siblings . A new drug-resistant virus will therefore be forced to mix its capsid proteins with those of its ‘family’ members , who are all drug-sensitive . These hybrid capsids will remain vulnerable to the drug—and in this way , the resistant strains do not become the dominant form of the virus . Tanner et al . also discovered a way to screen for drugs that have a similar resistance-blocking effect . These drugs would target capsids , or other viral structures made up of a mix of proteins from different virus strains . Such drugs might be useful against other viruses including the ones that cause the common cold , hepatitis C , or dengue fever . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2014 | Dominant drug targets suppress the emergence of antiviral resistance |
In order to produce movements , muscles must act through joints . The translation from muscle force to limb movement is mediated by internal joint structures that permit movement in some directions but constrain it in others . Although muscle forces acting against constrained directions will not affect limb movements , such forces can cause excess stresses and strains in joint structures , leading to pain or injury . In this study , we hypothesized that the central nervous system ( CNS ) chooses muscle activations to avoid excessive joint stresses and strains . We evaluated this hypothesis by examining adaptation strategies after selective paralysis of a muscle acting at the rat’s knee . We show that the CNS compromises between restoration of task performance and regulation of joint stresses and strains . These results have significant implications to our understanding of the neural control of movements , suggesting that common theories emphasizing task performance are insufficient to explain muscle activations during behaviors .
The question of how the central nervous system ( CNS ) determines motor commands is usually answered in terms of task performance ( Todorov and Jordan , 2002; Guigon et al . , 2007; Scott , 2004 ) : the CNS activates muscles according to how they contribute to task goals such as grasping an object or escaping a predator . Additional criteria based on energetics ( Todorov and Jordan , 2002; Izawa et al . , 2008; Kurtzer et al . , 2006 ) , dynamics ( Uno et al . , 1989 ) , or kinematics ( Flash and Hogan , 1985 ) can further constrain muscle activations , and the specific criteria used by the CNS to specify muscle activations underlying behavior has been the topic of considerable research ( Scott , 2004; Todorov , 2004; Alessandro , 2016; Kistemaker et al . , 2014; Prilutsky , 2000 ) . This previous work , however , has generally ignored another critical aspect of muscle actions: how muscles act on internal joint structures such as ligaments or articular cartilage between bones ( Herzog et al . , 2003 ) . Although muscle actions on internal joint structures have been considered in clinical biomechanics or sports medicine in the context of injuries ( Farrokhi et al . , 2013; Pal et al . , 2012; Felson , 2013; Konrath et al . , 2017 ) , their more general role in the context of the neural control of movement and how they are balanced with task goals remains poorly understood . We examine these issues in this study , evaluating the hypothesis that the nervous system chooses muscle activations to achieve task performance while minimizing internal joint stresses and strains . Evaluating whether the CNS regulates stresses and strains within internal joint structures is challenging , primarily because of the complexity of joint mechanics . Since muscles can only affect task performance by acting through joint structures ( Herzog et al . , 2003 ) , any manipulation affecting joint structures will usually also affect task performance . For example , damage to a ligament ( Gutierrez et al . , 2009; Needle et al . , 2014; O'Connor et al . , 1993 ) will alter both the distribution of stresses and strains within the joint ( Chen et al . , 2017; Gardinier et al . , 2013 ) as well as how muscle forces are transmitted across the joint ( Boeth et al . , 2013 ) . Similarly , although removal of sensory feedback from joint structures might compromise neural control of joint stresses and strains ( O'Connor et al . , 1993; O'Connor et al . , 1992; Salo et al . , 2002; Solomonow , 2006; Solomonow and Krogsgaard , 2001; Johansson et al . , 1991 ) , this sensory feedback can also provide information about task performance ( Ferrell et al . , 1985; Sjölander et al . , 2002 ) . These challenges have made it difficult to reach clear conclusions as to whether the CNS controls internal joint stresses and strains . In the experiments reported here , we overcome these challenges by taking advantage of the unique properties of the rat knee joint ( Figure 1a ) . Activation of the quadriceps muscles vastus lateralis ( VL ) and vastus medialis ( VM ) in the rat produces very similar forces at the distal tibia and so will produce similar joint torques ( Sandercock et al . , 2018 ) . These muscles therefore have redundant contributions to task performance . On the other hand , VL and VM produce opposing mediolateral forces on the patella ( Sandercock et al . , 2018; Lin et al . , 2004; Wilson and Sheehan , 2010 ) . In the rat , these mediolateral patellar forces are not transmitted to the tibia but are balanced by contact forces between the patella and the femur within the trochlear groove ( Sandercock et al . , 2018 ) . To avoid overloading patellofemoral contact forces and prevent joint pain or injury , the CNS should therefore balance activation of VM and VL to minimize net mediolateral forces on the patella while producing the knee torque necessary to achieve task goals . Since VM and VL in the rat have redundant contributions to task performance but opposing contributions to joint stresses , we can use the rat knee joint to design tractable experiments to evaluate whether the CNS regulates joint stresses and strains . We examined neural adaptation strategies following paralysis of VL ( Figure 1b ) . If the CNS prioritizes task performance , the easiest adaptation strategy would be to increase activation of VM . Because VM and VL have redundant contributions to task performance , this strategy would restore joint kinematics without requiring changes in other muscle activations . This straightforward restoration of task performance , however , would come at the cost of unbalanced mediolateral patellar forces . On the other hand , if the CNS prioritizes joint integrity , it should decrease activation of VM and increase activation of rectus femoris ( RF ) since RF produces minimal mediolateral force on the patella . This strategy to minimize joint stresses , however , would come at the cost of a more complex , potentially incomplete restoration of task performance: increased activation of RF will restore the knee torque lost by VL paralysis but will also introduce an additional hip flexion torque . This strategy will therefore either require compensatory activity in hip extensor muscles or introduce residual deviations in joint kinematics . Finally , if the CNS considers both task performance and internal joint stresses , it might preferentially increase RF activation while maintaining activation of VM . This strategy would minimize mediolateral forces on the patella while also limiting deviations of joint kinematics , reflecting a compromise between control of task performance and internal joint state . Observing either of the last two strategies would demonstrate that the CNS considers internal joint stresses since the preferential increase of RF activation reflects minimization of mediolateral patellar forces . Note that vastus intermedius ( VI ) is very small in the rat and so is unlikely to contribute substantially to the adaptation following VL paralysis . Our results show that rats compensated for the loss of VL by preferentially increasing activation of RF , consistent with the hypothesis that the CNS chooses muscle activations to minimize stresses and strains to internal joint structures .
Cutting the branch of the quadriceps nerve innervating VL abolished the large majority of VL activation during locomotion , as illustrated in Figure 2a–b . The loss of EMG persisted over the 7 week adaptation period , demonstrating the absence of VL reinnervation . Consistent with this loss of activation , the mass of the denervated VL measured at the end of the experiment ( i . e . 8 weeks after nerve cut ) was significantly lower ( p<0 . 001 ) than the mass of the intact contralateral VL ( Figure 2c ) , reflecting atrophy of the paralyzed muscle . These results demonstrate that our nerve cut procedures effectively paralyzed VL . Paralysis of VL will cause a reduced ability to balance the medial forces on the patella produced by VM , resulting in aberrant mechanical stresses within the knee joint . As illustrated in Figure 1b , we predicted that if the CNS acts to minimize these internal joint stresses , it should compensate for VL paralysis by preferentially increasing activation of RF . An example of RF and VM activations before and after VL paralysis is shown in Figure 3a–b for one animal . In the first week after nerve cut , there was a large increase of the activation of both RF and VM during stance . Over subsequent weeks of adaptation , the activation of both muscles decreased . By 7 weeks after the nerve cut , the activation of VM during stance returned to levels similar to those observed in baseline conditions . At this same time point , however , the activation of RF remained elevated as compared to baseline conditions , consistent with our predictions based on the neural control of internal joint stresses . This preferential increase in RF activation during stance was consistent across animals , as illustrated in Figure 3c . While the activity of VM was not significantly different from baseline at any time after VL nerve cut ( pweek1 = 1 , pweek2 = 1 , pweek7 = 0 . 154 ) , RF activity was significantly increased at all time points ( pweek1 = 0 . 007 , pweek2 < 0 . 001 , pweek7 < 0 . 001 ) . More importantly , the change in RF activation was significantly larger than the change in VM activation at all time points ( pweek1 < 0 . 001 , pweek2 < 0 . 001 , pweek7 < 0 . 001 ) . These results demonstrate that the CNS preferentially increased activation of RF following VL paralysis , consistent with the neural regulation of internal joint stresses . The persistent increased activation of RF that we observed following nerve cut should induce use-dependent muscle growth . To evaluate this possibility , we measured quadriceps muscle masses 8 weeks after VL paralysis ( Figure 4 ) . We found that there was a significant increase in RF muscle mass in the affected limb as compared to the mass of RF in the unaffected limb ( p=0 . 047 ) . In contrast , the masses of VM and VI were not significantly different between the two limbs ( VM: p=1; VI: p=1 ) . VL paralysis caused initial changes in overall joint kinematics that were compensated for over the subsequent weeks of adaptation . Figure 5b shows the average joint angle trajectories for one animal before and after VL paralysis . There was an initial reduction in the stance phase duration and in the range of motion at the knee , consistent with the reduction of knee extensor torque during stance caused by VL paralysis . The range of motion at the ankle , on the other hand , increased . In this animal , the range of motion at the hip was minimally affected . At later time points , all of these measures of overall joint kinematics recovered to levels similar to those observed before VL paralysis , although there were persistent differences in the specific joint angle trajectories ( see below ) . The changes in these measures of overall joint kinematics , averaged across all animals , are shown in Figure 5c . The duration of the stance phase , expressed as a percentage of the entire gait cycle , was significantly reduced in the first week after VL paralysis ( pweek1 <0 . 001 ) , but at later time points returned to levels that were not significantly different from baseline ( pweek2 = 0 . 457; pweek7 = 0 . 192 ) . The ranges of motion of the hip and knee were significantly smaller in the first week after VL paralysis ( hip: pweek1 = 0 . 008; knee: pweek1 <0 . 001 ) , while the range of motion at the ankle was significantly larger ( pweek1 <0 . 001 ) . At week 2 , the hip and the knee ROM remained significantly lower than baseline ( hip: pweek2 = 0 . 007; knee: pweek2 <0 . 001 ) whereas the ankle was no longer significantly increased ( pweek2 = 0 . 059 ) . By week 7 , the ROM of the hip and ankle joints were not significantly different from baseline ( hip: pweek7 = 0 . 502; ankle: pweek7 = 0 . 781 ) but there was a modest , though significant , decrease in the knee range of motion ( pweek7 = 0 . 009 ) . These results show that overall joint kinematics are initially impaired following VL paralysis . Over the subsequent weeks of adaptation , these deficits largely disappear and overall locomotor function is restored to baseline levels , although there is a persistent deficit in knee function . The preferential increase in RF activation shown in Figure 3 introduces an extra flexion torque at the hip . If the CNS preserves individual joint angles , this extra torque should be compensated by increased activation of hip extensor muscles ( e . g . BFp , SM , and GRc ) . However , we found no evidence for increased hip extensor muscle activation at any point of the adaptation ( Figure 3—figure supplement 1 ) , suggesting that the CNS does not preserve individual joint angle trajectories . To evaluate this possibility more directly , we examined the detailed joint kinematics through the adaptation period . The effects of increased RF activity on joint kinematics should be most apparent at the end of the stance phase since quadriceps muscle activations are largest in late stance ( Figures 2 and 3 ) . As illustrated in Figure 6a–b , the hip angle was significantly decreased after the first week ( i . e . more flexed; pweek1 = 1 , pweek2 = 0 . 003 , pweek7 = 0 . 032 ) , and the ankle angle was significantly increased at all time points ( i . e . more extended; pweek1 = 0 . 030 , pweek2 = 0 . 004 , pweek7 = 0 . 038 ) as compared to baseline . On the other hand , although the knee angle was significantly reduced in the first two weeks after VL paralysis ( pweek1 <0 . 001 , pweek2 = 0 . 031 ) , it was not significantly different from baseline at 7 weeks ( pweek7 = 0 . 167 ) . We found equivalent results for joint angles measured earlier in the stance phase ( Figure 6—figure supplement 1 ) , at another point within the quadriceps burst . These results show that increased activation of RF , although compensating for the knee extension lost after VL paralysis , introduced persistent deviations at other joints . In particular , the increased hip flexion is consistent with the extra hip flexor torque produced by RF . Although the results of Figure 6a–b demonstrate persistent changes in local joint kinematics , the CNS might restore global features of limb kinematics such as overall limb length or limb angle . As illustrated in Figure 6c , both limb length and limb angle measured at the end of stance were initially altered after VL paralysis ( limb length: pweek1 <0 . 001; limb angle: pweek1 = 0 . 002 ) . However , these alterations disappeared over the adaptation period so that these measures were not significantly different from baseline levels ( limb length: pweek2 = 0 . 142 , pweek7 = 1; limb angle: pweek2 = 0 . 141 , pweek7 = 1 ) . We found equivalent results for global kinematics measured earlier in the stance phase ( Figure 6—figure supplement 1 ) . These results suggest that the CNS adopts a control strategy that limits internal joint stresses and restores global aspects of locomotion .
We hypothesized that the CNS activates muscles to minimize internal joint stresses while achieving task requirements . We found that animals compensated for VL paralysis by increasing activation of RF while maintaining the activation of VM at levels similar to those observed before the perturbation . This strategy restores the knee extension torque lost after VL paralysis while causing minimal additional mediolateral patellar forces . The increased RF activation resulted in a persistent perturbation in joint kinematics due to the extra hip flexor torque from RF , but global aspects of limb kinematics were restored . These results suggest that the CNS considers internal joint stresses and strains when choosing muscle activations . The main result of this study is the observation that animals compensated for VL paralysis by preferentially increasing activation of RF . If one only considers muscle contributions to task performance , this adaptation strategy is counterintuitive; because VM and VL have redundant contributions to knee extension movements ( Sandercock et al . , 2018 ) , restoring task performance would be most easily accomplished by increasing activation of VM ( Figure 1b ) . This strategy would restore the same joint kinematics observed before VL paralysis without requiring any additional compensation . In contrast , increased activation of RF , while restoring the knee extension torque lost by VL paralysis , also introduces an additional flexion torque at the hip ( Johnson et al . , 2008; Greene , 1963 ) ( Figure 1 ) . Thus , this strategy would either require additional changes in other muscles across the limb , or introduce residual deviations in joint kinematics . The increased activation of RF we observed is therefore difficult to reconcile with strategies for muscle coordination focused solely on considerations of task performance . This strategy is consistent , however , with the hypothesis that the CNS takes into account the effect of muscle activity on the stresses and strains within a joint . Increasing activation of RF restores knee extension function while producing negligible additional mediolateral forces on the patella ( Figure 1 ) . On the other hand , higher activation of VM would increase the medial force on the patella and cause excess patellofemoral loading . Hence , our results suggest that the CNS considers these mediolateral forces when selecting muscle activations to compensate for VL paralysis . We also observed a selective increase in the mass of RF that likely reflects a use-dependent strengthening of that muscle , although we did not confirm this by measuring peak forces or myosin content ( Blaauw et al . , 2013 ) . This result indirectly confirms the preferential increase of RF activity after VL paralysis . Furthermore , it suggests that at late time points of the adaptation , the loss of VL was compensated both by higher neural activity and by additional force generating capability of RF . Finally , this result suggests that the increased RF activation was not specific to treadmill locomotion , but was also used in behaviors we did not measure here; if there was increased VM or VI activation in other behaviors , we would have observed an increase in the mass of those muscles . The observation of no change in the mass of VI also suggests that there was no consistent change in the activation of this muscle during the adaptation period , though we did not record from VI in these experiments due its small size and inaccessibility for implantation . We did not observe a complete silencing of VM , even after 7 weeks of adaptation: VM activation levels were on average similar to those observed before VL paralysis . This implies that a residual medial force on the patella persisted throughout the adaptation period . Such residual VM activity might reflect a compromise between unbalanced mediolateral patellar forces and perturbations in joint kinematics caused by the increased activation of RF . In the context of optimization approaches to motor control ( Todorov and Jordan , 2002; Scott , 2004 ) , this result suggests a cost function that includes terms for task performance as well as internal joint stresses . Alternatively , the residual activation of VM might reflect habitual persistence of the strategy used by the CNS prior to VL paralysis ( de Rugy et al . , 2012; Loeb , 2012 ) . Although our study cannot distinguish between these two explanations , in either case the fact that RF and not VM activity was increased demonstrates the influence of criteria related to the state of internal joint structures . The particular aspects of task performance that drive adaptation are potentially complex . Although muscle paralysis caused persistent changes at individual joint kinematics , we observed restoration of global limb kinematics . This result is consistent with ideas such as those used in uncontrolled manifold analyses ( Latash et al . , 2007 ) and with previous results examining adaptation following paralysis of triceps surae in rats and cats ( Chang et al . , 2009; Bauman and Chang , 2013 ) . In this context , our results suggest a hierarchical control of movement , in which higher level variables related to task goals ( i . e . global limb variables ) are preserved after VL paralysis while lower level variables related to task execution ( i . e . individual joint angles ) are allowed to vary . Many criteria have been proposed to influence the activation of muscles during behavior ( Alessandro , 2016; Prilutsky , 2000 ) , but they generally do not consider the internal state of joints . Although some of these criteria might indirectly minimize joint stresses and strains ( e . g . minimum jerk [Flash and Hogan , 1985] or torque change [Uno et al . , 1989] ) , they would not predict the selective increase in RF activation observed here . Including criteria that explicitly consider the consequences of muscle activation on joint stresses and strains might improve the predictions of such optimization approaches . Alternatively , minimization of internal joint stresses and strains in intact subjects might be accomplished without explicit control . The co-evolution of neural control and musculoskeletal anatomy ( Valero-Cuevas , 2015 ) might result in a neuromechanical system such that muscle activations optimizing task performance criteria also optimize internal joint state criteria . For example , when VM and VL are intact , a criterion that minimizes muscle metabolism ( Kistemaker et al . , 2014; Prilutsky , 2000 ) , or that considers signal dependent variability ( Harris and Wolpert , 1998; Haruno and Wolpert , 2005 ) would predict a distributed activation of VM and VL . This distributed activation of VM and VL might result from optimization over evolutionary time scales ( Giszter , 2015 ) and be implemented in coordinative structures such as muscle synergies ( Tresch and Jarc , 2009; Alessandro et al . , 2013a; Alessandro et al . , 2013b ) within the spinal cord ( Levine et al . , 2014; Hart and Giszter , 2010; Takei et al . , 2017 ) or other brain regions ( Overduin et al . , 2015 ) . This distributed activation could in turn balance net mediolateral patellar forces , even though such a balance was not explicitly considered . After injuries that alter limb properties ( Gutierrez et al . , 2009; O'Connor et al . , 1993 ) , this implicit regulation of internal joint state might be compromised ( Needle et al . , 2014 ) . In these situations , the CNS might be forced into a solution that requires a tradeoff between task performance and internal joint stresses and strains , such as that observed in our experiments . These results have implications to clinical syndromes affecting joint health , such as knee osteoarthritis ( Felson , 2013; Mahmoudian et al . , 2017 ) and patellofemoral pain ( Smith et al . , 2018 ) ( PFP ) in humans . Although many factors have been shown to contribute to PFP ( Smith et al . , 2018; Fagan and Delahunt , 2008 ) , one common suggestion is that PFP results from an imbalance in mediolateral patellar forces ( Pal et al . , 2012 ) . Our results show that although the CNS regulates these mediolateral forces , this regulation is imperfect since VM activation persisted after VL paralysis . PFP might result from similarly imperfect regulation of patellar forces , potentially reflecting a compromise between criteria related to task performance and internal joint states . More broadly , our experiments highlight the importance of considering the consequences of rehabilitation on internal joint stresses and strains following injuries ( Farrokhi et al . , 2013; Shull et al . , 2013; Simic et al . , 2011 ) , rather than solely focusing on restoration of task performance ( Hollands et al . , 2012 ) . Regulation of internal joint stresses and strains , either in intact subjects or following injury , might be accomplished by feedforward or feedback mechanisms . Sensory receptors in the joint can convey information about strains in ligaments or stresses between bones . Activity in these receptors can be conveyed to spinal interneuronal populations ( Solomonow et al . , 1987; Iles et al . , 1990 ) and other areas ( Sjölander et al . , 2002 ) to drive rapid feedback control of muscles ( Ferrell , 1980 ) or longer term adaptation of muscle activations in order to avoid potential injury . The CNS could also derive information about internal joint structures indirectly from muscle proprioceptors ( Wilmink and Nichols , 2003 ) , combining information about muscle forces and lengths with ‘internal’ models of muscle anatomy to estimate joint stresses and strains . Similarly , predictive models of muscle actions , combined with efference copy of motor commands , could be used to make feedforward predictions of internal joint stresses and strains . The strong coherence between VM and VL across frequencies ( Laine et al . , 2015 ) observed in humans is consistent with the importance of coordinated activation of these muscles ( Brøchner Nielsen et al . , 2017 ) , whether by feedforward or feedback processes . This coordinated muscle activation is also similar to ideas of muscle synergies ( Tresch and Jarc , 2009; Alessandro et al . , 2013a; Alessandro et al . , 2013b ) , so that the balanced activation of muscles within a synergy would reflect regulation of joint stresses and strains . Note that neural control of internal joint states might be relatively coarse , only responding to large deviations in joint states inducing discomfort or pain ( Grigg and Greenspan , 1977 ) . Such a coarse control of joint state might reflect the ‘good enough’ habitual strategy ( de Rugy et al . , 2012 ) described above , potentially resulting in the residual activation of VM observed in our experiments . Further experiments will be necessary to better understand the neural control of joint stresses and strains . Although the muscle paralysis by denervation used in these experiments has been used in many studies to examine neural adaptation strategies ( Bauman and Chang , 2013; Dambreville et al . , 2016; Frigon and Rossignol , 2007; Bennett et al . , 2012; Bouyer et al . , 2001; Maas et al . , 2007 ) , the loss of sensory information from the paralyzed muscle might contribute to the observed changes in muscle activation ( Akay et al . , 2014 ) . Perturbations directly manipulating patellar forces or paralyzing VM or RF might provide complementary experiments to further evaluate this hypothesis . Cutting the nerves to VM and RF selectively , however , is difficult because they branch extensively immediately after leaving the common quadriceps nerve , making them hard to isolate ( Greene , 1963 ) . The nerve to VL was simpler to cut because it persists as a single trunk for a long distance from the common quadriceps nerve . Our experiments exploited the unique mechanical properties of the rat knee joint: the clear distinction between muscle contributions to task performance and to patellar forces in the rat allowed us to make simple predictions about adaptation strategies consistent or inconsistent with neural regulation of internal joint stresses and strains . Such distinctions are more difficult to make in humans because the patellar mobility at extended knee angles complicates how quadriceps muscles affect both patellar mechanics and limb kinematics ( Farahmand et al . , 1998 ) . The rat knee joint might therefore provide a tractable experimental model to examine the neural control of joint stresses and strains in both healthy subjects and following injury .
We performed experiments on female Sprague Dawley rats ( n = 6 ) . All the procedures were approved by the Animal Care Committee of Northwestern University . We first trained rats to maintain a stable gait pattern during treadmill locomotion . We then implanted chronic EMG electrodes in multiple hindlimb muscles and allowed animals to recover for at least 10 days following implantation . We recorded kinematic and EMG activity during treadmill locomotion before ( baseline ) and at different times after VL paralysis: within one week after the procedure ( week 1 ) , within the second week ( week 2 ) , and within the seventh week ( week 7 ) . At the end of data collection , animals were euthanized , muscle masses weighed , and electrode location verified . The EMG channels corresponding to misplaced electrodes were not included in the analysis . We anesthetized animals with isoflurane ( 3% in O2 ~ 2 l/min ) , and prepared them for aseptic surgery . We implanted pairs of electrodes in seven muscles: vastus lateralis ( VL ) , vastus medialis ( VM ) , rectus femoris ( RF ) , semimembranosus ( SM ) , biceps femoris posterior ( BFp ) , tibialis anterior ( TA ) , and caudal gracilis ( GRc ) . Knots placed on either side of the muscle secured the exposed electrode sites within the muscle belly . The electrode leads were tunneled subcutaneously to a connector on the back of the animal . These methods have been described in more details previously ( Tysseling et al . , 2013 ) . During aseptic surgery ( isoflurane , 3% in O2 ~ 2 l/min ) , we exposed the quadriceps nerve plexus , and isolated the branch to VL . We anesthetized this branch with lidocaine , tied it off in two locations with silk suture to prevent re-innervation , and then cut between the sutures . Before each recording session , we applied retro-reflective markers on the shaved skin ( see Data acquisition and processing ) under brief isoflurane anesthesia ( 2–3% in O2 ~ 2 l/min ) . Markers were placed on bony landmarks on the hindlimb skin ( rostral and caudal tips of the pelvis , hip , knee , ankle and first digit , see Figure 5a ) . The EMG connector was attached via cable to the amplifier , and the animal placed on the treadmill . We waited at least 30 min after anesthesia before collecting behavioral data . Each recording session consisted of two minutes of locomotion at the maximum comfortable speed ( 12–15 m/min ) and incline ( usually +25% ) to induce strong quadriceps activations . One animal was unable to walk at an incline of +25% , and so for that animal we recorded data from level treadmill walking . The 3D position of markers was tracked using a motion capture system ( Vicon ) at a frequency of 200 Hz . These signals were low-pass filtered offline at a cut-off frequency of 10 Hz ( 5th order Butterworth ) . In order to reduce errors due to differential movements of the skin ( Filipe et al . , 2006; Bauman and Chang , 2010 ) , we estimated the 3D position of the knee by triangulation , using the lengths of the femur and the tibia ( Bauman and Chang , 2010 ) . Differential EMG signals were amplified ( 1000X ) , band-pass filtered ( 30–1000 Hz ) and notch filtered ( 60 Hz ) , and then digitalized ( 5000 Hz ) . The digitalized signals were further band-pass filtered offline to remove motion artifacts ( 5–500 Hz , 4th order Butterworth ) , rectified , and envelopes were computed from the rectified signals . We segmented both kinematic and EMG envelopes into separate strides , defining the beginning of each stride as the onset of activation of TA during bouts of stable walking . To obtain a consistent dataset that reflected the behavioral condition of the experiment , we only considered strides with durations within 1 . 5 standard deviations from the mean . We also excluded strides with clear EMG artifacts that could occur when the cable hit the side of the treadmill , as identified using Tukey outlier analysis ( i . e . identifying EMG values that were 1 . 5 interquartiles above the upper quartile of the maxima across steps ) . Further , we discarded strides that had more than 20% of missing values in the kinematic signals due to malfunctioning of the acquisition system or due to occlusions of the retro-reflective markers . Application of these inclusion criteria resulted in data sets with , on average , 65 strides for each recording session . Strides were then time-normalized to 100 samples for further analysis . We characterized each stride by means of several kinematic and EMG features . For kinematic analysis , we calculated the percentage of stance ( i . e . duration of the stance phase divided by the duration of the entire stride , with stance phase defined as the interval between foot-strike and foot-off ) . We calculated local joint angles ( hip , knee , and ankle ) as illustrated in Figure 5a . Furthermore , we calculated more global features describing overall limb kinematics as illustrated in Figure 6c: the length of the global limb vector ( defined as the vector from the hip to the toe ) , and the limb angle ( defined as the angle between the limb vector and the vector from the hip to the rostral pelvis ) . Finally , we calculated the range of motion of each joint ( ROM , i . e . the difference between maximum and minimum joint angles for each stride ) . All angles were calculated using the projections of the markers onto the sagittal plane of the treadmill , and were defined such that zero degrees reflected complete joint flexion and increasing angles reflected increasing amounts of extension . For the EMG analysis , we computed the integral of the EMG envelopes’ extension burst ( i . e . muscle activity related to the preparation and the execution of the stance phase ) , which represents the muscle activity contributing to limb extension . We normalized this measure to the pre-paralysis mean integrated activity for each muscle and subject , obtaining the relative change of activation with respect to this baseline . We employed Linear Mixed Effect Models ( LMEM ) to analyze both kinematic and EMG data using the nlme package ( Pinheiro et al . , 2017 ) in the R environment . The use of LMEMs allows us to exploit the fact that we have large numbers of observations for each animal across the adaptation period , obtaining the maximal statistical power from a relatively small number of animals . LMEMs also allow us to analyze samples of different size ( e . g . unequal numbers of strides across days and animals ) , to cope with missing data ( potentially due to the inclusion criteria described above ) , and to take into account variability at different levels of the dataset ( e . g . across strides , across days , and across subjects ) . To confirm that our dataset met the assumption of Gaussian distribution and independence of residuals and random effects ( Pinheiro and Bates , 2000 ) , we visually inspected the distributions using qq-plots and histograms . After fitting the LMEMs , we tested our specific hypotheses of interest by performing post-hoc tests on the model parameters and using Bonferroni corrections to adjust the obtained p-values . Note that because of this correction , the adjusted p-values can be equal to 1 . We considered tests to be statistically significant if their p-values were lower than the 0 . 05 significance level . For the kinematic analysis , we fit a LMEM for each measure , using the value of the feature as the dependent variable and the timing after VL paralysis as the independent variable . In order to take inter-subject variability into account , we considered subject as a random effect on both slope and intercept ( Pinheiro and Bates , 2000 ) . This method is essentially a more powerful version of a repeated measures ANOVA , and it has been proven effective at estimating degrees of freedom and standard errors , hence enabling accurate statistical tests ( Barr et al . , 2013 ) . We then performed post-hoc tests to evaluate whether the values of the features at a given week were different from their baseline values . To analyze muscle activity , we initially log-transformed the data to reduce the skewness of EMG features . This rendered the distribution of the dataset more symmetrical and the distribution of residuals approximately Gaussian . For the comparison between VM and RF , we fit a LMEM with normalized integrated EMG as the dependent variable , and the timing after VL paralysis as well as muscle identity ( i . e . whether the muscle was VM or RF ) as independent variables , including in the model also an interaction term between these independent variables . We considered subject as a random effect on both intercept and slope , and we nested the variability across strides within each subject and recording session . We performed post-hoc tests to evaluate: ( 1 ) whether the normalized activity in RF at any week after VL paralysis was greater than the activity in VM; and ( 2 ) whether the normalized activity of each muscle at a given week was different from its baseline value . For the analysis of the other muscles ( Figure 3—figure supplement 1 ) , we fit a LMEM for each signal using the normalized integrated EMG as the dependent variable , the timing after VL paralysis as the independent variable , and considering subject as a random effect on both slope and intercept . Then , we used post-hoc tests to evaluate whether the normalized activity at a given week was different from its baseline value . To compare the masses of the quadriceps in the denervated limb to those in the intact limb , we performed paired t-tests for each muscle , and we adjusted the obtained p-values with Bonferroni corrections . | Although most of us will never achieve the grace and dexterity of professional ballerina Misty Copeland , we each make sophisticated , complex movements every day . Even simple movements often involve coordinating many muscles throughout the body . Moreover , because we have so many muscles , there are often multiple ways that we could use them to make the same movement . So which ones do we use , and why ? Many studies into muscle control focus on how the muscles activate to perform a task like kicking a soccer ball . But muscles do more than just move the limbs; they also act on joints . Contracting a muscle exerts strain on bones and the ligaments that hold joints together . If these strains become excessive , they may cause pain and injury , and over a longer time may lead to arthritis . It would therefore make sense if the nervous system factored in the need to protect joints when turning on muscles . The quadriceps are a group of muscles that stretch along the front of the thigh bone and help to straighten the knee . To investigate whether the nervous system selects muscle activations to avoid joint injuries , Alessando et al . studied rats that had one particular quadriceps muscle paralyzed . The easiest way for the rats to adapt to this paralysis would be to increase the activation of a muscle that performs the same role as the paralyzed one , but places more stress on the knee joint . Instead , Alessando et al . found that the rats increase the activation of a muscle that minimizes the stress placed on the knee , even though this made it more difficult for the rats to recover their ability to use the leg in certain tasks . The results presented by Alessando et al . may have important implications for physical therapy . Clinicians usually work to restore limb movements so that a task is performed in a way that is similar to how it was done before the injury . But sometimes repairing the damage can change the mechanical properties of the joint – for example , reconstructive surgery may replace a damaged ligament with a graft that has a different strength or stiffness . In those cases , performing movements in the same way as before the surgery could place abnormal stress on the joint . However , much more research is needed before recommendations can be made for how to rehabilitate rats after injury , let alone humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Adaptation after vastus lateralis denervation in rats demonstrates neural regulation of joint stresses and strains |
A LOV ( Light , Oxygen , or Voltage ) domain containing blue-light photoreceptor ZEITLUPE ( ZTL ) directs circadian timing by degrading clock proteins in plants . Functions hinge upon allosteric differences coupled to the ZTL photocycle; however , structural and kinetic information was unavailable . Herein , we tune the ZTL photocycle over two orders of magnitude . These variants reveal that ZTL complexes with targets independent of light , but dictates enhanced protein degradation in the dark . In vivo experiments definitively show photocycle kinetics dictate the rate of clock component degradation , thereby impacting circadian period . Structural studies demonstrate that photocycle dependent activation of ZTL depends on an unusual dark-state conformation of ZTL . Crystal structures of ZTL LOV domain confirm delineation of structural and kinetic mechanisms and identify an evolutionarily selected allosteric hinge differentiating modes of PAS/LOV signal transduction . The combined biochemical , genetic and structural studies provide new mechanisms indicating how PAS/LOV proteins integrate environmental variables in complex networks .
Organisms have developed signaling networks to measure and respond to daily ( circadian ) and seasonal ( photoperiodic ) alteration in environmental variables . Key to these circadian and photoperiodic responses is measurement of day length through complicated feedback loops involving sensory proteins . These sensory proteins involve members of the Period-ARNT-Singleminded and its Light , Oxygen , or Voltage ( PAS/LOV respectively ) subclass that couple photic-input to metabolic and stress pathways ( Lokhandwala et al . , 2015; Sahar and Sassone-Corsi , 2009; Taylor and Zhulin , 1999 ) . Currently , how these signaling components are integrated is poorly understood due to a difficulty in decoupling photochemistry from allosteric protein changes and signal transduction . In plants and fungal species , LOV domain containing proteins act at signaling nodes to integrate sensory responses into circadian , reproductive and stress pathways ( Lokhandwala et al . , 2015; Imaizumi et al . , 2003; Somers et al . , 2000; Zoltowski et al . , 2007 ) . Central to their function are two key elements of the PAS family: ( 1 ) The ability to sense and respond to diverse environmental stimuli , and ( 2 ) The presence of multiple interaction surfaces that engage targets in a selective manner ( Figure 1 , Figure 1—figure supplement 1 ) . The ability to trigger these elements with light has positioned LOV proteins as an allosteric model and enabled the development of LOV optogenetic tools ( Pudasaini et al . , 2015 ) . However , limited understanding of how chemistry is linked to allostery and downstream signaling hampers our understanding of these systems . 10 . 7554/eLife . 21646 . 003Figure 1 . Models of ZTL photochemistry and regulation of circadian period . ( A ) In the light , ZTL associates with both GI and degradation targets ( PRR5/TOC1 ) . During the day , the strong affinity for GI allows GI , ZTL , TOC1 and PRR5 levels to rise . Upon dusk , the adduct form of ZTL decays with a rate constant k3 , leading to a conformational change . The conformation change decreases GI affinity and leads to ubiquitination of ZTL targets . ( B ) Modeling PRR5 degradation as a function of k3 ( see Equation 1 and Materials and methods for model generation and parameter selection ) . Using k3 for WT ( black ) , G80R ( red ) , V48I ( blue ) , G46S:G80R ( magenta ) and V48I:G80R ( green ) leads to predictable changes in PRR5 degradation patterns . ( Figure 1—supplements 1 , 2 and 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 00310 . 7554/eLife . 21646 . 004Figure 1—figure supplement 1 . LOV chemistry and signal transduction . ( A ) PAS/LOV domains signal through four elements . Multi-domain proteins reorient Ncap ( yellow ) and Ccap ( red ) elements to affect activity of signaling domains or recruit additional proteins to the PAS/LOV surface . The β-sheet ( blue ) and helical surface ( magenta ) act as protein:protein interactions motifs . ( B ) ZTL , and FKF1 domain architecture . The N-terminal LOV domain engages multiple targets . Light-activation then regulates Fbox activity to target proteins for degradation . ( C ) Superposition of VVD ( light orange ) , AsLOV2 ( grey ) and ZTL . The ZTL LOV core ( blue ) is similar to existing LOV structures . Significant deviations exist at the Ncap ( ZTL: Yellow ) , Ccap ( ZTL: Red ) and the E-F and H-I loops . These elements are believed to be important for signaling . ( D ) LOV proteins are activated by blue-light absorption to form an excited singlet state . The singlet rapidly undergoes intersystem crossing to generate a triplet intermediate . Proton coupled electron transfer from the conserved Cys residue and subsequent C4a adduct formation activates LOV proteins . The adduct state then decays in the dark or presence of UV-light . ( E ) Sequence alignment of ZTL family . Predicted helical regions ( red ) , β-sheet ( blue ) and an extended E-F loop ( green ) are noted . * implies conserved , ∧ residues tuning ZTL and FKF1 kinetics . The Ncap and Ccap regions are labeled . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 00410 . 7554/eLife . 21646 . 005Figure 1—figure supplement 2 . Schematic diagram of the daily protein abundance profile of ZTL and its function in the circadian clock . ZTL function in the circadian clock involves interactions with two core clock proteins , GI and TOC1 . The protein abundance patterns of ZTL , GI , and TOC1 proteins are based on previously published results ( Más et al . , 2003a; Kim et al . , 2007 ) . The circadian clock regulates the expression profile of GI protein with its peak near the end of the light period . The ZTL-GI interaction is enhanced when the ZTL LOV domain absorbs blue light . This interaction stabilizes both ZTL and GI proteins toward the end of the day , thus ZTL protein level also peaks at that time , although ZTL transcript levels are constitutive . TOC1 expression is also clock regulated with expression mainly during the nighttime . ZTL interacts with TOC1 in a light-independent manner , however ZTL mediated TOC1 degradation is enhanced at night . ZTL mutants lead to enhanced accumulation of TOC1 protein , leading to period lengthening of the circadian clock . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 00510 . 7554/eLife . 21646 . 006Figure 1—figure supplement 3 . Kinetics of ZTL Variants ( A–D ) Kinetics of ZTL adduct decay are determined from the absorbance recovery at 450 ( black ) and 478 nm ( red ) . G80R; τ = 6 . 6 hr ( A ) and V48I; τ = 10 . 7 hr ( B ) recover on similar timescales . The G46S:G80R; τ = 21 hr ( C ) variant recovers with a 2-fold slower time constant . A V48I:G80R variant recovers τ >65 hr ( see Table 1 for more information ) . Due to the long time constant for V48I:G80R , full recovery cannot be observed without the presence of an imidazole base catalyst ( Pudasaini and Zoltowski , 2013 ) . The recovery data shown ( D ) is under base catalyzed conditions ( 150 mM imidazole ) to allow full recovery . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 006 Our current understanding of LOV signaling has benefited from detailed structural studies that revealed amino acid sites that tune LOV allostery without affecting LOV photocycle kinetics ( Zoltowski et al . , 2007; Harper et al . , 2004; Jones et al . , 2007; Salomon et al . , 2000 ) . A consensus mechanism is summarized . LOV domains are typified by blue-light induced formation of a C4a adduct between a conserved Cys residue and a bound flavin ( FAD , FMN or riboflavin ) cofactor ( Figure 1—figure supplement 1D ) ( Zoltowski and Gardner , 2011 ) . C4a adduct formation then drives rotation of a conserved Gln residue to initiate conformational changes within N- or C-terminal extensions ( Ncap/Ccap ) to the LOV core ( Zoltowski and Gardner , 2011 ) . These N/Ccap elements in turn regulate activity of effector domains or recruit proteins to Ncap , Ccap or β-sheet surfaces ( Figure 1—figure supplement 1A , C , E ) . Upon incubation in the dark , or in the presence of UV-A light , the signaling adduct state spontaneously decays on a timescale of seconds to days ( Zoltowski and Gardner , 2011; Kennis et al . , 2004; Pudasaini and Zoltowski , 2013 ) . In this manner , LOV proteins dictate transiently stable signaling states capable of switching between a distinct on and off state depending on lighting conditions ( Zoltowski et al . , 2007; Halavaty and Moffat , 2007; Harper et al . , 2003; Zoltowski et al . , 2013 ) . These photoswitchable functions have made LOV proteins targets for optogenetic devices; however , currently we have limited understanding of the role of LOV photocycle kinetics for in vivo function . Genetic and photochemical studies of plant circadian and photoperiodic timing indicate Arabidopsis may function as a model organism for delineating the roles of photochemistry and allostery in LOV function . In Arabidopsis thaliana , three LOV domain containing proteins ZTL , FLAVIN-BINDING , KELCH REPEAT , F-BOX 1 ( FKF1 ) , and LOV KELCH PROTEIN 2 ( LKP2 ) act in circadian timing and seasonal flowering ( Baudry et al . , 2010 ) . Among them , the genetic functions of ZTL and FKF1 are highly characterized . Recent research indicates that their divergent roles in the circadian clock and photoperiodic flowering may enable interrogation of how chemistry regulates PAS/LOV protein function to select for signaling pathways ( Figure 1—figure supplement 1B ) ( Zoltowski and Imaizumi , 2014 ) . These proteins retain analogous elements , where an N-terminal LOV domain regulates activity of a C-terminal E3 ubiquitin ligase to target clock proteins for degradation in a time dependent manner ( Baudry et al . , 2010; Más et al . , 2003a ) . In addition , the LOV scaffold engages multiple proteins in a selective manner to regulate clock function ( Zoltowski and Imaizumi , 2014 ) . Despite conservation of domain elements , ZTL and FKF1 differ in their subcellular localization , degradation targets , and fundamental chemistry , thereby differentiating ZTL and FKF1 in circadian and photoperiodic timing ( Zoltowski and Imaizumi , 2014 ) . Central to their function are key differences in photocycle kinetics . In FKF1 , day-specific expression and a long-lived light-state species ( τ = 62 hr ) enable light-specific functions ( Imaizumi et al . , 2003; Pudasaini et al . , 2015 ) . In FKF1 , photon absorption facilitates complex formation with GIGANTEA ( GI ) through its LOV domain ( LOV-GI ) to mediate degradation of CYCLING DOF FACTOR 1 ( CDF1 ) and stabilization of CONSTANS ( CO ) ( Imaizumi et al . , 2005; Sawa et al . , 2007; Song et al . , 2012 ) . Here , selection of degradation or stabilization appears to center on the domain involved in target recruitment , where the Kelch repeat domain ( CDF1 ) specifies degradation and the LOV domain ( CO ) specifies stabilization ( Zoltowski and Imaizumi , 2014; Song et al . , 2014 ) . Importantly , both functions require both light and GI . In contrast , constitutive mRNA levels and a fast photocycle for ZTL ( τ = 1 . 4 hr ) suggest possible day ( light ) and evening ( dark ) functions ( Pudasaini and Zoltowski , 2013 ) . Light-state ZTL has two primary functions , formation of a LOV-GI complex to allow protein accumulation during the day via enhanced ZTL/GI stability ( Kim et al . , 2013 , 2007 ) , and ZTL/GI-dependent destabilization of CO in the morning ( Song et al . , 2014 ) . The latter results in antagonistic roles of ZTL and FKF1 in the photoperiodic response , and may contribute to a late-flowering phenotype in some ztl mutants ( Somers et al . , 2004 ) . Throughout the night , ZTL mediates rapid degradation of the clock components , TIMING OF CAB EXPRESSION 1 ( TOC1 ) and PSEUDO RESPONSE REGULATOR 5 ( PRR5 ) ( Figure 1—figure supplement 2 ) ( Kiba et al . , 2007; Fujiwara et al . , 2008; Más et al . , 2003b ) . TOC1 protein levels contribute to the control of period length of the circadian clock ( Zoltowski and Imaizumi , 2014 ) . Impaired degradation of TOC1 by ZTL mutants lead to accumulation of TOC1 and PRR5 protein and a long-period phenotype consistent with strains harboring additional copies of TOC1 ( De Caluwé et al . , 2016 ) . Interestingly , TOC1 degradation appears to be in competition with GI , occurs with an approximately 2 hr delay following dusk , and is enhanced in the dark ( Zoltowski and Imaizumi , 2014; Más et al . , 2003a; Song et al . , 2014; Kim et al . , 2007 ) . Current models explain such behavior in TOC1/PRR5 turnover by imparting a differential function between day and night conditions ( De Caluwé et al . , 2016; Pokhilko et al . , 2013 ) , where TOC1/PRR5 are targeted for degradation by ZTL regardless of lighting conditions , but degradation activity is enhanced in the dark . How this is achieved at the molecular and chemical level is not well understood . Herein , we present a comprehensive study of ZTL signaling to develop a broad understanding of blue-light regulation of circadian and photoperiodic timing and to understand the evolutionary basis for divergent functions of ZTL and FKF1 . Therein , we tackle two outstanding questions in the PAS/LOV field; ( 1 ) What purpose , if any , does the LOV photocycle lifetime play in biological function ? Previous research proposed divergent roles for ZTL and FKF1 in the measurement of light intensity and day length ( Pudasaini and Zoltowski , 2013 ) , however no experimental validation of such a model has been available . Further , with the exception of a fungal system ( Dasgupta et al . , 2015 ) , the role of LOV lifetimes in biology remains elusive . ( 2 ) How do PAS/LOV proteins signal to multiple interaction surfaces to allow signal integration ? Based on structural studies of ZTL we have developed protein variants that decouple photocycle lifetimes from signal transduction . In this manner , we show a definitive role of LOV lifetime in circadian timing . Further , we provide an allosteric model of LOV signal transduction enabling selection of diverse protein:protein interactions .
All ZTL variants form solution dimers , crystallize in the same space group ( P3121 ) and demonstrate topologically equivalent structures consistent with PAS/LOV proteins ( Figure 2 , Figure 2—figure supplements 1 and 2 ) . These structures reveal functional differences from all known LOV structures that differentiate the effects of V48I and G80R on ZTL signaling mechanisms . Namely , G80R structures are analogous to WT in all manners . The primary difference is the formation of a π-cation interaction that stabilizes the Dα/Eα linkage and C82 that is necessary for C4a adduct formation ( Figure 2B ) . Increased rigidity of C82 imposed by the π-cation interaction is consistent with having an effect only on photocycle kinetics . In contrast , examination of V48I containing structures reveals distinct differences that identify allosteric signal transduction mechanisms to the N/Ccap . These mechanisms identify V48I as a residue that disrupts allosteric regulation of ZTL function . Below , we provide detailed analysis of ZTL in dark and light states to highlight these functional differences . We focus on two reported aspects of ZTL group function , LOV:LOV mediated homodimerization and allosteric regulation of Ncap and Ccap elements . 10 . 7554/eLife . 21646 . 009Figure 2 . Structural analysis and LOV dimer formation in ZTL . ( A ) G80R ( dark-state , black; light-state ( grey ) , WT ( red ) , V48I:G80R ( green ) , G46S:G80R ( blue ) all elute as dimers with apparent MWs of 38–41 kDa compared to the expected monomer of 16 kDa . Multi-Angle-Light-Scattering ( MALS ) confirms dimer formation in WT 29–165 ( absolute MW 33 ± 2 kDa ) and 16–165 constructs ( See Figure 2—figure supplement 1 ) . Introduction of an I151R abolishes dimer formation ( magenta; apparent MW = 22 kDa ) . ( B ) Structure of ZTL active site ( yellow ) and residues involved in structural or kinetic modulation of signaling ( purple ) . R80 within Dα forms a π-cation interaction with F84 directly above the photoreactive C82 , resulting in steric stabilization of adduct formation . The observed steric stabilization through the π-cation interaction is consistent with the longer photocycle in G80R . V48I positions the additional methyl group into a pocket adjacent to N5 , C82 and Q154 . Comparisons of AsLOV2 structures ( white; buried conformation ) , dark-state ZTL ( yellow; exposed conformation ) demonstrates that V48I can impact the position of Q154 between buried and exposed conformations . Movement of Q154 correlates with movement of F156 in Iβ . ( C–F ) ZTL monomers are defined by an antiparallel β-sheet flanked by a series of α-helices ( Cα , Dα , Eα , Fα ) . The helices cradle the photoreactive FMN adjacent to C82 ( Eα helix ) . ZTL contains a 9-residue insert linking the E-F helices that accommodates the adenine ring of FAD in some LOV proteins ( black ) ( Zoltowski et al . , 2007 ) . N- and C-terminal extensions ( Ncap/Ccap; yellow ) are largely disordered; however , a short helix within the Ncap reaches across a dimer interface in some molecules to form contacts between the Cα and Dα helices . Two dimer interfaces are formed through the β-scaffold in ZTL . The compact dimer ( C , D ) differs from the elongated dimer ( E , F ) by a 2 . 0 Å translation along the β-sheet . Key residues in the dimer interface are shown in yellow . The translation disrupts a network of sulfur-π and π-π interactions involving C45 and F47 , centered around I151 . ( Figure 2—figure supplements 1 and 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 00910 . 7554/eLife . 21646 . 010Figure 2—figure supplement 1 . Dimerization of ZTL/FKF1/LKP2 LOV domains . ( A ) ZTL ( 16-165; apparent MW = 57 kDa; absolute MW from MALS ~40 kDa , panel B ) and FKF1 ( 28–174; apparent MW from SEC 62 kDa; absolute MW from MALS ~42 kDa ) elute as dimers on SEC as confirmed by multi-angle light scattering ( B ) . LKP2 ( 16–165 apparent MW = 33 kDa ) adopts a much smaller hydrodynamic radius indicating distinct differences in oligomeric structure and/or affinity . ZTL 16–165 was used to allow comparison of similar length constructs and allow use of a stable FKF1 construct characterized previously ( Nakasako et al . , 2005 ) , ( B ) Multi-angle light scattering of ZTL 16–165 . The expected monomer MW was ~18 kDa , indicating ZTL exists as a constitutive dimer . ( C ) Introduction of I160R variants ( equivalent to ZTL I151R ) in FKF1 renders the protein monomeric . It is also more susceptible to proteolysis . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01010 . 7554/eLife . 21646 . 011Figure 2—figure supplement 2 . Dark-state structure of ZTL 29–165 and helical dimer . ( A ) ZTL crystallizes as a tetramer , where the helical interfaces are buried within the crystal lattice . ( B ) A parallel helical dimer is observed in the crystal lattice that is characterized by slightly asymmetric contacts between the E and F helices ( yellow ) and associated loops ( blue ) . ( C ) Contacts between R95 in the E-F loop ( blue ) and the phosphate of the bound FMN of the neighboring molecule stabilize the helical dimer . ( C ) Introduction of a R95A variant does not affect in vitro dimerization of ZTL 29–165 . The elution profile of R95A is not concentration dependent and reflects a dissociation constant unable to be determined by SEC ( Kd <0 . 2 µM ) . Traces depicted and apparent MWs are: dark-state ( Black- 41 . 3 kDa at 130 µM and Blue- 39 . 6 kDa at 13 µM ) , light-state ( Red-40 . 8 kDa at 130 µM and Magenta- 37 . 8 at 13 µM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 011 Previous solution studies of ZTL group proteins suggest they function as obligate dimers ( Ito et al . , 2012; Nakasako et al . , 2005 ) , however ZTL group dimers have not been observed in vivo and their function is unknown ( Han et al . , 2004 ) . Consistent with solution studies , the crystallographic lattice is defined by two sets of anti-parallel dimers formed by extensive contacts along the β-scaffold . We term these interfaces the ‘compact’ and ‘elongated’ dimers based on a 2 Å translation that is coupled to the degree of order present within the Ncap ( Figure 2 ) . In the compact dimer , clear density for Ncap residues ( 31-43 ) is observed in one molecule that contacts the helical interface of the adjacent monomer ( Figure 2C , D ) . In contrast , the elongated dimer contains no density for Ncap residues ( Figure 2E , F ) . A stabilizing element in both dimers is a hydrophobic core composed of a tetrad of Ile residues ( l151 and I153 ) . Additional contacts along the helical interface define a possible secondary dimer . The helical dimer is parallel in orientation and involves contacts between the E-F helices and associated loop . Specifically , R95 forms a salt bridge with the phosphate moiety of FMN in the neighboring molecule ( Figure 2—figure supplement 2 ) . However , based on two lines of evidence , we conclude that the β-scaffold interface represents the solution ZTL and FKF1 homodimers . First , FKF1 lacking the entire E-F loop and ZTL variants that disrupt R95 contacts remain dimeric ( Figure 2—figure supplement 2 ) ( Nakasako et al . , 2005 ) . Second , I151R ( and I160R in FKF1 ) variants abolish dimer formation in vitro ( Figure 2A and Figure 2—figure supplement 1 ) . Thus , ZTL and FKF1 solution dimers are formed by equivalent anti-parallel contacts along the β-scaffold , similar to other PAS/LOV proteins ( Card et al . , 2005 ) . Given the unknown role of ZTL dimers in vivo and the known role of Ncap and Ccap elements in LOV allostery , we turned our attention to the structural differences between the different dimers . Close examination of residues linking the Ncap and Ccap to the active site FMN identifies distinguishing interactions that may be involved in signal transduction . The loss of Ncap density in the elongated dimer directly follows a CGF motif ( C45-G46-F47- ( V48 ) ) that links the Ncap , FMN binding pocket and the V48 position ( Figure 3 ) . An analogous hinge involving a Cys residue is known to mediate signal transduction in fungal LOV proteins , where the hinge directs both a conformational change and integration of oxidative and osmotic stress ( Lokhandwala et al . , 2015; Zoltowski et al . , 2007 , 2009; Lokhandwala et al . , 2016; Lamb et al . , 2009; Zoltowski and Crane , 2008 ) . Further , the CGF motif differentiates ZTL and FKF1 , where FKF1 contains the G46S and V48I ( equivalent ) mutations , thereby highlighting the region as a possible factor regulating the divergent functions of these closely related proteins . 10 . 7554/eLife . 21646 . 012Figure 3 . Q154 links Ncap , Ccap and helical elements . ( A ) Q154 exists in multiple conformations in WT ZTL structures . They differ in interactions with the active site flavin . An exposed conformation forms strong H-bonds to the O4 position ( black dotted line ) . A buried conformation forms weaker interactions ( 3 . 6 Å ) with O4 that leads to closer interactions at N5 ( 4 . 0 Å; red dotted line ) . The altered conformation is coupled to movement of F156 into the active site and multiple conformations of F155 , forming a QFF motif . The altered conformations define ZTL signaling as distinctly different than existing LOV structures . ( B ) The unusual orientations of Q154 differ from other LOV proteins that typically show strong interactions near N5 ( VVD; magenta ) . The heterogeneous conformations of Q154 directly abut G46 in a CGF motif allowing formation of the sulfur-π and π-π interactions involving C45 and F47 . F156 then adopts a buried conformation in contrast to the equivalent residue in VVD ( E184 ) ( C ) Sequence alignments of LOV proteins depict conserved elements within the CGF motif ( red ) and QFF motif ( blue ) in Arabidopsis thaliana ZTL , LKP2 , FKF1 , phototropin 1 LOV1 and LOV2 . Sequence conservation indicates divergent signaling mechanisms within the ZTL/FKF1 family compared to existing LOV allostery models . ( D ) Comparisons of ZTL ( yellow , black lettering ) and Arabidopsis thaliana phototropin 1 LOV1 ( PDB: 2Z6C; blue ) . The altered conformation of Q154 draws F156 into the active site . The buried conformation of F156 , leads to movement of Cα ( F66 , V69 ) . ( Figure 3—figure supplement 1 ) . . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01210 . 7554/eLife . 21646 . 013Figure 3—figure supplement 1 . Heterogeneous orientations of Q154 . WT ( left ) and G80R ( right ) conformations of Q154 compared to buried conformations in VVD ( magenta; PDBID 2PD7 ) and exposed conformations in ZTL V48I:G80R ( yellow ) . In both WT and G80R , Q154 samples orientations covering the range between exposed and buried conformations . One orientation in WT ( salmon ) lies close to the buried conformation . These heterogeneous conformations may contribute to ZTL retaining light and dark-state functions . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 013 In ZTL , signal transduction diverges from known LOV signaling mechanisms ( Figures 3 and 4 ) . In contrast to other LOV proteins , where a conserved Gln ( Q154 ) acts as a molecular bridge linking the FMN to a residue in Aβ , the dark-state structure of ZTL contains a heterogeneous orientation of the active site Q154 ( Figure 3 and Figure 3—figure supplement 1 ) ( Zoltowski et al . , 2007; Halavaty and Moffat , 2007; Möglich and Moffat , 2007 ) . The altered orientation is coupled to contacts in the Ncap , Ccap and helical interface . Given that all three sites have been implicated in PAS/LOV signaling ( Harper et al . , 2004; Card et al . , 2005; Zoltowski and Crane , 2008; Partch et al . , 2009 ) , we examine each interaction below . We define them as the Aβ-Ncap hinge , Iβ-Ccap hinge and Cα ( helical interface ) to specify structural elements that may be involved in signal transduction . 10 . 7554/eLife . 21646 . 014Figure 4 . Structural effects on ZTL chemistry and signaling . ( A ) Comparison of dark-state V48I:G80R ( yellow ) , light-state V48I:G80R ( cyan ) and dark-state AtLOV1 phototropin 1 ( purple: PDB ID: 2Z6C ) molecules . 2Fo-Fc ( 2 . 0 σ grey mesh ) and Fo-Fc ( 3 . 0 σ green mesh ) are depicted for dark-state V48I:G80R . Lack of density for an adduct is consistent with minimal population of the light-state species . Density shows clear selection of the I48 methyl group towards Q154 . Some residual density is present in the buried conformation of Q154 that indicates either partial occupancy of the site in the dark , or residual light-state conformations . The buried conformation correlates with the orientation of Q154 in all other LOV structures as depicted by AsLOV2 . Electron density for FMN is excluded to allow clear observation of electron density for active site side-chains . ( B ) Light state crystal structure of V48I:G80R , 2Fo-Fc data shown at 2 . 0 σ show ( grey mesh ) clear electron density for C4a adduct formation . ( C ) Rotated view of the active site of the light-state ZTL molecule . Modeling of Q154 in the exposed conformation ( yellow; panel C ) results in Fo-Fc ( blue mesh; panel C ) density at 3 . 0 σ for the light-state molecule . Electron density for the FMN is excluded for clarity . The data confirms rotation of Q154 to a buried conformation ( cyan ) following adduct formation . Rotation of Q154 is coupled to rearrangement of V/I48 . In V48I , the additional methyl groups blocks rotation , partially inhibiting population of the buried conformation of Q154 ( cyan conformation ) . ( D ) Predicted divergent ZTL model of allostery and signal transduction based on the integrated structural , mutational and in vivo data . Orientations of WT/G80R Q154 are derived from the dark-state G80R ZTL structure with buried and exposed conformations shown for reference from V48I:G80R . WT/G80R and V48I:G80R deviate from typical LOV models ( derived from VVD: light PDBID 3RH8 , dark PDBID 2PD7 ) on the position of Q154 . WT/G80R ZTL retains a heterogeneous orientation of Q154 . We propose that Q154 is heterogeneous regardless of lighting conditions , but biased towards the buried conformation in the light . At dusk adduct decay , with rate constant k3 , causes the Q154 conformation to be biased towards an exposed conformation , accelerating ubquitination of protein targets . For V48I , I48 selects the exposed conformation in the dark and leads to only partial burying of Q154 ( shown in C , ( D ) , leading to constitutively high ubquitination activity that mimics the dark-state of WT-ZTL . ( Figure 4—figure supplements 1 and 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01410 . 7554/eLife . 21646 . 015Figure 4—figure supplement 1 . Effect of ZTL variants on protein complex formation . ( A ) In vivo interactions between GI and ZTL variants were detected by transient co-expression in N . benthamiana under continuous light or dark conditions . WT and ZTL variants all show enhanced interactions with GI in the presence of light . Ncap variants lead to altered GI interactions . V48I disrupts light-state GI interactions , consistent with altered conformational changes . G46S enhances dark-state binding to GI due to partial dark-state activation . These identify the Ncap as the interaction surface for GI . ( B ) Effect of ZTL LOV domain mutations on ASK2 . Proteins were precipitated with anti-FLAG ( B ) or protein A antibody , and the presence of ZTL variants was detected by anti-HA antibody . All variants indicate that ASK2 interacts with ZTL regardless of lighting conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01510 . 7554/eLife . 21646 . 016Figure 4—figure supplement 2 . Predicted differences in G46 mutants . ( A–C ) Predicted models of G46 mutations and hypothetical effect based on WT ZTL structures . ( A ) The WT ZTL ( grey cartoon ) structure cannot accommodate a sidechain at position 46 . Addition of a Ser residue ( G46S ) at this position leads to steric clashes ( 2 . 2 Å ) between the Cβ position and the exposed conformation ( grey ) of Q154 . Additional clashes are present from the OH group ( dashed lines ) . G46S can be accommodated by very modest movement of Aβ ( yellow ) and rotation of Q154 to the buried position ( yellow ) . Such reorientation favors the putative light-state conformation of Q154 . S46 can adopt to orientations that do not clash with any residues in the ZTL structure ( shown ) . One orientation H-bonds with Q154 . Such interactions should not disrupt the ZTL fold . ( B ) The equivalent orientation of the Glu sidechain in a G46E that minimizes steric clashes . ( C ) A G46E mutation requires rotation of the Glu sidechain out of the active site pocket . The orientation with minimum steric clashes retains a van der Waals contact ( 2 . 5 Å ) with V65 in Cα ( dashed line ) . In addition , charged E46 is forced into a hydrophobic pocked lined by V65 , V69 and F156 . We predict these combined interactions destabilize the LOV fold in G46E . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 016 In ZTL , Q154 does not adopt a specific orientation as observed in all other LOV structures , rather it varies in all four molecules in the asymmetric unit of both WT-ZTL and G80R ( Figure 3A and Figure 3—figure supplement 1 ) . The altered conformations contact the Aβ-Ncap hinge through G46 in the CGF motif ( Ncap; Figure 3B ) . Here , Q154 abuts ( 3 . 4 Å ) the Cα carbon of G46 ( Ser in FKF1 ) . The lack of a side chain at G46 permits favorable interactions for Q154 at the flavin O4 position and allows insertion of F156 into the flavin active site . In this manner an unusual orientation of Q154 links the Iβ-Ccap hinge and helical interface through a QFF motif ( Q154-F155-F156 ) . In most LOV proteins , the equivalent residue to F156 is hydrophilic and adopts a solvent exposed position ( Figure 3 ) . The resulting cavity occupied by F156 is filled by a Met or Leu residue ( V69 in ZTL ) from Cα ( Figure 3B , D ) . In ZTL , the altered orientation of Q154 clashes with typical orientations of Cα . Steric clashes lead to movement of conserved F66 and rotation of V69 away from Q154 and the FMN binding pocket . The combined interactions stemming from Q154 positions it ideally to have concerted conformational changes within Ncap and Ccap elements . Although , the Gln locus has been cited as the primary source of signaling in LOV proteins ( Zoltowski et al . , 2007; Halavaty and Moffat , 2007; Freddolino et al . , 2013 ) , the lack of interactions near N5 and Aβ is atypical . Thus , it is likely that ZTL does not involve an H-bonding switch in signal transduction and that canonical mechanisms of LOV signaling may not be universally conserved . Rather , we propose that steric interactions involving the flavin O4 position and G46 shift a dynamic equilibrium in the Q154 position to stabilize the ZTL protein in a light-state configuration . We posit that the heterogeneous orientation of Q154 destabilizes ZTL in the absence of covalently attached FMN , consistent with in vitro studies demonstrating a tendency to lose flavin rapidly . Upon light activation , ZTL is stabilized by covalent attachment of FMN ( through C82 ) and results in movement of Q154 that necessitates rearrangement of the CGF , QFF and Cα sites to elicit signal transduction . Dark and light state structures of V48I:G80R confirm concerted movement of Q154 that is linked to conformational changes in V/I48 . The Q154 conformation is not heterogeneous in V48I:G80R , rather , the Ile side chain sterically directs Q154 to an exposed conformation near O4 that has not been observed in other LOV proteins . Comparing the dark-state to V48I:G80R grown in the light , confirms density for the C4A adduct and confirms direct crystallization of the light-state ( Figure 4 ) . Difference density maps of the light-state molecule indicate that adduct formation is coupled to rotation of Q154 to a buried position ( Figure 4C ) . Unexpectedly , rotation of Q154 requires concerted movement of I48 that reorients the Ncap . Thus , the presence of G46 selects for a heterogeneous conformation of Q154 and V48I biases the heterogeneous Q154 towards the exposed conformation in the dark . In the light , V48I disrupts rotation of Q154 to the buried conformation , thereby leading to a predominantly exposed conformation regardless of lighting conditions . Thus , unexpectedly , V48I is both an allosteric variant and alters photocycle kinetics . In particular , V48I retains an ability to form the covalent C4a adduct , which is more stable than WT ( V48I k3 = 0 . 094 hr−1; WT k3 = 0 . 71 hr−1 ) . However , in V48I allosteric regulation of the Ncap through Q154 is disrupted leading to V48I selecting for a distinct exposed Q154 conformation despite adduct formation ( Figure 4 ) . Based on our structural results we can refine our models on how ZTL rate altering variants will perturb ZTL function . G80R does not impact ZTL structure or allosteric regulation of Ncap or Ccap elements . In this manner , G80R acts as a photocycle variant only and allows direct testing of LOV photocycle kinetics on targeted degradation of PRR5 and TOC1 ( see Equation 1 ) . In contrast , despite being photochemically active and stabilizing the C4a adduct , V48I blocks allosteric regulation of the Ncap through incomplete rotation of Q154 and selection of a distinct exposed conformation of Q154 . Thus , we propose that V48I is an allosteric variant that mimics the degradation-active dark state . Combining our proposed mechanism with Equation 1 above and existing literature on light-dark formation of ZTL:GI complexes we make the following testable hypotheses: ( 1 ) V48I should demonstrate weaker interactions with GI . ( 2 ) V48I should show constitutive activity regardless of lighting conditions in targeting PRR5 and TOC1 for degradation . Thus , PRR5 and TOC1 levels should be constitutively low in variants containing the V48I mutation . We examined ZTL variants for their ability to form light and dark regulated complexes with GI and ARABIDOPSIS SKP1-LIKE 2 ( ASK2 ) of the SCF complex . CoIP results confirm that G80R retains light regulated complex formation with GI comparable to WT . In contrast , Ncap variants G46S and V48I both alter light driven complex formation with GI ( Figure 4—figure supplement 1 ) . Whereas , V48I leads to dampened light-driven formation of the ZTL:GI complex , G46S enhances GI complex formation in both the dark- and light-states ( Figure 4—figure supplement 1A ) . These results support our allosteric model of ZTL regulation and demonstrate that Ncap variants decouple allosteric regulation of signal transduction from photochemical formation of the C4a adduct . Further , the data implicates light-driven conformational changes near the Ncap in dictating GI interactions . Specifically , where V48I mimics the weak GI interacting dark-state and G46S mimics the strong GI interacting light state . We note that the G46S results reported here diverge from the effects of the G46E mutation reported by Kim et al . ( 2007 ) , where G46E abolishes GI interactions due to apparent misfolding of the LOV domain . These results are informative on the nature of mutations at the G46 site , namely the long side chain present in a G46E variant leads to steric clashes and likely would force E46 into a hydrophobic pocket ( see Figure 4—figure supplement 2 ) . We contend such clashes leads to the destabilization of the LOV domain and abolition of GI interactions as reported previously for G46E ( Kim et al . , 2007 ) . In contrast , the shorter sidechain in G46S , can be easily accommodated by a subtle rotation of the active site Q154 towards the proposed light-state buried conformation ( modeled in Figure 4—figure supplement 2 ) . Such results are consistent with enhanced GI interactions in G46S . Combined , the results implicate the G46/V48 locus in ZTL allostery and regulation . In contrast to GI , where G46/V48 mutations affect function in a light/dark manner , all ZTL variants complex with ASK2 in a light-independent manner comparable to WT ( Figure 4—figure supplement 1B ) . The protein:protein interaction data confirms G80R behaves as WT in known biochemical functions of ZTL and only differs in photocycle kinetics . In contrast , V48I acts as an allosteric variant mimicking the dark-state conformation . Based on these results and previously published in vivo studies showing enhanced degradation activity in the dark ( Más et al . , 2003a ) , we have tools to test the effect of photocycle kinetics ( G80R ) and Ncap allostery ( G46S/V48I ) . In this manner , G80R variants isolate effects of photocycle kinetics allowing testing of predictions based on equation 1 . The decrease in GI affinity in the V48I variant could complicate in vivo phenotypes for these variants . Literature indicates ZTL stability is dictated by GI interactions ( Kim et al . , 2007 ) . Thus , we could expect low ZTL levels in V48I variants . These low ZTL levels would act in opposition to any increased targeted degradation of PRR5/TOC1 and could mask allosteric phenotypes . Based on our experimental conditions , we do not expect any complications to result . Prior studies indicate that ZTL and GI reciprocally stabilize each other ( Kim et al . , 2013 , 2007 ) , but increased ZTL stability occurs in a circadian phase dependent manner ( Kim et al . , 2003 ) . Light does enhance ZTL:GI affinity and reciprocal stability by three fold , but ZTL retains some stabilization during the subjective circadian day regardless of lighting conditions ( Kim et al . , 2007 , 2003 ) . These previous findings suggest that in consideration of the ZTL:GI equilibrium , GI is limiting except during the subjective circadian day . During the day , GI levels rise sufficiently to shift the equilibrium to saturate the ZTL:GI complex regardless of lighting conditions . Given that our V48I variant retains light-state affinity comparable to WT dark-state protein , GI expression during the subjective day under LL conditions should rescue the decrease in affinity . Concomitant with ZTL-ox , reciprocal stabilization of ZTL/GI should lead to high ZTL levels and ZTL should no longer oscillate . Indeed , under our experimental conditions cycling of ZTL protein is lost and V48I variants show enhanced stability in vivo ( Figure 5—figure supplement 1 ) , thereby the role of GI in ZTL stability is masked under our conditions and our results likely reflect the effect of ZTL photochemistry ( G80R ) and allosteric activation of PRR5/TOC1 degradation ( V48I ) independent of the effects on GI binding . Based on the data above we make the following predictions . In comparing G80R-ox to WT-ox under LD conditions , the fast decay of WT-ox will lead to higher populations of the active-dark state early in the evening . In contrast , slower adduct decay in G80R will lead to prolonged population of the less active light-state and delays in ZTL degradation activity . As a result , we should observe an enhanced delay in PRR5/TOC1 degradation in G80R variants as shown in Figure 1 and Figure 5—figure supplement 2 . For V48I and V48I:G80R , the allosteric effects should enhance degradation of PRR5 and TOC1 leading to constitutively low PRR5/TOC1 levels . To test our predictions , we constructed Arabidopsis transgenic ZTL overexpression lines ( ZTL-ox ) containing WT , V48I , G80R and V48I:G80R . G46S variants were excluded due to poor yields from E . coli that render solution or structural information intractable for the isolated G46S variant ( see Materials and methods ) . These constructs were then examined for their effect on PRR5 and TOC1 degradation as well as circadian period and amplitude . All transgenic lines were selected based on similar ZTL transcript levels and ZTL protein levels were measured ( Figure 5—figure supplement 1 ) . Consistent with our predictions , G80R-ox ( #22 ) leads to delayed degradation of PRR5 and TOC1 under LL and LD ( Figure 5 ) . Specifically , despite being overexpressed ( Figure 5—figure supplement 1A ) , G80R-ox variants demonstrate peak amplitudes of TOC1 consistent with WT ( Figure 5A , B ) . Further , comparison of apparent rate constants for PRR5 degradation confirms a direct effect of ZTL photocycle kinetics on PRR5 degradation , where G80R #22 ( 0 . 34 hr−1 LD , 0 . 14 hr−1 LL ) exhibits smaller rate constants compared to WT-ox ( 0 . 5 hr−1 LD , 0 . 13 hr−1 LL ) under LD conditions where k3 is most relevant ( Table 3 ) . These results are consistent with a more active dark-state ZTL . Thus , ZTL photocycle kinetics regulate degradation of TOC1/PRR5 . 10 . 7554/eLife . 21646 . 017Figure 5 . Diurnal and circadian expression profiles of PRR5 and TOC1 proteins in ZTL variant overexpressors . ( A ) PRR5 and TOC1 protein levels were analyzed in WT , 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) under 12L/12D conditions . Actin ( ACT ) was used as a loading control for PRR5 . Arrowhead indicates the band corresponding to TOC1 protein , while an asterisk indicates a nonspecific cross-reacting band , which is used as a loading control . ( B ) Relative expression level of PRR5 and TOC1 were determined in WT , 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) under 12L/12D conditions . Actin and the TOC1 nonspecific bands were used for normalizing protein loadings for quantification of PRR5 and TOC1 . The data represent the averages ±SEM obtained from three biological replicates . ( C ) PRR5 and TOC1 protein levels were analyzed in WT , 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) during the subjective night under constant light conditions . ( D ) Relative levels of PRR5 and TOC1 proteins were determined in WT , 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) . Dashed lines represent protein levels under 12L/12D conditions , while solid lines represent protein levels under constant light conditions . The data represent the averages ±SEM obtained from three biological replicates . ( Figure —figure supplement 1 and 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01710 . 7554/eLife . 21646 . 018Figure 5—figure supplement 1 . Relative expression levels of clock proteins . Expression levels of total ZTL and endogenous ZTL transcripts ( A ) , and also ZTL protein ( B ) were analyzed in WT , 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) plants harvested at ZT8 under 12L/12D conditions . ( A ) Relative expression levels of ZTL were determined using IPP2 as the internal control . ( B ) Representative blot images of HA-ZTL and ACT are shown . Relative expression levels of ZTL protein were determined from three biological replicates . ( C ) Expression levels of PRR5 and TOC1 transcripts were analyzed in WT , 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) plants harvested at ZT8 under 12L/12D conditions . Relative expression levels of PRR5 and TOC1 were determined using IPP2 as the internal control . ( D ) Diurnal expression levels of ZTL protein were analyzed in 35S: HA-ZTL , 35S: HA-ZTL ( V48I ) , 35S: HA-ZTL ( G80R ) and 35S: HA-ZTL ( V48I:G80R ) plants under 12L/12D conditions . Representative blot images of HA-ZTL and ACT are shown . Relative expression levels of ZTL protein were determined using ACT as the internal control . All data represent averages ±SEM obtained from three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01810 . 7554/eLife . 21646 . 019Figure 5—figure supplement 2 . Comparison of model to PRR5 degradation in vivo . Comparison of PRR5 degradation model ( Equation 1 and Equation S10 ) with experimental data . The red curve ( simulated for G80R ) shows reasonable precision in predicting the observed delay in PRR5 degradation for G80R data ( red circles ) compared to simulated WT ( black ) and WT data ( black squares ) . See methods for model generation and parameter estimation . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 01910 . 7554/eLife . 21646 . 020Table 3 . Period length estimates of CCA1:LUC activity in WT and ZTL variants overexpression plants . See Figure 5—figure supplement 1 for expression levels . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 020GenotypeEstimated Period length ( hrs ) HA-ZTL protein abundance*Estimated kdeg PRR5 LD ( hrs−1 ) *Estimated kdeg PRR5 LL ( hrs−1 ) WT24 . 36 ± 0 . 40ND0 . 3 ± 0 . 10 . 13 ± 0 . 135S:HA-ZTL #723 . 69 ± 1 . 2310 . 5 ± 0 . 10 . 13 ± 0 . 135S:HA-ZTL #1723 . 65 ± 0 . 721 . 5 ± 0 . 3NDND35S:HA-ZTL ( V48I ) #7ND11 ± 40 . 8 ± 0 . 10 . 5 ± 0 . 235S:HA-ZTL ( V48I ) #22ND14 ± 5NDND35S:HA-ZTL ( G80R ) #2224 . 17 ± 1 . 083 ± 0 . 50 . 34 ± 0 . 050 . 14 ± 0 . 135S:HA-ZTL ( G80R ) #2319 . 33 ± 1 . 0612 ± 5NDND35S:HA-ZTL ( V48I:G80R ) #45ND10 ± 30 . 8 ± 0 . 20 . 3 ± 0 . 235S:HA-ZTL ( V48I:G80R ) #49ND7 . 6 ± 2NDND*Estimated kdeg values were extracted by fitting Equation S10 ( below ) to the PRR5 protein levels in vivo . For LD conditions , an average k is obtained by treating the system as only containing dark-state protein . Thus , the LD values are accurate as comparative terms between variants only . Examining V48I and V48I:G80R variants confirms a role of V48I in altering light-dark regulation of ZTL activity . Strains harboring V48I and V48I:G80R show constitutively low levels of PRR5 and TOC1 , consistent with high degradation activity regardless of lighting conditions ( Figure 5 ) . Both variants show degradation rate constants for PRR5 of ~0 . 8–1 hr−1 under LD conditions , similar to maximum rate constants ( 0 . 8 hr−1 ) predicted from computation models of clock function ( De Caluwé et al . , 2016; Pokhilko et al . , 2013 ) ( see Materials and methods and Table 3 ) . V48I and V48I:G80R also exhibit high activity under LL conditions , demonstrating a partial loss of light-dark regulation ( Figure 5 and Table 3 ) . Combined , G80R confirms a direct role of LOV photocycle kinetics on ZTL activity and that V48I acts in an allosteric switch to enable light-state V48I ( adduct formation ) to mimic the more degradation active and less GI-binding competent dark-state . Given the effect of V48I , G80R and V48I:G80R on PRR5 and TOC1 degradation , these variants should have predictable effects on circadian period . Previous studies of ZTL-ox variants demonstrate a dose dependent effect of ZTL on circadian period , where high protein levels lead to short-period phenotypes progressing to arrhythmicity and ZTL-nulls having a long-period phenotype ( Somers et al . , 2004 ) . Similarly , TOC1 overexpression strains have a long-period phenotype and TOC1-null strains have short periods ( Más et al . , 2003b; Gendron et al . , 2012 ) . Thus , we predict that despite overexpression the defect in degradation of TOC1 by G80R-ox should lead to WT periods . In contrast , low TOC1 levels in V48I and V48I:G80R should lead to arrhythmic phenotypes under expression levels comparable to WT-ox strains demonstrating normal periods . Indeed examination of circadian periods in WT and mutant-ox strains confirms predicted effects of photocycle kinetics ( G80R ) and the signaling defect ( V48I ) on circadian period . All strains containing V48I lead to an arrhythmic phenotype consistent with heightened degradation activity in these variants ( Figure 6 and Table 3 ) . In contrast , G80R-ox strains harboring 3-times more ZTL protein than WT has a circadian period indistinguishable from WT and WT-ox ( see Table 3; 35S:HA-ZTL ( G80R ) #22; 24 hr , compared to WT; 24 hr and 35S:HA-ZTL #17; 24 hr ) . Only when protein levels exceed 10-fold higher than WT-ox is the period shortened to 19 hr ( 35S:HA-ZTL ( G80R ) #23 ) ( Figure 6 and Table 3 ) . These results confirm that ZTL photocycle kinetics are coupled to selection of circadian period through PRR5/TOC1 degradation . 10 . 7554/eLife . 21646 . 021Figure 6 . Circadian clock phenotypes of ZTL variant overexpressors . ( A–D ) CCA1:LUC activity was analyzed in WT , 35S:HA-ZTL ( A ) , 35S:HA-ZTL ( V48I ) ( B ) , 35S:HA-ZTL ( G80R ) ( C ) and 35S:HA-ZTL ( V48I:G80R ) ( D ) lines under continuous light conditions . CCA1:LUC traces represent the averages ±SEM of the results obtained from eight individual seedlings . Period length estimation and relative amplification errors of 16 individual measurements are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 021
In Arabidopsis thaliana , ZTL and FKF1 diverge in their diurnal pattern of transcription and function . Whereas , FKF1 function and transcription is specific to the day , ZTL demonstrates constitutive transcription and distinct functions during the day and night ( Song et al . , 2014 ) . These distinct functions allow ZTL to impact photoperiodic flowering through morning specific and GI-dependent destabilization of CO , and circadian function through night specific degradation of TOC1 and PRR5 ( Más et al . , 2003a; Song et al . , 2014; Kiba et al . , 2007 ) . These results suggest that both light ( CO destabilization/GI interaction ) and dark ( TOC1/PRR5 degradation ) activate ZTL for function . Such observations seem paradoxical; however , examination of our structural and kinetic results indicates ZTL may use a divergent allosteric mechanism to enable dual light/dark functions . The ability of LOV domains to function as photoswitchable proteins is predicated on LOV proteins specifying a distinct light- and dark-state configuration . In all dark and light-state LOV structures currently in the protein data bank , the active site Gln ( Q154 in ZTL ) adopts a distinct buried conformation in the dark , with the NH moiety of the Gln side chain near flavin-N5 ( Halavaty and Moffat , 2007; Zoltowski et al . , 2007 ) . Existing light-state structures indicate that C4a adduct formation and N5 protonation induces rotation of the Gln side chain to favor an H-bond between flavin-HN5 and the O moiety of the Gln side chain ( Vaidya et al . , 2011 ) . In those structures , the Gln residue maintains the buried conformation and is only distinguished by the nature of H-bonds . In our ZTL structures this is not the case , rather dark-state structures have a heterogeneous orientation of Q154 that samples orientation near both the buried conformation and an exposed conformation ( Figure 3—figure supplement 1 ) . A heterogeneous conformation of the key signaling switch would suggest poor regulation of function under dark-state conditions . This is exactly what is observed in ZTL . ZTL retains fairly robust activity for GI interactions and TOC1/PRR5 degradation in both the light and dark , with light enhancing GI interactions and repressing TOC1/PRR5 degradation by 3–5 fold ( Kim et al . , 2007 ) and ( Table 3 ) . As such , GI-dependent function in CO destabilization is likely not a light-regulated event , but rather is a combined result of constitutive transcription of ZTL , day-specific expression of GI and poor signal amplification following light-dark interconversion due to the heterogeneous orientation of Q154 . Thus , the unusual Gln orientations appear to be evolutionarily selected to permit light- and dark-state functions of ZTL and to differentiate ZTL and FKF1 . Combining our dark- and light-state structures with in vivo data allows construction of a putative signaling mechanism differentiating ZTL from other LOV proteins . Examination of the ZTL structures identifies two key residues involved in regulating ZTL allostery . The heterogeneous conformation of Q154 is permitted by the lack of a sidechain in G46 ( model in Figure 4—figure supplement 2 ) . Introduction of V48I , directs Q154 to a single exposed conformation in the dark . Biasing the Q154 to the exposed conformation results in disrupted interactions with GI and high TOC1/PRR5 degradation , in vivo , consistent with selecting for a distinct dark-state conformation . Further , the light-state V48I:G80R molecule indicates partial Q154 rotation to a buried conformation that is impeded by V48I . This impediment does not allow robust sampling of the buried conformation under any lighting conditions and coincides with constitutively high TOC1/PRR5 degradation in V48I containing variants . Combined we propose a putative model of ZTL signaling ( Figure 4 ) . ZTL retains functionality in the light and dark due to a heterogeneous orientation of Q154 permitted by G46 and V48 . Differences in functionality between dark/light result from subtle biases in the orientation of Q154 between buried and exposed conformations . Based on V48I:G80R structures and activity , we propose that biasing towards the exposed conformation as the dark-state configuration ( enhanced PRR5/TOC1 degradation ) and biasing towards the buried conformation as the light-state configuration ( enhanced GI binding ) . Based on our proposed mechanism distinguishing ZTL signal transduction from other LOV proteins , one would predict evolutionary selection of G46 and V48 in ZTL proteins to permit the heterogeneous orientation of Q154 . Phylogeny of LOV sequences in planta demonstrates evolutionary selection of residues G46 , V48 and F156 to differentiate ZTL-like , FKF1-like and phototropin-like ( LOV2 ) proteins and putative signal transduction pathways ( Figure 7 and Figure 7—figure supplement 1 ) . Specifically , in monocots and dicots ZTL-like proteins conserve the G46 that is necessary for selection of the exposed conformation of Q154 . In FKF1 ( monocots: A46 and dicots: S46 ) , phototropins ( N46 ) and all other structurally characterized LOV proteins , an H-bond or sterically directing residue occupies this position . 10 . 7554/eLife . 21646 . 022Figure 7 . Phylogenetic Analysis of FKF1/ZTL family members in plants . Residue identity at position 46 ( Colored Bar ) distinguishes ZTL-like , LKP2-like and FKF1-like proteins consistent with evolutionary diversification of signaling mechanisms . LKP2 is isolated to a clade containing Brassica rapa members that all contain a Q154L substitution . Al ZTL members contain G46 which is necessary to promote the alternative conformation of Q154 . Spikemoss and liverwort FKF1’s are isolated indicating a possible intermediate function . Accession numbers for all sequences are shown after the protein name . The evolutionary history was inferred using the Minimum Evolution method ( Rzhetsky and Nei , 1992 ) . The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test ( 500 replicates ) are shown next to the branches ( Felsenstein , 1985 ) . The tree is drawn to scale , with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree . The evolutionary distances were computed using the Poisson correction method ( Zuckerkandl and Pauling , 1965 ) and are in the units of the number of amino acid substitutions per site . The ME tree was searched using the Close-Neighbor-Interchange ( CNI ) algorithm ( Nei and Kumar , 2000 ) at a search level of 1 . The Neighbor-joining algorithm ( Saitou and Nei , 1987 ) was used to generate the initial tree . The analysis involved 28 amino acid sequences . All positions containing gaps and missing data were eliminated . There were a total of 535 positions in the final dataset . Evolutionary analyses were conducted in MEGA7 ( 68 ) . ( Figure 7—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 02210 . 7554/eLife . 21646 . 023Figure 7—figure supplement 1 . Sequence Alignment of FKF1/ZTL family members in plants . ZTL/FKF1/LKP2 homologs from dicots; ( Arabidopsis thaliana ( At ) , Brassica rapa ( Bra ) , Glycine max ( Gm ) , Cucumis sativus ( Cs ) , Citrus clementine ( Cc ) , Gossypium hirsutum ( Gh ) , Avena sativa ( As ) , monocots; Oryza sativa japonica ( Os ) , Setaria italica ( Si ) , Zea mays ( Zm ) , Brachypodium distachyon ( Bd ) , Liverwort; Marchantia polymorpha ( Mp ) and Spikemoss; Selaginella moellendorffii ( Sm ) . ZTL and FKF1 cluster in reference to conserved CGF and QFF motifs . See Figure 7 for corresponding accession numbers . All ZTL proteins conserve G46 and V48 ( blue ) . In FKF1 the position corresponding to G46 contain Ala ( FKF1-like monocots ) or Ser ( FKF1-like dicots ) ( red ) ; Phototropins contain an Asn at the equivalent position ( green ) . * denote conserved residues through all proteins . All proteins conserve the canonical GXNCRFLQ motif ( magenta ) as well as residues leading into the E-F loop . The FKF1 species differ in the residues immediately following the LOV consensus sequence in the beginning of the E-F loop ( ZTL: C87 and G89; FKF1: F87 and D89 ) . Liverwort and spikemoss sequences diverge containing elements consistent with both ZTL and FKF1 ( G46 and ZTL E-F loop but I48 for Mp; S46 and F87 but G89 for Sm ) , indicating an evolutionary transition . The QFF ( blue ) motif is more divergent . All ZTL/FKF1 contain a Phe at position 156 that occupies an alternative buried position compared to solvent exposed hydrophilic residues in other LOV proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 21646 . 023 Residue selection at position 46 is coupled to V48 . V48 is conserved in ZTL , but substitutions are permitted in FKF1 , phototropins and fungal LOVs ( Figure 7—figure supplement 1 ) ( Lokhandwala et al . , 2015; Zoltowski et al . , 2007 ) . The altered signaling mechanism in ZTL is supported by examination of LOV proteins in Brassica rapa . ZTL-like proteins in Brassica and AtLKP2 contain a Q154L substitution , resulting in an LXF consensus sequence at that locus . The presence of a Q154L substitution is unexpected , since these substitutions abrogate blue light signaling in other LOV proteins due to an inability to undergo the Gln-flip mechanism ( Zoltowski et al . , 2007; Nash et al . , 2008; Nozaki et al . , 2004 ) . In the proposed ZTL mechanism a Gln-flip is not necessitated and Q→L substitutions are permitted . In this manner , our understanding of a canonical model of LOV signal transduction is incomplete . We propose that evolutionary selections at G46 and V48 tune allostery to the Ncap through the Q154 orientation to differentiate modes of LOV signal transduction . Given the evolutionary selection of A/S residues and permission of V48 substitutions in FKF1 , we propose that FKF1 functions in a manner analogous to other LOV proteins , where light activates the biological function ( CO stabilization/CDF degradation ) of the primarily nuclear FKF1 protein . In this manner , evolutionary selection at position 46 may dictate day functional ( FKF1 ) and day/night functional ( ZTL ) differentiation , thereby implicating LOV photocycle kinetics as being imperative for proper signal transduction . Currently , delays in TOC1 and PRR5 degradation are explained using competition between GI and TOC1/PRR5 for the available ZTL pool ( Más et al . , 2003a; Kim et al . , 2007; Kiba et al . , 2007; Fujiwara et al . , 2008 ) . In these models , GI expression during the day both stabilizes and sequesters ZTL . In the evening , GI pools decline leading to active free ZTL . In light of our data , the competition model must be incomplete . In the competition model there should be no differences in the LL/LD characteristics in WT and G80R strains . In the competition model GI expression profiles should dictate delays and these would be unaffected by rate-altering ZTL variants . WT and G80R interact with GI to comparable levels , yet G80R demonstrates enhanced delays in PRR5 and TOC1 degradation ( Figure 5D ) . These results are inconsistent with a competition model alone . In contrast , modeling PRR5 degradation using the time dependent conversion of ZTL-light to ZTL-dark can predict with reasonable certainty the extended delay in G80R and predicts distinct differences under LL conditions between the two proteins ( Figure 5—figure supplement 2 ) . In this manner , it is clear that ZTL photocycle kinetics are instrumental in dictating delays in PRR5/TOC1 turnover and circadian period . Based on all these elements we propose that circadian period is regulated in a complex manner involving a ZTL:TOC1:GI circuit , whereby competitive inhibition and ZTL photocycle kinetics act in concert to dictate ZTL protein levels and ZTL activity in a circadian phase dependent manner . Adduct decay in ZTL then enhances degradation of PRR5 and TOC1 , impacting circadian period through two methods: ( 1 ) Degradation of PRR5 , helps diminish TOC1 levels through increased cytosolic retention and accessibility to ZTL ( Wang et al . , 2010 ) and ( 2 ) Degradation of cytosolic TOC1 . Both factors require proper ZTL photocycle kinetics . Thus , LOV photocycle kinetics are instrumental in evolutionary selection for a 24 hr period .
Examination of recent mathematical models of plant circadian clocks reveals a common method of incorporating light-dark dependent degradation of PRR5 and TOC1 by ZTL . In Pokhilko et al . , PRR5 and TOC1 levels are treated as follows ( Pokhilko et al . , 2013 ) : ( S1 ) dcp5dt=p10cp5m− ( m17+m24D ) ∗cP5 ( S2 ) dcTOC1dt=p4 ( cTOC1m+n16 ) − ( m6+m7D ) *cTOC1* ( cZTL*p5+cZG ) −m8cTOC1 Where Cx represent protein concentrations , Cxm , mRNA levels , px , nx and mx are parameters fit to data sets . ZG represents the ZTL:GI complex . D represents darkness , where D = 1 at night and D = 0 during the day . Both equations are constructed of similar elements , a protein synthesis term defined by mRNA levels and a degradation term defined by PRR5/TOC1 protein levels , and in the case of TOC1 the total ZTL protein pool . We note , that ZTL protein levels are not incorporated into existing models of PRR5 degradation ( De Caluwé et al . , 2016; Pokhilko et al . , 2013 ) suggesting that overexpression has a weak enough effect on the overall rate of PRR5 degradation that models lacking ZTL concentration can accurately predict degradation kinetics . Although biologically such an analysis is incomplete , for modeling purposes the accuracy of these prior models suggests this assumption is reasonably valid . Therefore , they normalize the ZTL concentration to one and it does not appear in the PRR5 degradation term . Further , in the fit parameters , p5 = 1 , which allows simplification as shown in Equation S4 , where cZTL now equals the total pool of ZTL protein regardless of whether free or in complex with GI . For our purposes , ZTL rate altering variants should not perturb the transcription rates , therefore differences in PRR5/TOC1 degradation should be limited to the degradation terms . For PRR5 and TOC1 the degradation terms are as follows ( S3 ) dcp5dt=− ( m17+m24D ) *cP5 ( S4 ) dcTOC1dt=− ( m6+m7D ) *cTOC1cZTL−m8cTOC1 The above equations do not incorporate ZTL photochemistry , rather , the term m*D to enhance the degradation rate constant under dark state conditions . In the dark , D = 1 and the degradation rate constant ( k2 ) becomes ( m17 + m24 ) or ( m6 + m7 ) for PRR5 and TOC1 respectively . In the light , D = 0 and the rate constants ( k1 ) reduce to m17 and m6 . To add ZTL photochemistry to these equations we rewrite these degradation equations in terms of light and dark-state ZTL . We also note additional complications in the TOC1 equation . The term m8*cTOC1 is a non-ZTL dependent degradation term , presumably accounting for nuclear degradation of TOC1 . Because of the additional complexities in TOC1 expression and non-ZTL dependent degradation we do not model TOC1 degradation in vivo , we do however show below that the rate constant for LOV adduct decay , k3 , will impact TOC1 in a manner analogous to PRR5 . ( S5 ) dcp5dt=− ( k1*cZTL−L +k2*cZTL−D ) *cP5 ( S6 ) dcTOC1dt=− ( k1*cZTL−L +k2*cZTL−D ) *cTOC1−m8*cTOC1 Where k1 , k2 are the rate constant for degradation in the light and dark respectively . Similarly , cZTL-L and cZTL-D represent the light and dark-state levels of ZTL . We also normalize the total ZTL concentration to 1 . Doing so allows calculation of cZTL-L and cZTL-D as a function of time during the dark-phase of LD cycles , by incorporating the rate constant for adduct decay , k3 . ( S7 ) cZTL−D=1−cZTL−L ( S8 ) cZTL−L ( t ) =e−k3t It follows that: ( S9 ) dcp5dt=−[ k1*cZTL−L +k2* ( 1−cZTL−L ) ]*cP5 ( S10 ) dcp5dt=−[ ( k1−k2 ) *e−k3t +k2]cP5 Similary: ( S11 ) dcTOC1dt=−[ k1*cZTL−L +k2* ( 1−cZTL−L ) ]*cTOC1−m8*cTOC1 ( S12 ) dcTOC1dt=−[ ( k1−k2 ) *e−k3t +k2]cTOC1−m8*cTOC1 We note , that a compact model of clock function by De Caluwe et al . demonstrates that existing data can be adequately fit for both PRR5 and TOC1 levels using an analogous equation for both PRR5 and TOC1 ( only the degradation term has been extracted ) ( De Caluwé et al . , 2016 ) . ( S13 ) dcPTdt=− ( kDD+kLL ) *cPT Where , kD , kL are the light and dark degradation rate constants respectively , and L and D reference the lighting conditions , where L = 1 , D = 0 during the day and L = 0 , D = 1 during the evening . CPT represents either the PRR5 or TOC1 level . Applying our method above to the De Caluwe et al . model ( Equation S13 ) , results in an analogous expression as Equation S10 . Thus , in terms of ZTL chemistry , PRR5 and TOC1 degradation should be dictated by k3 . Using mutants that affect only k3 , see Table 1 , one should be able to accurately model PRR5 decay rates in vivo . Again we note that the expression pattern of TOC1 and complex regulation of TOC1 mRNA complicates TOC1 levels during the circadian cycle . For these reasons , we use the simpler PRR5 degradation data to mathematically test our model above and use qualitative differences in degradation patterns to examine effects of k3 on TOC1 . To predict how k3 may affect delays in PRR5 degradation we required estimates of k1 and k2 . Examining Pokhilko et al . and De Caluwe et al . provides reasonable bounds for parameter estimation . First , PRR5 models suggest that PRR5 can be accurately modeled excluding differences in ZTL expression levels ( De Caluwé et al . , 2016; Pokhilko et al . , 2013 ) , this is consistent with our data showing that WT , WT-ox and G80R-ox all have the same decay rate under LL conditions despite differences in protein levels . Second , in De Caluwe et al . the maximum degradation rate constant for PRR5 degradation was fit as 0 . 78 hr−1 . Similarly , in Pokhilko et al . a max value of 0 . 5 hr−1 was obtained for PRR5 and a kmax of 0 . 8 hr−1 for TOC1 by combining all degradation terms . Given the similarity of these maximum values we chose a kmax for the dark-state of 0 . 8 hr−1 . The lowest fit rate constant in mathematical models is m6 ( 0 . 2 hr−1 ) for TOC1 degradation under light-state conditions . Thus , we used 0 . 2 hr−1 as an estimate of PRR5/TOC1 degradation in the light . Data was then simulated by numerically solving Equation S9 in matlab using the following parameters and plotted in Figure 1B . k1 = 0 . 2 hr−1 k2 = 0 . 8 hr−1 k3 = 0 . 7 hr−1 ( WT ) , 0 . 15 hr−1 ( G80R ) , 0 . 09 hr−1 ( V48I ) , 0 . 05 hr−1 ( G46S:G80R ) , 0 . 02 hr−1 ( V48I:G80R ) PRR5 ( 0 ) =1 Values for k3 are derived from experimental values for the adduct decay time constant ( τ ) present in Table 1 , where k3 = 1/τ . To estimate the relative accuracy of our model depicted in Equation S10 , we extracted improved estimates of k1 and k2 from the experimental data shown in Figure 5D . Our structural data indicates that V48I mimics the dark-state regardless of lighting conditions , thus , V48I degradation kinetics under LD serves as a reasonable estimate for k2 , the dark-state rate constant . We note , that our experimental value for V48I ( 0 . 8 hr−1 ) is identical , to the kmax values present in both Pokhilko et al . and De Caluwe et al , which should reference the more active dark-state value . Under LL conditions we observe a rate constant for PRR5 degradation of 0 . 13–0 . 14 hr−1 for WT , WT-ox and G80R . Under our lighting conditions ( broad spectrum; 40–50 µmol m−2s−1 ) , the light-state should be near saturated , and thus in the absence of allosteric effects should report the light-state degradation rate constant . We use 0 . 14 hr−1 then as an estimate for k1 the light-state rate constant . We note the similarity of values between the three strains despite different expression levels . These results suggest that the models by Pokhilko et al . and De Caluwe et al . demonstrating a weak dependence of PRR5 degradation on ZTL expression levels is reasonably valid under our conditions . Further , it provides a reasonable estimate for k1 the light-state degradation term . Thus , to test the accuracy of our model we numerically solved Equation S9 in matlab using: k1 = 0 . 14 hr−1 k2 = 0 . 8 hr−1 k3 = 0 . 7 hr−1 ( WT ) , 0 . 15 hr−1 ( G80R ) Initial PRR5 levels were taken from Figure 5D at t = 12 hr ( initial dark ) . Values for k3 are derived from experimental values for the adduct decay time constant ( τ ) present in Table 1 , where k3 = 1/τ . Results were plotted in Figure 5—figure supplement 2 . ZEITLUPE construct ZTL-S composed of residues 29–165 were cloned into both p-His and pGST parallel vectors using NcoI and XhoI restriction sites . Proteins were purified as reported previously ( Pudasaini and Zoltowski , 2013 ) . These DNA sequences were verified by GENEWIZ sequencing service . The plasmids were then isolated and tested for protein expression . All constructs were transformed into E . coli ( JM109 or JM109DE3 ) cells and grown overnight at 37°C as starter culture till O . D600 ~0 . 6–0 . 7 . The rich culture was then transferred into 1 . 0 L of LB media for large-scale expression . These cultures were grown for 2–3 hr at 37°C till O . D600 ~0 . 5–0 . 6 then the temperature was lowered to 18°C . The culture was then induced with 200 µM Isopropyl-β-D-thio-galactopyranoside ( IPTG ) after 1 hr incubation at 18°C . After induction , the cells were grown at 18°C for about 18–20 hr . Finally , the cells were centrifuged at 4000 rpm to collect and store the cells in stabilizing buffer ( 50 mM Tris pH 7or 50 mM Hepes pH 8 with 100 mM and 10% glycerol ) . The harvested cell-pellets were stored for later use at −80°C . We note that ZTL is difficult to express and purify from E . coli as most WT proteins are confined to inclusion bodies . Typical experiments required cell pellets from 18 L of cells . Such difficulties make studies of variants such as G46S difficult in the absence of G80R . The G80R variant , likely due to the stabilizing effects of the R80-F84 π-cation interaction ( Figure 2B ) , enhances protein yields in E . coli substantially . This allows access to G46S:G80R , despite intractable yields of the isolated G46S variant . ZTL-S was purified using affinity chromatography followed by size exclusion chromatography . Prior to purification the cells were lysed by sonication at 4°C . After sonication the protein solution was centrifuged at 18 k rpm at 4°C for 60 min . The supernatant was then purified using Ni-NTA or GST affinity columns . 6xHis and GST tags were cleaved using 6His-TEV-protease followed by an additional round of Ni-NTA affinity chromatography to remove the 6His tag and 6His-TEV-protease . The final eluted protein was subjected to Fast Protein Liquid Chromatography ( FPLC ) using a Hiload Superdex 200 16/60 gel filtration column equilibrated with stabilizing buffer . Solution characterization of purified proteins was done using a Superdex 200 10/300 analytical column ( GE Lifesciences ) . Protein concentrations were determined using the absorbance at 450 nm ( ext . coefficient 12 , 500 ) . Apparent molecular weights were calculated by comparing the elution volume of known standards ( sweet potato β-amylase ( 200 kDa ) −12 . 4 ml; yeast alcohol dehydrogenase ( 150 kDa ) −13 . 31 ml; bovine albumin ( 66 kDa ) −14 . 61 ml; carbonic anhydrase ( 29 kDa ) −17 . 03 ml; horse heart cyctochrome c ( 12 . 4 kDa ) −18 . 32 ml ) ) ( Sigma Aldrich ) . Absolute molecular weights of ZTL 16–165 , ZTL 29–165 and FKF1 28–174 were determined by subjecting samples to refractive index and light-scattering detectors on a Wyatt Minidawn light-scattering instrument following a Superdex 200 10/300 analytical column . MW’s were determined using ASTRA software from Wyatt Technologies ( Santa Barbara , California ) . All SEC and multi-angle light scattering experiments were conducted in stabilizing buffer ( see above ) . Site specific protein variants of ZTL-S constructs were obtained using the quickchange protocol . Following PCR amplification samples were treated with 1 µL of DpnI and incubated at 37°C for 2 . 5–3 hr to cleave the methylated template DNA . A single colony of DH5α E . coli was grown at 37°C overnight and plasmid DNA was isolated and verified by DNA sequencing ( Genewiz ) . Rate altering variants of ZTL were expressed , purified and characterized using the method described above . ZTL-S and its variants were initially screened with Hampton Screens ( HR2-110 and HR2-112 ) via the hanging drop methods using 1 . 5 µL well solution with 1 . 5 µL of ZTL-S at various concentration range of 5–10 mg/ml . Optimum crystallization conditions for WT ( 0 . 1 M Tris pH 8 . 5 , 0 . 1 M Magnesium Chloride hexahydrate , 28% w/v PEG 4k ) , G80R ( 0 . 1 M Tris pH 8 . 5 , 0 . 2 M Magnesium Chloride hexahydrate , 30% w/v PEG 4k ) , V48I:G80R ( 0 . 1 M Tris ph 8 . 5 , 0 . 2 M Sodium Acetate trihydrate , 30% w/v PEG 4K ) . Protein for crystallographic studies was purified in the same stabilizing buffer . Light state crystals of V48I:G80R were obtained as follows . Prior to setting screens V48I:G80R protein was exposed to a broad spectrum white flood light ( 150 W ) , while on ice for two minutes . Saturation of the light-state was confirmed by UV-vis spectra analysis , by verifying disappearance of the 450 and 478 nm absorption bands characteristic of the dark-state protein . The light-state protein was then crystallized directly using the hanging drop method outlined above . Crystals appeared within 24 hr , and trays were exposed to white light once a day to maintain population of the light-state species . Diffraction data was collected at the F1 beamline at the Cornell High-Energy Synchrotron Source ( CHESS ) . Data for WT and all variants was collected at 100 K . The following cryoprotectants were added: V48I:G80R dark ( 25% Glycerol v/v ) , V48I:G80R light ( 25% Glycerol v/v ) , G80R ( 25% ethylene glycol v/v ) , and WT ( 25% ethylene glycol v/v ) . Data was scaled and reduced in HKL2000 ( Otwinowski and Minor , 1997 ) ( see Equation S2 for refinement statistics ) . The initial WT structure was solved using molecular replacement in PHASER ( McCoy et al . , 2007 ) and PHENIX ( Adams et al . , 2010 ) with the LOV1 domain of Arabidopsis phototropin 2 ( PDBID 2Z6D ) as a search model . Structures of ZTL variants were solved using the same method with WT ZTL as the search model . Rebuild cycles were completed in COOT ( Emsley and Cowtan , 2004 ) and refinement with REFMAC5 ( Murshudov et al . , 1997 ) and PHENIX ( Adams et al . , 2010 ) . All ZTL structures contain four molecules per asymmetric unit that is composed of two anti-parallel LOV dimers . Residues 29–31 are not visible in the electron density in any molecule . In several molecules residues 29–43 and 164–165 are unable to be resolved and have not been built . For light-state structures clear density is observed for the adduct state in two of four molecules . Although electron density suggests an adduct for the remaining two molecules , we do not model them with an adduct and the reduced density likely reflects reduction by x-rays during data collection as has been observed previously ( Zoltowski et al . , 2007 ) . Purified protein fractions were concentrated to 30–60 µM for UV-Vis absorption spectroscopy measurements and kinetics . Samples were exposed to a broad spectrum white flood light ( 150 W ) , while on ice for two minutes . An Agilent UV3600 spectrophotometer was used to characterize the absorption spectra of all constructs in both light and dark states . The light state peak of 378 nm ( ext . coefficient 8500 ) and dark state peak 450 nm ( ext . coefficient 12 , 500 ) were used to estimate the protein concentrations for experiments . Photocycle recovery kinetics were analyzed by measuring the absorption at 450 and 478 nm as a function of time . Spectra were collected at intervals ranging from 100 seconds-2 hours to ensure minimal repopulation of the light-state by the probe source . Time intervals were chosen to maintain approximately 10–20 measurements per half life . Kinetic traces at 450 and 478 nm were then fit with a monoexponential decay of the form y = y0 + A*e-k*x . The rate constant k and time constant ( 1/k ) were abstracted . Results are presented in Table 1 as the average of three independent measurements . The Columbia-0 ( Col-0 ) plant that possesses CCA:LUC reporter was previously described ( Pruneda-Paz et al . , 2009 ) . To generate overexpressors of HA-ZTL , HA-ZTL ( V48I ) , HA-ZTL ( G80R ) and HA-ZTL ( V48IG80R ) , the nucleotide sequences encoding HA tag was incorporated into the ZTL forward primer ( 5’-CACCATGTACCCATACGATGTTCCAGATTACGCTGAGTGGGACAGTGGTTC-3’ , the underline indicates the sequences that encodes HA tag ) . Amino acid substitutions on ZTL coding region were generated by using megaprimer-based PCR amplification method ( Burke and Barik , 2003 ) . The primers used for generating the mutated ZTL coding sequence are followings; ZTL ( V48I ) _R ( 5’- AACGGCATCAGTAACAATGAATCCACAAGGCGC-3’ ) , ZTL ( G80R ) _R ( 5’- CAAGAAGCGGCAATTTCGTCCGAGAACTTCCTC-3’ ) , ZTL_R ( 5’- TTACGTGAGATAGCTCGCTA-3’ ) . Amplified PCR fragments were cloned into pENTR/D-TOPO vector ( Invitrogen ) . After verifying sequences , HA-ZTL and mutated HA-ZTL coding regions were transferred into pB7WG2 or pH7WG2 binary vector ( Karimi et al . , 2002 ) using LR Clonase II enzyme mix ( Invitrogen ) to generate 35S:HA-ZTL , 35S:HA-ZTL ( V48I ) , 35S:HA-ZTL ( G80R ) and 35S:HA-ZTL ( V48I:G80R ) . The binary vectors were introduced into the CCA1:LUC plants by conventional Agrobacterium-mediated transformation method . The T3 generations of transgenic lines in which the expression levels of ZTL variant mRNAs were similar were chosen for the circadian analysis . The plants were grown on soil or Linsmaier and Skoog ( LS ) media ( Caisson ) in plant incubator ( Percival Scientific; Perry , Iowa ) set at 22°C under full-spectrum white fluorescent light ( 70–90 µmol m−2s−1: F017/950/24’ , Octron Osram Sylvania ) in long days ( 16 hr light/8 hr dark ) . Bioluminescence Imaging and analysis were performed as previously described with minor modifications ( Fenske et al . , 2015 ) . Seedlings were grown on LS media in the plant incubator ( Percival Scientific ) in 12 hr light/12 hr dark photoperiod for 10 days before being transferred to continuous light ( 40–50 µmol m−2s−1 ) conditions . 9-day-old seedlings were pretreated with 5 mM D-luciferin ( Biosynth ) in 0 . 01% Triton X-100 solution , and incubated one day before imaging . The bioluminescence generated from the CCA1:LUC reporter was imaged for 15 min at every 2 hr using NightOwl system ( Berthold; Germany ) and analyzed using IndiGO software ( Berthold ) . Period length estimation was performed using fast Fourier Transform-Nonlinear Least Squares ( FFT-NLLS ) analysis in the Biological Rhythms Analysis Software System ( BRASS ) ( http://millar . bio . ed . ac . uk/PEBrown/BRASS/BrassPage . htm ) . For gene expression analyses , 14-day-old seedlings grown on LS agar plates were harvested at ZT16 and used for RNA extraction using illustra RNAspin Mini kit ( GE Healthcare; Chicago , Illinois ) . 2 µg of RNA was reverse-transcribed using iScript cDNA synthesis kit ( Bio-Rad; Hercules , California ) . The cDNA was diluted by adding 40 µL of water , and 2 µL of cDNA was used for quantitative polymerase chain reaction ( q-PCR ) using MyiQ real-time thermal cycler ( Bio-Rad ) . IPP2 expression was used as an internal control to normalize cDNA amount . Primers and PCR conditions used for IPP2 , PRR5 , TOC1 and ZTL amplification were previously described ( Baudry et al . , 2010 ) . Expression of ZTL , PRR5 and TOC1 was calculated from three biological replicates . To analyze the expression levels of HA fused ZTL protein , endogenous PRR5 , and TOC1 proteins , seedlings were grown in 12 hr light/12 hr dark conditions for 14 days . Total protein was extracted using extraction buffer [50 mM Na-P pH7 . 5 , 150 mM KCl , 1 mM DTT , 1 mM EDTA , 0 . 05% Sodium deoxycholate , 0 . 1% SDS , 50 µM MG-132 , Protease inhibitor cocktail ( Pierce ) } . Protein was separated in 9% SDS-PAGE gels and transferred to Nitrocellulose membrane ( Bio-Rad ) . HA-ZTL and Actin were detected using anti-HA ( 3F10 , Roche ) and anti-actin ( C4 , Millipore ) antibodies , respectively . Western procedure for detecting TOC1 and PRR5 proteins was previously described ( Baudry et al . , 2010 ) . For protein quantification , western blot images were analyzed using Image J ( Schneider et al . , 2012 ) . For Figure 4—figure supplement 1A , Agrobacteria containing both GI and ZTL variants were coinfiltrated into 4-week-old N . benthamiana leaves . The infiltrated plants were incubated under LD for two days and transferred to continuous light or dark with additional 24 hr incubation . Co-IP was performed according to Fujiwara et al . ( 2008 ) . The immuno-complexes were resuspended in SDS sample buffer and heated briefly . GFP antibody ( Invitrogen , A11120 ) was used for immunoprecipitation of GI-GFP protein . ZTL variants were detected by western blotting with HA antibody . For Figure 4—figure supplement 1B , Agrobacterium harboring each overexpression construct was mixed according to combinations indicated and was infiltrated into 3-week-old N . benthamiana leaves and incubated in LD and either light or darkness as described above . Sample preparation , the IP buffer condition , the IP method and immunoblot procedure were described previously ( Song et al . , 2012 ) . | Many living organisms track the 24-hour cycle of day and night via collections of proteins and other molecules that together act like an internal clock . These clocks , also known as circadian clocks , help these organisms to predict regular changes in their environment , like light and temperature , and adjust their activities according to the time of day . Plants use circadian clocks to predict , for example , when dawn will occur and get ready to harness sunlight to fuel their growth . A plant called Arabidopsis thaliana has a light-sensitive protein called ZEITLUPE ( or ZTL for short ) that helps it keep its circadian clock in sync with the cycle of night and day . Previous studies have shown that light activates this protein causing part of it to change shape and then revert back after a period of about an hour and a half . However , it was unclear if this timing was important for ZEITLUPE to allow plants to keep track of time . To help answer this question , Pudasaini et al . set out to identify a specific chemical event behind ZEITLUPE’s changes in shape . A chemical bond forms when light activates ZEITLUPE , and it turns out that how long this bond lasts before it breaks plays an important role in allowing plants to maintain a 24-hour circadian clock . This chemical bond controls the shape changes that guide the protein’s activities and , when Pudasaini et al . modified ZEITLUPE so that it took much longer for this bond to break , they could tune how fast the plant’s internal clocks run . In essence , the time between the bond forming and breaking breaks acts like a countdown on a stopwatch , and it must be precisely timed to keep the clock in pace with the environment . These findings improve our understanding of how light can regulate an internal biological clock . This improved understanding could , in the future , allow researchers to manipulate how plants and other organisms respond to their environment . This in turn could change how these organisms develop , and how much they grow . As such , extending these findings into agricultural crops may one day lead to new ways to increase crop yields . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | Kinetics of the LOV domain of ZEITLUPE determine its circadian function in Arabidopsis |
Successfully completing the S phase of each cell cycle ensures genome integrity . Impediment of DNA replication can lead to DNA damage and genomic disorders . In this study , we show a novel function for NDE1 , whose mutations cause brain developmental disorders , in safeguarding the genome through S phase during early steps of neural progenitor fate restrictive differentiation . Nde1 mutant neural progenitors showed catastrophic DNA double strand breaks concurrent with the DNA replication . This evoked DNA damage responses , led to the activation of p53-dependent apoptosis , and resulted in the reduction of neurons in cortical layer II/III . We discovered a nuclear pool of Nde1 , identified the interaction of Nde1 with cohesin and its associated chromatin remodeler , and showed that stalled DNA replication in Nde1 mutants specifically occurred in mid-late S phase at heterochromatin domains . These findings suggest that NDE1-mediated heterochromatin replication is indispensible for neuronal differentiation , and that the loss of NDE1 function may lead to genomic neurological disorders .
The developmental formation of complex multicellular organs requires the impeccable integration of cell division with differentiation . The precise control of the DNA synthesis ( S ) phase of each cell division cycle warrants the faithful replication of the entire genome and at the same time establishes the epigenetic state that defines the differentiation identity of individual cells ( McNairn and Gilbert , 2003; Nordman and Orr-Weaver , 2012 ) . Concurrent with genome duplication , the protein components of chromatin are disassembled and re-assembled throughout the S phase into higher order structures that characterize the specific gene expression status of daughter cells ( Alabert and Groth , 2012; Budhavarapu et al . , 2013 ) . The task of assuring error-free S phases is especially challenging in the generation of organs with extraordinarily high cell number and diversity , such as the cerebral cortex , during which billions of functionally specialized neurons are produced following a tightly controlled developmental program of cell cycle progression and step-wise neural progenitor fate restriction ( Rakic , 1995; Desai and McConnell , 2000; Rakic , 2009; Florio and Huttner , 2014 ) . S phase-mediated genome regulation is recently shown to be essential not only for rapid expansion of neural progenitor pool but also for neural differentiation , as the amount of time neural progenitors spend in S phase is highly correlated to the cell fate of cerebral cortical neural progenitor cells ( Arai et al . , 2011 ) . Longer S phase appears more necessary for self renewing than neuron-producing cell cycles , suggesting an S phase specific quality control in maintaining the identity of neural progenitors before their terminal neurogenic division . The importance of genome maintenance in corticogenesis is also underscored by large amounts of clinical and experimental observation , which have shown that the functional impairment of genes important for DNA metabolism frequently leads to brain developmental pathology ( McKinnon , 2009; Ciccia and Elledge , 2010; Zeman and Cimprich , 2014 ) . However , as the genes involved are also essential for genome surveillance outside of the brain , pathogenic lesions of the brain genome have been believed to associate with the lack of effective DNA damage repair , improper checkpoint signaling , rapid cell proliferation , or increased metabolic , chemical , and physical stresses . It is unclear whether a genome quality control associated specifically with neuronal differentiation is required to ensure the correct genetic and epigenetic identity of both neural progenitors and daughter neurons . NDE1 is a multifunctional molecular scaffold fundamental for CNS development . It was originally identified as the central nervous system ( CNS ) specific partner of LIS1 ( known as PAFAH1B1 ) ( Feng et al . , 2000 ) , whose haploinsufficiency results in lissencephaly ( smooth brain ) ( Reiner et al . , 1993 ) . Homozygous mutations of NDE1 were found recently to cause microlissencephaly ( small and smooth brain ) with up to 90% reduction in brain mass , while the affected individuals showed normal development of non-CNS organs ( Alkuraya et al . , 2011; Bakircioglu et al . , 2011; Guven et al . , 2012 ) . Moreover , copy number variants ( CNVs ) in the NDE1 locus are increasingly shown to associate with a wide spectrum of neuropsychiatric disorders with complex genetic traits ( Ullmann et al . , 2007; Hannes et al . , 2009; Heinzen et al . , 2010; Mefford et al . , 2010; Nagamani et al . , 2011; Tropeano et al . , 2013 ) . Genetic epistasis studies in mice demonstrated that Nde1 and Lis1 function synergistically in a dose-dependent manner in governing the generation of late-born cortical neurons that comprise the upper cortical layers II and III . Layer II/III neurons were found specifically reduced in both Nde1−/− and Nde1+/− Lis1+/− mice , and they were abolished almost completely in Nde1−/−Lis1+/− mice along with severe brain hypoplasia but insignificant change in the body size ( Feng and Walsh , 2004; Pawlisz et al . , 2008 ) . Neurons in cortical upper layers are evolutionarily novel and undergo great expansion in mammalian evolution ( DeFelipe et al . , 2002; Molnar et al . , 2006 ) . They are highly diverse projection neurons and essential for cognitive functions including perception , emotion , attention , and memory through making functional connections among various cortical areas and between the two cerebral hemispheres ( DeFelipe et al . , 2002; Fame et al . , 2011; Greig et al . , 2013 ) . The essential requirement of Nde1 in the generation of cortical layer II/III neurons underscores the gene dosage dependency of NDE1 in brain cognition . However , current molecular information on NDE1 ( Nde1 ) is limited by its previously identified association with the cytoskeleton , which does not fully explain its CNS specific phenotype and function . In the present study of murine models of Nde1 mutations , we report that Nde1 is indispensible for the successful completion of S phase , specifically during the early neuronal fate restrictive differentiation of multipotent neural progenitors . The most profound phenotype that resulted from Nde1 mutations was severe DNA damage that occurred during mid to late S phase heterochromatic DNA replication . Stalled DNA replication led to DNA replicative catastrophe and the activation of tumor suppressor p53 ( encoded by Trp53 or Tp53 ) via DNA damage responses ( DDRs ) . Abrogating p53-suppressed apoptosis rescued the size and structure of Nde1 mutant brain but failed to mitigate the genomic stress . We also identified a nuclear pool of Nde1 and the interaction of Nde1 with the cohesin complex as well as the chromatin remodeler SNF2h . These findings suggest that Nde1 is essential for CNS specific genomic quality control , that chromatin remodeling during heterochromatic replication facilitated by Nde1 is essential for generating cortical layer II/III neurons , and that reduced fidelity of S phase choreography during the early phase of neural progenitor differentiation and identity establishment may lead to mosaic genomic lesions and developmental brain disorders .
The reduction of cortical layer II/III neurons in both Nde1−/− and Nde1−/−Lis1+/− mutants was previously found to result from a failure in cell division and precocious neurogenesis of the mutant neural progenitors ( Feng and Walsh , 2004; Pawlisz et al . , 2008 ) . However , massive apoptosis detected by TUNEL was also observed predominantly in the newly formed cortical plate ( CP ) of Nde1−/−Lis1+/− embryos at E12 . 5 ( Pawlisz et al . , 2008 ) . To understand the mechanism of apoptosis and evaluate its contribution to the loss of layer II/III neurons in Nde1 mutant brains , we examined its associated cellular processes and found that the apoptosis in both Nde1−/−Lis1+/− and Nde1−/− mutants corresponded with increased DNA damage . A substantial number of cells in the neocortex of Nde1−/−Lis1+/− and Nde1−/− mutants showed abnormally high γH2AX immunosignals , a hallmark for DNA double strand breaks ( DSBs ) ( Thiriet and Hayes , 2005; Figure 1A ) . While γH2AX foci associated with DNA replication were also widely detected in wild-type and mutant cortices , the γH2AX+ signals observed in the mutant were several orders of magnitude higher , showed pan-nuclear pattern , and were often co-stained by antibodies to cleaved caspase 3 ( CC3 ) , the marker for apoptosis . While apoptosis was wide-spread and observed in both CP neurons and ventricular zone ( VZ ) neural progenitors ( Pawlisz et al . , 2008 ) , γH2AX+ cells were confined in the VZ ( Figure 1A ) . This suggested that DNA damage occurred prior to apoptosis and that apoptosis was one of the endpoints of DNA damage caused by Nde1 deficiency . 10 . 7554/eLife . 03297 . 003Figure 1 . The correlation of DNA damage and apoptosis with neural progenitor early fate restriction in Nde1 mutant brains . ( A ) Immunohistological analysis of γH2AX ( red ) and cleaved caspase 3 ( CC3 , green ) reveals the co-existence of DNA damage and apoptosis in the neocortex of Nde1 mutants at E12 . 5 . Higher-magnification views indicate the high level of γH2AX pan-nuclear signals associated with severe DNA damage and low γH2AX signals associated with normal replication foci . ( B ) Developmental analysis of the temporal correlation between immunosignals of γH2AX ( red ) and cleaved caspase 3 ( CC3 , green ) from E10 . 5 to P0 . ( C ) The lack of DNA damage and apoptosis in the cortical hem ( a region where neural progenitors divide but do not undergo neuronal differentiation ) of Nde1−/− and Nde1−/−Lis1+/− brains at E12 . 5 . ( D ) Immunohistological analysis of serial coronal sections of Nde1−/− brains to demonstrate the spatial correlation of γH2AX ( red ) with TNG revealed by DCX abundance ( green ) . ( E ) Immunohistological analysis of γH2AX ( red ) and cleaved caspase 3 ( CC3 , green ) on sagittal sections of an Nde1−/−Lis1+/− embryo and its wild-type littermate at E12 . 5 . Nuclei DNA was stained with Hoechst 33342 and shown in blue in all fluorescent images . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 00310 . 7554/eLife . 03297 . 004Figure 1—figure supplement 1 . The correlation of DNA damage and apoptosis with neural progenitor early fate restriction in Nde1 mutants . Immunohistological analysis with antibodies to γH2AX ( red ) and cleaved caspase 3 ( CC3 , green ) shows the spatiotemporal concurrence of DNA damage and apoptosis with early stages of neuronal differentiation in both Nde1−/− and Nde1−/−Lis1+/− brains . Nuclei DNA was stained with Hoechst 33342 and shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 00410 . 7554/eLife . 03297 . 005Figure 1—figure supplement 2 . DNA damage and apoptosis along the transverse neurogenetic gradient ( TNG ) in Nde1−/− brains . Immunohistological analysis of serial coronal sections of Nde1−/− brains at early and late E12 , respectively , to demonstrate the spatial correlation of cleaved caspase 3 ( CC3 , green ) with TNG revealed by Tuj1 ( red ) . Nuclei DNA was stained with Hoechst 33342 and shown in blue in all fluorescent images . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 00510 . 7554/eLife . 03297 . 006Figure 1—figure supplement 3 . The correlation of DNA damage and apoptosis with neuronal differentiation in the Nde1 mutant spinal cord from E9 . 5 to E13 . 5 . Immunohistological analysis of γH2AX ( red ) and cleaved caspase 3 ( CC3 , green ) of the developing spinal cord of Nde1−/−Lis1+/− ( M ) and control ( C , Nde1+/− or wild type ) littermates from E9 . 5 to E13 . 5 . Nuclei DNA was stained with Hoechst 33342 and shown in blue in all fluorescent images . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 006 The DNA damage and apoptosis in Nde1 mutant brain were found to correlate spatiotemporally with the early fate restrictive differentiation of neural progenitors . In the Nde1−/−Lis1+/− brain , γH2AX+ and CC3+ cells were detected as early as the onset of neocortical neurogenesis at E10 . 5 ( Figure 1B ) . At this stage , a majority of the neural progenitors progress through the cell cycle rapidly to expand the progenitor pool , while some start to differentiate into intermediate neural progenitors or neurons . From E10 . 5 to E12 . 5 , as more neural progenitors become fate restricted , both DNA damage and apoptosis in the Nde1 mutant cortex increased and peaked at E12 . 5 . With the progression of cortical neurogenesis , γH2AX+ and CC3+ cells declined gradually after E13 . 5 when more cortical neurons were produced . Apoptosis in both Nde1−/−Lis1+/− and Nde1−/− mutants became lower during the later stages of cortical neurogenesis after E15 , even though a majority of cortical layer II/III neurons are generated through terminal mitosis between E15 and E17 . In contrast to the low level of apoptosis , γH2AX+ cells remained detectable until birth ( Figure 1B , Figure 1—figure supplement 1 ) . While γH2AX and CC3 signals were at peak levels in the neocortex of Nde1−/−Lis1+/− and Nde1−/− brains at E12 . 5 , they were absent in the mutant cortical hem , the highly proliferating but non-neurogenic hippocampus organizer ( Figure 1C ) . This suggested that DNA damage and apoptosis caused by Nde1 mutation were associated with neural progenitor differentiation as opposed to proliferation . During neocortical development , neurogenesis is known to proceed along the transverse neurogenic gradient ( TNG ) , which initiates rostrolaterally and propagates caudomedially ( Caviness et al . , 2009 ) . To better reveal the correlation of DNA damage and apoptosis with neocortical neurogenesis , we examined serially sectioned Nde1−/− brains ( n = 6 ) by double immunostaining of γH2AX and CC3 with newborn neuron markers DCX or Tuj1 . Compared to Nde1−/−Lis1+/− brains , the well-preserved cytoarchitecture of Nde1−/− brains allows for a better assessment of the spatial distribution of DNA damage and apoptosis with respect to the TNG . At earlier E12 , more γH2AX+ cells were observed in the rostrolateral than the caudomedial neocortex . Similarly , CC3+ cells were predominantly detected rostrolaterally in these brains . Towards later E12 , as TNG progresses caudomedially , CC3+ cells also increased in the caudolateral and rostromedial cortical regions . In all cortical sections examined , γH2AX+ and CC3 signals were correlated spatially with the presence and abundance of newborn neurons ( Figure 1D , Figure 1—figure supplement 2 ) . Together , these data demonstrated that DNA damage and apoptosis in Nde1 mutant cortices had little correlation with self-renewal proliferation or neurogenic terminal divisions but was well in line with early neural progenitor fate restrictive differentiation around E12 . 5 in the developing neocortex . The link of DNA damage and apoptosis with early progenitor differentiation was further shown in the developing spinal cord , where neurogenic domains are better spatiotemporally defined ( Figure 1—figure supplement 3 ) . We also examined DNA damage and apoptosis on sagittal sections of E12 . 5 Nde1−/−Lis1+/− embryos and found only a small number of cells outside of the developing CNS were affected ( Figure 1E ) . These developmental analyses demonstrate that the most profound phenotype caused by Nde1 mutations is characterized by DNA damage and apoptosis , and that this phenotype is spatiotemporally in parallel to the early differentiation period of neural progenitors . Thus , the indispensible role of Nde1 in generating layer II/III neurons appears to be at the time when the fate of these neurons is initially established rather than when they are generated through terminal mitosis . To confirm that the apoptosis in Nde1 mutants was the result of DNA damage but not vice versa , we detected DNA damage directly by performing the comet assay on primary cortical cells ( Collins , 2004; Olive and Banath , 2006 ) . Compared to cells from Nde1+/− embryos , more Nde1−/−Lis1+/− and Nde1−/− cells showed longer comet tails , indicating that loss of Nde1 function indeed resulted in an increase in DNA lesions in the developing cortex ( Figure 2A ) . 10 . 7554/eLife . 03297 . 007Figure 2 . Co-activation of γH2AX with p53 in Nde1 mutant neocortices . ( A ) Results from comet assay with cortical cells isolated from Nde1−/− or Nde1−/−Lis1+/− mutants and their Nde1+/− littermates at E12 . 5 . The distribution , mean , and median values of comet tail length ( T ) from over 300 randomly selected and photographed cells are presented . Nde1−/− and Nde1−/−Lis1+/− cells showed increased comet tail length compared to Nde1+/− cells , respectively; p < 0 . 0001 by the Wilcoxon rank-sum two sample test . ( B ) Immunohistological analysis of the γH2AX ( red ) and phospho-p53 Ser18 ( p-p53 , green ) in the neocortex of wild-type and Nde1 mutants at E12 . 5 . ( C ) Immunofluorescence analysis of the co-activation of γH2AX ( green ) and phospho-p53 Ser18 ( red ) in primary cortical cells isolated at E12 . 5 . Note the lack of DNA condensation and fragmentation of cells with high phospho-H2AX and p53 signals . ( D ) Immunoblotting analyses of phospho-p53 Ser18 in embryonic cortical lysates at E12 . 5 . Nuclei DNA was stained with Hoechst 33342 and shown in blue in all fluorescent images . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 007 DNA damage is known to initiate DNA damage responses ( DDRs ) primarily through the activation of ATM/ATR kinases , which in turn directly phosphorylate p53 on Ser18 ( Ser15 in human ) ( Banin et al . , 1998; Canman et al . , 1998; Khanna et al . , 1998 ) in addition to H2AX . Double immunostaining of γH2AX and phospho p53 at Ser18 showed their co-activation in mutant neocortical neural progenitors ( Figure 2B , C ) . Many mutant cells showed elevated γH2AX and p-p53S18 but lacked DNA condensation and fragmentation , supporting that γH2AX and p53 activation in Nde1 mutants occurred prior to the programmed cell death ( Figure 2C ) . We found that the level of p53 S18 hyper-phosphorylation correlated to Nde1 gene dosage as well as to the degree of brain malformations caused by Nde1 mutations ( Figure 2D ) . Similarly , the loss of Nde1 resulted in an elevated basal level of p53 , which is known to be low in healthy cells , supporting the DDR-mediated p53 activation . These results indicate that Nde1 mutation primarily results in DNA damage , and apoptosis is caused by DDR with p53 activation . Thus , in contrast to the prevailing view of Nde1 being a cytoskeletal regulator , these data revealed a novel function of Nde1 in protecting the genome . To further establish Nde1's essential role in safeguarding the genome during cortical neural progenitor differentiation , we abrogated p53 in Nde1-deficient mice . As expected , the loss of p53 abolished apoptosis in Nde1−/− progenitors and restored the size of the Nde1 mutant brain ( Figure 3A , B , Figure 3—figure supplement 1A ) . The brain of Nde1−/−Trp53−/− double mutant mice was found to be nearly the same as those of wild-type in both size and structure at the weaning age . Histological and immunohistological analyses showed all anatomical features , especially the thickness of layer II/III neurons was fully restored ( Figure 3C , D ) . Quantitative analysis of the fraction of Cux1 immunolabeled upper layer projection neurons ( with respect to total NeuN+ or NeuN+Cux1-neurons ) in the Nde1−/−Trp53−/− double mutant brain indicated that they were at the wild-type-level ( Figure 3E , Figure 3—figure supplement 1B , C ) . These results demonstrate that microcephaly and reduced layer II/III neurons caused by the loss of Nde1 is a result of p53-dependent apoptosis elicited by DDR and that NDE1 is specifically essential for the genomic integrity of neurons that conduct higher order brain functions . 10 . 7554/eLife . 03297 . 008Figure 3 . Restoration of the size and structure of the Nde1−/− brain by abrogating p53 . ( A ) Immunohistological analyses of cleaved caspase 3 ( CC3 , green ) in wild-type , Nde1−/− , Nde1+/−Trp53−/− , and Nde1−/−Trp53−/− neocortices at E12 . 5 . ( B ) Brain weight of Nde1−/− , Nde1−/−Trp53−/− mutant mice and their littermates at post-natal for 3 to 4 weeks . Data are mean ± SD . Significant overall differences were found among wild-type and Nde1–Trp53 double mutants by ANOVA ( p < 0 . 0001 ) . Pairwise comparisons showed that the brain mass of Nde1−/−Trp53−/− mice ( n = 8 ) was significantly increased compared to that of Nde1−/− mice ( *p < 0 . 0001 ) , but not significantly different from that of the wild-type mice ( n . s . , p = 0 . 06 ) . ( C ) H&E stained brain sections of wild-type , Nde1−/− , Nde1+/−Trp53−/− , and Nde1−/−Trp53−/− mice reveal normal size and anatomical structure of the Nde1−/−Trp53−/− brain . ( D ) H&E stained cortical sections of wild-type , Nde1−/− , and Nde1−/−Trp53−/− brains , showing restored layer II/III cortical neurons in the Nde1−/−Trp53−/− brains . ( E ) Immunohistological and quantitative analyses of the number and distribution of Cux1+ ( red ) superficial layer cortical neurons . Cortical neurons were identified by NeuN immunoreactivities ( green ) ; nuclei DNA was stained with Hoechst 33342 and shown in blue . Data are presented as mean ± SD in percentage of total NeuN+ neurons ( n = 5 ) . Significant overall differences were found among wild-type , Nde1−/− , and Nde1−/−Trp53−/− brains by ANOVA ( p < 0 . 0001 ) . Pairwise comparisons indicated that compared to the wild-type , Cux1+ neurons were significantly decreased in the Nde1−/− brains ( *p < 0 . 0001 ) , but not significantly changed in Nde1−/−Trp53−/− brains ( n . s . , p = 0 . 94 ) . Compared to Nde1−/− , Cux1+ neurons were significantly increased in Nde1−/−Trp53−/− brains ( p < 0 . 0001 ) . ( F ) Immunohistological and quantitative analyses of the number and distribution of Foxp2+ ( red ) deep layer cortical neurons . Significant overall differences were found among wild-type , Nde1−/− , and Nde1−/−Trp53−/− brains by ANOVA ( p < 0 . 0001 ) . Pairwise comparisons indicated that compared to the wild-type , Foxp2+ neurons were significantly increased in the Nde1−/− brains ( *p<0 . 005 ) , but not significantly changed in Nde1−/−Trp53−/− brains ( n . s . , p = 0 . 77 ) . Compared to Nde1−/− , Foxp2+ neurons were significantly decreased in Nde1−/−Trp53−/− brains ( p = 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 00810 . 7554/eLife . 03297 . 009Figure 3—figure supplement 1 . Restoration of brain size and structure of Nde1−/− mutants by abrogating p53 . ( A ) Representative brain images of wild-type , Nde1−/− , and Nde1−/−Trp53−/− mice at post-natal day 23 . ( B ) Immunostaining with parvalbumin antibody ( PV ) showing the normal number and distribution of GABAergic interneurons in the Nde1−/−Trp53−/− cortex . Nuclei DNA was stained with Hoechst 33342 and shown in blue . ( C ) Quantitative analyses of Cux+ superficial and Foxp2+ deep layer cortical neurons . Data are presented as mean ± SD ( n = 5 ) . Significant overall differences were found among wild-type , Nde1−/− , and Nde1−/−Trp53−/− brains by ANOVA ( p < 0 . 0001 for Cux1+; p = 0 . 0009 for Foxp2+ neurons ) . Pairwise comparisons indicated that compared to the wild-type , Cux1+ neurons were significantly decreased in the Nde1−/− brains ( *p < 0 . 0001 ) but not significantly changed in Nde1−/−Trp53−/− brains ( n . s . , p = 0 . 95 ) . Compared to Nde1−/− , Cux1+ neurons were significantly increased in Nde1−/−Trp53−/− brains ( p < 0 . 0001 ) . Compared to the wild-type , Foxp2+ neurons were significantly increased in the Nde1−/− brains ( *p = 0 . 0045 ) , but not significantly changed in Nde1−/−Trp53−/− brains ( n . s . , p = 0 . 72 ) . Compared to Nde1−/− , Foxp2+ neurons were significantly decreased in Nde1−/−Trp53−/− brains ( p = 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 009 Consistent with an essential role of Nde1 in early neuronal fate restriction , we found that active DSBs and DDR in the Nde1−/− brain occurred in the primary multipotent neural progenitors in the VZ but was rarely detected in the intermediate neural progenitors of the sub-ventricular zone ( SVZ ) , and not in the post-mitotic neurons in the CP ( Figure 4A , Figure 4—figure supplement 1A , B ) . To identify the source of DSBs caused by Nde1 mutations , we examined their occurrence with respect to the cell cycle . After BrdU transient labeling , we found about 20% of γH2AX+ progenitors were also BrdU+ ( Figure 4B ) . In contrast , co-immunostaining of γH2AX and phospho-histone H3 ( PH3 ) failed to show cells with elevated γH2AX during mitosis ( Figure 4—figure supplement 1C ) . Because of the transient nature of both γH2AX and the BrdU pulse , these results demonstrate that DNA damage caused by Nde1 deficiency can occur during DNA replication . 10 . 7554/eLife . 03297 . 010Figure 4 . DNA damage caused by Nde1 mutation occurs concurrently with DNA replication . ( A ) Co-immunostaining of γH2AX ( red ) and multipotent/primary progenitor marker Pax6 ( green ) on cortical sections at E12 . 5 . ( B ) Co-immunostaining of γH2AX ( red ) and BrdU ( green ) ; cortical sections were prepared from BrdU pulse ( 30 min ) labeled embryos at E12 . 5 . Quantification of γH2AX+ cells showed that 18 . 3 ± 6 . 1% ( mean ± SD ) of γH2AX+ Nde1−/− and 23 . 4 ± 3 . 1% ( mean ± SD ) of γH2AX+ Nde1−/−Lis1+/− cells were also BrdU+ . ( C ) S phase duration ( Ts , hr ) measurement indicates significant delay of DNA replication in Nde1−/− neural progenitors . Data are mean ± SD , p < 0 . 0001 by Student's t test . ( D ) Representative images of B44 ( green , recognizes both IdU and BrdU ) and BU1/75 ( red , recognizes only BrdU ) double immunohistological staining in an IdU ( 2 hr ) , BrdU ( 30 min ) sequential labeling experiment . Cells that have finished S phase and progressed to G2/M are indicated by pink arrows; cells that remained in the S phase zone but stopped incorporating BrdU are indicated by white arrows . A diagram to indicate the cell cycle dependent nuclei position through INM is included . ( E ) B44 ( green ) , BU1/75 ( blue ) , and γH2AX ( red ) triple immunostaining of cortical sections of Nde1−/−Lis1+/− mutant after IdU ( 2 . 5 hr ) , BrdU ( 30 min ) sequential labeling . 47 . 5 ± 0 . 1% ( mean ± SD ) of total γH2AX+ cells were B44+BU1/75− , indicating the association of DNA damage with stalled DNA replication ( white arrows ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 01010 . 7554/eLife . 03297 . 011Figure 4—figure supplement 1 . The cell type and cell cycle specificity of DNA damage in Nde1 mutant brains . ( A ) Double immunohistological staining E12 . 5 cortical sections with γH2AX ( green ) and the new neuron marker DCX ( red ) . ( B ) Double immunohistological staining E13 . 5 cortical sections with γH2AX ( red ) and the intermediate progenitor marker Tbr2 ( green ) . ( C ) Double immunohistological staining E12 . 5 cortical sections with γH2AX ( red ) and the G2/M marker PH3 ( green ) . Nuclei DNA was stained with Hoechst 33342 and shown in blue in all fluorescent images . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 01110 . 7554/eLife . 03297 . 012Figure 4—figure supplement 2 . Stalled or delayed DNA replication in Nde1 mutant neural progenitors . ( A ) Triple immunofluorescence histological staining of neocortical sections with B44 ( green ) , BU1/75 ( red ) , and PH3 ( blue ) , showing the lack of PH3 immunoreactivity in the arrow-indicated B44+BU1− cells of Nde1−/− and Nde1−/−Lis1+/− cortices . ( B ) Representative images of B44 ( green ) , BU1/75 ( red ) , and PH3 ( PH3 ) triple immunohistological staining after IdU ( 2 . 5 hr ) , BrdU ( 30 min ) sequential labeling . Note the B44+BU1/75−PH3− cells ( circles ) in the S phase region of Nde1−/− and Nde1−/−Lis1+/− cortices . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 012 The concurrence of DNA damage and DNA replication in Nde1 mutant progenitors suggested an essential function of Nde1 in S phase . To confirm this , we measured the duration of S phase ( Ts ) of primary progenitors by sequential IdU-BrdU labeling and analysis ( Martynoga et al . , 2005 ) . In this experiment , IdU was used to label progenitors in S phase , and BrdU was added 1 . 5 hr later to assess if IdU+ cells had completed DNA replication . Our data showed that the Ts of Nde1−/− progenitors was significantly longer than that of the wild-type cells ( Figure 4C ) . Furthermore , we also noticed that in both Nde1−/−Lis1+/− and Nde1−/− mutants , many nuclei that stopped incorporating BrdU ( IdU+BrdU− ) remained in the S phase zone of the basal VZ instead of descending apically as seen in the wild-type progenitors ( Figure 4D , white arrows ) . Neural progenitors in the cortical VZ are polarized apical-basally and known to undergo interkinetic nuclear migration ( INM ) by moving their nuclei basally during S phase . After S phase is completed , the nuclei are moved apically for mitosis to occur at the ventricular surface ( Takahashi et al . , 1993 ) . As the basally localized IdU+BrdU− nuclei in the mutant were not recognized in G2/M by PH3 ( Figure 4—figure supplement 2A ) , it suggested that these mutant cells might arrest in the basal S phase zone due to the stalled DNA replication . To confirm the faulty S phase prior to G2/M , we increased the time between IdU and BrdU pulses to 2 . 5 hr to observe if the IdU+BrdU− cells in the basal S phase zone of the mutant may progress to mitosis over time . However , we observed that rather than progressing to mitosis , these cells remained stalled and that about half ( 47 . 5% ) of the γH2AX+ cells were positively stained by antibodies to IdU but not to BrdU , indicating that S phase-stalled cells underwent DDR by activating γH2AX ( Figure 4E , Figure 4—figure supplement 2B ) . This explains why only 20% of γH2AX+ cells were actively incorporating BrdU ( Figure 4B ) , since a majority of γH2AX+ cells were stalled in the S phase . Therefore , our data demonstrate that Nde1 is functionally essential during S phase and that DNA damage in Nde1 mutant progenitors results from stalled and catastrophic DNA replication . During the S phase of metazoan cells , various chromosomal domains are known to replicate in a well-defined spatiotemporal sequence and can be visualized experimentally by labeling and immunostaining with nucleotide analogs ( Dimitrova and Gilbert , 1999; Dimitrova et al . , 2002; Maison et al . , 2010 ) . To better understand the function of Nde1 in S phase , we studied the dynamic progression of S phase and analyzed IdU and CldU sequentially labeled chromosomal domains by taking advantage of the non-overlapping immunosignals between IdU and CldU . Embryos were pulse labeled by IdU for 1 . 5 hr before being labeled again by CldU transiently . Cells in early S phase were identified by CldU but not IdU incorporation ( IdU−CldU+ ) as well as a more evenly distributed CldU immunosignals , as new replication foci emerge continuously throughout the transcriptionally active euchromatin . Whereas cells in mid to late S phase were labeled by both IdU and CldU ( IdU+CldU+ ) and recognized by enhanced nucleotide incorporation in various transcriptionally silent heterochromatic domains . Cells that had left S phase could be noted by IdU+ only ( IdU+CldU− ) immunosignals , since they stopped nucleotide incorporation before CldU pulse labeling . By quantitative comparison of IdU-CldU labeled progenitors in the cortical VZ of wild-type and Nde1−/− mutant , we failed to detect a significant difference in the early S phase population but found more mutant progenitors in mid to late S phase ( Figure 5A ) . In a large number of Nde1−/− progenitors , CldU signals in IdU+CldU+ cells were predominantly detected along the nuclear periphery , the rim of nucleoli , and on particles of satellite repeats which are characteristic for heterochromatic domains , indicating replication difficulties in these domains during mid to late S phase ( Figure 5B–D ) . Triple immunostaining of IdU , CldU , and PH3 allowed us to access the S to G2/M progression; we found that Nde1−/− progenitors showed a reduction in the intensity and the number of IdU labeled foci in PH3+ cells ( Figure 5B , arrows ) , suggesting reduced nucleotide incorporation into mutant progenitors before S phase completion . Collectively , the unaltered early S phase population , significant increase in CldU labeling in various heterochromatic domains , reduced IdU labeling in PH3+ cells , combined with prolonged duration of S phase , suggest that DNA replication in Nde1−/− progenitors progresses normally through the euchromatin , but stalls during heterochromatin replication in mid-late S phase . 10 . 7554/eLife . 03297 . 013Figure 5 . Stalled DNA replication during mid-late S phase at heterochromatic domains in Nde1−/− mutant neural progenitors . ( A ) Quantitative analysis of IdU+CldU− ( red ) , IdU+CldU+ ( yellow ) , and IdU−CldU+ ( green ) cell fractions ( % ) by IdU ( 2 hr ) and CldU ( 30 min ) sequential labeling . Data are presented as mean ± SD . n . s . : p > 0 . 05; **: p < 0 . 001 by Chi–Square tests . A diagram that shows spatiotemporal patterns of early , mid , and late S phase DNA replication is included . ( B ) Representative images of B44 ( red ) , BU1/75 ( green ) , and PH3 ( blue ) triple immunostaining from IdU ( 2 hr ) and CldU ( 30 min ) sequential labeling experiments . Note the CldU ( BU1/75 , green ) signals in IdU+CldU+ progenitors highlight predominantly heterochromatic structures ( circled cells , better revealed in C ) and the low IdU ( B44 , red ) signals in PH3+ cells of the Nde1−/− mutant ( arrows ) . ( C ) Higher-magnification views of selected Nde1−/− mutants progenitors ( circled in B ) that show DNA replication at heterochromatin ( nuclear periphery , the rim of the nucleoli , and large foci of repeated heterochromatic sequences ) . ( D ) Double immunohistological stain of CldU ( green ) and nucleolar marker Nucleophosmin 1 ( Npm1 , red ) to view stalled DNA replication at perinuclear heterochromatin , which is known to comprise centromeres and pericentromeres . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 013 The high level of p53-dependent apoptosis in Nde1−/− brain suggested that the DNA replication catastrophe activated DDR to eliminate neural progenitors or neurons with genomic lesions . In this case , the restored size and structure of Nde1−/− brain by p53 abrogation is not expected to show genome rescue even though DDR-induced apoptosis was suppressed . This would also suggest that genomic lesions incurred during development could be structurally tolerated and not necessarily manifest in changed brain anatomy . To test this hypothesis , we measured the Ts of Nde1−/−Trp53−/− progenitors and found the lengthened Ts in Nde1−/− progenitors was not mitigated by p53 removal ( Figure 6A ) . Likewise , the IdU-CldU sequential labeling study also indicated delayed DNA replication at heterochromatic domains in Nde1−/−Trp53−/− neural progenitors ( Figure 6B ) . Reduced BrdU or IdU incorporation was found when Nde1−/−Trp53−/− progenitors progressed from S to G2/M in 1 . 5 hr , which was in agreement with hindered nucleotide incorporation before S phase completion ( Figure 6B , C , arrows ) . These observations suggest that replication stress and genomic impairments remained with the absence of apoptosis . 10 . 7554/eLife . 03297 . 014Figure 6 . Persistent cell cycle stress and genotoxicity in Nde1 mutants after p53 abrogation . ( A ) Representative images of B44 ( green , recognizes both IdU and BrdU ) and BU1/75 ( red , recognizes only BrdU ) double immunostained neocortical sections from IdU ( 2 hr ) , BrdU ( 30 min ) sequential labeling experiments to measure S phase duration ( Ts ) . Cells that were stalled in the S phase zone ( IdU+ BrdU−; green ) are indicated by white arrows . Significant overall difference in Ts was found among wild-type , Nde1−/− , and Nde1−/−Trp53−/− progenitors by ANOVA ( p < 0 . 0001 ) . Pairwise comparisons indicated that Ts of Nde1−/−Trp53−/− progenitors was significantly longer than that of wild-type progenitors ( *p < 0 . 0001 ) but not significantly different from that of the Nde1−/− progenitors ( n . s . , p = 0 . 74 ) . Data are presented as mean ± SD . ( B ) Representative images of B44/IdU ( red ) , BU1/75/CldU ( green ) , and PH3 ( blue ) triple immunostaining of neocortical sections from embryos sequentially labeled by IdU ( 2 hr ) and CldU ( 30 min ) . Note the enhanced CldU immunosignals at heterochromatin structures ( higher magnification views ) and reduced IdU immunosignals in PH3+ cells ( arrows ) in Nde1−/− and Nde1−/−Trp53−/− cortical neural progenitors . ( C ) Immunohistological analysis of BrdU ( red ) –PH3 ( green ) co-labeled cells 1 . 5 hr after BrdU pulse . Arrows indicate PH3+ cells with very few BrdU foci , suggesting hindered BrdU incorporation at the end of S phase . Higher magnification views of a BrdU–PH3 double positive Nde1−/−Trp53−/− cell with uncondensed DNA are also shown . ( D ) Double immunostaining with G2/M marker PH3 ( green ) and M phase marker phospho-vimentin 4A4 ( red ) showing increased PH3+4A4− G2 population in Nde1−/− and Nde1−/−Trp53−/− progenitors ( arrows ) . Higher magnification views of selected Nde1−/−Trp53−/− cells are included to show uncondensed DNA in PH3+4A4− but condensed DNA in PH3+4A4+ cells . ( E ) Immunoblotting analysis of neocortical lysates showing elevated PH3 and increased Cdc25A degradation in the Nde1−/−Trp53−/− mutant brain . β-actin ( Actin ) and dynein intermediate chain ( Dynein IC ) were used as loading controls . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 01410 . 7554/eLife . 03297 . 015Figure 6—figure supplement 1 . Lymphomagenesis and remarkably increased p53 in Nde1−/−Trp53+/− thymic lymphoma . ( A ) Representative T cell thymic lymphoma of Nde1−/−Trp53+/− mice at 3 months . ( B ) Representative image of H&E stained Nde1−/−Trp53+/− lymphoma . Apoptotic cells with fragmented nuclei are indicated by green arrows; mitotic figures are indicated by yellow arrows . ( C ) Immunoblotting analysis of Nde1−/−Trp53+/− lymphoma and normal thymus tissues show remarkable elevation of p53 in the tumor tissue . Note that the level of p53 in control thymus tissue was almost undetectable . ( D ) . Northern blot analysis of Nde1 and Ndel1 mRNA from adult mouse tissues . Note the high Nde1 expression in hematopoietic tissues . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 015 The lack of apoptosis as well as the restored cortical structure and number of layer II/III neurons in the Nde1−/−Trp53−/− brain did not appear to associate with a fully rescued genome . As somatically generated chromosomal aberrations can happen in neocortical progenitors and neurons of the normal brain ( Rehen et al . , 2001; Bushman and Chun , 2013 ) , and Nde1 mutation affects only selected cells in a heterogeneous pool of neural progenitors , such genomic mosaicism and cell heterogeneity of the developing cortex precluded us from obtaining direct information on the genomic lesions in the Nde1−/−Trp53−/− brain . Instead , we looked into further evidence of DNA damage by examining cell cycle stress in Nde1−/−Trp53−/− embryonic cortices . After completing S phase , neural progenitors of the VZ normally move replicated chromosomes through INM towards the apical ventricular surface where chromosome condensation and mitosis occur . In both Nde1−/− and Nde1−/−Trp53−/− cortices , a substantial number of PH3+ progenitors were found mislocalized basally ( Figure 6C , D ) . PH3 is an indicator of cells in both late-G2 and M phases; we found chromosomes of the mislocalized PH3+ mutant progenitors were not condensed ( Figure 6C , D higher magnification views ) , indicating cell cycle arrest in G2 instead of M phase . The G2 arrest in Nde1−/− and Nde1−/−Trp53−/− progenitors was confirmed by co-immunostaining PH3+ cells with the phospho-vimentin antibody 4A4 , which exclusively recognizes progenitors in M phase ( Weissman et al . , 2003 ) . Most PH3+4A4+ progenitors in Nde1−/− and Nde1−/−Trp53−/− mutants were correctly localized at the ventricular surface , whereas most PH3+4A4− progenitors in the mutants were ectopically localized , indicating that the basally mislocalized PH3+ cells were those arrested in late-G2 and that INM in the mutant was largely normal ( Figure 6D ) . Consistent with the G2 arrest , we found that the level of PH3 was elevated in the Nde1−/−Trp53−/− brain ( Figure 6E ) . The hallmark for DNA damage-induced G2/M checkpoint involves the degradation of Cdc25A , a phosphatase essential for Cdk1 activation ( Mailand et al . , 2000 ) . Both Nde1−/− and Nde1−/−Trp53−/− brains showed reduced levels of Cdc25A at E12 . 5 ( Figure 6E ) , supporting the notion that prolonged G2 in Nde1−/−Trp53−/− progenitors was a result of persistent DNA damage . We followed a cohort of Nde1 Trp53 double mutant mice ( n > 80 ) over a 4-month period , and none of the Nde1−/−Trp53−/− mice ( n > 20 ) was found to have altered brain structures or tumors of CNS origin , though they showed obvious tumors in multiple organs outside of the CNS . We did not find tumors in Nde1+/−Trp53+/− mice ( n > 30 ) , but approximately one third of the Nde1−/−Trp53+/− mice ( ∼10 out of >30 ) developed tumors predominantly of hematopoietic lineages including lymphoma and tumor of the thymus ( Figure 6—figure supplement 1A ) . Apoptosis and mitotic figures along with remarkably elevated p53 protein levels were observed in the Nde1−/−Trp53+/− thymus tumor , suggesting malignancy and increased stress in the tumor tissue ( Figure 6—figure supplement 1B , C ) . The tumor formation in the Nde1−/−Trp53+/− mice is in agreement with Nde1's selective expression and functional requirement in hematopoietic cells in addition to neural progenitors ( Figure 6—figure supplement 1D ) . Together , the experimental evidence also supported that although p53 abrogation could restore the size and number of cortical layer II/III neurons of Nde1−/− brain , it unlikely ameliorated the genomic lesion caused by the Nde1 deficiency . To understand the molecular mechanism by which Nde1 safeguards the heterochromatin replication , we re-examined its subcellular localization and molecular interactions . Nde1 is a dynamic scaffold regulated by cell cycle dependent post-translational modifications , and it is capable of participating in various cellular compartments ( Feng and Walsh , 2004; Stehman et al . , 2007; Pawlisz and Feng , 2011 ) . By varying conditions of immunodetection , we found Nde1 was localized in the nucleus of a subset of primary progenitors identified by Pax6 immunoreactivity in the neocortical VZ at E12 . 5 ( Figure 7A ) . The possible nuclear targeting of Nde1 was confirmed by transfecting and examining GFP-tagged Nde1 in HeLa cells using confocal microscopy ( Figure 7B ) . Furthermore , co-immunostaining of Nde1 and Pcna , a DNA polymerase accessory factor which marks the replication fork with a distinctive punctate pattern , indicated that the nuclear pool of Nde1 was elevated when the progenitors were undergoing DNA replication ( Figure 7C ) , which suggested that Nde1 may translocate into the nucleus during S phase to play a direct role in DNA replication . 10 . 7554/eLife . 03297 . 016Figure 7 . Identification of a nuclear pool of Nde1 that interacts with the cohesin complex . ( A ) Double immunohistological staining with antibodies to Nde1 ( red ) and Pax6 ( green ) reveals the presence of Nde1 in the nucleus of cells in the neocortical VZ . Notice the detection of Ndel1 in the cortical neurons of the Nde1−/− brain due to the cross reactivity of anti-NDE1/Nde1 to Ndel1 . ( B ) Immunofluorescence confocal image of GFP-Nde1 ( green ) transfected HeLa cells showing nuclear targeting . Cell nuclei are highlighted by co-staining with Nucleoporin p62 at the nuclear envelope ( red ) as well as with Hoechst ( blue ) . ( C ) Immunohistological analysis reveals enhanced nuclear Nde1 ( red ) in S phase neural progenitors ( identified by Pcna foci in green ) . ( D ) Co-immunoprecipitation of the cohesin complex with Nde1 . Myc-Nde1 was transfected in 293T cells and immunoprecipitated by the anti-myc 9E10 antibody or mouse IgG . The presence of SMC3 , SMC1 , and RAD21 in the Myc-Nde1 immunocomplex is shown by immunoblotting . ( E ) Binding of SMC3 NBD with Nde1 . Flag-NBD of SMC3 was co-transfected with GFP-Nde1 in 293T cells and immunoprecipitated by the Flag antibody and mouse IgG . The presence of GFP-Nde1 and the absence SMC3 in the Flag-NBD immunocomplex are shown by immunoblotting . A diagram of the cohesin complex and the Nde1 binding domain ( NBD ) of SMC3 is included . ( F ) Flow cytometry analysis of cell cycle DNA content of 293T cells transfected with the vector control , Flag-NBD , and MycNde1-N1 , a Nde1 N-terminal fragment that was previously shown to induce G2/M arrest by blocking Nde1 dimerization . ( G ) Nde1 co-complexes with cohesin and its associated chromatin remodeler SNF2h . Myc-Nde1 transfected in 293T was immunoprecipitated by the anti-myc 9E10 antibody using the mouse IgG as negative control . After the positive identification of core subunits of cohesin in the Myc-Nde1 immunocomplex , the immunoblots were re-probed with antibodies to several cohesin interacting proteins . SNF2h was consistently found to co-complex with Nde1 and cohesin . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 01610 . 7554/eLife . 03297 . 017Figure 7—figure supplement 1 . Blocking Nde1–cohesin interaction results in apoptosis . Immunofluorescence analysis of 293T cells transfected by SMC3 Flag-NBD ( red ) showing increased apoptosis indicated by cleaved caspase 3 immunoreactivities ( CC3 , green ) and micronuclei indicated by Hoechst ( blue , arrow ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 01710 . 7554/eLife . 03297 . 018Figure 7—figure supplement 2 . Increased genomic instability of Nde1 mutant MEFs . ( A ) Karyotyping analysis shows the rapid development of aneuploidy in Nde1−/− and Nde1−/−Trp53−/− MEFs . Chromosome number distributions ( % of total diploid cells ) of wild-type and Nde1−/− MEF lines from passage 3 to passage 10 , as well as Nde1−/−Trp53−/− MEFs from passage 3 to passage 5 are presented . ( B ) Examples of Giemsa stained wild-type , Nde1−/− and Nde1−/−Trp53−/− mitotic spreads; as well as high magnification images of chromosome aberrations observed in Nde1−/− and Nde1−/−Trp53−/− MEFs at passage 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 03297 . 018 To further assess the involvement of Nde1 in DNA replication , we searched for Nde1 binding proteins and identified structural maintenance of chromosomes 3 ( Smc3 ) as a novel Nde1 partner from an Nde1 yeast two-hybrid screen . Smc3 is a core subunit of cohesin , an essential guardian of genome integrity . Composed of SMC1 , SMC3 , SCC1 ( RAD21 ) , and SCC3 ( STAG ) , cohesin forms a ring-like structure to embrace chromatin fibers from the moment they arise following the replication fork until the onset of anaphase to ensure the proper segregation of genetic materials to daughter cells ( Nasmyth , 2011 ) . The physical interaction of Nde1 and Smc3 was confirmed by the co-immunoprecipiation studies ( Figure 7D ) . Two other core subunits of the cohesin complex , Smc1 and Rad21 , were also detected in the Nde1–Smc3 immunocomplex , indicating the direct interaction of Nde1 with the entire cohesin complex . We also performed the reciprocal co-immunoprecipitation analysis to detect Nde1 in the immunoprecipitate of the SMC3 fragment recovered from the Nde1 yeast two-hybrid screen . This Nde1 binding domain of Smc3 ( NBD ) , which corresponds to amino acids R746 to R903 in the coiled-coil region between the C-terminal globular and central hinge domains of Smc3 , was immunoprecipitated and found to form a specific immunocomplex with Nde1 ( Figure 7E ) . To determine the functional significance of the Nde1–Smc3 interaction , we over-expressed the NBD to block the binding of Nde1 with cohesin and found that NBD arrested the cell cycle preferentially in S phase ( Figure 7F ) . We also noted that many cells expressing the NBD showed fragmented nuclei or committed to apoptosis as seen in Nde1 mutant neural progenitors ( Figure 7—figure supplement 1 ) . These data suggested that the lack of proper interaction of Nde1 with cohesin may underlie the S phase genotoxicity in Nde1 mutant neural progenitors . Besides a canonical role in maintaining the fidelity of chromosome segregation , cohesin has emerged as an important regulator for higher order chromatin structures , essential for DNA replication through densely packed heterochromatic repetitive sequences . Cohesin can also serve as a platform for DSB repair or act as a nuclear global controller for gene expression on both transcriptional and epigenetic levels . The diverse functions of cohesin are accomplished through recruiting various molecules to the core cohesin ring complex ( Peters et al . , 2008; Imakaev et al . , 2012; Baranello et al . , 2014 ) . To define the functional cooperation of Nde1 with cohesin , we tested if Nde1 co-complexes with cohesin partners for DNA repair , chromatin remodeling , and transcriptional regulation through co-immunoprecipitation studies . The chromatin remodeler SNF2h was consistently found in the Nde1–cohesin immunoprecipitates ( Figure 7G ) . SNF2h is an ATPase chromatin remodeler that interacts with RAD21 ( Hakimi et al . , 2002 ) . During S phase , SNF2h is recruited to the replication foci and facilitates heterochromatic remodeling ( Zhou et al . , 2009; Guetg et al . , 2010; Sugimoto et al . , 2011; Postepska-Igielska et al . , 2013 ) . The participation of Nde1 in the cohesin–SNF2h complex suggests that Nde1 may participate in mid to late S phase heterochromatin remodeling . Karyotyping analysis of mouse embryonic fibroblast ( MEFs ) also revealed that Nde1 mutation led to increased genomic instability similar to those that have been seen in cohesinopathy , and that Nde1 Trp53 double mutation accelerated the aneuploidy caused by the Nde1−/− single mutation . At passage 3 , when Nde1−/− MEFs only showed a modest increase in aneuploidy compared to the wild-type MEFs , more than 50% of the Nde1−/−Trp53−/− MEFs were aneuploid . About 70% of the Nde1−/− MEFs became aneuploid by passage 10 , but a similar degree of aneuploidy developed in the Nde1−/−Trp53−/− MEFs at passage 5 ( Figure 7—figure supplement 2A ) . While some Nde1−/− cells showed chromosomal aberrations with visible chromosome breaks on crossed chromosome arms , which was reminiscent of what have been observed in cohesin mutants ( Peters et al . , 2008 ) , more severe chromosomal aberrations including breaks , fragmentation , detachments , and cross of chromosome arms were frequently observed in Nde1−/−Trp53−/− MEFs ( Figure 7—figure supplement 2B , C ) . The increased chromosomal instability of Nde1−/−Trp53−/− over Nde1−/− MEFs was also in agreement with what was observed in Nde1−/−Trp53−/− neural progenitors ( Figure 6 ) and supported that abrogating p53 increased genotoxic lesions in the Nde1−/− mutant by allowing mutant cells with DSBs to escape DDR-induced p53-dependent apoptosis , enabling the genetically altered mutant cells to survive and propagate .
One of the most important observations presented in this study is that DNA damage and apoptosis in Nde1 mutant brain were spatially and temporally associated with the early neuronal differentiation of multipotent neural progenitors . The peak detection of γH2AX and CC3 in the developing cortex of Nde1 mutant mice was around E12 . 5 . Levels of DNA damage and apoptosis were low , both during the early phase neural progenitor self renewal before E12 and when the bulk of cortical neurons were generated after E15 . During the period when genomic lesions were detected at the peak level between E12 and E13 , their abundance spatially correlated with the level of neurogenesis following the TNG . According to a recent genetic fate mapping study , at least a subset of progenitors become fate restricted to become upper layer cortical neurons before E13 . 5 , even through these neurons are generated through terminal mitosis 3–4 days later ( Franco et al . , 2012 ) . Therefore , the temporal-specific DNA damage that we observed in the Nde1 mutant brain matches closely with the time when primary neural progenitors become fate restricted to be layer II/III neurons . This phenotype is different from what have been seen in conditional brain mutation of genes known to play global roles in DNA metabolism and genome surveillance , such as Brca1 , ATR , DNA ligase IV , XRCC4 , TopBP1 , and Cdh1 ( Frank et al . , 2000; Gao et al . , 2000; Pulvers and Huttner , 2009; Gatz et al . , 2011; Lee et al . , 2012; Eguren et al . , 2013 ) . In these mutants , endogenous or gamma irradiation-induced DNA damage and p53-dependent apoptosis were shown to occur non-selectively in all progenitors that undergo active proliferation; abrogating p53 ameliorated the microcephaly but made little improvement to the structure of the mutant brain , which agree well with the wide-spread loss-of-function defects of these genes in many other organs . These previous studies demonstrated the importance of genome maintenance in brain development but did not reveal the brain specific endogenous origins of DNA damage and its underlying molecular mechanisms . In contrast , experimental findings of this study not only identified Nde1 as a key player for brain specific genome maintenance but also highlighted the principal source of DNA damage in brain development . Our data suggest that the strongest demand for genomic surveillance occurs in S phase and at heterochromatic DNA domains when neural progenitors undergo fate restrictive differentiation . Heterochromatin constitutes a significant portion of the mammalian genome and is preferentially formed at chromosomal regions with high density of repetitive DNA elements , such as rRNA genes , centromeres , pericentromeres , and telomeres ( Lander et al . , 2001; Grewal and Jia , 2007; Politz et al . , 2013 ) . Compared to the actively transcribed euchromatin , heterochromatin is densely packed and traditionally considered transcriptionally inert . However , heterochromatin is known to produce RNA transcripts necessary for the establishment of heterochromatic states ( Bierhoff et al . , 2014 ) , and these RNA transcripts were preferentially generated during the replication of the heterochromatin in late S phase ( Li et al . , 2011 ) . Heterochromatin may not only influence gene expression profiles through regulating the higher-order chromatin structure , the repetitive sequences of heterochromatin are also hotspots for recombination which represent serious challenges to genomic integrity ( Peng and Karpen , 2008; Almouzni and Probst , 2011 ) . The proper maintenance of heterochromatin following the replication fork is indispensible for genomic stability by suppressing recombination and somatic mutations . The newly identified Nde1 partners cohesin and SNF2h are both known to participate in heterochromatin formation and function in multiple tissues . Therefore , further delineating the genomic target specifically co-regulated by Nde1and cohesin complex in neural progenitors will provide critical insight into how genomic quality is specifically controlled during neuronal differentiation . In addition to the S phase , previous observation has shown that neocortical neurogenesis is governed by the length of G1 phase of the cell cycle ( TG1 ) ( Lange et al . , 2009; Pilaz et al . , 2009 ) , which progressively increases over fourfold in mice from E11 to E17 ( Takahashi et al . , 1995; Miyama et al . , 1997 ) . While the lengthened G1 may permit increased gene expression and protein synthesis required for neurogenesis , another important task of G1 is to determine the timing of DNA replication and ‘license’ the replication origins through loading pre-replication complexes of DNA helicase at various chromosomal domains ( Dimitrova and Gilbert , 1999; Gilbert , 2010; Nordman and Orr-Weaver , 2012; Renard-Guillet et al . , 2014 ) . In mammalian cells , over 30 , 000 origins are unutilized in each cell cycle; the spatiotemporal pattern of replication origin activation may depend on , as well as determine , a cell's differentiation state ( Gilbert , 2010; Nordman and Orr-Weaver , 2012; Rhind and Gilbert , 2013 ) . As the peak of S phase defect in Nde1 mutants at E12 to E13 overlaps with the developmental period , when Ts is at the top value and TG1 increased most drastically ( Takahashi et al . , 1995; Miyama et al . , 1997 ) , it is not unreasonable to speculate that the early phase of neuronal fate determination involves the reprogramming of chromatin state that initiates in G1 and executes in S phases . Thus , cell fate in cortical development may not merely be determined by turning on or off a handful of genes , but rather controlled at the genomic level by a nuclear global alteration of the chromatin state . The core Nde1 protein is approximately 40 kDa and comprises an N-terminal extended coiled-coil that shows remarkable evolutionary conservation . However , Nde1 is diverged structurally and functionally from its paralog Ndel1 in mammals outside of the coiled-coil and is primarily expressed in neural progenitors and hematopoietic cells ( Feng et al . , 2000 ) . We have found that the native Nde1 protein in the developing mouse tissue is very insoluble and exists at a relatively low level , but genetic studies have shown that Nde1/NDE1 is an essential player for various pivotal aspects of the neural progenitor biology . It is likely that all native Nde1 molecules are fully occupied through binding to other cellular proteins . The multifaceted function of NDE1 as a molecular scaffold might be fulfilled by post-translational modifications or through undergoing reversible transitions between monomeric and polymeric forms ( Feng and Walsh , 2004; Soares et al . , 2012 ) , which facilitate its dynamic partition and reassembly into different molecular complexes . DNA replication is one of the most challenging processes that requires efficient and accurate molecular re-arrangements . During S phase , thousands of replication origins are utilized in a spatiotemporally defined pattern . At each licensed origin , helicase is loaded , replisome is assembled , nucleosomes are disassembled ahead of each replication fork , and resembled onto the two daughter genomes . Meanwhile , multiple histone chaperones , modification enzymes , and numerous chromatin remodelers are recruited to ensure the precise duplication of both the genome and the epigenome ( Alabert and Groth , 2012; Whitehouse and Smith , 2013 ) . Such a dynamic process requires a protein ( s ) like NDE1 to serve as a versatile platform to facilitate various protein–protein interactions . Thus , an evolutionarily advanced dynamic scaffold function during heterochromatin replication is consistent with a functional involvement of Nde1 in neural progenitor biology . Nde1 has been implicated in the regulation of the cytoskeleton and dynein motors ( Lipka et al . , 2013 ) . However , it is unclear how such regulation could underlie the tissue specific and spatiotemporally dependent phenotype of both human and mouse NDE1/Nde1 mutations nor the DNA replication defects described in this study . Our data suggest that instead Nde1 may be involved in assisting ATP-utilizing chromatin remodeling proteins such as SNF2h to move and reorganize nucleosomes along the DNA strand during heterochromatin replication . Replication forks in Nde1 mutant neural progenitors are paused or stalled at heterochromatic domains composed of centric/pericentric repeats . The integrity of centric/pericentric heterochromatic domains is essential not only for defining the chromatin state but also for the recruitment and the establishment of kinetochore complexes required for mitotic spindle formation and chromosome segregation in mitosis ( Ishii et al . , 2008; DeLuca and Musacchio , 2012; Mankouri et al . , 2013 ) . While a direct role of Nde1 in mitotic spindle organization remains plausible , we also believe that the skewed mitosis previously observed in Nde1−/− neural progenitors may , at least in part , originate from unresolved centromeric and pericentromeric structural defects arising during S phase . It remains to be further tested whether the mitotic spindle defect observed in Nde1 mutant progenitors is a primary cause for the defective cortical neurogenesis or a consequence of an ill-replicated genome . The selective expression of Nde1 in neural progenitors and cells of hematopoietic lineage implies specific functional requirement of Nde1 in these cells . It is intriguing that the loss of NDE1 impairs the developing CNS much more profoundly than the hematopoietic system and other non-CNS tissues , though NDE1 mutations are associated with leukemogenesis ( Cavazzini et al . , 2009; Van der Reijden et al . , 2010 ) . In the developing CNS , the role of NDE1 is more essential in humans than in mice . Patients who lose both functional copies of NDE1 show severe malformation of many brain structures , but the detectable brain anatomical defect of Nde1−/− mice is largely limited to the neocortex and specifically to layer II/III cortical neurons ( Feng and Walsh , 2004; Alkuraya et al . , 2011; Bakircioglu et al . , 2011; Guven et al . , 2012; Paciorkowski et al . , 2013 ) . While the mild phenotype in many key brain structures , such as the hippocampus , of Nde1 mice may be due to the functional redundancy of Nde1 with Ndel1 , the common etiology of NDE1 mutation-induced genotoxicity in both neural and hematopoietic progenitors suggests that NDE1 may be important for establishing the evolutionarily increased cell diversity and function in both brain and blood . The cerebral cortex is the largest organ in terms of both cell number and functional diversity , and cortical layer II/III is expanded exponentially in evolution ( DeFelipe et al . , 2002; Molnar et al . , 2006 ) . The small and medium-size pyramidal neurons of cortical layer II/III are essential for functional connectivity between two cerebral hemispheres and among various cortical regions; they are essential for evolutionarily advanced higher order brain activities ( DeFelipe et al . , 2002; Fame et al . , 2011; Greig et al . , 2013 ) . It is evident that even a slight perturbation to these neurons is likely to result in cognitive deficits . Therefore , it is conceivable that advanced novel molecular controls are gained in mammalian evolution to protect the genome of these neurons , although further study is required to elucidate the mechanism underlying such genome maintenance . Our data show that genomic damages during neurogenesis are primarily responsible for the cortical neuronal loss in Nde1 mutant brains . However , depending on the activity of DDRs , genomic damages may not necessarily result in altered brain structure as evidenced by the restored structure and layer II/III neurons in the Nde1−/−Trp53−/− mutant . We believe that these data highlight the importance of genomic control of neuronal functions . Nde1 mutant cells presented an entire array of genotoxic features including DSBs and chromosomal lesions that would ordinarily lead to carcinogenesis . Nonetheless , even the Nde1−/−Trp53−/− brain , which presumably carries the unresolved genomic lesions to adulthood , was cancer-free . While the absence of brain tumor may be in line with our observation that genotoxicity in Nde1 mutant neural progenitors occurred when they become fate restricted along the path of differentiation towards post-mitotic neurons , it is still interesting to learn whether cortical neurons can truly forgive the genomic lesions they inherit from their precursors . With the advent of comprehensive genome-wide analysis , somatic mutations , especially copy number variants ( CNVs ) are found to be specifically abundant in human neurons and increasingly linked to neurodevelopmental diseases ( Grayton et al . , 2012; McConnell et al . , 2013; Moreno-De-Luca et al . , 2013 ) . The NDE1 gene is one of the ‘hot spots’ for CNVs at chromosome 16p13 . 11 , of which deletions and duplications have been found to associate with a wide spectrum of developmental brain disorders including intellectual disability , epilepsy , autism , schizophrenia , and attention-deficit hyperactivity disorder ( ADHD ) ( Hebebrand et al . , 1994; Gillberg , 1998; Sharp et al . , 2006; Ullmann et al . , 2007; Hannes et al . , 2009; Heinzen et al . , 2010; Mefford et al . , 2010 ) . The genomic region span 16p13 . 11 encodes eight transcripts including MPV17L , C16orf45 , KIAA0430 , MYH11 , C16orf63 , ABCC1 , and ABCC6 in addition to NDE1 . Among these NDE1 is the only gene that is known to be important for brain formation . Findings described in this study suggest that NDE1 dosage alteration may result in secondary genomic lesions in cortical neurons . Therefore , even with the lack of neoplastic over-proliferation , brain developmental diseases may have a commonality with cancers as genomic mosaic genomic disorders . Although neurons with altered genome do not become cancerous due to their post-mitotic nature , they can be manifested by functional deficits of the brain in various forms . While further studies are required for delineating the causal relationship between NDE1 gene dosage and compromised cortical functions , results from this study predict that genomic insults incurred during heterochromatic DNA replication during neural progenitor differentiation may underlie a large variety of developmental neurological and psychiatric disorders .
The Nde1 and Lis1 knockout mice have been described previously ( Feng and Walsh , 2004; Pawlisz et al . , 2008 ) . The Trp53 knockout mice ( Trp53tm1Tyj/J ) were obtained from JaxMice ( Bar Harbor , ME , stock # 002101 ) . The Nde1–Trp53 double mutant mice were generated by standard genetic crosses . Mice used for this study were housed and bred according to the animal study protocol ( protocol number 2012-1655 ) approved by IACUC committee of Northwestern University . All procedures were in compliance with the NIH Guide for Care and Use of Animals . For timed matings , the day of vaginal plug was considered E0 . 5 . Immunohistology studies were carried out as described ( Pawlisz et al . , 2008 ) on 12-μm frozen or 5-μm paraffin tissue sections . Neocortical coronal or transverse sections matched spatially at the mid-hemisphere level using the ganglionic eminence , the midline choroidal fissure , and the roof of the third ventricle as references were compared among littermates . The following antibodies were used: γH2AX , PH3 , p53 , NeuN ( Millipore , Billerica , MA ) ; phospho-p53 ( Ser15 ) , Rad50 , γH2AX ( Cell Signaling Tech , Beverly , MA ) ; SMC3 , Tbr2 , BU1/75 , Foxp2 , NPM1 ( Abcam , Cambridge , MA ) ; PCNA , SMC1 , 53PB1 ( Novus Biologicals Littleton , CO ) ; Cux1 , DCX , BRCA1 , Cdc25A ( Santa Cruz , Dallas , TX ) ; Pax6 ( Thermo , Waltham , MA ) ; B44 ( BD Biosciences , San Jose , CA ) ; phospho-vimentin 4A4 ( MBL International , Woburn , MA ) ; SNF2h ( Active Motif , Carlsbad , CA ) ; RAD21 ( Bethyl Lab , Montgomery , TX ) ; parvalbumin ( Sigma , St . Louis , MO ) ; Tuj1 ( Covance , Princeton , NJ ) . To detect Nde1 in mouse embryonic cortical progenitors , freshly dissected mouse embryos were embedded in OTC , frozen on dry ice , cryosectioned , fixed with 4% paraformaldehyde in PBS for 10 min , permeabilized with 0 . 1% Triton X-100 , then immunostained with the NDE1 antibody ( ProteinTech , Chicago , IL ) . As the NDE1 antibody cross-reacts with Ndel1 that is expressed highly in neurons but absent in neural progenitors , brain sections from Nde1−/− embryos were used as negative control so that Nde1-specific immunosignals in VZ progenitors could be distinguished . All experiments were repeated with at least three litters of mouse embryos; images from a representative experiment are shown . BrdU , IdU , or CldU ( all from Sigma ) were injected intraperitoneally ( i . p . ) to pregnant mice at 50 mg/kg to label embryos at E12 . 5 . Labeled embryos were fixed , embedded , sectioned transversely , and processed for immunohistology with antigen retrieval in citric acid-based antigen unmasking solution ( Vector lab , Burlingame , CA ) . Double or triple immunofluorescence-stained coronal brain sections were imaged with a Leica CTR5000 fluorescence microscope equipped with a Qimage RETIGA 2000R digital camera under 20× or 40× objectives . Photographs from sections of the dorsal–medial cerebral wall at the mid-hemisphere level were taken . Fluorescent images were co-stained with Hoechst 33342 to identify cell nuclei . All images taken at different fluorescent channels were from the same focal-plane . At least three serial sections from three different litters for each genotype were analyzed . Quantitative analyses of immunosignals were performed with ImageJ or Adobe Photoshop CS4 . Cell counts and comparisons between different genotypes were performed on mid-hemisphere neocortical sections . To avoid errors introduced by variable brain size , embryo shape , and embedding angles , sections were spatially matched using the ganglionic eminence and the midline choroidal fissure or the roof of the third ventricle , the anterior commissure , and the hippocampus as references for E12 . 5 or weaning age , respectively . The area of interest was specified in an approximately 400-micron length derived by measuring the distance along the ventricular surface in the dorsal and lateral pallium ( between the medial pallium and the pallial–subpallial boundary ) at E12 . 5 or in the neocortex at an approximate 45°-angle from the dorsal–ventral axis of mid-hemisphere level coronal sections at weaning age . All cells from the pial to ventricular surfaces were included in the analysis . Double positive cells were overlaid manually by color-coded dots in different layers . The number of cell counts was recorded using the measurement and analysis tools of either ImageJ or Photoshop CS4 and imported to Excel for quantitative statistical analysis and presentation . All results shown are mean ± SD from a minimum of three independent biological replicates . Statistical significance was estimated using the Student's t test between two groups . Analysis of variance ( ANOVA ) was used for the comparison among groups with ≥3 categories ( with SAS 9 . 4 ) . When the overall F test from the ANOVA was significant ( p < 0 . 05 ) , Tukey-Kramer simulation-based adjusted p values were used for pairwise comparisons between the categories of the groups . Chi-Square test was used to compare proportions among categorical groups . Differences were considered significant with a p < 0 . 05 . The plasmid encoding the Nde1 binding domain ( NBD ) of Smc3 was PCR amplified with primers ATG GATTAC AAGGATGACGACGATAAGAGACAGCAATCAGAA AAG and TTA GCGCTCCATACTTTTCTG , and a Flag tag was incorporated to the N-terminal end of the construct . The PCR product was cloned to pCR II TA cloning vector ( Invitrogen ) , sequenced , and subcloned to pcDNA3 . 0 for mammalian cell expression . Cell culture , transfection , and immunofluorescence were performed as described ( Pawlisz and Feng , 2011 ) . Cerebral cortices were dissected from mouse embryos and analyzed as described ( Pawlisz and Feng , 2011 ) . Experiments were repeated with samples from at least three litters; results from a representative experiment are shown . Comet assay for the assessment of DNA damage was performed according to the alkaline single-cell gel electrophoresis method as previously described ( Singh et al . , 1988 ) with minor modifications . Briefly , cortical progenitors were isolated in cold PBS with 20 mM EDTA from E12 . 5 mouse embryos , resuspended in 1% low melting point agar and PBS , and spread onto agar-coated slides . Cells on the slides were incubated in lysis buffer ( 2 . 5 M NaCl , 100 mM EDTA , 10 mM Tris base , 1% Triton X-100 , and 10% DMSO ) for 30 min at 4°C , followed by alkaline buffer ( 0 . 3 N NaOH and 1 mM EDTA ) for 20 min at room temperature . After washing three times with 0 . 5× TBE , the cell-coated slides were electrophoresed at 0 . 7 volt/cm in 0 . 5× TBE , then neutralized , stained with DAPI , and analyzed under fluorescence microscope . Cells from Nde1 mutants and littermate controls were analyzed in parallel under the same experimental conditions , and comet tail length of over 300 randomly selected cells was scored . The assay was performed with more than three litters of Nde1 , Lis1 double mutant embryos; results from representative experiments are shown . Immunoprecipitation was performed as described ( Pawlisz and Feng , 2011 ) . In a buffer with 25 mM HEPES , 150 mM NaCl , 10 mM NaF , 100 μM Na3VO4 , 0 . 5% NP40 , and 10% glycerol with the addition of protease inhibitors and 1 mM ATP . Flow cytometry analyses of cell cycle DNA content were performed as described ( Feng and Walsh , 2004 ) . Northern blotting analysis of Nde1 and Ndel1 expression was performed essentially as described ( Feng et al . , 2000 ) . Briefly , 10 µg of total RNA extracted from various tissues of 3-month-old wild-type mice were loaded into each lane . Full-length coding cDNAs of mouse Nde1 and Ndel1 were used to probe the blot , respectively . Loading was normalized by the amount of 18S rRNA in each sample . Mouse embryo fibroblasts were isolated and cultured as described ( Alkuraya et al . , 2011 ) . To make the mitotic chromosome spreads , cells at passage 3 , 5 , and 10 grown on coverslips were arrested by 0 . 05 μg/ml colcemid for 5 hr . Karyotyping analyses were then performed as described ( Eves and Farber , 1981 ) . WT , Nde1−/− and Nde1−/−Trp53−/− MEFs were cultured and analyzed in parallel . 80–120 mitotic spreads from MEFs derived from three different embryos were scored for each genotype . The seeding trypsinized embryo was counted as passage 0 and the first replating as passage 1 . | The brain is a complex organ with many different cell types that each have specialized functions . Mutations in genes that control how the brain develops can have serious consequences—and one of the most important genes involved in the development of the human brain is called NDE1 . Individuals who inherit mutated copies of the NDE1 gene from both parents have brains with fewer folds than a healthy brain , and their brains are up to 90% smaller than normal . These individuals may have serious developmental disabilities , struggle with basic functions like swallowing , and die early . Moreover , having just one defective copy of the NDE1 gene has been linked to the development of cancers such as leukemia . The protein encoded by the NDE1 gene acts as a scaffold for many protein complexes and can be found throughout the cell in various cellular compartments ( from the cell periphery to the nucleus ) . However , it was unclear where the NDE1 protein's activity was most needed in developing brain cells . Feng and Houlihan now provide a new explanation for NDE1's role in brain development . Looking for molecules that interact with the mouse version of the Nde1 protein revealed that it binds to proteins that control how DNA is packaged inside the nucleus of a cell . In doing so , Nde1 appears to protect the genome of brain stem cells , while these cells' DNA is copied and before they divide to form new cells destined to become neurons . For a cell to divide , its genetic information must be accurately copied and then segregated between the two new cells to ensure that each receives all the genetic instructions needed to develop and function properly . In brain cells from mice without functional Nde1 , the DNA frequently breaks as it is copied . Brain cells that inherit damaged DNA might not function correctly , while serious breaks that affect both strands of the DNA can trigger a response that kills the mutated cell . This means , there are fewer cells that make up the outer layer of the brain , making it less wrinkled than normal . This part of the brain—called the cerebral cortex—is important for many processes including thought and memory , and it helps different areas of the brain communicate with one another . This may explain why mutations in the NDE1 gene can contribute to a variety of brain disorders . Errors during DNA replication can also cause cancer to develop . As such , the findings of Feng and Houlihan may also help to explain why some genetic mutations are associated with both cancer and brain disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"developmental",
"biology"
] | 2014 | The scaffold protein Nde1 safeguards the brain genome during S phase of early neural progenitor differentiation |
How different helicase families with a conserved catalytic ‘helicase core’ evolved to function on varied RNA and DNA substrates by diverse mechanisms remains unclear . In this study , we used Mss116 , a yeast DEAD-box protein that utilizes ATP to locally unwind dsRNA , to investigate helicase specificity and mechanism . Our results define the molecular basis for the substrate specificity of a DEAD-box protein . Additionally , they show that Mss116 has ambiguous substrate-binding properties and interacts with all four NTPs and both RNA and DNA . The efficiency of unwinding correlates with the stability of the ‘closed-state’ helicase core , a complex with nucleotide and nucleic acid that forms as duplexes are unwound . Crystal structures reveal that core stability is modulated by family-specific interactions that favor certain substrates . This suggests how present-day helicases diversified from an ancestral core with broad specificity by retaining core closure as a common catalytic mechanism while optimizing substrate-binding interactions for different cellular functions .
Helicases of superfamilies ( SFs ) 1 and 2 use ATP or other NTPs to bind , unwind , or remodel RNA or DNA in essentially all facets of nucleic acid metabolism ( Preugschat et al . , 1996; Tanaka and Schwer , 2005; Singleton et al . , 2007; Fairman-Williams et al . , 2010; Jarmoskaite and Russell , 2014 ) . They contain a conserved ‘helicase core’ of two RecA-like domains but act on varied substrates by different mechanisms . SF1 and SF2 helicases can be grouped into families with distinct variations in specificity , mechanism , function , and appended domains ( Figure 1A ) ( Gorbalenya and Koonin , 1993; Singleton et al . , 2007; Fairman-Williams et al . , 2010 ) . The mechanisms by which SF1 and SF2 helicases act on RNA or DNA include non-processive unwinding of short duplexes ( e . g . , DEAD-box RNA helicases [Jarmoskaite and Russell , 2011; Linder and Jankowsky , 2011] ) , unwinding coupled to directional movement ( ‘translocation’ ) along the unwound single strand ( e . g . , DEAH/RHA , NS3/NPH-II , and RecQ-like helicases [Pyle , 2008] ) , and binding or translocation along a duplex without unwinding ( e . g . , RIG-I-like and Swi/Snf helicases [Durr et al . , 2005; Myong et al . , 2009; Rawling and Pyle , 2014] ) ( Figure 1A ) . How helicases that share a conserved catalytic core evolved such functional diversity remains unknown . 10 . 7554/eLife . 04630 . 003Figure 1 . Structure , specificity , and mechanisms of the helicase core of Mss116 and other SF1 and SF2 helicases . ( A ) Domain architecture and characteristics of helicases belonging to different SF1 and SF2 families ( Fairman-Williams et al . , 2010 ) . Two other SF1 ( Pif1-like and Upf1-like ) and four other SF2 ( Ski2-like; RecG-like; T1R; and Rad3/XPD ) families have been identified ( Fairman-Williams et al . , 2010 ) . Helicase core domains 1 and 2 are colored light blue and green , respectively , while appended domains and insertions , which vary in size , composition , and function , are colored orange; domains are not to scale . ( B ) Schematic of the domain architecture of the helicase core of Mss116 ( D1 , blue; D2 , green; C-terminal extension of D2 [CTE] , orange ) showing the location of conserved DEAD-box sequence motifs ( Fairman-Williams et al . , 2010 ) . Full-length Mss116 contains additional unstructured N-terminal ( residues 37–87 ) and C-terminal ( residues 598–664 ) extensions that are not required for helicase activity ( Cao et al . , 2011; Mohr et al . , 2011 ) . ( C ) Structure of the closed-state helicase core of Mss116 ( PDB accession 3I5X ) ( Del Campo and Lambowitz , 2009 ) bound to ssRNA ( U10-RNA; yellow ) and adenosine nucleotide ( AMP-PNP; black ) . ( D ) Model for RNA duplex binding and unwinding by Mss116 . The helicase core domains of Mss116 have modular roles in substrate loading ( Mallam et al . , 2012 ) . D1 captures ATP in the open-state enzyme using the Q-motif , which coordinates the adenine base , and motifs I and II , which are the conserved triphosphate-binding loop and Mg2+-binding aspartic acid motifs , respectively , present in many other ATP-binding enzymes ( Walker et al . , 1982; Rudolph et al . , 2006; Schutz et al . , 2010; Mallam et al . , 2012 ) . D2 recognizes duplex RNA ( Mallam et al . , 2012 ) . When ATP and dsRNA are bound to D1 and D2 , respectively , core closure occurs , leading to unwinding of the dsRNA bound to D2 by bending one RNA strand and displacing the other . During unwinding and formation of the closed-state helicase core complex , ATP bound to D1 makes additional interactions with motifs Va and VI in D2 . The closed-state helicase core bound to ssRNA and ATP represents the ‘post-unwound’ state of the enzyme ( Figure 1C ) . ATP hydrolysis occurs in the closed state , followed by dissociation of Pi and ADP , which leads to the reopening of the core and the release of the bound ssRNA , thereby regenerating the enzyme ( Henn et al . , 2010; Cao et al . , 2011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 00310 . 7554/eLife . 04630 . 004Figure 1—figure supplement 1 . Crystal structures of helicases belonging to different SF1 and SF2 families . Examples are taken from the helicase families shown in Figure 1A . Most are in complex with nucleic acid ( yellow ) and domains are colored as in Figure 1A . The composition and PDB accession code are given for each structure . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 004 Here , we use the yeast DEAD-box protein Mss116 ( Figure 1B , C ) as a model system to pinpoint the molecular basis for the specificity and mechanism of the conserved helicase core . Mss116 functions as a general RNA chaperone in mitochondrial intron splicing by locally unwinding and disrupting stable but inactive RNA structures that impede RNA folding ( Huang et al . , 2005; Del Campo et al . , 2009; Potratz et al . , 2011 ) . As a general RNA chaperone , Mss116 binds diverse RNA substrates non-specifically and has high RNA helicase activity in the absence of partner proteins ( Halls et al . , 2007; Del Campo et al . , 2009 ) . This makes it an ideal model system to study the properties of an isolated helicase core . The helicase core of Mss116 consists of two RecA-like domains ( D1 and D2 ) that are in an extended ‘open state’ in the absence of substrates ( Mallam et al . , 2011 ) and recognize ATP and duplex RNA in a modular manner ( Mallam et al . , 2012 ) ( Figure 1D ) . Upon substrate binding , the two core domains join to form a ‘closed state’ containing an ATPase active site , while conserved DEAD-box protein motifs in D1 promote the unwinding of short duplexes bound to D2 by excluding one RNA strand and bending the other ( Figure 1D ) . The closed-state complex bound to ssRNA and ATP represents the ‘post-unwound’ state of the helicase core ( Figure 1C ) . ATP hydrolysis is required for core reopening and enzyme turnover ( Liu et al . , 2008; Cao et al . , 2011 ) . In this study , we determined the structural and biochemical factors that govern how analogues of NTPs ( ATP , CTP , GTP , and UTP ) and different nucleic acids ( single-stranded [ss] RNA , ssDNA , double-stranded [ds] RNA , A-form dsDNA , and B-form dsDNA ) interact with the helicase core . In this way , we identify the core–substrate interactions that dictate the physiological specificity and mechanism of Mss116 . Our results define the structural and biochemical determinants for the substrate specificity of a DEAD-box protein . Furthermore , they demonstrate that Mss116 has surprisingly ambiguous substrate binding and unwinding properties . Considered in the context of other SF1 and SF2 helicases , our findings show how small structural changes within conserved regions of these protein families can facilitate the emergence of specialized enzymes with new activities and cellular functions .
We investigated how Mss116 specifies for ATP during local unwinding by comparing the ability of the helicase core ( D1D2 , residues 88–597 ) to use different nucleotides to catalyze RNA unwinding . First , we measured the concentration of different NTP analogues required by the helicase core to unwind an RNA duplex under equilibrium conditions ( Figure 2A ) . This was done by using a 12-base pair ( bp ) dsRNA , which was labeled with a fluorophore and quencher at its 5′ and 3′ ends , respectively . A native gel-based assay was then used to monitor unwinding by the increase in fluorescence in a closed-state core containing a bound single strand ( Figure 2—figure supplement 1 ) . We find that all of the non-hydrolyzable analogues NDP-BeFx , where N = A , C , G , or U , can promote the unwinding of a dsRNA . However , ADP-BeFx is the most efficient with at least sixfold higher concentrations of C- , G- , or U-analogues required for RNA duplex unwinding ( K1/2 = 0 . 14 , 0 . 8 , 0 . 8 , and 2 . 4 mM , respectively; Figure 2A and Figure 2—figure supplement 1B–E ) . 10 . 7554/eLife . 04630 . 005Figure 2 . The biochemical basis for the ATP specificity of the helicase core of Mss116 . ( A ) dsRNA unwinding by the MBP-tagged helicase core measured under equilibrium conditions using a gel-based fluorescence assay to monitor the formation of a closed-state complex containing bound ssRNA at increasing concentrations of NDP-BeFx , N = A , C , G , or U ( Figure 2—figure supplement 1 ) . The fraction of unwound duplex was obtained by normalizing the band intensities separately for each gel using the parameters from the fit to a one-site binding model , as the change in fluorescence upon unwinding is different under each condition . The extent of unwinding with UDP-BeFx was less than that for the other nucleotide analogs , and the maximum concentration of UDP-BeFx used in this assay was insufficient to drive unwinding to completion ( Figure 2—figure supplement 1 ) . This could be because UDP-BeFx bound at saturating concentrations to D1 cannot efficiently induce a closed state . ( B ) Equilibrium binding of A10-RNA to the MBP-tagged helicase core determined by fluorescence anisotropy measurements at increasing concentrations of NDP-BeFx , N = A , C , G , or U . ( C ) Equilibrium binding of A10-RNA to the MBP-tagged helicase core determined as in ( B ) at increasing concentrations of ADP-BeFx , AMP-PNP , ADP , and ADP + Pi . Error bars in ( A–C ) represent the standard error for at least three independent measurements , and the error in the K1/2 or Kd represents the standard error of the non-linear regression . NB , no appreciable binding . In ( B and C ) , the fraction of A10-RNA bound was calculated by normalizing against the anisotropy signal for unbound and fully bound substrate obtained from the fit to a one-site binding model . ( D ) Normalized SEC profiles monitored by absorbance at 260 nm ( red ) and 280 nm ( black ) for the helicase core in the absence of all substrates and in the presence of A10-RNA + NDP-BeFx , N = A , C , G , or U . An A260/A280 >1 at the maximum absorbance indicates the formation of a closed-state complex . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 00510 . 7554/eLife . 04630 . 006Figure 2—figure supplement 1 . RNA unwinding measured by using a gel-based fluorescence assay to monitor the formation of a closed-state complex containing bound ssRNA . ( A ) Schematic representation of the equilibrium unwinding reaction measured in this assay . Unwinding was probed by using a 12-bp dsRNA substrate labeled with a fluorophore ( 6-carboxyfluorescein; FAM ) and quencher ( Iowa Black FQ; IBFQ ) probes at the 5′ and 3′ ends , respectively . An increase in fluorescence of this substrate occurs when the helicase core unwinds the dsRNA and forms a closed-state bound to ssRNA . ( B–E ) Representative unwinding assays for dsRNA ( 100 nM ) by the helicase core of Mss116 ( 2 μM ) measured at increasing concentrations of NDP-BeFx with N = A , C , G , and U for B–E , respectively . Samples were loaded in the reaction medium and resolved in a non-denaturing 6% polyacrylamide gel run at 4°C in 0 . 5× Tris/Borate/EDTA buffer ( pH 8 . 3 ) . Arrows mark complexes corresponding to the open- ( in the absence of NDP-BeFx ) and closed-state protein bound to RNA . Proteins have an N-terminal MBP tag to increase solubility under the EMSA conditions . The double band seen in some lanes could be the result of one or two protein molecules bound to a partially unwound duplex or to a closed-state with or without a partially unwound second strand . ( F ) Control unwinding assay using an equivalent 12-bp 5′ FAM-dsRNA with no quencher to demonstrate that , under the assay conditions , the RNA is always bound to the helicase core and widely separated from free substrate . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 00610 . 7554/eLife . 04630 . 007Figure 2—figure supplement 2 . Kinetic assay of the unwinding of dsRNA by Mss116 with different NTPs . ( A ) Schematic representation of the unwinding reaction measured in this assay . Unwinding was probed by using a 12-bp dsRNA substrate labeled with a fluorophore ( 6-carboxyfluorescein; FAM ) and quencher ( Iowa Black FQ; IBFQ ) probes at the 5′ and 3′ ends , respectively ( IDT ) . An increase in fluorescence of this substrate occurs upon unwinding and re-annealing to an unlabeled strand from a duplex of the same sequence that is present in excess . ( B ) Representative unwinding time course for labeled dsRNA ( 125 nM ) by the helicase core of Mss116 ( 2 μM ) measured at 5 mM ATP-Mg2+ . After the addition of stop buffer to remove any bound protein , duplex samples were resolved in a non-denaturing 20% polyacrylamide gel run at 4°C in 1× Tris/Borate/EDTA buffer ( pH 8 . 3 ) . ( C ) Representative unwinding time course for labeled dsRNA ( 125 nM ) by the helicase core of Mss116 ( 2 μM ) measured at 5 mM CTP-Mg2+ with samples resolved as in ( B ) . The last lane represents the same duplex unwound by ATP after 60 min . ( D ) Kinetic unwinding profiles of dsRNA by Mss116 for NTP , N = A , C , G , or U . Error bars represent the standard error for at least three independent measurements , and the error in k1 represents the standard error of the non-linear regression . NU , no appreciable unwinding . Unwinding data for ATP were normalized using the parameters obtained from the fit to a first-order reaction with a single exponential . In the case of other nucleoside triphosphates where no unwinding was observed , data were normalized against the signal for a duplex fully unwound by ATP at the same concentration ( see panel C , final lane ) . Assays were performed in a buffer containing 5 mM free Mg2+ . Additional assays were performed at 0 . 5 mM Mg2+ , as previous data indicate that the unwinding activity of Mss116 increases at lower Mg2+ concentrations ( Halls et al . , 2007 ) . These gave similar results . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 007 Kinetic unwinding assays were also performed using the same dye-labeled dsRNA in the presence of an unlabeled duplex . In these experiments , an increase in fluorescence occurs upon unwinding of a labeled duplex and subsequent re-annealing to an unlabeled strand . This was measured by isolating the duplexes using native gel electrophoresis at various times after unwinding was initiated by the addition of NTP , where N = A , C , G , or U ( Figure 2—figure supplement 2 ) . These assays show that only ATP , and not other NTPs , catalyzes the unwinding of the dsRNA ( Figure 2—figure supplement 2B–D ) . This indicates that under our assay conditions , the diphosphate beryllium fluoride analogue is necessary to promote unwinding with nucleotide bases other than adenine . This difference likely reflects that the NDP-BeFx analogues form longer-lived , more stable complexes with RNA than do the corresponding NTPs ( Liu et al . , 2014 ) . We next examined how the stability of the ternary closed-state complex with ssRNA and the same NTP analogues correlates with the efficiency of duplex unwinding . Equilibrium fluorescence anisotropy binding assays with a fluorescein ( FAM ) -labeled A10-RNA were used to monitor formation of the closed state with increasing concentrations of NDP-BeFx ( N = A , C , G , or U; Figure 2B ) . These assays show that the closed-state complex is most stable with ADP-BeFx ( Kd = 0 . 022 mM ) , while CDP-BeFx , GDP-BeFx , and UDP-BeFx promote formation of the closed state only at significantly higher concentrations of nucleotide analogue ( Kd = 0 . 09 , 0 . 11 , and 0 . 63 mM , respectively ) . Similarly , analytical size-exclusion chromatography ( SEC ) shows that a closed-state helicase core with A10-RNA is maintained during elution for complexes containing ADP-BeFx , CDP-BeFx , or GDP-BeFx but not those containing UDP-BeFx , consistent with the latter complex having a lower stability ( Figure 2D and Table 1 ) . Together , these findings indicate that the unwinding efficiencies and closed-state core stabilities with different NTP analogues follow the same order of A > C , G > U from higher to lower efficiency and stability . 10 . 7554/eLife . 04630 . 008Table 1 . Size exclusion chromatography analysis of the helicase core of Mss116DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 008SampleElution volume at maximum absorbance/mlA260/A280 of peak at maximum absorbanceLikely predominant state of coreFree protein D1D2 ( Mss116 helicase core ) 11 . 40 . 6OpenFree nucleic acid dsRNA16 . 32 . 1− A-DNA duplex15 . 11 . 6− B-DNA duplex15 . 31 . 9− A10-RNA18 . 63 . 0− A10-DNA16 . 63 . 4−Protein–RNA–nucleotide complexes* D1D2–dsRNA–ADP-BeFx9 . 61 . 1Closed D1D2–A-DNA-duplex–ADP-BeFx9 . 61 . 2Closed D1D2–B-DNA-duplex–ADP-BeFx11 . 40 . 6Open D1D2–A10-RNA–ADP-BeFx11 . 02 . 2Closed D1D2–A10-RNA–CDP-BeFx11 . 12 . 2Closed D1D2–A10-RNA–GDP-BeFx11 . 22 . 3Closed D1D2–A10-RNA–UDP-BeFx11 . 41 . 0Open D1D2–A10-DNA–ADP-BeFx11 . 40 . 6Open*Parameters are quoted for the peak containing protein as determined by A214 . Additional fluorescence anisotropy assays show that a closed-state complex with A10-RNA forms at significantly lower concentrations of ADP-BeFx compared to AMP-PNP ( Kd = 0 . 022 and 0 . 12 mM , respectively; Figure 2C ) . This indicates a more stable closed state and accounts for the higher unwinding activity observed for ADP-BeFx compared to AMP-PNP for several DEAD-box proteins ( Liu et al . , 2008 ) . Further , neither ADP nor ADP + Pi in large excess led to the formation of a stable closed state in our assays ( Figure 2C ) , suggesting that the effective concentration of the ATP γ-phosphate is critical for the stability of the closed-state . This finding explains energetically why ATP hydrolysis leads to core re-opening and enzyme turnover in DEAD-box proteins ( Henn et al . , 2010; Cao et al . , 2011 ) and perhaps other SF1 and SF2 helicases . Together , our results show the unwinding efficiency of Mss116 with different nucleotides is directly correlated with the stability of the post-unwound closed-state complex . To investigate the structural basis for the difference in stability of the closed state with different NTP analogs , we determined crystal structures of the closed-state helicase core with A10-RNA and either ADP-BeFx , CDP-BeFx , GDP-BeFx , or UDP-BeFx at 2 . 2 , 2 . 7 , 2 . 4 , and 3 . 2 Å resolution , respectively ( Figure 3 and Table 2 ) . These structures show that the ATP-binding motifs I and VI make similar direct contacts to the phosphate groups of all four NTP analogs ( Figure 3C ) . Motif II ( DEAD ) is positioned identically in all structures and interacts indirectly via waters with the BeF3 moiety , which corresponds to the ATP γ-phosphate ( Figure 3B ) . However , each base interacts differently in the ATP-binding pocket . The purine bases ( A and G ) are stacked optimally with F126 in the Q-motif , which primarily confers ATP specificity in DEAD-box proteins ( Linder and Jankowsky , 2011 ) , whereas the pyrimidine bases ( C and U ) adopt a less favorable stacking orientation with this residue ( Figure 3B ) . Also , fewer direct contacts are made to the C , G , and U bases than to A ( Figure 3C ) . In particular , compared to the closed-state structure with ADP-BeFx , two hydrogen ( H ) -bonds from G128 and Q133 in the Q-motif to the base are absent in the complex with GDP-BeFx , and all of the direct interactions of the Q-motif with the base are missing in the structures with CDP- or UDP-BeFx . The fewer contacts of all other bases relative to adenine and the less favorable stacking of pyrimidine bases in the ATP-binding pocket explain the relative stabilities of the closed-state complexes and reveal how the helicase core of Mss116 adapted to unwind RNA most efficiently using ATP . 10 . 7554/eLife . 04630 . 009Figure 3 . The structural basis for the ATP specificity of the helicase core of Mss116 . ( A ) Crystal structures of the closed-state helicase core of Mss116 bound to ssRNA and different nucleotide analogues ( D1D2–A10-RNA–NDP-BeFx for N = A , C , G , or U ) . Structures are colored according to the scheme in Figure 1C . ( B ) Comparisons of the protein–substrate interactions in the NDP-BeFx binding pockets of the structures shown in ( A ) . Side chains that make direct contacts with the NDP are shown as ball and stick models . A 2Fo − Fc electron density map contoured at 1 . 0 σ for the NDP-BeFx ligand is shown in gray . Mg2+ ions and water molecules are shown as green and red spheres , respectively , and the atoms of BeF3 are shown in purple ( Be ) and yellow ( F ) . Motif II ( ‘DEAD’ ) makes indirect contacts via water molecules to the BeF3 moiety , which corresponds to the γ-phosphate of ATP . ( C ) Schematics of direct NDP–protein interactions for the structures shown in ( A ) . See also Table 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 00910 . 7554/eLife . 04630 . 010Table 2 . Crystallographic data and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 010ComplexD1D2–A10-RNA–ADP-BeFxD1D2–A10-RNA–CDP-BeFxD1D2–A10-RNA–GDP-BeFxD1D2–A10-RNA–UDP-BeFxD1D2–A10-DNA–ADP-BeFxData collection Space groupP21212P21212P21212P21212P21212 Unit cell a , b , c ( Å ) 89 . 83 , 126 . 26 , 55 . 5589 . 64 , 126 . 84 , 55 . 0389 . 99 , 126 . 61 , 55 . 5589 . 76 , 126 . 51 , 55 . 5190 . 39 , 126 . 19 , 55 . 23 α , β , γ ( ° ) 90 , 90 , 9090 , 90 , 9090 , 90 , 9090 , 90 , 9090 , 90 , 90 Wavelength ( Å ) 1 . 00001 . 00001 . 00001 . 00001 . 0000 Total reflections222 , 375129 , 478565 , 72986 , 580 Unique reflections32 , 64216 , 98227 , 14810 , 51413 , 111 Resolution* ( Å ) 50 − 2 . 20 ( 2 . 24 − 2 . 20 ) 50 − 2 . 60 ( 2 . 64 − 2 . 60 ) 50 − 2 . 35 ( 2 . 39 − 2 . 35 ) 50 − 3 . 30 ( 3 . 36 − 3 . 30 ) 50 − 3 . 00 ( 3 . 05 − 3 . 00 ) Redundancy6 . 8 ( 5 . 4 ) 6 . 1 ( 5 . 4 ) 19 . 2 ( 14 . 1 ) 8 . 2 ( 8 . 1 ) 5 . 5 ( 4 . 9 ) Completeness ( % ) 99 . 4 ( 97 . 7 ) 99 . 7 ( 98 . 3 ) 99 . 5 ( 94 . 9 ) 99 . 5 ( 95 . 0 ) 96 . 7 ( 88 . 8 ) Overall I/σ ( I ) 19 . 0 ( 1 . 5 ) 12 . 1 ( 1 . 5 ) 26 . 4 ( 2 . 4 ) 11 . 1 ( 2 . 5 ) 7 . 1 ( 1 . 5 ) Rmerge† ( % ) 9 . 7 ( 60 . 3 ) 13 . 8 ( 77 . 0 ) 13 . 3 ( 99 . 7 ) 19 . 8 ( 66 . 5 ) 19 . 8 ( 61 . 2 ) Refinement Resolution ( Å ) 47 . 24 − 2 . 2047 . 07 − 2 . 7447 . 28 − 2 . 3547 . 22 − 3 . 2144 . 15 − 2 . 96 No . of reflections32 , 64216 , 98227 , 14610 , 51413 , 111 Rwork ( % ) 21 . 622 . 323 . 0922 . 1619 . 6 Rfree§ ( % ) 25 . 426 . 726 . 0027 . 4324 . 4 No . atoms Protein79117519772375287774 Nucleic acid232298230232147 Ligands4542434044 Water115286100 Rmsd bonds ( Å ) 0 . 0030 . 0030 . 0030 . 0040 . 003 Rmsd angles ( ° ) 0 . 6960 . 6290 . 7100 . 9680 . 751 Ramachandran favored# ( % ) 97 . 2396 . 0196 . 8498 . 4097 . 04 Ramachandran allowed ( % ) 2 . 301 . 942 . 191 . 401 . 98 PDB ID4TYW4TYY4TZ04TZ64TYN*The numbers in parentheses refer to the highest resolution shell . †Rmerge = ∑hkl ∑i |Ihkl , i − 〈Ihkl〉|/∑hkl ∑〈Ihkl〉 . §Rfree was calculated with 5% of reflections that were excluded from refinement . #Analysis by MolProbity ( Chen et al . , 2010 ) . D2 of Mss116 ( residues 342–597 ) functions as an RNA-duplex recognition domain in the open-state enzyme ( Mallam et al . , 2012 ) ( Figure 1D ) . To determine how Mss116 specifies for dsRNA , we first examined how 12-bp RNA and DNA duplexes of different geometries ( Figure 4 ) interact with D2 in the absence of nucleotide . EMSAs using fluorescein amidite ( FAM ) -labeled duplexes show that D2 has surprisingly similar affinities for dsRNA ( K1/2 = 400 nM ) and A-DNA and B-DNA duplexes ( K1/2 = 410 and 510 nM , respectively ) ( Figure 5A and Figure 5—figure supplement 1A–C ) . Circular dichroism ( CD ) measurements confirmed that the geometry of the A-DNA and B-DNA duplexes is maintained upon binding to D2 , and that binding does not induce a B- to A-form transition ( Figure 4E and Figure 4—figure supplement 1 ) . The B-DNA duplex also competitively displaces dsRNA bound to D2 ( Ki = 1700 nM ) ( Figure 5B ) . These results indicate that D2 can bind dsRNA and dsDNA of A- or B-form geometry in the dsRNA binding pocket even with the different spacing of the backbone phosphate groups ( Mallam et al . , 2012 ) . Our findings are consistent with recent studies showing that several DEAD-box proteins can interact with dsDNA ( Kammel et al . , 2013; Tuteja et al . , 2014 ) . D2 of Mss116 is therefore a general and flexible nucleic acid duplex binding domain . 10 . 7554/eLife . 04630 . 011Figure 4 . Model nucleic acid substrates . ( A–C ) 12-bp model substrates of ( A ) dsRNA ( yellow ) ; ( B ) A-DNA duplex ( pink ) ; and ( C ) B-DNA duplex ( red ) . The duplex geometry of the DNA substrates has been previously characterized in solution by CD measurements ( Basham et al . , 1995 ) and X-ray crystallography ( Verdaguer et al . , 1991 ) . The duplexes are predicted to have similar stabilities ( predicted melting temperatures are 61 . 0°C , 59 . 4°C , and 63 . 9°C for the dsRNA , A-DNA , and B-DNA duplexes , respectively [Owczarzy et al . , 2008] ) . ( D ) CD spectra of A-DNA ( pink ) and B-DNA ( red ) duplexes , which are consistent with previously reported spectra of identical duplexes ( Basham et al . , 1995; Kypr et al . , 2009 ) . The CD-spectrum of the A-DNA duplex has a characteristic strong positive peak at 260 nm and negative peaks at 240 and 210 nm ( Ivanov et al . , 1973 ) . The B-DNA duplex is characterized by a positive peak at 260–280 nm and a negative peak at ∼245 nm ( Kypr et al . , 2009 ) . ( E ) CD spectra of the B-DNA duplex ( 100 μM ) in the absence ( solid red line ) and presence ( dashed black line ) of D2 ( 120 μM ) . Spectra are shown in units of molar circular dichroism ( Δε ) and are background subtracted for the presence of protein . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 01110 . 7554/eLife . 04630 . 012Figure 4—figure supplement 1 . CD spectra of A-DNA duplex ( 80 μM ) in the absence ( solid pink line ) and presence ( dashed black line ) of MBP-D2 ( 100 μM ) . Spectra are shown in units of molar circular dichroism ( Δε ) and are background subtracted for the presence of protein . The characteristic strong positive and negative peaks in the CD-spectrum of the A-DNA duplex at 260 nm and 240 nm , respectively , remain in the presence of protein . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 01210 . 7554/eLife . 04630 . 013Figure 5 . The biochemical basis for the RNA specificity of the helicase core of Mss116 . ( A ) Equilibrium binding of duplex substrates to MBP-tagged D2 in the absence of nucleotide determined by EMSA . ( B ) Competitive displacement from MBP-tagged D2 of 5′ FAM-B-DNA duplex ( 250 nM ) by unlabeled dsRNA ( 0–6 μM , yellow , Ki = 860 ± 40 nM ) and of 5′ FAM-dsRNA ( 250 nM ) by unlabeled B-DNA duplex ( 0–6 μM , red , Ki = 1700 ± 200 nM ) . ( C ) Unwinding of duplex substrates by the MBP-tagged helicase core measured under equilibrium conditions by using a gel-based fluorescence assay to monitor the formation of a closed-state complex at increasing concentrations of ADP-BeFx ( see also Figure 2—figure supplement 1 ) . NU , no appreciable unwinding . ( D ) Equilibrium binding of A10-DNA to the MBP-tagged helicase core determined by fluorescence anisotropy measurements at increasing concentrations of ADP-BeFx . The binding of A10-RNA under the same conditions is shown for comparison ( taken from Figure 2B ) . In ( A–D ) , data were normalized using the signal obtained from the fit to the appropriate model outlined in the ‘Materials and methods’ . ( E ) Normalized SEC profiles monitored by the absorbance at 260 nm ( red ) and 280 nm ( black ) for the helicase core in the absence of substrates ( top ) and in the presence of either A10-RNA + ADP-BeFx ( middle ) and A10-DNA + ADP-BeFx ( bottom ) . An A260/A280 >1 at the maximum absorbance indicates the formation of a stable closed-state complex ( Table 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 01310 . 7554/eLife . 04630 . 014Figure 5—figure supplement 1 . EMSA binding assays of model duplexes . Representative binding assays of Mss116 MBP-D2 ( 0–6 μM ) to a 5′ FAM-labeled 12-bp duplex substrate ( 100 nM ) for ( A ) dsRNA; ( B ) an A-DNA duplex; and ( C ) a B-DNA duplex . Samples were loaded in the reaction medium and resolved in a non-denaturing 6% polyacrylamide gel run at 4°C in 0 . 5× Tris/Borate/EDTA buffer ( pH 8 . 3 ) . Arrows mark the positions of free and bound duplex substrate . Proteins have an N-terminal MBP tag to increase solubility under the EMSA conditions . The binding data show cooperativity for all duplex substrates ( Hill coefficients are 2 . 2 , 1 . 5 , and 1 . 6 for the dsRNA , A-DNA , and B-DNA duplexes , respectively , Figure 5A ) , which suggests that multiple molecules of D2 can bind to a single duplex substrate . The second band corresponding to bound substrate seen in some lanes could also be indicative of this . ( D ) Representative binding assays of Mss116 MBP-D1 ( 0–6 μM ) to a 5′ FAM-B-DNA duplex ( 100 nM ) to demonstrate minimal binding of D1 to the B-DNA duplex under these experimental conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 01410 . 7554/eLife . 04630 . 015Figure 5—figure supplement 2 . Duplex unwinding measured by using a gel-based fluorescence assay to monitor the formation of a closed-state complex containing bound ssRNA or ssDNA . Unwinding was probed by using the duplex substrates shown in Figure 4A–C , which were labeled with fluorophore ( FAM ) and quencher ( IBFQ ) probes at the 5′ and 3′ ends , respectively ( IDT ) . A change in fluorescence of these substrates occurs when the helicase core unwinds the duplex and forms a closed-state bound to a single-stranded region of RNA ( Figure 2—figure supplement 1A ) . ( A–C ) Representative unwinding assays for ( A ) dsRNA; ( B ) A-DNA; and ( C ) B-DNA duplexes by the MBP-tagged helicase core ( 2 μM ) measured at increasing concentrations of ADP-BeFx ( 0–4 mM ) , as described in Figure 2—figure supplement 1A . Arrows mark the complexes corresponding to the open and closed state protein bound to nucleic acid . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 01510 . 7554/eLife . 04630 . 016Figure 5—figure supplement 3 . Kinetic assay of unwinding of duplex substrates by ATP . Kinetic unwinding profiles of dsRNA , A-DNA , and B-DNA duplexes catalyzed by D1D2 ( 2 μM ) and ATP ( 5 mM ) . Error bars represent the standard error for at least three independent measurements , and the error in k1 represents the standard error of the non-linear regression . NU , no appreciable unwinding . Data were normalized using the parameters obtained from the fit to a first-order reaction with a single exponential . In the case of B-DNA when no unwinding was observed , data were normalized using a signal for a fully unwound duplex . This was obtained unwinding and re-annealing a control sample containing the same amount of labeled and unlabeled duplex by heating to 94°C for 3 min and cooling to room temperature on the bench . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 01610 . 7554/eLife . 04630 . 017Figure 5—figure supplement 4 . Characterization of the helicase core in the absence and presence of duplex substrates using size-exclusion chromatography . SEC was performed using a Superdex 75 10/300 GL column ( GE Healthcare ) and a BioLogic DuoFlow chromatography system ( Bio-Rad ) in a buffer of 20 mM Tris–HCl ( pH 7 . 5 ) , 200 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 . Complexes were assembled as outlined in the ‘Materials and methods’ and SEC data were measured using absorbance at 260 nm ( red ) and 280 nm ( black ) . Example elution profiles are shown for D1D2 in the absence of substrates; in the presence of dsRNA only; in the presence of dsRNA and ADP-BeFx; in the presence of A-DNA-duplex and ADP-BeFx; and in the presence of B-DNA duplex and ADP-BeFx . The ratio of A260/A280 , which is approximately 0 . 5 for free protein and >1 for protein–nucleic acid complexes , was used as an indicator of the formation of a closed-state complex that contains nucleic acid . However , the smaller elution volume at maximum A260 for dsRNA–D1D2–ADP-BeFx and A-DNA duplex-D1D2–ADP-BeFx ( Ve = 9 . 6 ml for both ) suggests the formation of some higher-order closed-state complexes , possibly with two protein molecules bound on either side of a partially unwound duplex . This is consistent with the cooperativity in duplex unwinding reactions previously observed for Mss116 ( Halls et al . , 2007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 017 We next examined the ability of Mss116 to unwind the same RNA and DNA model duplexes in the presence of increasing concentrations of ADP-BeFx ( Figure 5C and Figure 5—figure supplement 2 ) . Equilibrium duplex unwinding assays ( Figure 2—figure supplement 1A ) show that Mss116 can unwind dsRNA and an A-DNA duplex , although a lower concentration of ADP-BeFx is required to unwind dsRNA ( K1/2 = 0 . 14 and 0 . 25 mM , respectively ) . Notably , we did not observe any appreciable unwinding of the B-DNA duplex under these conditions ( Figure 5C and Figure 5—figure supplement 2 ) . In this case , kinetic unwinding assays demonstrate the same trend . They show that Mss116 can unwind dsRNA and the A-DNA duplex in the presence of ATP with observed first-order rate constants ( k1 ) of 0 . 46 and 0 . 15 min−1 , respectively , but does not unwind the B-DNA duplex ( Figure 5—figure supplement 3 ) . Similarly , analytical SEC showed elution profiles for D1D2 that are consistent with closed-state complexes when measured with ADP-BeFx and dsRNA or the A-DNA duplex but not the B-DNA duplex ( Table 1 and Figure 5—figure supplement 4 ) . These data indicate that Mss116 selectively unwinds A-form duplex nucleic acids . Further , contrary to what was previously thought ( Fairman-Williams et al . , 2010 ) , they demonstrate that a DEAD-box protein can unwind an all DNA duplex in a nucleotide-dependent manner if it has A-form geometry . Although D2 can bind a B-DNA duplex , a closed-state complex does not readily form with B-form DNA and unwinding of this substrate does not occur . To further investigate why Mss116 preferentially unwinds RNA duplexes , we compared the characteristics of the closed-state helicase core with equivalent ssRNA ( A10-RNA ) and ssDNA ( A10-DNA ) substrates . Equilibrium fluorescence anisotropy assays in the presence of increasing concentrations of ADP-BeFx indicate that the closed-state complex forms with both substrates , but at a much lower concentration of ADP-BeFx for ssRNA than for ssDNA ( Kd = 0 . 022 and 0 . 79 mM , respectively; Figure 5D ) . SEC data also demonstrate that a closed-state complex with A10-RNA and ADP-BeFx remains intact during elution , whereas an identical complex with A10-DNA dissociates on the SEC column ( Figure 5E and Table 1 ) . Thus , the closed-state core is significantly more stable and long-lived with ssRNA than with ssDNA . To probe the structural basis for the difference in stability of the closed-state complex with ssRNA compared to ssDNA , we determined crystal structures of the closed-state helicase core with ADP-BeFx and either A10-RNA or A10-DNA at 2 . 5 and 2 . 9 Å resolutions , respectively ( Figure 6 and Table 2 ) . These structures confirm that Mss116 can form the same closed-state complex with ssRNA and ssDNA and allow a direct comparison of the interactions made by these substrates with the same helicase core . The structures show that trajectories of the bound ssRNA and ssDNA are very similar ( Figure 6B ) and that most of the interactions between the conserved nucleic acid binding motifs IV–V and the phosphate backbone are identical in both complexes ( Figure 6C ) . However , the closed-state complex with ssRNA contains protein contacts to RNA 2′-OH groups that are not present in the closed-state complex with ssDNA . These include four from residues in motifs Ia and Ic in D1 that form during core closure and account for the higher stability of the closed-state with ssRNA ( Figure 5D , E ) . 10 . 7554/eLife . 04630 . 018Figure 6 . The structural basis for the RNA specificity of the helicase core of Mss116 . ( A ) Closed state crystal structures of the helicase core of Mss116 with the ATP analogue ADP-BeFx and A10-RNA ( yellow ) or A10-DNA ( red ) . The helicase core is colored as in Figure 1C . ( B ) A comparison of the binding trajectory of equivalent nucleotides of A10-RNA ( yellow ) and A10-DNA ( red ) bound in the closed state . ( C ) A schematic comparing the interactions of A10-RNA ( yellow ) and A10-DNA ( red ) with the closed-state helicase core , colored blue and green to D1 and D2 , respectively . Interactions unique to each structure are colored black . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 018
Collectively our results elucidate the basis for the physiological preference of the DEAD-box protein Mss116 for ATP and RNA , but also show that the helicase core has a surprising degree of substrate ambiguity . This is a consequence of the ability of conserved helicase motifs to interact with the phosphate groups of different NTPs or nucleic acids and promote the formation of the same closed-state complex ( Figure 3A and Figure 6A ) . The preference of Mss116 for ATP is dictated by optimal base-stacking and H-bonding interactions between the Q-motif and adenine base ( Figure 3B , C ) . However , interactions between conserved motifs I , II , and VI and nucleotide phosphate moieties are sufficient to promote duplex unwinding at lower efficiency irrespective of the nucleotide base ( Figure 2A and Figure 3B , C ) . The specificity of Mss116 for unwinding RNA duplexes is dictated by both A-form geometry ( Figure 5C ) and interactions by motifs Ia and Ic in D1 with 2′-OH groups of ssRNA in the closed state ( Figure 5D and Figure 6C ) . Additionally , Mss116 belongs to a subclass of DEAD-box proteins that has a CTE appended to D2 ( Figure 1B ) ( Mohr et al . , 2008 ) . This CTE makes additional 2′-OH contacts to dsRNA in the open state ( Mallam et al . , 2012 ) that may favor its binding to D2 ( Figure 5B ) . Nevertheless , the interactions of nucleic acid-binding motifs with the phosphate backbone are sufficient to enable Mss116 to unwind A-form DNA duplexes at lower efficiency ( Figure 5C and Figure 5—figure supplement 3 ) . Mss116 cannot unwind a B-form DNA duplex ( Figure 5C and Figure 5—figure supplement 3 ) , and a model of the closed state with a B-DNA duplex indicates that the helicase motifs in D1 that clash with dsRNA ( Mallam et al . , 2012 ) ( Figure 7A ) are not positioned to catalyze the unwinding of longer , thinner B-form duplexes ( Figure 7B ) . 10 . 7554/eLife . 04630 . 019Figure 7 . Models and crystal structures of closed-state complexes of SF2 helicases . ( A ) Surface representation of closed-state Mss116 with dsRNA modeled in the duplex RNA-binding pocket of D2 . Sterically incompatible regions of D1 are highlighted in red , and these indicate how D1 promotes RNA unwinding upon core closure by disrupting the base pairing in the dsRNA . In particular , helicase motifs Ia , Ib , and Ic and the DEAD-box specific post-II motif in D1 displace one RNA strand and bend the other during RNA duplex unwinding ( Mallam et al . , 2012 ) . ( B ) Surface representation of closed-state Mss116 with a B-DNA duplex , which is longer and thinner than an A-form duplex ( Dickerson et al . , 1982 ) , modeled in the duplex RNA-binding pocket of D2 . There are no appreciable clashes between dsDNA and the core in this model , which suggests why core closure does not promote unwinding of a B-DNA duplex ( Figure 5C and Figure 5—figure supplement 3 ) . ( C ) Closed-state structure of D1-D3 of human RIG-I helicase ( PDB = 3TMI ) bound to dsRNA ( Jiang et al . , 2011 ) . dsRNA is accommodated in the closed-state of RIG-I , which explains how it functions by binding and/or translocating along a duplex RNA substrate ( Myong et al . , 2009; Rawling and Pyle , 2014 ) . ( D ) Closed-state model of Sulfolobus solfataricus Swi2/Snf2 helicase core and a B-DNA duplex adapted from Durr et al . ( 2005 ) . This model suggests that the Swi2/Snf2 helicase core can accommodate a B-form DNA duplex in a closed-state conformation and explains how helicases in this family function by translocating along DNA duplexes ( Figure 1A ) . Proteins and nucleic acids are colored as in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04630 . 019 Importantly , the substrate ambiguity of Mss116 suggests an evolutionary scenario for how SF1 and SF2 helicases diverged from an ancestral helicase core with broad specificity into specialized enzymes . In each case , core closure was retained as a catalytic mechanism using the interactions common to all NTP or nucleic acid substrates predicted from our results . However , the stability of the closed-state was further modulated by family-specific interactions that favor a particular NTP and nucleic acid . Thus , helicase families that display the most substrate ambiguity by utilizing all four NTPs and function on either DNA or RNA ( for example the DEAH/RHA [Tanaka and Schwer , 2005] and NS3/NPH-II [Preugschat et al . , 1996] families; Figure 1A ) may contain a core that functions similarly to that of an ancestral helicase . Helicases that preferentially use ATP maintained the conserved interactions with nucleotide phosphate groups but acquired additional interactions with the adenine base that further stabilize the closed-state complex . Similarly , DEAD-box proteins , which act preferentially on RNA ( Fairman-Williams et al . , 2010 ) , maintained conserved interactions with the nucleic acid backbone but evolved specificity for A-form duplexes and additional stabilizing interactions with RNA 2′-OH groups in the closed state , as demonstrated here for Mss116 . The lack of unwinding activity in some DEAD-box proteins may stem from structural changes in the helicase core that mitigate RNA bending or strand displacement ( Young et al . , 2013 ) . Helicase families that function on DNA ( for example , the Swi/Snf , RecQ-like , and UvrD/Rep families ) could have diversified by the preservation of conserved interactions with the nucleic acid backbone combined with the selection of additional interactions that favor B-form duplexes and/or disfavor nucleic acids with 2′-OH groups . Similar inferences can be made from our data about the evolution of distinct mechanisms in SF1 and SF2 families ( Figure 1A ) . We propose that although core-closure was retained as a mode of catalysis , the differences in the stability of the closed-state complex between helicase families allowed the diversification of the observed helicase mechanism . Thus , the localized unwinding mechanism used by DEAD-box proteins ( Yang et al . , 2007 ) likely evolved by the selection of a helicase core that is able to ‘clamp’ ssRNA and form a highly stable closed-state complex ( Figure 5D , E ) . This mode of interaction compensates for the energy cost to locally unwind an RNA duplex , which is critical for DEAD-box protein function ( Del Campo et al . , 2009 ) . In comparison , helicase cores that diverged to form less stable , more transient closed states with ssRNA or ssDNA would favor a mechanism that involved loading and translocating along a single strand ( for example , NS3/NPH-II and RecQ-like helicases; Figure 1A ) . Our data also demonstrate that the stability of the closed state depends upon interactions with nucleotides as well as nucleic acids ( Figure 2 ) . The DEAH family of helicases are a potential example of a case where a sequence change in motif II compared to DEAD-box proteins ( ‘DEAH’ instead of ‘DEAD’ ) might result in a weaker interaction with the ATP γ-phosphate and favor the observed switch from localized to translocation-based unwinding ( Figure 1A ) . More generally , ATP-dependent core closure to form a ternary complex with nucleic acid may have evolved from tighter to weaker binding as the helicase mechanism concurrently evolved from localized to translocation-based . This is in addition to structural features , such as extra terminal domains or β-hairpins within the helicase core , which favor translocation-based unwinding in some helicase families ( Fairman-Williams et al . , 2010 ) . Protein cofactors may also play a role in helicase substrate specificity , as illustrated for the DEAD-box protein Rok1 , whose cofactor Rrp5 increases the specificity of the helicase core 10-fold for a pre-rRNA duplex ( Young et al . , 2013 ) . Finally , other SF2 helicases have evolved to optimally accommodate dsRNA ( e . g . , RIG-I ) or dsDNA ( e . g . , Sulfolobus solfataricus Swi2/Snf2 ) in a closed state complex and translocate with no observable unwinding ( Figure 1A , Figure 7C–D ) ( Durr et al . , 2005; Myong et al . , 2009; Jiang et al . , 2011 ) . In these cases , subtle changes in the closed-state core , perhaps combined with additional flanking domains , enable the helicase to bind duplex nucleic acid without the need to overcome the energetic barrier to unwinding and lead to this distinct mechanism of action . It has been hypothesized that during evolution , progenitor enzymes of low activity and broad specificity diverge into families of more potent and highly specialized enzymes ( Jensen , 1976; Khersonsky and Tawfik , 2010 ) . Taken together , our findings suggest how a progenitor helicase core that had broad specificity and used conserved motifs to recognize the phosphate groups of NTPs and the backbone of nucleic acids diverged to present day SF1 and SF2 helicases with different cellular functions .
Unlabeled self-complementary RNA or DNA oligonucleotides ( Integrated DNA Technologies , IDT , Coralville , IO; Figure 4A–C ) were annealed to form 12-bp RNA or DNA duplexes by heating solutions at 6 mM single strands in 100 mM potassium acetate , 30 mM HEPES ( pH 7 . 5 ) at 94°C for 1 min and then slowly cooling to room temperature over 1 hr . Labeled duplexes for unwinding and binding assays were annealed similarly at 200 μM single strands . Sequences for 12-bp dsDNA substrates were chosen based upon previous studies which indicated that they adopt either A-form or B-form geometry ( Basham et al . , 1995; Kypr et al . , 2009 ) . We further characterized these substrates by using circular dichroism ( CD ) to confirm that they retained the required duplex geometry under our experimental conditions in the absence and presence of protein ( Figure 4D , E and Figure 4—figure supplement 1 ) . The helicase core of Mss116 ( D1D2 ) and separate domains D1 and D2 were expressed as N-terminal MalE fusions in Escherichia coli Rosetta 2 ( EMD Biosciences , Germany ) , grown in ZYP-5052 auto-inducing medium for 24 hr at 22°C , and purified at 4°C , as described ( Del Campo and Lambowitz , 2009; Mallam et al . , 2011 , 2012 ) . Proteins for binding and unwinding assays were exchanged into a storage buffer of 20 mM Tris–HCl ( pH 7 . 5 ) , 200 mM KCl , 1 mM dithiothreitol ( DTT ) , 10% glycerol during a final SEC purification step . D1D2 for crystallization was dialyzed into 10 mM Tris–HCl ( pH 7 . 5 ) , 250 mM NaCl , 1 mM DTT , 50 mM arginine + glutamine , 50% glycerol . All proteins were stored at −80°C before use . Equilibrium unwinding of 12-bp dsRNA , A-form DNA , and B-form DNA duplexes was measured in increasing concentrations of NDP-BeFx ( N = A , C , G , or U ) using a gel-based fluorescence assay to monitor the formation of a closed-state complex containing a bound single-stranded substrate . Duplexes were labeled with a fluorescent probe ( FAM ) and quencher ( Iowa Black FQ ) at the 5′ and 3′ ends , respectively . These substrates gave a change in fluorescence upon unwinding and formation of a closed state ( Figure 2—figure supplement 1A ) . NDP-BeFx ( N = A , C , G , or U ) was prepared as described ( Del Campo and Lambowitz , 2009 ) . Measurements were performed using MBP-tagged D1D2 to increase protein solubility under the experimental conditions . MBP-D1D2 ( 2 μM ) was incubated with the appropriate duplex substrate ( 100 nM ) and increasing concentrations of NDP-BeFx-Mg2+ ( ranging from 0 to 20 mM ) at 22°C for at least 1 hr in a reaction medium containing 20 mM Tris–HCl ( pH 7 . 5 ) , 100 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 , and 0 . 1 mg/ml of bovine serum albumin . The protein concentration was chosen so that all of the duplex substrate is bound in the open state at equilibrium ( Figure 5A ) . Samples were analyzed in a non-denaturing 6% polyacrylamide gel run at 4°C for 60 min . The fluorescence signal of the bound duplex substrate was quantified by using a Typhoon imager ( GE Healthcare , UK ) to measure the formation of a closed-state complex containing a single-stranded nucleic acid region , indicating duplex unwinding ( Figure 2—figure supplement 1 ) . The apparent fraction of unwound duplex at increasing concentrations of NDP-BeFx was quantified by using ImageJ and fit to a one-site binding model to estimate the concentration of nucleotide at the midpoint ( K1/2 ) of the unwinding reaction . In all cases , equilibrium was verified by additional assays for samples that were incubated for extended times ( up to approximately 4 hr ) , which gave the same unwinding profiles as those incubated for 1 hr . Kinetic-unwinding assays of 12-bp dsRNA , A-form DNA , and B-form DNA duplexes by the helicase core were performed with the same fluorophore–quencher labeled probes ( Figure 2—figure supplement 1A ) in the presence of 5 mM NTP ( N = A , C , G , or U ) . In these assays , a change in the fluorescence of the labeled duplex was seen upon unwinding and subsequent re-annealing to form a duplex with an unlabeled strand of the same sequence without a quencher present in excess ( Figure 2—figure supplement 2 ) . Annealing of these duplexes occurs within the dead time of mixing at the concentration of substrates used in these experiments . D1D2 ( 2 μM ) was mixed with NTP-Mg2+ ( 5 mM ) , labeled duplex ( 125 nM ) , and unlabeled duplex ( 500 nM ) at 22°C in a reaction medium containing 20 mM Tris–HCl ( pH 7 . 5 ) , 100 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 . Reactions were terminated at appropriate time points with 1 volume of stop buffer ( 50 mM EDTA , 1% SDS , 10% glycerol ) and run in a non-denaturing 20% polyacrylamide at 22°C for 60 min . The fluorescence signal of duplex substrate was quantified by using a Typhoon imager ( GE Healthcare ) to measure the extent of unwinding/re-annealing . The apparent fraction of unwound duplex at various time points was quantified by using ImageJ and ( where appropriate ) fit to a first-order reaction to estimate an observed first-order rate constant ( k1 ) . Equilibrium binding of A10-RNA and A10-DNA to D1D2 in increasing concentrations of NDP-BeFx was measured by fluorescence anisotropy using MBP-tagged protein to increase the change in anisotropy upon binding . 5′ FAM-labeled A10-RNA or A10-DNA ( 10 nM; IDT ) was incubated with protein ( 2 μM ) and increasing concentrations of NDP-BeFx ( N = A , C , G , or U; 0 to 10 mM ) at 22°C for at least 1 hr in a reaction medium containing 20 mM Tris–HCl ( pH 7 . 5 ) , 100 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 , and 0 . 1 mg/ml of bovine serum albumin . The observed fluorescence anisotropy at increasing concentrations of protein was measured by using an EnVision Microplate Reader ( Perkin Elmer , Waltham , MA ) and was fit to a one-site binding model with a Hill coefficient to estimate the Kd of single-stranded nucleic acid in the presence of increasing nucleotide . Equilibrium was verified by carrying out assays on samples incubated for extended times up to 4 hr , which gave the same binding profiles as those incubated for 1 hr . Equivalent experiments were performed to measure the binding of A10-RNA to D1D2 in increasing concentrations of AMP-PNP or ADP ( 0–10 mM ) and ADP + Pi ( 0–100 mM Pi in the presence of 10 mM ADP ) . Equilibrium binding of 12-bp RNA ( A-form ) and DNA ( A-form and B-form ) duplexes to D1 or D2 was measured by EMSA using MBP-tagged proteins to increase protein solubility as described ( Mallam et al . , 2012 ) . 5′ FAM-labeled 12-bp duplexes ( 100 nM; IDT; Figure 4A–C ) were incubated with increasing concentrations of protein ( 0–6 μM ) at 22°C for at least 1 hr in a reaction medium containing 20 mM Tris–HCl ( pH 7 . 5 ) , 100 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 , and 0 . 1 mg/ml of bovine serum albumin to stabilize the protein at low concentrations . Samples were then analyzed in a non-denaturing 6% polyacrylamide gel run at 4°C for 60 min , and the fluorescence signal of the bound duplex substrate was quantified by using a Typhoon imager . The fraction of bound duplex with increasing concentrations of MBP-tagged protein was quantified by using ImageJ and fit to a one-site binding model with a Hill coefficient to estimate a Kd . Competition assays were performed similarly by measuring the competitive displacement from MBP-D2 ( 500 nM ) of 5′ FAM-B-DNA duplex ( 250 nM ) by unlabeled dsRNA ( 0–6 μM , Ki = 860 ± 40 nM ) and of 5′ FAM-dsRNA ( 250 nM ) by unlabeled B-DNA duplex ( 0–6 μM , Ki = 1700 ± 200 nM ) . In these cases , the fraction of free substrate was quantified and a Ki was estimated from a one-site binding model . Binding of nucleotide and nucleic acid substrates to D1D2 was examined by size-exclusion chromatography . The helicase core of Mss116 does not contain tryptophan residues and its calculated extinction coefficient is small ( ε280 = 18 , 255 M−1 cm−1; ExPASy Proteomics Server ProtParam tool [Wilkins et al . , 1999] ) . The formation of a closed-state complex in the presence of nucleic acid and NDP-BeFx therefore gives rise to a large change in A260 compared to A280 . Protein samples ( 10 μM ) were incubated at 22°C for 30 min in NDP-BeFx-Mg2+ ( 5 mM , N = A , C , G , or U ) and single-stranded ( A10-RNA or A10-DNA; 20 μM ) or duplex ( dsRNA , A-DNA duplex or B-DNA duplex; 10 μM ) nucleic acid and loaded onto a Superdex 75 column ( GE Healthcare ) pre-equilibrated in a buffer containing 20 mM Tris–HCl ( pH 7 . 5 ) , 200 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 . The absorbance and elution volume of the protein complexes above the background signal of the buffer were measured at 260 and 280 nm ( Table 1 ) . Control samples of protein alone , substrate alone , or protein and either nucleotide or nucleic acid were also measured; closed-state complexes were not detected in these cases . All measurements were performed in 20 mM Tris–HCl ( pH 7 . 5 ) , 100 mM KCl , 10% glycerol , 1 mM DTT , 5 mM MgCl2 buffer using a thermostatically controlled 0 . 01-cm path-length cuvette at 25°C and a Jasco J-815 spectrometer ( Jasco Inc . , Easton , MD ) . Scans were taken between 200 and 325 nm at a scan rate of 0 . 5 nm s−1 with 30 accumulations . Measurements were made on samples of SEC-purified A-form DNA or B-form DNA duplexes ( 100 μM ) in the absence or presence of Mss116 D2 or MBP-D2 ( 120 μM ) . For the D1D2–A10-RNA–NDP-BeFx complexes , protein ( ∼350 μM ) was incubated with A10-RNA ( 600 μM ) , NDP-BeFx-Mg2+ ( 5 mM; N = A , C , G , or U ) and MgCl2 ( 1 mM ) for 30 min on the desktop . Sitting drops were assembled using 0 . 5 μl of complex and 0 . 5 μl of a well solution of 0 . 1 M HEPES , pH 7 . 5 , 2% tacsimate , pH 7 . 0 , 20% PEG ( polyethylene glycol ) 3350 for D1D2–A10-RNA–ADP-BeFx; 0 . 2 M sodium malonate , pH 5 . 0 , 20% PEG 3350 for D1D2–A10-RNA–CDP-BeFx; 4% tacsimate , pH 8 . 0 , 12% PEG 3350 for D1D2–A10-RNA–GDP-BeFx; and 0 . 1 M DL-malic acid , pH 7 . 0 , 12% PEG 3350 for D1D2–A10-RNA–UDP-BeFx ( Hampton Research , Aliso Viejo , CA ) . Drops were stored at 22°C and plate-like crystals appeared within 1–2 weeks . Crystals were removed from sitting drops and flash cooled immediately in liquid N2 . Crystals of D1D2–A10-DNA–ADP-BeFx were obtained similarly and drops were assembled with a well solution of 0 . 2 M ammonium acetate , 20% PEG 3350 . X-ray diffraction data were collected at the Advanced Light Source ( ALS ) , Lawrence Berkeley National Laboratory ( mail-in service on beamlines 5 . 0 . 2 or 5 . 0 . 3; wavelength = 1 . 00003 Å ) . Details of data collection and refinement are in Table 2 . Diffraction intensities were indexed , integrated , and scaled with HKL-2000 ( Otwinowski and Minor , 1997 ) . Initial space groups were determined by using Pointless ( Evans , 2006 ) and confirmed by decreases in both Rwork and Rfree after refinement of molecular replacement solutions . Molecular replacement was performed with Phaser ( McCoy et al . , 2007 ) , using the previously determined structure of Mss116 D1D2 in the closed state ( PDB 3I5X ) as a search model . Structures were completed with cycles of manual model building in Coot ( Emsley et al . , 2010 ) and refinement in Phenix ( Adams et al . , 2010 ) . Validation of protein and nucleic acid models and their contacts was done by using MolProbity ( Chen et al . , 2010 ) and indicated that at least 98% of residues are located in the most favorable region of the Ramachandran plot . Structural figures were prepared by using the PyMOL Molecular Graphics System , Version 1 . 4 , Schrödinger , LLC . Coordinates and structure factors were deposited in the Protein Data Bank under accessions 4TYW ( D1D1–A10-RNA–ADP-BeFx ) , 4TYY ( D1D1–A10-RNA–CDP-BeFx ) , 4TZ0 ( D1D1–A10-RNA–GDP-BeFx ) , 4TZ6 ( D1D1–A10-RNA–UDP-BeFx ) , and 4TYN ( D1D1–A10-DNA–ADP-BeFx ) . | Living cells store their genetic material as DNA , which can be copied to make another molecule called RNA . DNA consists of two strands that are wound around each other in a double helix . RNA is made in a similar way to DNA , but it is usually present as a single strand that folds into a three-dimensional structure that is held in shape by regions of the molecule interacting with each other . Before DNA and RNA can perform their essential tasks in cells , enzymes called helicases must separate the interacting strands . A large group of helicases , known as superfamily 1 and 2 , are involved in virtually all aspects of the control of RNA and DNA structure . All of these helicases contain a region called the ‘helicase core’ , but they work in different ways . For example , some move along the DNA or RNA strand whilst they unwind it , while others can unwind RNA without moving . It remains unclear how these helicases have evolved different ways to unwind DNA and RNA structures using the same helicase core . Mallam et al . have now analyzed a helicase from yeast called Mss116 , which belongs to superfamily 2 . It is known from previous work that Mss116 binds to many different RNA molecules and—unlike most other helicases—it does not require any extra proteins to help . This makes it an ideal model to study the properties of a helicase core on its own . Helicases use the energy released from breaking down molecules called nucleotides to pull apart the bonds that hold DNA and RNA strands together . The experiments found that for Mss116 , a nucleotide called ATP is the best for providing the energy needed to unwind RNA but other nucleotides can work less efficiently . The experiments also show that in addition to RNA , Ms116 is able to unwind double-stranded DNA molecules that have a certain shape . Using a technique called X-ray crystallography , Mallam et al . observed the structure of the Mss116 core when it is bound to RNA and DNA . While there are some shared points of contact between the helicase and the DNA or RNA , there are more points of contact between Mss116 and RNA than between Mss116 and DNA . Mallam et al . propose that present-day helicases have diversified from enzymes that had broad specificity for RNA and DNA , by optimizing interactions that favor the binding of particular nucleotides and nucleic acids . These changes enabled the helicases to become a versatile set of tools that control the structure of RNA or DNA in different ways . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Molecular insights into RNA and DNA helicase evolution from the determinants of specificity for a DEAD-box RNA helicase |
Alvinocaridid shrimps are emblematic representatives of the deep hydrothermal vent fauna at the Mid-Atlantic Ridge . They are adapted to a mostly aphotic habitat with extreme physicochemical conditions in the vicinity of the hydrothermal fluid emissions . Here , we investigated the brain architecture of the vent shrimp Rimicaris exoculata to understand possible adaptations of its nervous system to the hydrothermal sensory landscape . Its brain is modified from the crustacean brain ground pattern by featuring relatively small visual and olfactory neuropils that contrast with well-developed higher integrative centers , the hemiellipsoid bodies . We propose that these structures in vent shrimps may fulfill functions in addition to higher order sensory processing and suggest a role in place memory . Our study promotes vent shrimps as fascinating models to gain insights into sensory adaptations to peculiar environmental conditions , and the evolutionary transformation of specific brain areas in Crustacea .
The alvinocaridid shrimps were discovered in 1985 during a mission of the deep submersible vehicle ALVIN ( Rona et al . , 1986 ) and are now known to be widely distributed representatives of the deep hydrothermal vent fauna along the Mid-Atlantic Ridge ( MAR; Desbruyères et al . , 2001; Desbruyères et al . , 2000; Gebruk et al . , 1997; Segonzac et al . , 1993 ) . Active vents are dynamic environments , where geothermally heated seawater , the hydrothermal fluid , discharges from chimneys and cracks in the seafloor . At the MAR , vents occur from 850 to 4080 m depth and the pure hydrothermal fluid , which may be up to 350°C , is anoxic , acid , and enriched in potentially toxic minerals and dissolved gases ( Charlou et al . , 2010; Charlou et al . , 2002; Charlou et al . , 2000 ) . Hydrothermal vent habitats , in addition to high hydrostatic pressure and the complete absence of sunlight , are characterized by steep gradients of temperature and concentration of chemicals ( Bates et al . , 2010; Johnson et al . , 1988; Johnson et al . , 1986; Le Bris et al . , 2005 ) . Vent organisms are well adapted to these physicochemical conditions , and alvinocaridid shrimps colonize in high abundance the walls of active chimneys , where the hydrothermal fluid mixes with the surrounding cold ( 4°C ) and oxygenated seawater . Vent ecosystems rely on chemoautotrophic bacteria as primary producers , which convert reduced chemicals through oxidation , thus providing the energy to fix carbon and to produce organic matter that serves as a nutritional basis for primary consumers ( Fisher et al . , 2007; Jannasch and Mottl , 1985; Ponsard et al . , 2013; Van Dover , 2000 ) . The shrimp Rimicaris exoculata ( Williams and Rona , 1986 ) is the most intensely studied vent crustacean due to its high abundance at most sites along the MAR and its singular lifestyle ( Figure 1A , B; Desbruyères et al . , 2001; Gebruk et al . , 1997; Segonzac et al . , 1993; Van Dover et al . , 1988 ) . Specimens of R . exoculata are found from 1600 to 4000 m depth ( Lunina and Vereshchaka , 2014 ) and they form massive aggregations in the vicinity of the chimneys , with up to 3000 ind . m−2 ( Segonzac et al . , 1993 ) . This species is a strict primary consumer , relying on ectosymbiotic bacteria harbored in its enlarged branchial chambers , through a direct nutritional transfer of bacterial carbon products by trans-tegumental absorption ( Corbari et al . , 2008; Petersen et al . , 2010; Ponsard et al . , 2013; Zbinden et al . , 2004 ) . The associated bacterial metabolic activities include oxidation of sulfide , iron , methane and hydrogen , suggesting that R . exoculata symbionts could have both nutritional and detoxifying roles for the shrimp ( Hügler et al . , 2011; Zbinden et al . , 2008 ) . Hence , this species is strictly dependent on hydrothermal fluid emissions to supplement its symbionts with reduced compounds , and might possess specific sensory abilities for this purpose . Because R . exoculata preferentially lives close to the hydrothermal fluids , the shrimp constantly has to cope with steep temperature gradients ranging approximately from 4°C to 40°C ( Cathalot and Rouxel , 2018 ) , and its sensory system might be tuned to efficiently probe this dynamic thermal environment . A fundamental question regarding vent shrimp’s environment and lifestyle is how they detect hydrothermal emissions and further select their microhabitat . Both abiotic and biotic factors are important to determine the animal’s local distribution at hydrothermal vent sites ( Le Bris et al . , 2005; Luther et al . , 2001 ) . Several studies showed that R . exoculata possesses a range of morphological , anatomical and physiological adaptations to the hydrothermal environment , related for instance to ectosymbiosis with bacteria ( Casanova et al . , 1993; Ponsard et al . , 2013; Zbinden et al . , 2004 ) , respiration in hypoxic conditions ( Hourdez and Lallier , 2006; Lallier and Truchot , 1997 ) , or thermal stress ( Cottin et al . , 2010; Ravaux et al . , 2003 ) . However , the sensory mechanisms and adaptations used by the shrimps to perceive their habitat have only been partially investigated ( see references below ) despite their importance in understanding the lifestyle of vent shrimp species and their long-term evolution . Vision and chemoreception have been proposed to be the major sensory modalities used by vent shrimp to perceive environmental cues ( Chamberlain , 2000; Jinks et al . , 1998; Pelli and Chamberlin , 1989; Renninger et al . , 1995 ) . In vent shrimps , the stalked compound eyes that characterize most malacostracan crustaceans are modified to form enlarged sessile eyes , which in R . exoculata are located underneath the dorsal carapace ( Chamberlain , 2000; Gaten et al . , 1998; O'Neill et al . , 1995; Van Dover et al . , 1989 ) . The eyes cannot form images since the ommatidia lack a dioptric apparatus necessary to refract and focus rays of light , but the retina instead consists of hypertrophied rhabdoms and a reflective subjacent layer , structures that maximize the absorption of light . These anatomical features could represent an adaptation to detect very dim light sources . It was suggested that the animals may perceive the black body radiation emitted by the extremely hot fluid which exits the chimney ( Chamberlain , 2000; Pelli and Chamberlin , 1989; Van Dover et al . , 1989 ) . Furthermore , the animal’s antennal appendages respond to sulfide , suggesting that vent shrimps can detect key chemical components of the hydrothermal fluid ( Machon et al . , 2018; Renninger et al . , 1995 ) , but sulfide detection is not restricted to vent shrimps since antennal responses were also recorded from shallow-water palaemonid shrimp ( Machon et al . , 2018 ) . From structural descriptions of their antennae 1 and 2 and chemosensory sensilla , it is not clear whether their chemosensory system presents specific adaptations related to the hydrothermal environment ( Machon et al . , 2018; Zbinden et al . , 2017 ) . One specific feature of these organisms is the dense coverage of their antennal appendages by bacterial communities ( Zbinden et al . , 2018 ) , whose potential roles remain unknown . Nevertheless , their occurrence on the sensory organs suggests a functional significance for the shrimp sensory abilities ( Zbinden et al . , 2018 ) . Crustacean brain structure is best understood in crayfish , crabs , and clawed and spiny lobsters ( reviews for example Derby and Weissburg , 2014b; Harzsch and Krieger , 2018; Schmidt , 2016 ) . We are interested in exploring adaptive changes of crustacean brain structures that have occurred during their evolutionary radiation into particular habitats and their adoption of specific life styles ( e . g . Harzsch et al . , 2011; Kenning and Harzsch , 2013; Krieger et al . , 2015; Krieger et al . , 2012b; Krieger et al . , 2010; Meth et al . , 2017 ) . Differential investment in certain brain neuropils might reflect the sensory landscape which a certain crustacean species typically exploits , so that studying an animal’s brain anatomy may allow for predictions related to its ecology and lifestyle ( Sandeman et al . , 2014 ) . For example , in peracarid and remipedian cave crustaceans , the visual neuropils are absent whereas the central olfactory pathway is well developed , highlighting that these blind animals may rely on olfaction as a major sensory modality in their lightless habitat ( Fanenbruck et al . , 2004; Fanenbruck and Harzsch , 2005; Stegner et al . , 2015; Stemme and Harzsch , 2016 ) . In representatives of the genus Penaeus , the olfactory system is moderately developed , while sophisticated antenna two neuropils are present , suggesting that the detection of hydrodynamic stimuli is important for these animals ( Meth et al . , 2017; Sandeman et al . , 1993 ) . Hence , comparing the architecture of the sensory centers among divergent crustacean lineages , across wide evolutionary distances and across diverse life styles , can help to understand structural adaptations to specific sensory environments ( review in Sandeman et al . , 2014 ) . Studying crustaceans from extreme habitats is particularly informative in this respect ( Ramm and Scholtz , 2017; Stegner et al . , 2015 ) . However , the structure of the brain in vent shrimps remains poorly understood ( Charmantier-Daures and Segonzac , 1998; Gaten et al . , 1998 ) . Therefore , the present study sets out to provide a detailed description of the architecture of the R . exoculata brain against the background of the extreme conditions that characterize its habitat , and to ultimately discuss its contribution for crustacean brain evolution .
The wide cephalothorax of Rimicaris exoculata displays large branchiostegites ( bs ) which surround voluminous gill chambers ( Figure 1C ) . The animals do not possess eyestalks but rather have bilaterally paired , wing-shaped eyes with a conspicuous , whitish retina that is fused in the anterior region to form the ocular plate ( ocp ) . The lateral parts of the eye extend further dorsally and towards the posterior region of the cephalothorax ( Figures 1C and 2A ) . The first pair of antennae ( A1 ) is biramous , with two flagella of similar length ( Figure 1C ) . The second pair of antennae ( A2 ) consists of a basal element , the scaphocerite ( sc ) , and a long uniramous flagellum , slightly wider than those of the antennae 1 ( Figure 1C ) . Micro-CT scans show that the brain is located in the anterior region of the cephalothorax , and receives main sensory afferences from the antenna 1 ( A1Nv ) and antenna 2 ( A2Nv ) nerves anteriorly , from the eye nerves ( ENv ) posterodorsally , from the tegumentary nerves ( TNv ) laterally , and from the oesophageal connectives ( oc ) posteriorly ( Figure 1D ) . Decapod crustacean brains are subdivided into three successive neuromeres: proto- , deuto- and tritocerebrum . In R . exoculata , these regions form a single , medially located mass ( i . e . the median brain ) ( Figures 1D and 2A ) . The visual neuropils are closely associated with the lateral protocerebrum ( lPC ) , at a posterodorsal position ( Figure 2A ) . This arrangement contrasts with other shallow-water carideans and most decapod crustaceans ( see for example Cronin and Porter , 2008; Meth et al . , 2017 ) , in which the lateral protocerebrum is located at some distance from the median brain , in movable eyestalks ( Figure 2B ) . The deutocerebrum ( DC ) is associated with the antenna one nerves , and the tritocerebrum ( TC ) is associated with the antenna two nerves ( Figure 2 ) . The brain’s neuraxis is not aligned with body axis in R . exoculata , but is bent dorsally so that the protocerebrum is situated posterodorsally to the deutocerebrum ( Figure 2A ) . Data from micro-CT scans and aligned serial paraffin sections provided a consistent picture of the brain anatomy that we compiled in both three-dimensional reconstructions ( Figures 3A and 4A–D ) and a schematic drawing of the R . exoculata brain ( Figure 3B , C ) . In the following , for simplicity only one brain hemisphere is described , although mirror symmetrical structures are present in the contralateral hemisphere . R . exoculata presents three successive visual neuropils , which are the lamina ( La ) , medulla ( Me ) , and lobula ( Lo ) , from distal to proximal ( Figures 3 , 4 and 5 ) . The lamina is thin , flattened and elongated dorsally ( Figures 4A , D and 5B-D ) . The cell cluster ( 1 ) dorsally covers the lamina ( Figures 3C , 4B , D and 5C ) . Numerous axon bundles from the entire length of the retina ( ENv ) converge onto the lamina ( Figures 3 , 4A–D and 5D ) . The retina consists of photoreceptor organelles , the rhabdoms ( dR ) ( for which the degradation is ascribed to the damaging exposure to intense light during sampling and manipulation of the specimens at the surface ) , which overlie a white layer of reflecting cells , the tapetum ( T ) , and clusters of pigment cells ( pc ) ( Figure 5D , Figure 11A; Nuckley et al . , 1996; O'Neill et al . , 1995 ) . The medulla is spherical ( Figures 3 , 4A–D and 5B , C ) and is connected by thin fibers to the lamina ( Figure 5C ) , the lobula ( Figure 5B ) and by a dense fibers tract to the terminal medulla ( Figure 5B , white arrowhead ) . The lobula is slightly larger than the medulla , and is adjacent to the posterior side of the terminal medulla ( TM ) ( Figures 3 , 4A–D , F , 5A and 11A ) . The merged cell clusters ( 2 ) and ( 3 ) cannot be clearly separated and cover both the medulla and the lobula ( Figures 3C , 4B , D and 5A–C ) . The lateral protocerebrum dominates the R . exoculata brain , with the hemiellipsoid body ( HN ) , the terminal medulla ( TM ) , together with the cell clusters ( 4 ) and ( 5 ) representing about 25% of the brain volume . The hemiellipsoid body is well defined , with a voluminous , hemispherical cap region ( HNcap ) located dorsally ( Figures 3 , 4 , 5D and 6A–F ) and displaying synapsin-like immunoreactivity ( SYNir ) ( Figure 6G–I ) . The core region of the hemiellipsoid body ( HNcore ) is fused posteriorly with the terminal medulla ( Figures 4F and 6B , D , E ) . The cap and core regions are separated by an arcuate intermediate layer ( IL ) ( Figures 4E and 6A , B , D , E , G–I ) which receives parallel afferent fibers from the terminal medulla anteriorly ( namely the HN-TM tract ) ( Figures 3A , B , 4F and 6A–C , black asterisks ) and a massive bundle of neurites from somata in the cell cluster ( 5 ) at the medial side ( Figures 4E , 5D and 6C–E , white asterisks ) . Some of the intermediate layer fibers display allatostatin-like immunoreactivity ( ASTir ) near the cap region ( Figure 6G ) . The intermediate layer is devoid of SYNir ( Figure 6J ) . The cap region is characterized by synaptic sites forming microglomeruli ( Figure 6H’ ) and is also innervated by serotonergic neurons ( Figure 6H ) . The cell cluster ( 5 ) is voluminous ( Figures 3C , 4B and 6B–E , I , J ) and contains approximately 30 , 000 cell somata of the so-called globuli cells ( Wolff et al . , 2017 ) . The hemiellipsoid body receives input from the olfactory neuropils via the projection neuron tract ( PNT ) in the posterior region ( Figures 3A and 6B , F ) . The terminal medulla is a large and complex neuropil . Anteriorly , it is shaped like a sphere ( Figures 3 , 4A , C , F and 6A–C ) , and it connects to the intermediate layer of the hemiellipsoid body via the HN-TM tract ( Figures 3A , D , 4F and 6A–C , black asterisks ) . Posterior to this region , the terminal medulla is large , crossed by seemingly unstructured networks of fibers ( Figures 3A , 4C , F and 6D–F ) and displays SYNir ( Figure 6I , J ) . It is innervated by neurites from the cell cluster ( 4 ) ( Figure 6D , white arrowhead ) , and further connects again to the intermediate layer of the hemiellipsoid body via radiating fiber bundles ( Figure 6E , black arrowheads ) . The median protocerebrum ( mPC ) comprises two medially fused neuropils , the anterior ( AMPN ) and posterior ( PMPN ) medial protocerebral neuropils . The AMPN connects to the terminal medulla of the lateral protocerebrum anteriorly via the protocerebral tract ( PT ) ( Figures 3B and 6C , D ) and the PMPN via the posterior protocerebral tract ( PPT ) , the latter containing neurites with strong serotonin-immunoreactivity ( 5HTir ) ( Figure 7C ) and seemingly interconnecting the terminal medulla of both hemispheres . Both , the AMPN and PMPN are separated by the unpaired central body neuropil ( CB ) ( Figure 3 ) , which displays ASTir ( Figure 7B ) , weak SYNir ( Figure 7A ) and strong 5HTir ( Figure 7C ) . Overall , the median protocerebrum contains many fibers from serotonergic neurons , partly from the cell cluster ( x ) ( which likely refers to the cell clusters ( 12 , 13 ) and ( 17 ) according to Sandeman et al . , 1992 ) , which define well the elements of the central complex , that is the protocerebral bridge ( PB ) and the central body ( CB ) , and also the posterior region associated to the posterior protocerebral tract . Posteriorly to the central body , fiber bundles of the projection neuron tracts from both hemispheres meet in a region with strong SYNir , that we will call the projection neuron tract central neuropil ( PNTCN ) ( Figure 7A ) . In the deutocerebrum , a paired neuropil with a conspicuous structure is located laterally , the lobe-shaped olfactory neuropil ( ON ) ( Figures 3 , 4A–D , F , 5D and 8A–F ) . It is composed of approximately 180 wedge-shaped neuropil units , the olfactory glomeruli ( og ) , which are arranged radially around the periphery of a non-synaptic core ( Figures 3B , 4F , 5D and 8A–F ) . Each glomerulus shows strong SYNir ( Figure 8E–F ) , as well as ASTir which highlights a subdivision of each glomerulus into a cap ( c ) , subcap ( sbc ) and base ( b ) region ( Figure 8E’ ) . The sensory input of the olfactory neuropil comes from the olfactory sensory neurons innervating the aesthetasc sensilla on the lateral flagellum of the antenna 1 . The somata of olfactory interneurons located in the cell cluster ( 9/11 ) innervate fibers of the olfactory neuropil , some of which display ASTir ( Figure 8H ) . These fibers enter via the medial foramen ( mF ) into the core of the neuropil ( Figure 8C , E ) , from where they target the glomerular base region ( Figure 8E ) , or cross to the lateral foramen ( lF ) ( Figure 8A , C , E ) to spread out laterally and innervate the glomerular cap region ( Figure 8E , white arrowhead ) . The medial foramen is also the place where efferent fibers exit from the olfactory neuropil . These are the axons of the olfactory projection neurons that form the projection neuron tract ( Figure 8B , D ) . A projection neuron tract neuropil ( PNTN ) as known from other decapods ( e . g . Sandeman et al . , 1992; Harzsch and Hansson , 2008; Krieger et al . , 2012a ) is identifiable close to the ascending branch of the tract ( Figure 8E , F ) . The projection neuron tract then transverses the median protocerebrum and projects to the lateral protocerebrum ( see above ) . The lateral antenna one neuropil ( LAN ) is located medially to the olfactory neuropil . It is U-shaped ( Figures 3 and 8A , B , H ) and displays strong SYNir , as well as ASTir , which reveals a transversely stratified pattern ( Figure 8G ) . This neuropil connects posterodorsally to the median protocerebrum ( Figure 8G , H ) . The median antenna one neuropil ( MAN ) is small , poorly defined , and located in the center of the deutocerebrum , below the anterior region of the median protocerebrum and between the paired lateral antenna one neuropils ( Figure 3A , B ) . The tritocerebrum comprises the antenna two neuropil ( AnN ) , which has a cylindrical shape and lies in front of the oesophageal connectives ( Figures 3 and 4 , 8 HA ) . SYNir and ASTir show a transversely stratified pattern within this neuropil ( Figure 8I ) . Poorly differentiated from the antenna two neuropil , the tegumentary neuropil ( TN ) is located posterodorsally ( Figure 3 ) . The organ of Bellonci ( OB ) is typical for many crustaceans but its sensory function remains unclear ( Chaigneau , 1994 ) . In R . exoculata , this organ is conspicuous and comprises onion bodies ( Ob ) structures connected to a well-developed nerve tract ( OBNv ) . The onion bodies are situated on the anterolateral side of the brain , in front of the hemiellipsoid body ( Figures 3 , 4A–D , 5D and 9A ) . They represent a cluster of about fifty densely packed lobules ( Figure 9A , B ) , many of them containing elements of granular appearance ( Figure 9B’ , white arrowhead ) . Some lobules are further located in the proximal region of OBNv ( Figure 9B ) . This nerve is large in its proximal region , and progressively tapers as it draws away from the brain ( Figure 9A ) . Anterodorsally , the nerve extends through the retinal layers and connects underneath the cuticle of the ocular plate ( Figure 9C , C’ ) . The myoarterial formation ( maf ) ( or cor frontale , auxiliary heart ) underlies the dorsal carapace and is located between the paired eyes , above the brain ( Figure 10A , B , C ) . This organ is voluminous , being almost as long as the elongated retina , and extends ventrally towards the dorsal region of the brain ( Figure 10B’ , C’ ) . Two adjacent and parallel muscle bundles ( mafm ) penetrate through the myoarterial formation ( Figure 10A ) and attach to the cuticle via tendons located either anteriorly ( Ta ) or dorsally ( Td ) ( Figure 10B , C ) . Two thinner muscular bundles cross the myoarterial formation in its middle region , perpendicular to the main adjacent muscles , and are attached both to the dorsal and ventral cuticle of the cephalothorax , by secondary dorsal and ventral tendons ( T’d , Figure 10B , C; T’v , Figure 10B’ ) . Anteriorly , below the junction of the ocular plate with the dorsal carapace , the myoarterial formation gives rise to three conspicuous , large cerebral arteries , a central cerebral artery ( CA ) and two ophthalmic arteries ( OA ) , which all make a steep U-turn and extend parallel towards the dorsal side of the brain ( Figure 10B , C , arrowhead ) . In an anterodorsal position , median to the hemiellipsoid bodies , the central cerebral artery divides into three smaller arteries , one median ( CAM ) and two lateral ones ( CAL ) ( Figure 10B’ , C’ ) . The median artery passes over the brain , between the two spherical masses of the lateral protocerebrum , and then divides into two branches , one entering the brain posteriorly , at the level of the median protocerebrum , and a larger one merging with the ventral region of the myoarterial formation . This suggests a loop system wherein part of the hemolymph in the cerebral artery goes back into the myoarterial system . The lateral cerebral arteries are coil-shaped and enter the brain above the insertion of the antenna one nerve to target for instance the olfactory neuropils and the lateral antenna one neuropils in the deutocerebrum . The ophthalmic arteries enter the brain in a posterodorsal position ( Figure 10C’ ) and target the visual neuropils and the lateral protocerebrum . The cerebral vascular system of R . exoculata is considerably developed , with blood vessels supplying all brain neuropils and cell clusters , as in other crustaceans , including large vessels that irrigate the visual neuropils ( Figure 5A , C ) , the deutocerebrum ( Figure 8B ) and the lateral protocerebrum ( Figure 9B , B’ ) . The Azan staining reveals pink-to-purple cerebral arteries that enter the brain ( CA ) ( Figures 5C , D , 8A , B and 10A ) , and orange vessels inside the brain ( V ) ( Figures 5A , C , 8B and 9B ) . Table 1 presents a comparison of aesthetasc and olfactory neuropil characteristics in different taxa of crustaceans . The number of olfactory glomeruli and their unitary volume in R . exoculata fit within the ranges displayed by other decapods . However , this species presents relatively small olfactory neuropils ( excluding the fibrous core ) compared to other species with roughly the same body size , such as the caridean Palaemon elegans and the anomuran Coenobita clypeatus . Yet , other species of about the same body size ( e . g . the crayfish Procambarus clarkii and the isopod Saduria entomon ) possess olfactory neuropils of even smaller volume than those of R . exoculata . The higher integrative centers ( i . e . the hemiellipsoid body and the terminal medulla , associated with cell cluster 5 ) are especially well-developed in R . exoculata in relation to the relative size of the olfactory neuropils , compared to other crustaceans ( Figure 11 ) . As an example , from relative volumes obtained from 3D reconstructions , the higher integrative centers in R . exoculata occupy approximately 25% of the total brain volume , similarly to the caridean shrimp P . elegans ( 22% ) but twice more than in the giant robber crab Birgus latro ( 13% ) . However , the olfactory neuropils of P . elegans and B . latro occupy roughly two and six times more volume than those of R . exoculata , values that represent 4 . 2% , 17% and 2 . 7% of the total brain volume , respectively in these species .
The alvinocaridid shrimp Rimicaris exoculata is an endemic species to hydrothermal vent habitats , well adapted to these deep sea environments with peculiar physicochemical conditions . The present study sets out to gain insights into adaptations to specific features of the vent habitat ( e . g . low ambient light levels and steep variations of chemical concentrations ) . The analysis of the brain architecture in R . exoculata aims to highlight relative investments into certain neuronal subsystems , in relation with the animal’s habitat and lifestyle . The general anatomy of the brain of R . exoculata corresponds in many aspects to the ground pattern of the malacostracan crustacean brain ( Kenning et al . , 2013 ) , including the subdivision into proto- , deuto- and tritocerebrum , the location of main nerves and the presence of distinct cell clusters . However , the brain of R . exoculata also exhibits morphological differences to other malacostracans , especially at the level of the lateral protocerebrum . The organ of Bellonci is especially conspicuous ( also observed by Charmantier-Daures and Segonzac , 1998 ) , but its sensory function remains elusive . In the following , we will focus on the structure of major sensory centers ( i . e . the visual system , the olfactory system and the higher integrative centers ) . We will also discuss the evolution of the hemiellipsoid bodies as higher integrative brain centers , which are substantial in R . exoculata . We will begin our account by addressing the neurovascular system that supplies the brain . In crustaceans , the neurovascular system has been described mainly in crayfish ( Chaves da Silva et al . , 2013; Scholz et al . , 2018 ) , crabs ( McGaw , 2005; McGaw and Reiber , 2002; Sandeman , 1967 ) and spiny lobsters ( Steinacker , 1979 ) ( reviews in McMahon , 2001; Steinacker , 1979; Steinacker , 1978; Wilkens , 1999 ) . The brain and eyes are supplied with hemolymph via the anterior aorta system , which originates antero-medially from the heart and runs between the stomach and the dorsal integument ( Scholz et al . , 2018 ) . Anteriorly , a dilatation of the anterior aorta , the myoarterial formation ( Scholz et al . , 2018; also named the cor frontale in for example McGaw ( 2005 ) ; Steinacker ( 1978 ) which functions as an auxiliary heart , pumps the hemolymph specifically towards the anterior part of the central nervous system . In malacostracan crustaceans , the myoarterial formation above the brain gives rise to a descending cerebral artery , which supplies the median brain , and to two ophthalmic arteries that turn laterally and extend into the eyestalks to supply the visual neuropils ( Chaves da Silva et al . , 2013; McGaw , 2005; Scholz et al . , 2018 ) . In R . exoculata , consistent with the absence of eyestalks , the myoarterial formation and its arteries differ in shape , size and position from those previously described in other malacostracans ( Figure 10 ) . Among the potential corollaries for the pronounced neurovascular system in R . exoculata , one is the more efficient hemolymph pumping to the brain . In crustaceans , the perfusion of the brain is modulated by physiological or environmental factors , such as hypoxia ( Reiber and McMahon , 1998 ) . Because the pure hydrothermal fluid is anoxic , the mixing of the fluid with the surrounding seawater can create hypoxic conditions for vent animals ( Childress and Fisher , 1992; Schmidt et al . , 2008 ) . Known adaptations to hypoxia in vent crustaceans include an hemocyanin with a higher affinity for oxygen compared to shallow-water species ( Chausson et al . , 2004; Lallier and Truchot , 1997; Sanders et al . , 1988 ) . A very pronounced capillary network was also observed in hydrothermal vent alvinellid polychaetes ( Hourdez and Lallier , 2006 ) . Accordingly , the pronounced myoarterial formation and large cerebral arteries in R . exoculata could represent a particularly efficient system for oxygen delivery to the brain to cope with low availability of oxygen . Van Dover et al . ( 1989 ) described the R . exoculata eyes as a pair of large anteriorly fused organs that underlie the transparent dorsal carapace of the cephalothorax and demonstrated the presence of rhodopsin-like visual pigments in high quantity , with a maximum absorption at 500 nm . Subsequent analyses showed that the eyes comprise a smooth cornea located above a dense layer of hypertrophied rhabdoms , under which a white layer of reflective cells , the tapetum , is located and maximizes the absorption of light by the photoreceptors ( Chamberlain , 2000; Jinks et al . , 1998; Nuckley et al . , 1996; O'Neill et al . , 1995 ) . These elements of the retina were all discernible in our histological sections ( Figure 5D ) , although the rhabdoms were strongly degenerated , a process ascribed to the damaging exposure to intense light during sampling ( Herring et al . , 1999; Johnson et al . , 1995 ) . The eyes of R . exoculata lack the dioptric apparatus which characterizes the ommatidia of compounds eye of pelagic and shallow water crustaceans and thus cannot form images ( Chamberlain , 2000; Jinks et al . , 1998; Nuckley et al . , 1996; O'Neill et al . , 1995 ) , but their highly sensitive naked retina seems adapted for the detection of low ambient light levels , to the detriment of spatial resolution ( Chamberlain , 2000; Van Dover et al . , 1989 ) . In malacostracan crustaceans , the visual input from the compound eyes is processed by a suite of retinotopic visual neuropils , usually but not exclusively located within the moveable eyestalks ( Figure 2B ) ( Strausfeld , 2012; Loesel et al . , 2013 ) . The absence of eyestalks of R . exoculata coincides with a strong size reduction and fusion of the visual neuropils with the median brain ( Figure 2A ) . Nevertheless , already Charmantier-Daures and Segonzac ( 1998 ) and Gaten et al . ( 1998 ) differentiated three visual neuropils in R . exoculata , namely the lamina , medulla , and lobula , as in the ground pattern of the Malacostraca . However , in R . exoculata these neuropils are located posterodorsally to the enlarged lateral protocerebrum ( Figures 2A and 4A , C ) . The dorsal expansion of the flattened lamina , that extends in parallel to the retina , may indicate a retinotopic projection of photoreceptor input onto the lamina which could allow the animals to extract directional information from light sources above but this issue must be addressed in future experiments . In the medulla , immunohistochemistry revealed an outer layer ( Figure 6I ) ( which is also faintly visible in histological sections , Figure 5B , C ) , suggesting a subdivision of the medulla into an outer and inner region , as seen in crayfish ( Strausfeld and Nassel , 1981 ) . No such stratification was observed for the lobula ( Figure 5A , B ) , and synapsin immunoreactivity was weak in this most proximal neuropil ( Figure 6J ) , although in malacostracans with well-developed compound eyes , the lobula displays numerous , neurochemically diverse strata ( e . g . Brachyura and Anomura , Harzsch and Hansson , 2008; Krieger et al . , 2012b; Krieger et al . , 2010; Wolff et al . , 2012; Polanska et al . , 2007; Meth et al . , 2017; Strausfeld , 2005 ) . The simplified structure of the lobula , which in other malacostracans plays a role in motion detection ( Strausfeld , 2012 ) , might mirror the inability of the eye to form images . Also , the lobula plate , a fourth visual neuropil present in several malacostracan taxa ( e . g . Bengochea et al . , 2018; Harzsch and Hansson , 2008; Krieger et al . , 2012a; Krieger et al . , 2010; Sztarker et al . , 2009; Meth et al . , 2017; Strausfeld , 2005; Kenning et al . , 2013; Kenning and Harzsch , 2013; Sinakevitch et al . , 2003 ) could not be identified in R . exoculata . The lobula plate has been suggested to mediate optokinetic control , necessary to track moving objects ( e . g . conspecifics , preys , predators ) ( Sztarker et al . , 2005 ) . Such a role is consistent with the loss of the lobula plate in R . exoculata , which lacks the realization of image formation , necessary for tracking moving objects . Many eyeless representatives of Crustacea have partially or totally lost their central visual pathways while adapting to a life under dim-light conditions or complete darkness ( e . g . Ramm and Scholtz , 2017; Stegner et al . , 2015; Elofsson and Hessler , 1990; Stegner and Richter , 2011; Fanenbruck et al . , 2004; Brenneis and Richter , 2010 ) . It is likely that the reduction of these nervous tissues is promoted under the selective pressure of those conditions resulting in a less energy expenditure for organisms living in constant or partial darkness , since eliminating neuronal structures which are no longer useful saves considerable amounts of energy ( Klaus et al . , 2013; Moran et al . , 2015; Niven and Laughlin , 2008 ) . However , the fact that neuronal elements indicative for a functional visual system are present in R . exoculata must mean that there is light to exploit as an environmental cue . Also , the unusual nature of the visual system of R . exoculata suggests that it exploits a specific type of signal . One prominent hypothesis refers to the thermal black body radiation emitted by the hot hydrothermal fluid at the chimney’s exit with a temperature of up to 350°C , which peaks in the infrared but part of its spectrum extends into the visible light ( Pelli and Chamberlin , 1989; Van Dover et al . , 1996; Van Dover et al . , 1988; Van Dover and Fry , 1994 ) . The ability to localize this radiation could serve both to attract the shrimp to optimal areas for supplying its symbionts with vital , reduced compounds of the hydrothermal fluid , and to allow avoidance of scorching fluid ( Van Dover et al . , 1989 ) . Visual cues other than thermal radiation are likely to be also exploited by R . exoculata , related to turbulence , mixing and precipitation , such as chemi- , crystallo- , tribo- and sono-luminescence , for which the emission spectra lie between 450–800 nm ( Reynolds and Lutz , 2001; Tapley et al . , 1999; Van Dover et al . , 1996; Van Dover and Fry , 1994; White , 2000; White et al . , 2002 ) . Two modes of chemoreception , linked to distinct chemosensory pathways , are distinguished in malacostracan crustaceans ( Derby and Weissburg , 2014b; Schmidt and Mellon , 2010 ) : olfaction , which is mediated by the aesthetasc sensilla located on the lateral flagellum of the antenna 1 , and distributed chemoreception , which is mediated by the bimodal chemo- and mechanosensory sensilla located mainly on all antennal appendages , the mouthparts , and the walking appendages ( Garm et al . , 2005; Garm et al . , 2003; Garm and Watling , 2013; Mellon , 2014; Mellon , 2012; Schmidt and Gnatzy , 1984 ) . R . exoculata presents aesthetascs in similar number and dimensions to other caridean representatives ( Table 1 ) , as well as several bimodal sensilla with different morphologies on the antennal appendages ( Zbinden et al . , 2017 ) . Olfaction has been extensively studied in malacostracans ( e . g . Ache , 2002; Derby and Weissburg , 2014b; Schmidt and Mellon , 2010 ) , and the central olfactory pathway has received much attention in crustacean neuroanatomy ( e . g . Blaustein et al . , 1988; Harzsch and Krieger , 2018; Kenning et al . , 2013; Kenning and Harzsch , 2013; Krieger et al . , 2015; Krieger et al . , 2012b; Krieger et al . , 2010; Sandeman et al . , 1992; Schachtner et al . , 2005; Schmidt and Mellon , 2010 ) . The afferent olfactory input from the olfactory sensory neurons innervating the aesthetascs targets the conspicuous olfactory neuropils , which are lobe-shaped and bilaterally arranged in the deutocerebrum ( Figure 8A–F ) . They are composed of spherical or cone-shaped dense synaptic neuropils , namely the olfactory glomeruli , which are radially arranged around the periphery of a core of non-synaptic fibers . The olfactory glomeruli are subdivided into a cap , subcap and base regions in several decapod taxa ( e . g . Harzsch and Krieger , 2018; Schachtner et al . , 2005; Schmidt and Ache , 1997 ) . The glomeruli of R . exoculata appear to conform to this design principle , with an identical subdivision ( Figure 8E’ ) . Although the number of olfactory glomeruli is roughly in the same range to that of its close relative Palaemon elegans of about the same body size , the olfactory neuropils of R . exoculata are relatively small in terms of volume compared to P . elegans ( Table 1 ) . Notably , the olfactory neuropils of R . exoculata are moderately developed compared to other species ( Table 1 and Figure 11 ) . Hence , the dimensions and structural complexity of the olfactory neuropils in R . exoculata do not suggest , judging from comparative brain anatomy , that the loss of the eye’s capacity to form images is compensated by sophisticated olfactory abilities . Efficient olfactory abilities would have been especially relevant to probe the chemical environment of R . exoculata , which is dynamic , with strong concentration variations of hydrothermal fluid chemicals as the hydrothermal fluid dilutes with the surrounding seawater . Sulfide and other chemicals could serve as highly important environmental cues for R . exoculata ( Renninger et al . , 1995; Machon et al . , 2018 ) to locate active edifices as optimal areas to supply its chemoautotrophic symbionts with reduced compounds . However , sulfide detection is likely mediated by distributed chemoreception , or both distributed chemoreception and olfaction , rather than exclusively olfaction , as it can be detected by the flagella of the antenna two which does not bear aesthetascs ( Machon et al . , 2018 ) . Olfaction is also involved in the recognition of conspecifics ( Breithaupt and Thiel , 2011 ) and the localization of sexual partners ( Wyatt , 2014 ) , but there is to date no detailed information on the inter-individual interactions in and out of the swarms of R . exoculata . The detection of chemical cues produced by bacteria could also appear especially relevant since the sensory antennal appendages of vent shrimp are often covered by a dense bacterial layer , whose roles are currently unknown ( Zbinden et al . , 2018 ) . Malacostracan crustaceans display a rich repertoire of complex behavioral patterns related to finding food , shelter and mating partners , kin recognition and brood care , as well as orientation and homing . Decapod crustaceans are also known for complex social interactions such as communal defensive tactics , the occupation of common shelters , cooperative behavior during long-distance , offshore seasonal migration and the establishment of dominance hierarchies ( Breithaupt and Thiel , 2011; Derby and Thiel , 2014a; Duffy and Thiel , 2007; Thiel and Watling , 2015 ) . Because such complex behaviors most likely involve elements of learning and memory , higher integrative brain centers are suggested to provide the neuronal substrate for more sophisticated processing underlying such behaviors ( review in Sandeman et al . , 2014 ) . Such centers receive input exclusively from second or higher order neurons but not from any primary sensory afferents ( i . e . from the peripheral nervous system ) and contain interneurons responding to the stimulation of several different sensory systems . In the malacostracan brain , the ( bilaterally paired ) terminal medulla , hemiellipsoid body , and accessory lobe seem to function as higher integrative centers , all three distinct neuropil areas which display a high level of complexity and are notable for their substantial volume ( Sandeman et al . , 2014 ) . The terminal medulla and the closely associated hemiellipsoid body , are targeted by axons of the olfactory projection neurons as output pathway of the olfactory neuropil and accessory lobe ( where present; reviews Derby and Weissburg , 2014b; Harzsch and Krieger , 2018; Schmidt , 2016 ) . Because of these anatomical relationships , evolutionary ( Sullivan and Beltz , 2004; Sullivan and Beltz , 2001 ) and functional considerations ( Harzsch and Krieger , 2018; Sandeman et al . , 2014; Strausfeld , 2012 ) have focused on possible roles of these centers in higher order olfactory processing . In addition to the olfactory projection neuron axons , the terminal medulla also receives input from the visual neuropils in several malacostracans ( reviewed in Sandeman et al . , 2014 ) . A specific type of local interneuron associated to the medulla terminalis and the hemiellipsoid body are the parasol cells ( Mellon and Alones , 1997; McKinzie et al . , 2003; Mellon et al . , 1992a; Mellon , 2000; Mellon et al . , 1992b ) which respond to olfactory , tactile , and visual stimuli , thus highlighting their role as elements in higher order integration ( Mellon and Alones , 1997; Mellon , 2003; Mellon , 2000; Mellon and Wheeler , 1999 ) . Recent evidence obtained from a brachyuran crab suggests an involvement of the crustacean hemiellipsoid body/terminal medulla complex in memory processes ( Maza et al . , 2016 ) . Furthermore , considering anatomical similarities of the crustacean hemiellipsoid body and insect mushroom body , Wolff et al . ( 2017 ) suggested an involvement in place memory . During the evolutionary elaboration of malacostracan brains , substantial modifications occurred related to the relative proportion of types of input and investment in size of the various higher integrative centers ( Figure 11; Harzsch and Krieger , 2018; Sandeman et al . , 2014 ) . Because the terminal medulla has a highly complex and highly variable structure , being composed of several , partly confluent neuropil lobes with heterogeneous appearance containing both coarse and fine fibers ( e . g . Blaustein et al . , 1988 ) , its architecture so far has not been studied in a comparative context . We will focus in the following on the hemiellipsoid body whose structure is somewhat easier to grasp ( Figure 11 ) . In its simplest form , the hemiellipsoid body consists of a volume of fine neuropil with little texture that is closely associated with the terminal medulla ( Figures 11A , B , C , F and 12 ) . Such a phenotype is for example common in leptostracans , the presumably most basal branch of the Malacostraca ( Figures 11A and 12; Kenning et al . , 2013 ) , but also in representatives of the Dendrobranchiata ( Figures 11B and 12; Meth et al . , 2017; Sullivan and Beltz , 2004 ) , and several Brachyura ( Figures 11F and 12; Krieger et al . , 2010; Krieger et al . , 2012a; Krieger et al . , 2015 ) . Isopoda as representatives of the Peracarida also feature simple , dome-shaped hemiellipsoid bodies ( Figures 11C and 12; Kenning and Harzsch , 2013; Stemme and Harzsch , 2016 ) whereas in Amphipoda ( Ramm and Scholtz , 2017 ) and blind groups of peracarids from relict habitats ( Stegner et al . , 2015 ) , this center is poorly developed and may be entirely missing ( Figure 12 ) . A more complex phenotype features a separation of the hemiellipsoid body into two separated areas , an architecture present for example in the spiny lobsters ( neuropils I and Blaustein et al . , 1988 ) , the crayfish Procambarus clarkii and Orconectes rusticus ( neuropil I and II , Sullivan and Beltz , 2001 ) , and Cherax destructor ( Sullivan and Beltz , 2005 ) ( Figure 12 ) . The clawed lobster Homarus americanus also features two neuropil units , but these are stacked on top of each other as cap and core neuropils separated by an intermediate , non-synaptic layer ( Sullivan and Beltz , 2001 ) ( Figure 12 ) . Additional differences exist between the crayfish and the clawed lobster concerning the areas that are targeted by the axons of the projection neurons . ( Mellon et al . , 1992a; Mellon et al . , 1992a; Sullivan and Beltz , 2005; Sullivan and Beltz , 2001 ) . Hemiellipsoid bodies with a cap/core structure separated by an intermediate layer are also present in the brains of marine ( Krieger et al . , 2012a ) and terrestrial hermit crabs of the taxon Coenobitidae , Coenobita clypeatus ( Harzsch and Hansson , 2008; Polanska et al . , 2012; Wolff et al . , 2012 ) and Birgus latro ( Krieger et al . , 2010 ) . These animals all display a large hemiellipsoid body with a peripheral , dome-shaped cap neuropil enclosing two dome-shaped core neuropil areas Core one and Core 2 ( Figures 11G and 12 ) . Their hemiellipsoid body is associated with several thousands of small , intrinsic neurons ( Harzsch and Hansson , 2008; Krieger et al . , 2010 ) . The cap and core neuropils are separated by intermediate layers formed by the neurites of these intrinsic interneurons and the afferents of the projection neuron tract in a rectilinear arrangement ( Wolff et al . , 2012 ) . In the hemiellipsoid bodies of the stomatopod crustaceans Gonodactylus bredenii ( Sullivan and Beltz , 2004 ) and Neogonodactylus oerstedii ( Wolff et al . , 2017 ) , the cap/core motif is modified such that the cap layer ( termed ‘calyx’ in Wolff et al . , 2017 ) is much thinner than the core neuropil ( Figure 12 ) and that the cluster of intrinsic neurons expands over much of the surface of the cap neuropil . The additional stalked neuropils in the lateral protocerebrum of N . oerstedii ( Wolff et al . , 2017 ) will not be discussed here for simplicity . In Stenopus hispidus ( Stenopodidea ) , the hemiellipsoid body appears very complex in structure , with apparently three distinct lobular neuropils ( Figure 12; Sullivan and Beltz , 2004; Krieger et al . , unpublished ) . The hemiellipsoid body in the caridean species P . elegans and Palaemonetes pugio also presents three lobular neuropils , two of which present a cap layer and one or two core regions , and a third neuropil without clear subdivision ( Figures 11D and 12; Sullivan and Beltz , 2004 ) . The hemiellipsoid body of R . exoculata in many aspects , closely corresponds to the cap/core layout ( Figures 6 , 11E and 12 ) although it is slightly simpler than in Coenobitidae with only one core neuropil , similar to the arrangement observed in H . americanus ( Sullivan and Beltz , 2001 ) ( Figure 12 ) . In summary , the hemiellipsoid body displays more structural variations across the Malacostraca than many other elements of the crustacean brain areas which led Sandeman et al . ( 2014 ) to note that , within the Malacostraca , several different evolutionary trajectories are present to increase their brain’s capacity for integrating olfactory and multimodal stimuli . This diversity masks common motifs of hemiellipsoid body architecture , explaining why genealogical relationships of the crustacean and insect protocerebral multimodal centers have been discussed controversially for many years ( reviews e . g . Loesel et al . , 2013; Sandeman et al . , 2014; Strausfeld , 2012; Strausfeld , 2009; Strausfeld , 1998 ) . Recent evidence suggests that , despite many morphological differences , these protocerebral structures of insects and crustaceans nevertheless share common architectural , physiological and neurochemical features suggesting a homology of their very basic neuronal circuitry ( Brown and Wolff , 2012; Maza et al . , 2016; Wolff et al . , 2012; Wolff et al . , 2017; Wolff and Strausfeld , 2015 ) . Because the projection neuron tract provides a massive input to the lateral protocerebrum , recent comparative considerations have suggested that the structural elaboration and size of hemiellipsoid bodies largely mirror the importance of the central olfactory pathway in a given brain , thus emphasizing their role in higher order olfactory processing ( Harzsch and Krieger , 2018; Sandeman et al . , 2014 ) . Along these lines , Harzsch and Hansson ( 2008 ) and Krieger et al . ( 2010 ) noted that in representatives of the Coenobitidae , the architectural complexity and volume of the olfactory neuropil closely correlates to that of the hemiellipsoid body . The comparative plates ( Figures 11 and 12 ) demonstrate that R . exoculata dramatically deviates from this pattern in that their disproportionally large hemiellipsoid body contrasts with inconspicuous and moderately developed olfactory neuropils . The observation that visual input is likely to also play a subordinate role in these animals compared to shallow-water relatives with fully developed compound eyes makes us suggest that in R . exoculata , their impressive hemiellipsoid body may fulfill functions in addition of higher order sensory processing . Although the size alone does not qualify to be better performing ( Chittka and Niven , 2009 ) and caution must be taken to conclude about functional differences from differences in size of structures ( Striedter , 2005 ) , comparative anatomical studies may lead to functional hypotheses . Discussing anatomical similarities of the crustacean hemiellipsoid body and insect mushroom body , Wolff et al . ( 2017 ) suggested for these two neuropils a role in place memory , based on observations that insects with elaborate navigational skills display elaborate mushroom bodies . Considering recent experiments that suggest an involvement of the crustacean hemiellipsoid body/terminal medulla-complex in memory processes ( Maza et al . , 2016 ) , we here propose that the hemiellipsoid body in R . exoculata is involved in the formation of place memory . This hypothesis is further supported by the presence of serotonergic tracts within the hemiellipsoid bodies ( Figure 6H ) , since serotonin has a function for place memory and learning in the mushroom bodies of Drosophila melanogaster ( e . g . Sitaraman et al . , 2008 ) . Spiny lobsters Panulirus argus are renowned for their extensive offshore migrations and their ability to orient accurately towards their home sites over long distances by using the direction of water movement ( surge ) caused by wave action , learned local structural features , and geomagnetic cues for navigation ( reviewed in Sandeman et al . , 2014 ) . Using a GPS-based telemetric system , giant robber crabs , Birgus latro , were shown to form route memories and may use path integration as navigation strategy and in translocation experiments were shown to be capable of homing over large distances ( Krieger et al . , 2012b ) . The above mentioned crustacean species display hemiellipsoid bodies impressive in size or structure . For survival in the extreme , lightless habitat of R . exoculata , an excellent place memory may be essential for avoiding the dangerously hot vent chimneys and memorizing emission sites of hydrothermal fluids rich in those chemicals on which their endosymbiont bacteria depend . Our observations of the general brain architecture of R . exoculata highlight several unusual characteristics , which could be related to adaptations to the specific sensory landscape of the vent habitat . The well-developed neurovascular system could be particularly efficient for brain oxygenation , to cope with the low availability of oxygen in the close surroundings of active chimneys . The conservation of the visual pathway and neuropils in a mostly aphotic environment suggests that vision nevertheless is a relevant sense for vent shrimp . The olfactory system does not present unusual traits and olfaction is probably not a dominant sensory modality in this shrimp unlike what has been proposed so far ( Renninger et al . , 1995 ) . On the other hand , the higher integrative centers are well-developed . The hemiellipsoid bodies are disproportionally large relative to the visual and olfactory neuropils size , and could be involved in complex integrative processes such as place memory . Overall , vent shrimp appear to be especially interesting models to investigate both sensory adaptations to extreme environmental conditions , and the evolution of the sensory centers among Crustacea .
Specimens of alvinocaridid shrimp R . exoculata ( Williams and Rona , 1986 ) were collected on the TAG vent site ( MAR , 26°08’N-44°49’W , 3600 m depth ) during the BICOSE 2018 cruise on the Research Vessel ‘Pourquoi Pas ? ’ . Animals were sampled with the suction device of the Diving Support Vessel ‘Nautile 6000’ , and recovered at their in situ pressure using the PERISCOP isobaric recovery device ( Shillito et al . , 2008 ) . Immediately after retrieval , specimens were dissected to remove the hepatopancreas prior to fixation . The specimens for histology and x-ray micro-computed tomography ( micro-CT ) scans were stored in Bouin’s fixative ( 10% formaldehyde , 5% glacial acetic acid in saturated aqueous picrinic acid ) at 4°C until use . The specimens for immunohistochemistry were fixed 24 to 48 hr in 4% formaldehyde ( FA ) in 0 . 1 M Phosphate buffered saline ( PBS ) at 4°C for 24 hr , and then stored in 0 . 1 M PBS with NaN3 at 4°C until use . All specimens were sexed using the sexual dimorphism from the second pair of pleopods . Specimens are females for all micro-CT and histology experiments , and are both females and males for immunohistochemistry . Caridean shallow water shrimp Palaemon elegans ( Rathke , 1837 ) were collected from Saint-Malo Bay ( France; 48°64’N , −2°00’W ) , in January 2018 , using a shrimp hand net . Specimens were dissected and fixed as described above . Protocols for other species are described in the following: Nebalia herbstii , Kenning et al . ( 2013 ) ; Penaeus vannamei , Meth et al . ( 2017 ) ; Saduria entomon , Kenning and Harzsch ( 2013 ) ; Carcinus maenas , Krieger et al . ( 2012a ) ; Birgus latro , Krieger et al . ( 2010 ) . The heads of Bouin-fixed animals ( six specimens ) were dehydrated in a graded series of ethanol and embedded in paraffin wax mixed with 5% beeswax . Serial sections ( 7 µm ) were taken in the frontal or sagittal plane with a microtome ( Leica RM 2145; Leica Microsystems , Wetzlar , Germany ) . The sections were stained with Azan-novum according to Geidies ( 1954 ) using standard protocols ( Welsch and Mulisch , 2010 ) . The brains of fixed animals ( 4% FA; five specimens ) were dissected in PBS 0 . 1 M , pH 7 . 4 , embedded in low-gelling agarose ( Cat . A9414; Sigma-Aldrich Chemie GmbH , Munich , Germany ) and sectioned ( 100 µm ) with a vibratome ( Hyrax V50; Carl Zeiss , Oberkochen , Germany ) . The sections were preincubated for 1 . 5 hr in PBT ( PBS + 0 . 3% Triton X-100 +1% bovine serum albumine ) to improve antibody penetration . Two sets of combinations of markers were used: 1 . anti-synapsin +anti-allatostatin+nuclear marker; 2 . anti-synapsin +anti-serotonin+nuclear marker . The sections were first incubated overnight in the primary antisera at room temperature . The antisera used were: monoclonal anti-SYNORF1 synapsin antibody ( DSHB , 3C11; from mouse; 1:10 dilution; RRID: AB_2313867 ) ; polyclonal anti-A-allatostatin antiserum ( A-type Dip-allatostatin I; Jena Bioscience , abd-062; from rabbit; 1:1000 dilution; RRID: AB_2314318 ) ; polyclonal anti-Serotonin ( 5-HT , Immunostar , Cat . No 20080 , from rabbit , igG; 1:1000 dilution; RRID: AB_572263 ) . After incubation , the sections were washed in several changes of PBT for 1 hr and afterwards incubated in the secondary antibodies ( anti IgGs ) conjugated to Alexa Fluor 488 ( Alexa Fluor 488 goat anti-rabbit IgG Antibody , Invitrogen , Thermo Fisher Scientific; Waltham , MA , USA; RRID: AB_10374301 ) and Cy3 ( Cy3-conjugated AffiniPure Goat Anti-Mouse IgG Antibody , Jackson ImmunoResearch Laboratories Inc . ; West Grove , PA , USA; RRID: AB_2338000 ) overnight at room temperature . Additionally , HOECHST 33258 ( Cat . 14530; Sigma-Aldrich Chemie GmbH , Munich , Germany ) was used as a nuclear marker to show the cell clusters . The sections were finally washed in several changes of PBT for 2 hr and mounted in Mowiol 4–88 ( Cat . 0713 . 2; Carl Roth , Karlsruhe , Germany ) . The brain tissues processed for immunofluorescence were viewed with a Leica TCS SP5II confocal laser-scanning microscope equipped with DPSS , Diode- and Argon-lasers and operated by the Leica ‘Application Suite Advanced Fluorescence’ software package ( LASAF ) ( Leica Microsystems , Wetzlar , Germany ) . Digital images were processed with Adobe Photoshop CS4 or ImageJ . Only global picture enhancement features ( brightness and contrast ) were used . The head tissues processed for histology were viewed with a Nikon Eclipse 90i upright microscope and bright-field optics ( Nikon , Amstelveen , Netherlands ) . Serial images using a mounted digital camera ( Nikon DS-Fi3 ) were aligned manually with the 3D-reconstruction software Amira 5 . 6 . 0 ( FEI Visualization Science Group , Burlington , VT , USA; RRID: SCR_007353 ) . For frontal and sagittal sections , dorsal is always towards the top . In the figures , the following color-coded abbreviations were used to identify the markers: SYN , synapsin ( magenta ) ; AstA , allatostatin ( green ) ; 5HT , serotonin ( green ) ; NUC , nuclear counter stain ( cyan ) . Colors were chosen according to Color Universal Design for accessibility to colorblind readers . Micro-CT scans were performed using an X-ray microscope ( Xradia MicroXCT-200; Carl Zeiss Microscopy GmbH , Jena , Germany ) that uses a 90-kV/8 W tungsten X-ray source and switchable scintillator-objective lens units as described by Sombke et al . ( 2015 ) . The heads of fixed animals ( Bouin; two specimens ) were contrasted in iodine solution ( 2% iodine resublimated ( Cat . #X864 . 1; Carl Roth GmbH , Karlsruhe , Germany ) in 99 . 5% ethanol ) , critical point-dried using a fully automatic critical point dryer Leica EM CPD300 ( Leica Microsystems , Wetzlar , Germany ) and scanned dry ( scan medium air ) . Tomography projections were reconstructed using the reconstruction software XMReconstructor ( Carl Zeiss Microscopy GmbH , Jena , Germany ) , resulting in image stacks ( DICOM format ) with a pixel size of about 5 . 8 µm for the 4 × objective and 1 . 9 µm for the 10 × objective . The 3D reconstructions of brain and substructures are based on manual segmentation based on image stacks obtained either by the micro-CT scans or by the alignment of serial histological sections , and were performed using the software Amira ( FEI Visualization Science Group , Burlington , VT , USA ) as described in Sombke et al . ( 2015 ) . The computed 3D surfaces were slightly smoothed . The neuroanatomical nomenclature used in this manuscript for neuropils , clusters of cell bodies and tracts is based on Sandeman et al . ( 1993 ) and Richter et al . ( 2010 ) with some modifications adopted from Krieger et al . ( 2015 ) and Loesel et al . ( 2013 ) . The term ‘visual neuropils’ is used instead of ‘optic neuropils’ as suggested by Krieger et al . ( 2015 ) . The terms lamina , medulla and lobula are used for the visual neuropils instead of the lamina ganglionaris , medulla externa and medulla interna ( Harzsch , 2002 ) . The term ‘olfactory neuropil’ refers to the deutocerebral chemosensory lobe in Loesel et al . ( 2013 ) and Krieger et al . ( 2015 ) . The olfactory globular tract is named the projection neuron tract ( PNT ) according to Loesel et al . ( 2013 ) . Cell clusters are referred by their given numbers in parentheses . Because no border was detectable between the cell clusters ( 9 ) and ( 11 ) , they are collectively referred as cluster ( 9/11 ) ( Krieger et al . , 2015 ) , and accordingly are the cell clusters ( 2 ) and ( 3 ) , referred as cluster ( 2/3 ) . ( x ) refers likely to the fusion of the cell clusters ( 12 , 13 ) and ( 17 ) according to the nomenclature from Sandeman et al . ( 1992 ) . For the volumes of the HN-TM ( hemiellipsoid body and terminal medulla complex in both hemispheres ) and the olfactory neuropils relative to the total brain volume , measurements were made from 3D reconstructions of the relevant structures from micro-CT scans using the Amira software . For the calculation of total brain volume of P . elegans , the volume of the tissues connecting the lateral protocerebrum with the central brain in the eye peduncles was omitted , because by omitting the neurites connecting both brain regions , the total brain volume is better comparable with that of R . exoculata . Same calculations were applied on the Birgus latro data from Krieger et al . ( 2010 ) . The number of globuli cells ( i . e . cell somata in the cell cluster ( 5 ) ) was determined by estimation of the globuli cell densities in the cell cluster ( 5 ) , and the total volume of one cell cluster ( 5 ) . The globuli cell densities were estimated by direct counting of the somata within 0 . 02 to 0 . 04 mm² paraffin sections of 0 . 007 mm thickness ( 1 . 3 × 10−4 to 2 . 8 × 10−4 mm3 ) , with a density estimated to be approximately 1 . 3 × 106 globui cells per mm3 . The total volume of one cell cluster ( 5 ) was calculated from 3D-reconstructions with the Amira software . For the volume of the olfactory neuropils and the number of olfactory glomeruli , measurements and estimations were made from sections revealed by synapsin immunoreactivity as described in Beltz et al . ( 2003 ) . The morphological raw data of our contribution have been deposited in the online repository MorphDBase ( https://www . morphdbase . de/ ) under the following accession numbers : J_Machon_20190502 M-3 . 1 J_Machon_20190502 M-4 . 1 J_Machon_20190502 M-5 . 1 J_Machon_20190502 M-6 . 1 J_Machon_20190502 M-7 . 1 | Oceanic vents are areas where hot gases and liquids emerge from cracks and chimneys on the seafloor . These fluids can be as hot as 350°C and are rich in potentially toxic chemicals . Nevertheless , they are the key energy source of many animals that make the vents their home . Vents can be found thousands of meters under sea level , where no sunlight penetrates , so the animals living there must use senses other than vision . As an example , the vent shrimp Rimicaris exoculata , which is used as a vent model animal , was thought to orient itself by sensing chemicals in the vents through their sense of smell . Machon et al . investigate whether vent shrimps possess particular abilities to detect the chemical landscape of the hydrothermal environment , and describe the brain structure and associated sensory systems of R . exoculata . Since the brain in these shrimps is subdivided into regions devoted to different functions , if one of their senses were used more than the others the region devoted to this sense should be bigger or structurally different . When the anatomy of the brain centers in R . exoculata was compared to that of its shallow-water relatives , there was no suggestion that the vent shrimps had an advanced ability to sense chemicals . Rather , a striking feature of the brain of the vent shrimps is the volume and structure of their higher brain centers , which integrate all of their sensory information . It is possible that these regions are also involved in other brain functions as well , since they take up an especially high proportion of the brain . Machon et al . found similarities between R . exoculata and other crustaceans that have sophisticated navigation skills so they hypothesize that integrative brain centers in vent shrimps could play a role in place memory . The findings provide new insights for biologists studying animals associated with deep hydrothermal vents and are also important for neuroscientists interested in brain function and evolution . Future studies should focus on senses of the vent shrimp other than smell to ultimately understand the lifestyle and long-term survival of vent animals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Neuroanatomy of a hydrothermal vent shrimp provides insights into the evolution of crustacean integrative brain centers |
NRF2 is emerging as a major regulator of cellular metabolism . However , most studies have been performed in cancer cells , where co-occurring mutations and tumor selective pressures complicate the influence of NRF2 on metabolism . Here we use genetically engineered , non-transformed primary murine cells to isolate the most immediate effects of NRF2 on cellular metabolism . We find that NRF2 promotes the accumulation of intracellular cysteine and engages the cysteine homeostatic control mechanism mediated by cysteine dioxygenase 1 ( CDO1 ) , which catalyzes the irreversible metabolism of cysteine to cysteine sulfinic acid ( CSA ) . Notably , CDO1 is preferentially silenced by promoter methylation in human non-small cell lung cancers ( NSCLC ) harboring mutations in KEAP1 , the negative regulator of NRF2 . CDO1 silencing promotes proliferation of NSCLC by limiting the futile metabolism of cysteine to the wasteful and toxic byproducts CSA and sulfite ( SO32- ) , and depletion of cellular NADPH . Thus , CDO1 is a metabolic liability for NSCLC cells with high intracellular cysteine , particularly NRF2/KEAP1 mutant cells .
NRF2 ( Nuclear factor-erythroid 2 p45-related factor two or NFE2L2 ) is a stress-responsive cap’n’collar ( CNC ) basic region leucine zipper ( bZIP ) transcription factor that directs various transcriptional programs in response to oxidative stress . Under basal conditions , NRF2 is kept inactive through binding to its negative regulator KEAP1 ( Kelch-like ECH-associated protein ) , which is a redox-regulated substrate adaptor for the Cullin ( Cul ) 3-RING-box protein ( Rbx ) 1 ubiquitin ligase complex that directs NRF2 for degradation ( Kobayashi et al . , 2004 ) . KEAP1 is the major repressor of NRF2 in most cell types , which is supported by the evidence that disruption of Keap1 in the mouse increased the abundance and activity of Nrf2 ( Wakabayashi et al . , 2003 ) . NRF2 plays a critical role in tumor initiation and progression in response to oncogenic signaling and stress ( DeNicola et al . , 2011; Todoric et al . , 2017 ) . Further , NRF2 and KEAP1 mutations are common in many cancers and lead to impaired NRF2 degradation and constitutive NRF2 accumulation ( Ohta et al . , 2008; Shibata et al . , 2008 ) , thereby promoting glutathione ( GSH ) synthesis , detoxification of reactive oxygen species ( ROS ) and proliferation . While the role of NRF2 in ROS detoxification is well established , novel roles of NRF2 in the regulation of cellular metabolism have been recently identified . NRF2 promotes the activity of the pentose phosphate pathway to support the production of NADPH and nucleotides ( Mitsuishi et al . , 2012; Singh et al . , 2013 ) . Further , NRF2 promotes serine biosynthesis to support GSH and nucleotide production ( DeNicola et al . , 2015 ) . These metabolic programs support cell proliferation and tumor growth but not all metabolic consequences of NRF2 activation are favorable . Although uptake of cystine ( CYS ) 2 via the xCT antiporter ( system xc- ) promotes GSH synthesis and antioxidant defense ( Sasaki et al . , 2002 ) , it also induces glutamate export and limits glutamate for cellular processes ( Sayin et al . , 2017 ) . NRF2 suppresses fatty acid synthesis to conserve NADPH , which may antagonize proliferation ( Wu et al . , 2011 ) . Importantly , the activity of these metabolic pathways may be influenced by co-occurring mutations found in the model systems used for study , such as LKB1 mutations , which commonly co-occur with KEAP1 mutations and influence NADPH levels ( Jeon et al . , 2012; Skoulidis et al . , 2015 ) . Further , NRF2 directs distinct transcriptional programs under basal and stress-inducible conditions ( Malhotra et al . , 2010 ) , complicating the interpretation of its effects on cellular metabolism . To examine the immediate consequence of constitutive NRF2 stabilization on cellular metabolism in non-transformed cells , we generated a genetically engineered mouse model expressing the KEAP1R554Q loss-of-function mutation found in human lung cancer . Using this model , we have examined the control of cellular metabolism by NRF2 in mouse embryonic fibroblasts ( MEFs ) and find that NRF2 promotes the accumulation of intracellular cysteine ( CYS ) and sulfur-containing metabolites , including GSH and the intermediates of the taurine ( TAU ) biosynthesis pathway cysteine sulfinic acid ( CSA ) and hypotaurine ( HTAU ) . Entry of CYS into the TAU synthesis pathway was mediated by cysteine dioxygenase 1 ( CDO1 ) , which was elevated in KEAP1R554Q MEFs . TAU synthesis is initiated by the irreversible metabolism of CYS by CDO1 to CSA , which is then decarboxylated by cysteine sulfinic acid decarboxylase ( CSAD ) to HTAU . In turn , HTAU is non-enzymatically converted to TAU , or CSA is transaminated by the cytosolic aspartate aminotransferase ( GOT1 ) to produce β-sulfinyl pyruvate , which spontaneously decomposes to pyruvate and sulfite ( SO32- ) . At the organismal level , decarboxylation of CSA via CSAD predominates over transamination by GOT1 ( Weinstein et al . , 1988 ) . By contrast , lung cancer cell lines accumulated significant CYS due to epigenetic silencing of the CDO1 locus . CDO1 re-expression antagonized proliferation and promoted the metabolism of CYS to CSA , but surprisingly most CSA was exported from cells or transaminated to produce toxic SO32- . Further , continual ( CYS ) 2 reduction to replenish the CYS pool impaired NADPH-dependent cellular processes . These results demonstrate that CDO1 antagonizes the proliferation of lung cancer cells with high intracellular CYS and its expression is selected against during tumor evolution .
To evaluate how constitutive NRF2 activity reprograms metabolism , we generated a genetically engineered , conditional knock-in mouse model of the cancer mutation KEAP1R554Q ( Figure 1A ) . Mutations at this residue prevent the association of KEAP1 with NRF2 , thereby stabilizing NRF2 and inducing the expression of NRF2 target genes ( Hast et al . , 2014 ) . We inserted a loxP-flanked wild-type Keap1 cDNA upstream of the R554Q mutation in exon four in the endogenous Keap1 gene . Prior to exposure to Cre recombinase , wild-type Keap1 protein is expressed . Following Cre-mediated excision of the loxP-flanked cargo , mutant Keap1R554Q is expressed at physiological levels , thus recapitulating the genetic events of human NSCLC and allowing for the interrogation of the consequences of Keap1R554Q expression in an isogenic system . Mouse embryonic fibroblasts ( MEFs ) harboring this allele were derived to evaluate the consequence of Keap1R554Q expression in primary cells . The expression of homozygous Keap1R554Q led to Nrf2 accumulation and increased expression of the Nrf2 target Nqo1 ( Figure 1B ) . We performed non-targeted metabolomics to identify metabolite alterations in these cells and found that the most abundant metabolites following Nrf2 accumulation are sulfur-containing metabolites derived from CYS ( Figure 1C ) , while infection of wild-type MEFs with adenoviral Cre did not significantly alter metabolite levels ( Figure 1—figure supplement 1A ) . To interrogate cysteine metabolism in more detail , we performed targeted metabolomics to quantify the concentration of intracellular CYS and its downstream metabolites ( Figure 1—figure supplement 1B ) . As expected , Nrf2 promoted an increase in intracellular CYS and its downstream metabolite GSH ( Figure 1D–F ) , consistent with previous observations that NRF2 promotes the uptake of ( CYS ) 2 and the synthesis of GSH ( Sasaki et al . , 2002; Wild et al . , 1999 ) . Surprisingly , we also observed a significant increase in intermediates of the TAU biosynthesis pathway , including CSA and HTAU ( Figure 1G–I ) . Importantly , HTAU is a highly abundant metabolite and the increase of HTAU was similar to the increase of GSH in the millimolar range ( Figure 1F , H ) , suggesting that entry into the TAU biosynthesis pathway may represent a significant percentage of total CYS usage . Collectively , these results indicate that NRF2 promotes the accumulation of intracellular cysteine and entry of cysteine into multiple downstream pathways . The significant accumulation of intracellular CYS and TAU synthesis intermediates led us to hypothesize that Nrf2 promotes the accumulation of Cdo1 protein , which is stabilized following CYS accumulation due to a loss its ubiquitination and degradation ( Dominy et al . , 2006 ) . We observed a robust increase in Cdo1 protein in Keap1R554Q MEFs compared to Keap1WT MEFs in the absence of an increase in mRNA expression ( Figure 2A , B ) , consistent with the known mechanism of Cdo1 regulation . To examine whether Cdo1 mediates CYS metabolism to CSA and HTAU , and whether this limits the use of CYS for GSH synthesis , we deleted Cdo1 with CRISPR/Cas9 , followed by infection with empty or Cre expressing adenovirus to generate Cdo1-deficient , isogenic Keap1WT and Keap1R554Q MEFs . Western analysis of Cdo1 protein revealed a significant reduction of Cdo1 expression in Keap1R554Q MEFs , although the already low Cdo1 levels did not change significantly in Keap1WT MEFs ( Figure 2C ) . We performed quantitative 13C6-cystine [ ( CYS ) 2] tracing to examine the entry of CYS into GSH and TAU synthesis and found that depletion of Cdo1 inhibited HTAU synthesis from CYS ( Figure 2H , I ) . However , HTAU labeling was not completely abolished , which may be explained by incomplete Cdo1 deletion or Cdo1-independent HTAU synthesis from CYS via CoA breakdown , which cannot be distinguished by this method . By contrast , the total CYS and GSH levels as well as GSH labeling from 13C6 labeled ( CYS ) 2 were modestly increased by Cdo1 depletion , without any change in CYS2 levels ( Figure 2D–G ) . These results demonstrate that Cdo1 accumulation in Keap1R554Q MEFs promotes CYS entry in into the TAU synthesis pathway , and modestly limits CYS accumulation and GSH synthesis . The limitation of CYS availability by CDO1 suggests that this enzyme may antagonize NRF2-dependent processes in cancer . Thus , we hypothesized that the CDO1-mediated CYS homeostatic control mechanism might be deregulated in NSCLC , allowing enhanced CYS entry into GSH synthesis and other pathways . To evaluate this possibility , we examined the expression of CDO1 in NSCLC patient samples from The Cancer Genome Atlas ( TCGA ) . CDO1 mRNA expression was significantly lower in lung adenocarcinoma samples compared to normal lung ( Figure 3A ) , which was associated with CDO1 promoter methylation ( Figure 3B ) and poor prognosis ( Figure 3—figure supplement 1A ) . Methylation was strongly correlated with mRNA expression across patient samples ( Figure 3—figure supplement 1B ) . Interestingly , the incidence of CDO1 promoter methylation was significantly higher and its mRNA expression significantly lower in KEAP1 mutant lung adenocarcinoma compared to wild-type ( Figure 3A , B ) , and NRF2 activity high lung adenocarcinoma compared to NRF2 low ( Figure 3—figure supplement 1C ) , suggesting that CDO1 expression confers a selective disadvantage in the context of NRF2 accumulation . CDO1 protein expression was undetectable in a panel of NSCLC cell lines with the exception of H1581 cells ( Figure 3C ) , and treatment with the DNMT inhibitor decitabine restored CDO1 mRNA expression ( Figure 3—figure supplement 1D ) . These results indicate that CDO1 epigenetically silenced by promoter methylation in NSCLC cell lines and patient samples . To investigate the NRF2-dependent regulation of CDO1 protein in NSCLC , we generated a doxycycline-inducible lentiviral expression system to reintroduce GFP , CDO1WT or a catalytically inactive CDO1 mutant ( Y157F , Ye et al . , 2007 ) at single copy into the panel of NRF2LOW and NRF2HIGH NSCLC cell lines ( Figure 3D , Figure 3—figure supplement 2 ) . The level of CDO1 protein expression in these cells was similar with the physiological Cdo1 levels in mouse lung and liver ( Figure 3—figure supplement 2B ) , with liver being one of the highest CDO1-expressing tissues that is responsible for supplying TAU to the body ( Stipanuk et al . , 2015 ) . We find that CDO1 accumulated to higher levels in NRF2HIGH cells than NRF2LOW , although accumulation was observed in many NRF2LOW cell lines as well ( Figure 3D ) . We investigated the association with intracellular CYS levels across the panel of parental cell lines and found a strong association between the level of CDO1 accumulation and intracellular CYS levels but not with the level of CDO1 mRNA expressed from our inducible promoter system ( Figure 3D , Figure 3—figure supplement 2A ) , which is consistent with our findings in Keap1R554Q MEFs ( Figure 2A , B ) . Consistent with MEFs , we also find that deletion of endogenous CDO1 in H1581 cells , the only NSCLC line with detectable CDO1 expression , promoted CYS accumulation , demonstrating that CDO1 functions to limit intracellular CYS in lung cells as well ( Figure 3E ) . To directly examine the effect of NRF2 on CDO1 expression in NSCLC cell lines , we used multiple isogenic cell systems . First , we used NRF2-deficient A549 cells ( Torrente et al . , 2017 ) , in which we restored NRF2 expression in combination with CDO1WT , CDO1Y157F , or GFP . NRF2 restoration in these cells led to higher expression of endogenous xCT and accumulation of ectopically expressed CDO1 compared to GFP control ( Figure 3F ) . Next , we selected the two KEAP1WT NSCLC cell lines that had the lowest ectopic CDO1 accumulation and low intracellular CYS in our cell line panel , H1299 and H1975 . Using a NRF2T80K mutant that is unable to bind KEAP1 ( Berger et al . , 2017 ) , the effects of NRF2 on CDO1 accumulation were recapitulated in these KEAP1WT NSCLC cell lines ( Figure 3G ) . Interestingly , we observed that NRF2T80K could also promote the accumulation of endogenous CDO1 in H1299 cells . Further , NRF2 stabilization with the ROS inducing agent β-lapachone promoted NRF2 , CYS and CDO1 accumulation in H1299 and H1975 cells ( Figure 3—figure supplement 3A , B ) , which was most pronounced the day following the 4 hr treatment window . Consistently , NRF2 depletion in KEAP1MUT NSCLC cells following KEAP1WT expression led to CDO1 depletion ( Figure 3H ) , although the effects were more modest than what was observed with NRF2 activation . Notably , NRF2 expression promoted intracellular CYS accumulation , while NRF2 depletion impaired CYS accumulation ( Figure 3I ) , supporting a role for intracellular CYS in CDO1 stabilization . To directly assess the requirement for CYS , A549 cells were cultured in high or low ( CYS ) 2 and CDO1 levels were found to be dependent on CYS availability ( Figure 3—figure supplement 3C ) . Next , we examined the consequence of CDO1 expression on cellular proliferation . Using the isogenic NRF2 KO A549 cell system , we observed that CDO1 expression significantly impaired the proliferation of NRF2-expressing cells , while no effect was observed on NRF2 KO cells ( Figure 3J ) . Looking more broadly , we observed that CDO1 expression generally antagonized the proliferation of NSCLC cell lines and proliferation inhibition was strongly correlated with CDO1 protein expression , but not RNA expression ( Figure 3K and Figure 3—figure supplement 3D , E ) . Overall , these results demonstrate that NRF2 and other mechanisms of intracellular CYS accumulation promote CDO1 accumulation , which leads to a selective growth disadvantage in lung cancer cells . To evaluate the mechanism by which CDO1 expression impaired proliferation we interrogated CYS metabolism following CDO1 expression . CYS has multiple intracellular fates , including the synthesis of GSH . CDO1 metabolizes CYS to CSA , which is then decarboxylated to HTAU ( Figure 4A ) . CDO1WT and CDO1Y157F-expressing A549 cells were fed fresh ( CYS ) 2-containing medium and sulfur containing metabolites were quantified over time ( Figure 4B–D ) . CDO1WT , but not the enzyme-inactive CDO1Y157F , limited intracellular CYS levels and promoted the accumulation of CSA , which peaked at 4 hr ( Figure 4B ) . We observed a steady increase in both CDO1Y157F and CDO1WT protein over the 24 hr time period , although CDO1WT protein levels were lower , consistent with the intracellular CYS levels ( Figure 4—figure supplement 1A ) . Interestingly , unlike what was observed in MEFs , the levels of GSH , HTAU , and TAU were not changed over the time course of this assay ( Figure 4C ) . We also interrogated metabolite changes in the medium and observed that ( CYS ) 2 was rapidly depleted from the medium by 24 hr , while CSA steadily accumulated ( Figure 4D ) . Based on this time course , 4 hr was selected for all subsequent experiments to prevent ( CYS ) 2 starvation by CDO1 . Importantly , deletion of endogenous CDO1 in H1581 cells reduced CSA production and ( CYS ) 2 consumption ( Figure 4E , F ) . To examine the NRF2-dependence of these metabolite alterations , we utilized the NRF2KO cells , KEAP1WT cells , or KEAP1MUT cells from Figure 3 . NRF2 promoted CSA accumulation and ( CYS ) 2 consumption following CDO1WT expression in NRF2 KO cells ( Figure 4G , H ) . This effect was recapitulated by NRF2T80K expression in KEAP1WT cells ( Figure 4I , J ) . Consistently , NRF2 depletion by KEAP1WT expression in KEAP1MUT cells inhibited the CDO1-dependent production of CSA ( Figure 4K–M ) . Interestingly , KEAP1WT expression did not robustly affect CDO1 protein levels in H322 cells but significantly impaired CSA production by CDO1 . Intracellular CYS can also influence CDO1 activity promoting its catalytic efficiency ( Dominy et al . , 2008 ) , which may explain these results . Collectively , these results demonstrate that CDO1 expression promotes the production of CSA from CYS , leading to CSA accumulation both intracellularly and extracellularly , and enhanced ( CYS ) 2 consumption . We found that CSA only accounted for a fraction of CDO1-dependent ( CYS ) 2 depletion ( Figure 4—figure supplement 1B ) , suggesting that CSA is metabolized to an alternative product in NSCLC cell lines . To further characterize the consequence of CDO1 expression on CYS metabolism , we performed untargeted metabolomics and found significant accumulation of SO32- in CDO1-expressing cells ( Figure 5A ) . Importantly , CSA can be transaminated by the cytosolic aspartate aminotransferase ( GOT1 ) to produce β-sulfinyl pyruvate , which spontaneously decomposes to pyruvate and SO32- ( Singer and Kearney , 1956 ) ( Figure 5B ) . While HTAU and TAU are non-toxic molecules that have important physiological functions ( Aruoma et al . , 1988; Hansen et al . , 2010; Schaffer et al . , 2000; Suzuki et al . , 2002 ) , SO32- is toxic at high levels due to its cleavage of disulfide bonds in proteins and small molecules , including ( CYS ) 2 ( Clarke , 1932 ) . Thus , we hypothesized that in addition to accounting for the fate of CSA metabolism , SO32- production may also contribute to ( CYS ) 2 depletion through disulfide cleavage , also known as sulfitolysis . We next performed a quantitative analysis of SO32- levels following CDO1 expression . CDO1WT- , but not CDO1Y157F-expressing A549s demonstrated rapid accumulation of extracellular SO32- over the 24 hr time course following media replenishment ( Figure 5C ) . Further , the accumulation of the product of the sulfitolysis reaction , cysteine-S-sulfate ( CYS-SO3- ) , was also observed in the medium of CDO1WT-expressing cells ( Figure 5D , Figure 5—figure supplement 1A ) . We observed that CYS-SO3- appeared earlier than SO32- , and stopped accumulating once ( CYS ) 2 levels were depleted , suggesting that SO32- reacted with ( CYS ) 2 in a rapid and complete manner . To test this possibility , we incubated either CSA or sodium sulfite ( Na2SO3 ) with culture medium in the absence of cells , and observed rapid and robust conversion of ( CYS ) 2 to CYS-SO3- by Na2SO3 ( Figure 5E ) , but not CSA ( Figure 5—figure supplement 1B ) , within 5 min . Interestingly , similar depletion kinetics in this experiment could also be observed by substituting ( CYS ) 2 with oxidized glutathione ( GSSG ) ( Figure 5—figure supplement 1C , D ) , although intracellular and extracellular levels of GSSG in cell culture were much lower than ( CYS ) 2 ( Figure 5—figure supplement 1E , F ) , suggesting it is not the major target of SO32- . These results demonstrate that SO32- is generated downstream of CDO1 and rapidly reacts with ( CYS ) 2 , thereby depleting ( CYS ) 2 from the culture media . To evaluate whether SO32- production was a consequence of our CDO1 overexpression system , we transduced Keap1WT and Keap1R554Q MEFs with our inducible CDO1 vectors ( Figure 5—figure supplement 2A ) . Consistent with its regulation by intracellular CYS , ectopic CDO1 expression was significantly higher in Keap1R554Q MEFs compared to Keap1WT MEFs . While CDO1 overexpression promoted the accumulation of intracellular CSA and HTAU , and the depletion of CYS and GSH ( Figure 5—figure supplement 2B–G ) , we did not observe the production of SO32- or CYS-SO3- in MEFs ( data not shown ) . Interestingly , MEFs express lower Got1 but higher Csad protein compared to A549 cells ( Figure 5—figure supplement 2A ) , and NSCLC cell lines were uniformly low for CSAD and high for GOT1 ( Figure 5—figure supplement 3A ) suggesting that expression of CSA metabolic enzymes may be a key determining factor in the generation of HTAU vs . SO32- . Importantly , deletion of endogenous CDO1 in H1581 cells resulted in a dramatic reduction in CYS-SO3- production ( Figure 5F ) , thereby demonstrating that SO32- is generated by CDO1 under physiological expression levels in lung cells . To examine the NRF2-dependence of ( CYS ) 2 depletion via SO32- , we utilized the NRF2 KO cells , KEAP1WT , and KEAP1MUT cell lines . NRF2 promoted CYS-SO3- production following CDO1WT expression ( Figure 5G ) , which was accompanied by significant accumulation of both intracellular and extracellular SO32- ( Figure 5—figure supplement 3B , C ) . These findings were recapitulated in KEAP1WT and KEAP1MUT NSCLC cell lines expressing NRF2T80K or following KEAP1 restoration , respectively ( Figure 5H–K ) , with the exception of H322 , which maintained CYS-SO32- production following KEAP1 restoration ( Figure 5—figure supplement 3G ) . While KEAP1 restoration in these cells significantly reduced CSA production ( Figure 4M ) , unlike A549 and H1944 , intracellular CSA levels in H322 cells were still in the millimolar range ( Figure 5—figure supplement 3D–F ) , suggesting that CSA transamination by GOT1 was saturated . Further , NRF2 stabilization with β-lapachone promoted CDO1-dependent CSA and CYS-SO3- production and ( CYS ) 2 depletion in H1299 and H1975 cells ( Figure 5—figure supplement 3H–J ) . Collectively , these results suggest that NRF2 induces CDO1-mediated sulfitolysis , thereby depleting extracellular ( CYS ) 2 in NSCLC cells . Next , we examined the toxicity of CDO1 products to NSCLC cell lines . Treatment of cells with CSA and Na2SO3 , but not HTAU , led to cytotoxicity ( Figure 5—figure supplement 4A ) . We found that ( CYS ) 2 starvation and Na2SO3 treatment were universally toxic to NSCLC cell lines , which did not depend on NRF2 activity ( Figure 5—figure supplement 4B , C ) . In addition , CDO1 , CSA and Na2SO3 sensitized A549 cells to oxidative stress ( Figure 5—figure supplement 4D , E ) , consistent with their ability to deplete ( CYS ) 2 . Collectively , these results demonstrate that CSA and SO32- are toxic to NSCLC cells regardless of NRF2 activity , suggesting that resistance to ( CYS ) 2 starvation is not an inherent phenotype of NRF2HIGH cells . Rather , they are sensitive to CDO1 expression due to high intracellular CYS and CDO1 stabilization . To evaluate the role of GOT1 in CDO1-dependent sulfitolysis and cell growth inhibition , we generated GOT1 KO A549 and H460 cells ( Figure 6A ) . Two independent clones of each were generated , and a sgRNA-resistant GOT1 cDNA was expressed in each to restore GOT1 expression ( Birsoy et al . , 2015 ) . In support of GOT1 mediating the production of SO32- and ( CYS ) 2 depletion , GOT1 KO cells had significantly lower CDO1-dependent ( CYS ) 2 consumption and SO32- and CYS-SO3- production rates compared to each parental line , which was rescued by GOT1 restoration ( Figure 6B ) . Surprisingly , we observed that CDO1 antagonized cell proliferation independent of GOT1 expression and sulfitolysis , suggesting that the intracellular metabolism of CYS also contributes to the phenotype ( Figure 6C ) . To address this mechanism , we evaluated whether CDO1 could limit CYS-dependent processes in NSCLC cell lines , similar to what was observed in MEFs . However , unlike in MEFs , CDO1 expression did not affect CYS utilization as the rates of GSH , Coenzyme A ( CoA ) , and protein synthesis were similar ( Figure 6D–G ) . Next , we examined other consequences of CDO1 expression in cells . After ( CYS ) 2 enters cells through its transporter xCT , it must be reduced to two CYS molecules using NADPH as the electron donor . As such , we hypothesized that the continual reduction of ( CYS ) 2 to CYS in CDO1-expressing cells would consume a significant amount of cellular NADPH . Indeed , we observed that the NADPH/NADP+ ratio was lower following CDO1 expression in both GOT1 KO and GOT1 expressing cells ( Figure 7A ) . Consistently , ( CYS ) 2 starvation had the opposite effect on the NADPH/NADP+ ratio , which increased following starvation ( Figure 7—figure supplement 1A ) . NADPH is critical for both antioxidant defense and cellular biosynthetic processes . While we found that the decrease in the NADPH/NADP+ ratio did not dramatically influence the levels of GSH and GSSG in non-stressed cells ( Figure 7B ) , expression of CDO1 increased sensitivity to the lipid peroxidation inducer cumene hydroperoxide ( CuH2O2 ) independent of GOT1 expression ( Figure 7C ) , and CDO1 deletion in H1581 cells promoted CuH2O2 resistance ( Figure 7—figure supplement 1B ) . Next , we placed cells into detached conditions , which has been shown to increase reliance on IDH1-dependent reductive carboxylation to promote NADPH generation in the mitochondria ( Jiang et al . , 2016 ) . Consistently , CDO1 expression significantly impaired the ability of NSCLC cell lines to grow in soft agar ( Figure 7—figure supplement 1C ) . Next , we examined the consequence of the altered NADPH/NADP+ ratio on NADPH-dependent metabolic reactions . Using 13C5-glutamine tracing , we observed that glutamine readily entered the TCA cycle to produce M + 4 citrate , which was unaffected or increased following CDO1 expression ( Figure 7E , F ) , but CDO1 impaired both NADPH-dependent synthesis of proline from glutamate ( Figure 7D ) , and NADPH-dependent reductive carboxylation of α-ketoglutarate to produce M + 5 citrate ( Figure 7E , F ) . The antiproliferative effects of CDO1 were CYS-dependent , as either inhibition of ( CYS ) 2 uptake with erastin or low ( CYS ) 2 conditions resulted in loss of CDO1 expression and completely rescued the CDO1-induced proliferation defect ( Figure 7—figure supplement 1D–F ) . Collectively , these results suggest that CDO1 further inhibits cellular processes by limiting NADPH availability , thereby impairing cellular proliferation . We next examined the ability of NRF2 to promote CDO1 stabilization under physiological ( CYS ) 2 conditions . To this end , we generated KrasG12D; Trp53flox/flox lung tumor mice expressing either wild-type Keap1 ( Keap1WT ) , heterozygous for Keap1R554Q ( Keap1R554Q/WT ) , or homozygous for Keap1R554Q ( Keap1R554Q/R554Q ) . We chose this model because Keap1 deletion in the KrasG12D; Trp53flox/flox lung tumor model was recently shown to activate Nrf2 and promote cystine uptake ( Romero et al . , 2017 ) . We found that KrasG12D; Trp53flox/flox mice had similar survival regardless of Keap1 mutation status ( Figure 8A ) . However , survival was not related to tumor burden , as KrasG12D; Trp53flox/flox; Keap1R554Q/R554Q mice displayed significant lung hemorrhage at endpoint ( Figure 8B ) despite small tumors ( Figure 8C ) . Quantification of average lung tumor size revealed that KrasG12D; Trp53flox/flox mice expressing one copy of Keap1R554Q had modestly larger tumors , consistent with the findings of Romero et al . , and in agreement with the dominant negative activity displayed by select Keap1 mutants ( Suzuki et al . , 2011 ) . Surprisingly , complete loss of Keap1 function in the Keap1R554Q/R554Q mice led to significantly smaller tumors ( Figure 8C ) . Further , Keap1 loss of function led to a gradient of Nrf2 activation with Keap1WT tumors expressing little to no Nqo1 , Keap1R554Q/+ tumors displayed increased Nqo1 expression , and Keap1R554Q/R554Q tumors were strongly positive ( Figure 8D ) . Importantly , Keap1R554Q/R554Q tumors strongly expressed Cdo1 ( Figure 8D ) , demonstrating that Nrf2 activation promotes Cdo1 accumulation under physiological conditions in vivo , and suggesting that Cdo1 may impede tumor progression . Additional work is needed to understand whether induction of Cdo1 , or even Nrf2 , is responsible for the apparent block in tumorigenesis observed in the Keap1R554Q/R554Q mice , or whether an alternative Keap1 substrate may mediate these effects .
The carbon , nitrogen and sulfur molecules of CYS are used for diverse cellular processes that are required for both homeostasis and proliferation . The carbon , nitrogen and sulfur atoms are incorporated into protein , GSH , TAU and CoA . Further , the sulfur atom of cysteine is incorporated into iron-sulfur ( Fe-S ) clusters . CYS is generally thought to be more limiting than glycine or glutamate for GSH synthesis in most tissues ( Stipanuk et al . , 2006 ) . CDO1 plays a critical role in limiting CYS availability and toxicity ( Jurkowska et al . , 2014 ) . CYS promotes both the stability and activity of CDO1 ( Stipanuk et al . , 2009 ) , leading to the irreversible metabolism of CYS to CSA ( Stipanuk et al . , 2009 ) . This mechanism of regulation prevents toxicity associated with CYS accumulation . However , the contribution of this regulatory process to CYS availability in cancer was not well understood . Our findings implicate CDO1 as a metabolic liability for lung tumor cells with high intracellular CYS levels , particularly those with NRF2/KEAP1 mutations ( Figure 9 ) . We find that high intracellular CYS levels are a common feature of lung cancer cell lines , suggesting that NRF2-independent mechanisms exist to promote ( CYS ) 2/CYS uptake . Indeed , ( CYS ) 2 uptake is regulated by many signaling pathways , including EGFR , mTORC2 and p53 ( Gu et al . , 2017; Jiang et al . , 2015; Tsuchihashi et al . , 2016 ) . Further , de novo CYS synthesis via the transsulfuration pathway ( Prigge et al . , 2017 ) , direct CYS transport , or decreased CYS utilization may play a role in CYS accumulation . Notably , CDO1 promoter methylation is common across multiple cancer types ( Brait et al . , 2012; Jeschke et al . , 2013 ) raising the possibility that CDO1 antagonizes the proliferation or viability of other cancers through similar mechanisms . We find that CDO1 promotes the wasting of the carbon , nitrogen and sulfur molecules of CYS as CSA and SO32- , promotes ( CYS ) 2 depletion by SO32- , and induces depletion of NADPH . Any of these mechanisms could contribute to its growth suppressive function in vivo , but additional work is needed to evaluate the consequence of CDO1 loss in the relevant tumor microenvironment . Importantly , while limited GSSG is present in our cells and media , Sullivan et al . find that GSSG levels may exceed ( Cys ) 2 levels in murine tumor interstitial fluid ( Sullivan et al . , 2019 ) , suggesting that GSSG may be a major target of SO32-in vivo . Our findings suggest that oncogene-induced metabolic processes can be unfavorable , and warrant further investigation into the selection against metabolic processes in the context of specific driving oncogenes or nutritional states . A surprising finding from this study is that NRF2 stabilization promotes the accumulation of intracellular CYS to levels that far exceed those which are necessary for CYS-dependent metabolic processes in cancer cells . Normal intracellular CYS concentrations are approximately 100 μM ( Stipanuk et al . , 2006 ) , around the Km for GCLC for CYS and typically an order of magnitude lower than GSH levels , consistent with our observed concentrations in wild-type MEFs . However , in many cases KEAP1 mutant NSCLC cell lines and even some KEAP1 wild-type lines accumulated CYS to millimolar levels without any apparent toxicity . CYS toxicity is a poorly described phenomenon that has been attributed to the reactivity of the free thiol on the CYS molecule , the production of hydrogen sulfide ( H2S ) , autooxidation and free radical formation , and other mechanisms ( Stipanuk et al . , 2006 ) . More work is needed to understand whether high intracellular CYS levels are a vulnerability of CDO1-silenced cancer cells as a consequence of these mechanisms , or whether there is an advantage to high intracellular CYS . We observe that Cdo1 modestly limits GSH synthesis in MEFs , similar to what has been demonstrated in the liver , where deletion of Cdo1 in vivo resulted in the accumulation of CYS and GSH ( Roman et al . , 2013 ) . By contrast , CDO1 depleted intracellular CYS but did not limit GSH synthesis or other CYS-dependent processes in NSCLC cell lines . One potential explanation for this difference is that cancer cells are more efficient at maintaining intracellular CYS levels and/or GSH synthesis . While more efficient maintenance of intracellular CYS is an advantage for CYS-dependent metabolism , it is also a liability because considerable resources in the form of reducing power must be committed to continually reduce ( CYS ) 2 to replenish the CYS pool . Although we do not know the rate of NADPH production in our cell lines , comparison with NADPH production rates in other lines ( Fan et al . , 2014 ) suggest that CDO1-dependent ( CYS ) 2 consumption would consume a significant fraction of the cellular NADPH produced , consistent with our observations . Interestingly , we find that CSA is differentially metabolized in cancer cells and mouse embryonic fibroblasts , which correlates with differential expression of CSAD and GOT1 . Partitioning between the decarboxylation and transamination reactions is likely influenced by the levels and activities of CSAD and GOT1 , but this is not well studied . Evaluation of CSA metabolism at the organismal level has shown that decarboxylation of CSA via CSAD predominates over transamination by GOT1 ( Weinstein et al . , 1988 ) but little is known about this partitioning in individual tissues . Consistently , CSAD has a lower Km for CSA than GOT1 ( Recasens et al . , 1980; Wu , 1982 ) . Importantly , HTAU and TAU are non-toxic molecules that have important physiological functions in controlling osmolarity , mitochondrial function , cellular redox , and other processes ( Aruoma et al . , 1988; Hansen et al . , 2010; Schaffer et al . , 2000; Suzuki et al . , 2002 ) . TAU biosynthesis is not a required process in non-hepatic tissues , however , as adequate TAU is supplied from the liver via the blood supply ( Stipanuk , 2004 ) . By contrast , SO32- is toxic in large quantities ( Menzel et al . , 1986 ) and thus excessive CSA transamination is likely disadvantageous in most cell types . Interestingly , hemorrhagic pulmonary edema is a consequence of sulfur dioxide exposure in humans ( Charan et al . , 1979 ) , and additional work is needed to determine if Cdo1-mediated SO32- production is the cause of lung hemorrhage in the Keap1R554Q/R554Q mice . It is unclear what function the CSA transamination reaction serves for cells . Unlike the decarboxylation pathway , it retains carbon and nitrogen intracellularly as pyruvate and glutamate . Sulfite oxidase ( SUOX ) can reduce sulfite ( SO32- ) to sulfate ( SO42- ) , which is an important precursor for sulfation-based detoxification of phenols , hydroxylamines , or alcohols to sulfate esters . Sulfation is also an important post-translational modification of proteins . There are currently several approaches being developed to target aberrant CYS metabolism in cancer , including NRF2/KEAP1 mutant cancer . Despite their increased antioxidant capacity , KEAP1 mutant cells are still sensitive to ( CYS ) 2 starvation , which could be targeted with cyst ( e ) inase ( Cramer et al . , 2017 ) . Further , ( CYS ) 2 uptake via xCT results in a central carbon imbalance and dependence on glutamine that can be targeted with glutaminase inhibitors ( Romero et al . , 2017 ) . Our findings suggest excessive ( CYS ) 2 uptake can also impair NADPH-dependent processes and activating CDO1 expression may have unique therapeutic potential compared to ( CYS ) 2-depleting strategies . We find that Cdo1 accumulates in tumors from our Keap1R554Q/R554Q mutant lung cancer GEMM , which is correlated with a block in lung tumor formation . Further work is needed to determine whether Cdo1 impairs tumorigenesis in this model , and to evaluate the consequences of strategies to induce CDO1 expression on tumor growth and metabolism in vivo .
The Keap1 targeting vector was constructed to contain homology arms and a minigene containing a cDNA encoding wild-type exons 3–5 , followed by a SV40 polyA signal , inserted upstream of endogenous exon 3 of the Keap1 gene . Codon 554 in endogenous exon four was mutated from arginine to glutamine in the targeting vector and the endogenous Keap1 locus in C10 murine ES cells ( Beard et al . , 2006 ) was targeted and cells were selected with blasticidin . Positive clones were screened by copy number real-time PCR and injected into blastocysts . Mice were housed and bred in accordance with the ethical regulations and approval of the IACUC ( protocol # R IS00003893 ) . Keap1R554Q mice were crossed with LSL-KrasG12D and Trp53flox mice for lung tumor studies . Lung tumor formation was induced by intranasal installation of 2 . 5 × 107 PFU adenoviral-Cre ( University of Iowa ) as described previously ( Jackson et al . , 2001 ) . Viral infections were performed under isofluorane anesthesia , and every effort was made to minimize suffering . Tissues were fixed in 10% formalin overnight before embedding in paraffin and sectioning . Sections were de-paraffinized in xylene and rehydrated in a graded alcohol series . Antigen retrieval was performed in 10 mM citrate buffer ( pH 6 . 0 ) and endogenous peroxidase activity was quenched with 3% hydrogen peroxide . Immunohistochemical staining was performed with the ImmPRESS HRP anti-rabbit kit according to manufacturer’s instructions ( Vector Labs ) , followed by incubation with DAB substrate ( Vector Labs ) . Staining intensity was graded on a 0 ( no staining ) – 4 ( most staining ) scale . Slides were scanned with the Aperio imager and tumor volume calculations were performed with Image Scope software ( Aperio ) . MEFs were isolated from E13 . 5–14 . 5 day old embryos and maintained in pyruvate-free DMEM ( Corning ) supplemented with 10% FBS . MEFs were infected with empty adenovirus or adenoviral-Cre ( University of Iowa ) at an MOI of 500 and used within four passages . For lentivirus production , Lenti-X 293 T cells ( Clontech ) were transfected at 90% confluence with JetPRIME ( Polyplus ) . Packaging plasmids pCMV-dR8 . 2 dvpr ( addgene # 8455 ) and pCMV-VSV-G ( addgene #8454 ) were used . For retroviral production , Phoenix-AMPHO packaging cells ( ATCC CRL-3213 ) were used . To generate CDO1 KO MEFs , MEFs were first infected with empty pLenti-CRISPR-V2 , or pLenti-CRISPR-V2 encoding sgCDO1 #2 or #3 and selected with 1 μg/mL puromycin for 4 days , followed by infection with empty or cre-encoding adenovirus . To generate CDO1 overexpressing MEFs , MEFs were first infected with pRRL-CDO1 or GFP lentivirus and selected with 1 μg/mL puromycin for 4 days , followed by infection with empty or cre-encoding adenovirus , and then treated with doxycycline . GOT1 knockout A549 and H460 cells were generated using the plentiCRISPR-sgGOT1 vector and KO clones were verified by western blotting . Cell lines were infected with lentivirus at a MOI <0 . 2 to achieve single copy integration and CDO1 expression verified by Q-RT-PCR . Cell lines were routinely tested and verified to be free of mycoplasma ( MycoAlert Assay , Lonza ) . All lines were maintained in RPMI 1640 medium ( Hyclone or Gibco ) supplemented with 10% FBS without antibiotics at 37°C in a humidified atmosphere containing 5% CO2 and 95% air . Patient lung adenocarcinoma data from The Cancer Genome Atlas ( TCGA ) , with associated KEAP1 mutation and CDO1 methylation status ( Illumina HM450 Beadchip ) , was obtained via cBioPortal ( Cerami et al . , 2012; Gao et al . , 2013 ) . Patient normal lung and lung adenocarcinoma data from The Cancer Genome Atlas ( TCGA ) , containing CDO1 methylation status ( Illumina HM450 Beadchip , beta value 0 [least methylated] – 1 [most methylated] ) and mRNA expression data ( RNA-seq RPKM ) , was obtained via the MethHC database ( Huang et al . , 2015 ) . The CDO1 promoter was defined as the region from 1 . 5 kb upstream to 0 . 5 kb downstream of the RefSeq TSS . The NRF2 activity score was determined as described previously ( DeNicola et al . , 2015 ) . MEFs were seeded onto 6-well dishes and the cells were harvested at 70% confluence . NSCLC lines were harvested at 70% confluence , after treatment with 5 μM Decitabine for 3 days . During the treatment , medium was changed every 24 hr . For the NSCLC lines expressing CDO1 , cells were pre-treated with DOX for 48 hr , and harvested at 4 hr after medium change . RNA was isolated with the E . Z . N . A . Total RNA Kit I ( Omega Bio-Tek ) according to the manufacturer’s instructions . cDNA was synthesized from 500 ng of RNA using PrimeScript RT Master Mix ( Takara ) according to the manufacturer’s instructions , and analyzed with Taqman gene expression assays . The N-ethylmaleamide ( NEM ) derivatized , isotope labeled , [13C3 , 15N]-cysteine-NEM , [13C3]-cysteine-NEM , [13C2 , 15N]-GSH-NEM and [D5]-GSH-NEM which were prepared by derivatizing the [13C3 , 15N]-cysteine , [13C3]-cysteine , [13C2 , 15N]-GSH , and [D5]-GSH standards with 50 mM NEM in 10 mM ammonium formate ( pH = 7 . 0 ) at room temperature ( 30 min ) as previously described ( Ortmayr et al . , 2015 ) . [13C4 , 15N2]-GSSG was prepared from the oxidation of [13C2 , 15N]-GSH as described ( Zhu et al . , 2008 ) . [13C3 , 15N]-CSA was synthesized from [13C3 , 15N]-cysteine as previously described ( Santhosh-Kumar et al . , 1994 ) . Briefly , 1 . 23 mg of [13C3 , 15N]-cysteine was combined with 300 μL of 0 . 1 N NaOH , followed by addition of 300 nmol of CuCl2 . After a 12 hr incubation at 37°C , 400 μL of 0 . 1 N NaOH was added and the product purified by solid phase extraction ( SPE ) as following . The reaction mixture was passed through 100 mg of AG MP-1M strong anion exchange resin ( Bio-Rad ) , which was pre-equilibrated with 1 . 5 mL of 0 . 1N NaOH . After washing with 12 mL of HPLC-grade water , the isotope labeled [13C3 , 15N]-CSA was eluted with 1 . 8 mL of 1N HCl ( Figure 1—figure supplement 1B ) . In addition to [13C3 , 15N]-CSA , [13C6 , 15N2]-Cystine and [13C3 , 15N]-Cysteine S-sulfate ( CYS-SO3- ) ( Figure 5—figure supplement 1A ) were also obtained during [13C3 , 15N]-CSA synthesis as byproducts . The quantities of synthesized metabolites were determined using unlabeled CSA , CYS-SO3- , and Cystine standards and LC-MS/MS analysis as described ( Bennett et al . , 2008 ) . Following quantification , these internal standards were used to quantify concentrations of CSA , CYS-SO3- and cystine in cell and medium extracts . Cells were grown in 6-well dishes , quickly washed in cold PBS , and extracted in 80% methanol ( 0 . 5 mL for NSCLC cells , 0 . 2 mL for MEFs , −80°C ) . The extracts were cleared by centrifugation , and the metabolites in the supernatant were directly analyzed ( MEFs ) or dried by centrifugation under vacuum ( NSCLC cells , SpeedVac , Thermo Scientific ) . The dried pellets were re-dissolved in 20 μL of HPLC grade water and analyzed by liquid chromatography-high resolution mass spectrometry ( LC-HRMS ) . The extracellular metabolites from 10 μL of cell culture medium were extracted with 40 μL 100% methanol ( −80°C ) . The extract was cleared by centrifugation , and 5 μL of the supernatant analyzed by LC-HRMS . Cells were plated the day before extraction in 6-well dishes so they were 70% confluent at extraction . The medium was changed the next day and the cells harvested at the indicated time point . The metabolites were extracted and derivatized with 0 . 5 mL of ice-cold extraction solvent ( 80% MeOH:20% H2O containing 25 mM NEM and 10 mM ammonium formate , pH 7 . 0 ) containing 20 μM [13C3 , 15N]-cysteine-NEM , 36 . 4 μM [13C2 , 15N]-GSH-NEM , 0 . 13 μM [13C4 , 15N2]-GSSG , 10 μM [D4]-Cystine , 20 μM [13C2]-Taurine , and 20 μM [D4]-Hypotaurine , followed by incubation on ice for 30 min . The NEM-derivatized metabolite extracts were cleared by centrifugation and analyzed by LC-MS/MS via multiple reaction monitoring ( LC-MRM ) or LC-HRMS . Cell volumes and number were determined using a Scepter 2 . 0 cell counter ( Millipore ) and used to calculate the intracellular metabolite concentrations and to normalize metabolite consumption/production rates . The cell culture medium was aspirated after reserving 1 mL for extracellular metabolite analysis . Cells were quickly washed with ice cold PBS , and intracellular CSA was extracted with 80% MeOH including approximately 1 μM [13C3 , 15N]-CSA as an internal standard . The exact quantity was determined for each experiment using unlabeled CSA standard . After incubation for 15 min ( −80°C ) and scraping , the cell and metabolite mixture were transferred into a 1 . 5 mL tube . After clarification by centrifugation , the metabolite extracts were dried under vacuum and re-dissolved into HPLC grade water ( 20 μL for 6-well dishes or 40 μL for 100 mm dishes ) for analysis by targeted LC-MRM . 10 μL of cell culture medium was extracted at −80°C for at least 15 min with 40 μL of 100% ice cold MeOH containing approximately 25 μM [13C3 , 15N]-CSA , 10 μM [13C3 , 15N]-CYS-SO3- , and 300 μM [13C6 , 15N2]-Cystine internal standards . The exact quantity was determined for each experiment using unlabeled standards . Following centrifugation ( 16 , 000 g , 20 min , 4°C ) , the extracellular metabolite extracts were transferred into a vial and analyzed by LC-MRM-based quantification . A [34S]-Na2SO3 standard was purchased from Sigma Aldrich ( 99% purity ) . The cells were extracted in 80% methanol containing 1 μM of [34S]-Na2SO3 ( 0 . 5 mL for 6-well dish or 3 mL for 100 mm dishes , −80°C ) . The extracts were cleared by centrifugation , and the metabolites in the supernatant were dried by centrifugation under vacuum . The pellets were re-dissolved in HPLC grade water ( 20 μL for 6-well dishes or 40 μL for 100 mm dishes ) and analyzed by UPLC-Q-Exactive-HF ( Thermo Fisher Scientific , Waltham , MA ) . The extracellular metabolites from 10 μL of cell culture medium were extracted with 40 μL 100% methanol containing 96 . 5 μM [34S]-Na2SO3 . The extract was cleared by centrifugation , and the metabolites in the supernatant were analyzed by LC-HRMS . Medium containing 200 μM [13C6]-cystine were prepared as follows: For NSCLC cell line studies , cysteine/cystine , methionine and glutamine free RPMI ( MP Biomedicals ) was supplemented with 400 μM [13C3]-cysteine , 100 μM methionine , 2 mM glutamine and 10% dialyzed FBS ( dFBS ) . For MEF studies , cysteine/cystine , methionine , pyruvate and glutamine free DMEM ( Gibco ) was supplemented with 400 μM [13C3]-cysteine , 200 μM methionine , 4 mM glutamine and 10% dFBS . Medium was sterile filtered using a 0 . 2 um PES vacuum filter and incubated for 48 hr at 4°C to allow oxidation of cysteine to cystine . Prior to 13C-labeling , cells were preconditioned in medium containing dFBS and 12C-cystine for 24 hr . Cells were labeled with 13C-cystine for the indicated timepoints , and then harvested . The metabolites were extracted and derivatized with NEM in ice-cold extraction solvent ( 80% MeOH:20% H2O containing 25 mM NEM and 10 mM ammonium formate , pH = 7 . 0 ) which includes stable isotope labeled internal standards of 20 μM [13C3 , 15N]-cysteine-NEM , 6 . 4 μM [D5]-GSH-NEM , 10 μM [D4]-Cystine , and 20 μM [D4]-Hypotaurine followed by incubation on 4°C for 30 min . The NEM-derivatized metabolite extracts were cleared by centrifugation and analyzed by targeted LC-MRM . Cysteine/cystine , methionine and glutamine free RPMI ( MP Biomedicals ) was supplemented with 200 μM [13C6 , 15N2]-Cystine , 100 μM methionine , 2 mM glutamine and 10% dFBS as described above . NSCLC cell lines were preconditioned in medium containing dFBS and 12C-cystine for 24 hr in 10 cm dishes , followed by feeding with [13C6 , 15N2]-Cystine containing media . After 4 hr , the medium was aspirated , cells were quickly washed with ice cold PBS , and cellular metabolites extracted with 1 mL of 80% MeOH ( −80°C , 15 min ) . After scraping , the metabolite extract was transferred into an Eppendorf tube and cleared by centrifugation ( 17000 g , 20 min , 4°C ) . The supernatant was dried under vacuum overnight , and stored at −80°C . Dried sample pellets were resuspended in HPLC-grade water ( 20 μl ) and centrifuged at 20 , 000 g for 2 min to remove insoluble material . Supernatants ( 5 μl ) were injected and analyzed using a hybrid 6500 QTRAP triple quadrupole mass spectrometer ( AB/SCIEX ) coupled to a Prominence UFLC HPLC system ( Shimadzu ) via selected reaction monitoring ( MRM ) . ESI voltage was +4900V in positive ion mode with a dwell time of 3 ms per SRM transition . Approximately 10–14 data points were acquired per detected metabolite . The following SRM transitions were used: native Coenzyme A [M + 0] ( Q1 = 768 . 13 , Q2 = 261 . 13 , collision energy =+ 39V ) and 13C2 , 15N1 isotope labeled Coenzyme A [M + 3] ( Q1 = 771 . 13 , Q2 = 264 . 13 , collision energy =+ 39V ) . Incorporation of 13C3 , 15N1-cysteine into Coenzyme A yields 13C2 , 15N1-Coenzyme A due to a decarboxylation step during Coenzyme A synthesis . Samples were delivered to the mass spectrometer via hydrophilic interaction chromatography ( HILIC ) using a 4 . 6 mm i . d x 10 cm Amide XBridge column ( Waters ) at 400 μl/min . Gradients were run starting from 85% buffer B ( HPLC grade acetonitrile ) to 42% B from 0 to 5 min; 42% B to 0% B from 5 to 16 min; 0% B was held from 16 to 24 min; 0% B to 85% B from 24 to 25 min; 85% B was held for 7 min to re-equilibrate the column . Buffer A was comprised of 20 mM ammonium hydroxide/20 mM ammonium acetate ( pH = 9 . 0 ) in 95:5 water:acetonitrile . Peak areas from the total ion current for each metabolite SRM transition were integrated using MultiQuant v3 . 0 software ( AB/SCIEX ) . Cysteine/cystine , methionine , pyruvate and glutamine free DMEM ( Gibco ) was supplemented with 4mM [13C5]-glutamine , 200μM methionine , 200μM cystine and 10% dFBS . NSCLC cell lines were seeded onto 6-well dishes , preconditioned in DMEM containing 12C-glutamine and 10% dFBS for 24 hr , and subsequently the medium was changed to [13C5]-glutamine containing medium for the indicated time . The medium was aspirated , cells were quickly washed with ice cold PBS , and cellular metabolites were extracted with 500 μL of 80% MeOH ( -80°C , 15 min ) . After scraping , the metabolite extract was cleared by centrifugation ( 17000g , 20 min , 4°C ) . For [13C5]-glutamine → ( M+5 ) proline tracing , the supernatant was directly analyzed by LC-MS . For the [13C5]-glutamine → citrate tracing , the supernatant was dried under vacuum and derivatization was performed for GC-MS analysis ( Carey et al . , 2015 ) . The dried pellet was derivatized by adding 50 μL of methoxylamine hydrochloride ( 40 mg/mL in Pyridine ) at 30°C for 90 min . After mixing the derivatized solution with 70 μL of Ethyl-Acetate in a glass vial , the metabolite mixture was further derivatized by adding 80 μL of MSTFA + 1%TMCS solution ( 37°C for 30 min ) . The final derivatized solution was transferred into a glass vial insert , and analyzed using an Agilent 7890B/5977B MSD ( Agilent , Santa Clara , CA ) GC-MS system . Non-targeted metabolite profiling and sulfite quantification was conducted using Ultimate 3000 or Vanquish UPLC systems coupled to a Q Exactive HF ( QE-HF ) mass spectrometer equipped with HESI ( Thermo Fisher Scientific , Waltham , MA ) . The analytical condition was as following . A Luna 3 µm HILIC 200 Å , LC Column 100 × 2 mm ( Phenomenex , Torrance , CA , Part N: 00D-4449-B0 ) was applied as a stationary phase . The mobile phase A was 0 . 1 % formic acid in water , and the mobile phase B was 100% Acetonitrile ( ACN ) . The column temperature was set to 30°C , and the gradient elution was conducted as following at a flow rate of 0 . 35 mL/min: 0 to 3 min , 0 % of phase A; 3 to 13 min , linear gradient from 0% to 80% of phase A; 13 to 16 min , 80% of phase A . The MS scan was operated in negative mode and the mass scan range was 58 to 870 m/z . The FT resolution was 120 , 000 , and the AGC target was 3 × 106 . The capillary temperature was 320°C , and the capillary voltage was 3 . 5 kV . The injection volume was 5µL . The MS peak extraction , the chromatographic peak extraction and deconvolution , the peak alignment between samples , gap filling , putative peak identification , and peak table exportation for untargeted metabolomics were conducted with MZmine 2 ( 2 . 23 Version ) ( Pluskal et al . , 2010 ) . The identity of CSA , cysteine , GSH , and sulfite were further confirmed by authentic standards . For sulfite quantification , [34S]-HSO3- and [32S]-HSO3- peaks were extracted with a 20 ppm MS filter , and their peak areas were manually integrated using Thermo Xcaliber Qual Browser . The quantification was based on previous methods ( Bennett et al . , 2008 ) . The LC-HRMS method is modified from previous study ( Cantor et al . , 2017 ) using Vanquish UPLC systems coupled to a Q Exactive HF ( QE-HF ) mass spectrometer equipped with HESI ( Thermo Fisher Scientific , Waltham , MA ) . For chromatographic separation , a SeQuant ZIC-pHILIC LC column , 5 µm , 150 × 4 . 6 mm ( MilliporeSigma , Burlington , MA ) with a SeQuant ZIC-pHILIC guard column , 20 × 4 . 6 mm ( MilliporeSigma , Burlington , MA ) was used . The mobile phase A was 10mM ammonium carbonate and 0 . 05% ammonium hydroxide in water , and the mobile phase B was 100% Acetonitrile ( ACN ) . The column temperature was set to 30°C , and the gradient elution was conducted as following at a flow rate of 0 . 25 mL/min: 0 to 13 min , linear gradient from 80% to 20% of phase B; 13 to 15 min , 20% of phase B . The MS1 scan was operated in both positive and negative mode for non-targeted metabolomics , or only in positive mode for the NEM-derivatized cysteine quantification . The mass scan range was 60 to 900 m/z . The MS resolution was 120 , 000 , and the AGC target was 3 × 106 . The capillary temperature was 320°C , and the capillary voltage was 3 . 5 kV . The injection volume was 5µL for both positive mode and negative mode . For the non-targeted metabolomics , the data analysis was performed with EI Maven v0 . 3 . 1 ( https://elucidatainc . github . io/ElMaven/ ) ( Clasquin et al . , 2012 ) . Metabolite identification was based on comparison of both retention time and m/z value of sample peaks with an internal library ( MSMLS Library , Sigma Aldrich ) . For cysteine quantification , the Cysteine-NEM and [13C3 , 15N]-Cysteine-NEM peaks were extracted with a 10 ppm MS tolerance , and their peak areas were manually integrated using Thermo Xcaliber Qual Browser . The quantification was based on previous method ( Bennett et al . , 2008 ) . The targeted analysis of Cysteine-NEM , GSH-NEM , GSSG , cystine , hypotaurine , taurine , CSA , CYS-SO3- , GS-SO3- , Glutamic acid , and Proline were conducted by selected reaction monitoring ( MRM ) using an Ultimate 3000 UPLC system coupled to a Thermo Finnigan TSQ Quantum equipped with HESI ( Thermo Fisher Scientific , Waltham , MA ) . The chromatographic separation was performed as previously described ( Liu et al . , 2014 ) . As a stationary phase , an XBridge Amide Column 3 . 5µm ( 2 . 1 × 100mm ) ( Waters , Milford , MA ) was used . The mobile phase A was 97% Water and 3% ACN ( 20 mM NH4Ac , 15 mM NH4OH , pH = 9 . 0 ) and the mobile phase B was 100% ACN . The column temperature was set to 40°C , and the gradient elution was as following at 0 . 35 mL/mL of flow rate: 0 to 3 min , linear gradient from 15% to 70% of Phase A; 3 to 12 min: linear gradient from 70% to 98% of Phase A; 12 to 15 min , sustaining 98% of Phase A . The MS acquisition operated at the positive or negative mode . The capillary temperature was 305 °C , the vaporizer temperature was 200 °C . The sheath gas flow was 75 and the auxiliary gas flow was 10 . The spray voltage was 3 . 7 kV . The MRM conditions ( parent ion → fragment ion; collision energy ) of metabolites were as follows . Positive mode: Hypotaurine ( m/z 110 → m/z 92; 10 ) ; [13C2]-Hypotaurine ( m/z 112 → m/z 94; 10 ) ; [D4]-Hypotaurine ( m/z 114 → m/z 96; 10 ) ; Taurine ( m/z 126 → m/z 108; 11 ) ; [13C2]-Taurine – ( m/z 128 → m/z 110; 11 ) ; Cysteine-NEM ( m/z 247 → m/z 158; 30 ) ; [13C3]-Cysteine-NEM ( m/z 250 → m/z 158; 30 ) ; [13C3 , 15N]-Cysteine-NEM ( m/z 251 → m/z 158; 30 ) ; GSH–NEM m/z ( m/z 433 → m/z 304; 15 ) ; [13C2 , 15N]-GSH-NEM ( m/z 436 → 307m/z; 15 ) ; [D5]-GSH–NEM ( m/z 438 → m/z 304; 15 ) ; GSSG ( m/z 613 → m/z 355; 25 ) , [13C4 , 15N2]-GSSG ( m/z 619 → m/z 361; 25 ) ; Cystine ( m/z 241→ m/z 74; 30 ) ; [D4]-Cystine ( m/z 245 → m/z 76; 30 ) ; [13C6]-Cystine ( m/z 247 → m/z 76; 30 ) ; [13C6 , 15N2]-Cystine ( m/z 249 → m/z 77; 30 ) ; CYS-SO3- ( m/z 202 → m/z 74; 27 ) , [13C3 , 15N]- CYS-SO3- ( m/z 206 → 77; 27 ) ; [M+5]-Glutamate ( m/z 153 → m/z 88; 20 ) ; [M+5]-Proline ( m/z 121 . 1 → m/z 74 . 1 ) . Negative mode: CSA ( m/z 152 → m/z 88; 17 ) , [13C3 , 15N]-CSA ( m/z 156 → m/z 92; 17 ) , GS-SO3- ( m/z 386 → m/z 306; 20 ) . All peaks were manually integrated using Thermo Xcaliber Qual Browser . The quantification of metabolites was calculated by an isotope ratio-based approach according to published methods ( Bennett et al . , 2008 ) . For the [13C5]-glutamine → citrate labeling analysis , the TMS-derivatized samples were analyzed using previously established GC-MS conditions ( Kind et al . , 2009 ) . The inlet temperature was set to 250°C and the inlet pressure was 11 . 4 psi . The purge flow was 3 mL/min and the split ratio was 2:1 . The stationary phase was an Agilent J&W DB-5MS +10m Duraguard Capillary Column ( 30 m × 250 μm × 0 . 25 μm ) ( Part number: 122-5532G , Agilent Technology , Santa Clara , CA ) . The oven temperature gradient condition was set as following: 0 to 27 . 5 min , linear gradient from 60 to 325°C; 27 . 5 to 37 . 5 min , 325°C . The EI-MS scan range was 50 to 600 m/z and the scan speed was 2 . 7 scans/sec . The solvent delay was set to 5 . 90 min . The MSD transfer line temperature was 250°C , The EI energy was 70 eV , the EI source temperature was 230°C , and the MS Quadrupole temperature was 150°C . The ions for Selective Ion Monitoring ( SIM ) of derivatized citrate ( 4TMS ) were as following: 465 [M]+ , 466 [M+1]+ , 467 [M+2]+ , 468 [M+3]+ , 469 [M+4]+ , 470 [M+5]+ , and 471 [M+6]+ . All peaks were manually integrated using Agilent MassHunter Qualitative Analysis Software and the natural abundance isotopes correction was performed according to published methods ( Mullen et al . , 2011 ) using IsoCor software ( Millard et al . , 2012 ) . Cells were plated in 96-well plates at a density of 5 , 000–10 , 000 cells/well in 100 μL final volume . The next day , the medium was changed to 100 μL medium containing CSA , Na2SO3 or HTAU at the indicated concentrations . 3–5 days later , CellTiter-Glo ( Promega ) was used to measure cell viability . Alternatively , cells were fixed with 4% paraformaldehyde , stained with crystal violet , washed and dried . Crystal violet was solubilized in 10% acetic acid and the OD600 was measured . Relative cell number was normalized to untreated cells . Experiments were repeated more than twice . Prior to the start of the proliferation assay , cells were treated with 0 . 25 μg/mL doxycycline for 2 days to induce CDO1 or GFP expression . Cells were seeded at 5 , 000 cells/well in 96-well plates and allowed to proliferate for the indicated time points . Plates were fixed with 4% paraformaldehyde , stained with crystal violet , washed and dried . Crystal violet was solubilized in 10% acetic acid and the OD600 was measured . Lysates were prepared in RIPA buffer ( 20 mM Tris-HCl [pH 7 . 5] , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 1% NP-40 , 1% sodium deoxycholate ) containing protease inhibitors ( Roche complete ) . Protein concentrations were determined by the DC protein assay ( Bio-Rad ) . Lysates were mixed with 6X sample buffer containing β-ME and separated by SDS-PAGE using NuPAGE 4–12% Bis-Tris gels ( Invitrogen ) , followed by transfer to 0 . 45 μm Nitrocellulose membranes ( GE Healthcare ) . The membranes were blocked in 5% non-fat milk in TBST , followed by immunoblotting . During this study , the Abcam CDO1 antibody was discontinued . The CDO1 antibody from Sigma Aldrich provides similar results for human and mouse CDO1 ( data not shown ) . NSCLC cells were plated in 6 cm dishes and analyzed at 70% confluence . 4 hr prior to harvest , cell culture medium was changed to fresh media . Cells were extracted and the NADPH/NADP+ ratio was measured using the NADP/NADPH-Glo Assay kit ( Promega ) according to the manufacturer’s instructions . Protein synthesis analysis was performed using the Click-It HPG Alexa Fluor 488 Protein Synthesis Assay Kit ( Life Technologies ) . Briefly , cells were treated with doxycycline ( 0 . 25 μg/mL ) for 24 hr , then seeded into black walled , clear bottom 96-well tissue culture plates at 20 , 000 cells/well in doxycycline-containing media . The following day , cells were treated with HPG ( 1 μg/mL ) in methionine-free DMEM ( 10% dFBS , 1% Pen/Strep , 10 mM HEPES ) . Cycloheximide ( 50 μg/mL ) treatment was included as a positive control . After 4 hr , the cells were fixed with 3 . 7% formaldehyde in PBS for 15 min at room temperature , and stained according to the manufacturer’s instructions . The fluorescence signal ( ex/em: 475/520 nm ) was measured using a Glomax plate reader ( Promega ) , followed by staining with crystal violet . The fluorescence signal was normalized to the crystal violet absorbance value . Soft agar assays were performed in triplicate in 6-well dishes . A 1 mL base layer of 0 . 8% agar in RPMI was plated and allowed to solidify , then 5 , 000 cells/well were plated in 0 . 4% agar on top . The following day , 1 mL of RPMI was added to each well , and changed as needed . Colonies were allowed to form for 10–14 days , and wells were stained with 0 . 01% crystal violet in a 4%PFA/PBS solution . Plates were scanned on a flatbed scanner and colonies quantified with Image J . Data were analyzed using a two-sided unpaired Student’s t test unless otherwise noted . GraphPad Prism seven software was used for all statistical analyses , and values of p<0 . 05 were considered statistically significant ( *p<0 . 05; **p<0 . 01; ***p<0 . 001 , ****p<0 . 0001 . In some figures with multiple comparisons # is used in addition to * ) . The mean ± standard error of experiments performed in at least triplicate is reported . Similar variances between groups were observed for all experiments . Normal distribution of samples was not determined . | Cancers form in humans and other animals when cells of the body develop mutations that allow them to grow and divide uncontrollably . The set of chemical reactions happening inside cancer cells , referred to as “metabolism” , can be very different to metabolism in the healthy cells they originate from . Some of these differences are directly caused by mutations , while others are a result of the environment surrounding the cancer cells as they develop into a tumor . A protein called NRF2 is often overactive in human tumors due to mutations in its inhibitor protein KEAP1 . Previous studies have shown that NRF2 changes the metabolism of cancer cells by switching specific genes on or off . However , since cancer cells also have other mutations that could mask or amplify some of the effects of NRF2 , the precise role of this protein in metabolism remains unclear . To address this question , Kang et al . generated mice that could switch between producing the normal KEAP1 protein or a mutant version that is unable to inhibit NRF2 . The mouse model was then used to examine the immediate effects of activating the NRF2 protein . This revealed that NRF2 altered how mouse cells used a molecule called cysteine , which is required to make proteins and other cell components . When NRF2 was active , some of the cysteine molecules were converted into two wasteful and toxic particles by an enzyme called CDO1 . Kang et al . found that inactivating CDO1 in human lung cancer cells prevented these wasteful particles from being produced . This allows cancer cells to grow more rapidly , and may explain why human tumors generally evolve to shut down CDO1 . The findings of Kang et al . show that not all of the changes in metabolism caused by individual mutations in cancer cells help tumors to grow . As a tumor develops it may need to acquire further mutations to override the negative effects of these changes in metabolism . In the future these findings may help researchers develop new therapies that reactivate or mimic CDO1 to limit the growth of tumors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cancer",
"biology"
] | 2019 | Cysteine dioxygenase 1 is a metabolic liability for non-small cell lung cancer |
Do developmental systems preferentially produce certain types of variation that orient phenotypic evolution along preferred directions ? At different scales , from the intra-population to the interspecific , the murine first upper molar shows repeated anterior elongation . Using a novel quantitative approach to compare the development of two mouse strains with short or long molars , we identified temporal , spatial and functional differences in tooth signaling center activity , that arise from differential tuning of the activation-inhibition mechanisms underlying tooth patterning . By tracing their fate , we could explain why only the upper first molar reacts via elongation of its anterior part . Despite a lack of genetic variation , individuals of the elongated strain varied in tooth length and the temporal dynamics of their signaling centers , highlighting the intrinsic instability of the upper molar developmental system . Collectively , these results reveal the variational properties of murine molar development that drive morphological evolution along a line of least resistance .
Evolutionary developmental biology postulates that developmental mechanisms confer specific variational properties on a trait , and can thereby channel its evolutionary trajectory . In extreme cases , a trait may repeatedly evolve similar phenotypes . Although this conceptual framework is central to evo-devo , it lacks cohesive supporting evidence . Only rarely the different levels of variation are bridged , from developmental variation to adult variation , and from variation between individuals to variation between populations or species . In this study focused on mouse molar teeth , we bridged these levels and reveal particularities of the developmental system that explain the morphological variation produced and its repeated appearance . The idea that developmental mechanisms may channel and even direct the evolution of phenotypes is central to evo-devo ( Brakefield , 2006; Brakefield , 2011; Hendrikse et al . , 2007 ) . It relies on the concept that developmental mechanisms bias the direction and the amount of variation available to both natural selection and neutral drift . This was recognized early , mainly under the term of ‘developmental constraints’ ( Gould and Lewontin , 1979; Smith et al . , 1985 ) and studied from different viewpoints . In the field of quantitative genetics , the analysis of phenotypic variation in crosses provides the direction of the genetic correlation between the different traits characterizing a shape . It was found that the direction of the genetic correlation between traits can match the direction of phenotypic variation within species , that itself matches the phenotypic variation between divergent populations or species . This suggested that phenotypic evolution happens along ‘genetic lines of least resistance’ ( Schluter , 1996 ) . Because the structure of the genetic correlations itself also match over long time spans ( e . g . the G matrix was found to be similar among distant species ) , these ‘genetic lines’ are thought to be produced by developmental constraints , rather than by the persistence of specific genetic variants . This finding was recovered in a number of models including the molars of murine rodents ( Renaud and Auffray , 2013; Renaud et al . , 2006; Renaud et al . , 2011 ) . The study of developmental systems in terms of their evolution also argues for a role of development in orienting morphological diversification ( Smith et al . , 1985; Sears , 2014 ) , including in the tooth model . This is recognized under the more specific term of ‘morphogenetic constrains’ . The patterns of variation recovered following experimental perturbations of amphibian development ( Alberch and Gale , 1985; Oster et al . , 1988 ) or mouse molar development ( Kavanagh et al . , 2007 ) predicted the pattern of morphological variation seen among species . By experimentally manipulating the mouse tooth germ or tinkering with one or two parameters of a computational model of tooth morphogenesis , Harjunmaa et al . ( 2014 ) have reproduced evolutionary transitions seen in the fossil record , implying that the same construction rules have constrained morphogenesis since early mammals . Despite this long interest and recent advances , there is still active discussion about how much development really influences evolutionary trajectories ( Laland et al . , 2014; Smith et al . , 1985 ) . One difficulty is that the different levels of variation are rarely bridged in a single model: from variation in embryo to variation in adult , and from variation in populations to variation between well-diversified species . Mouse molars represent a rare opportunity to construct such a bridge: the Mus genus is well diversified , with many instances of repeated evolution and well-characterized trajectories of phenotypic variation in molar shape . Moreover , molar development is well known in the laboratory mouse . Mice are part of the larger group of murine rodents ( Old World mice and rats ) . In this group , the main direction of phenotypic variation in first molar shape divides species with narrower molars ( with narrower cusps , for example dwarf mice of Nannomys subgenus in Figure 1 ) from species with broader molars ( with broader , more roundish cusps , e . g . wood mice or grass mice of Arvicanthis genus , Figure 1 ) . These differences in tooth morphology have been associated with different diet preferences , narrow teeth being mostly found in faunivorous rodents while broad teeth are characteristics of herbivorous ones , because of the latter allowing the consumption of harder and more abrasive food ( Gómez Cano et al . , 2013 ) . Molar tooth morphology thus reflects adaptation in murine rodents as seen in many other mammals . However , whatever the mean morphology of a taxon , the variation within a taxon ( e . g . house mouse and wood mouse ) , including at the population level , seems to reproduce , to a lesser degree , the basic variation ranging from narrow to broad tooth ( Renaud et al . , 2009; Renaud et al . , 2011 ) . Such micro-scale variation is more likely to be shaped by developmental properties , rather than adaptation . The high integration between the variation of lower and upper first molars suggests that both evolved in a concerted manner under similar developmental constraints . In summary , this alignment of the main phenotypic variation across the taxonomic scale suggests that murine molars evolve of along a line of least resistance , with adaptation occurring along the line imposed by developmental properties . On top of that , in some species or populations , only the upper molar tends to elongate , specifically from its anterior part which may even form an additional small cusp ( Misonne , 1969; Renaud et al . , 2011 ) . This additional cusp is especially common in the Mus genus ( yet occasionally seen in other murine species Misonne , 1969 ) . For example , it is especially marked in some species of the Mus ( Nannomys ) subgenus , and also repeatedly seen in diverse house mouse populations ( Renaud et al . , 2011 , see later Figure 1 ) . In particular , it evolved independently in many Mus ( Mus ) musculus domesticus island populations ( e . g . on several Corsican islands , Marion Island [Renaud et al . , 2011] , Orkney islands [Ledevin et al . , 2016; Renaud et al . , 2018] , as well as on Kerguelen and the Canary islands , S . Renaud , [Ledevin et al . , 2016] ) . Interestingly , anterior elongation is found associated with increased body size in domestic mouse populations ( Renaud et al . , 2011 ) , being trapped on island ( and following ‘Foster’s rule’ , or ‘island rule’ where small mammals become gigantic ) or in cold environments . In conclusion , two intermingled developmental constraints seem to act here to channel phenotypic variation and evolution in murine rodents ( or said differently , to shape a line of least resistance ) : one acts on both first molars , and the other one acts on the upper molar only , favoring its repeated anterior elongation in many independent populations and species . Molar development is well known ( Balic and Thesleff , 2015; Peterkova et al . , 2014 ) , so that in a previous study , we had put forward some hypothesis for the developmental basis of this line of least resistance characteristic of the upper molar ( Renaud et al . , 2011 ) . During molar development , signaling centers , called ‘enamel knots’ , are positioned in the epithelium by activation-inhibition mechanisms , and determine the location where the crown , or later the cusps , form Jernvall et al . ( 1994 ) ; Sadier et al . ( 2019 ) . Although murine rodents lack premolar teeth , structures found transiently in mouse embryos are thought to correspond to their rudiments , each with their own signaling centers , notably secreting the signaling molecule Shh ( Lesot et al . , 1998; Peterková et al . , 1996; Prochazka et al . , 2010; Viriot et al . , 2000; Peterková et al . , 2002 ) . Following molar row initiation around 11 . 5 dpc , the dental epithelium progressively invaginates into the mesenchyme along the antero-posterior jaw axis , and Shh-signaling centers are patterned sequentially in this epithelial ridge ( Prochazka et al . , 2010; Sadier et al . , 2019 ) . In mandible , the first and most anterior signaling center , associated with a discrete epithelial swelling , has formed by 12 . 5 dpc but disappears soon after ( in CD1 mice ) . A second signaling center ( R2 signaling center ) has formed more posteriorly by 13 . 5 dpc , where the epithelial ridge locally enlarges and gives rise to a more prominent bud called R2 . This structure was interpreted as a rudiment of a suppressed premolar ( Prochazka et al . , 2010; Viriot et al . , 2000 ) , and can form a small tooth in mutant conditions interpreted as an atavistic premolar ( Cobourne and Sharpe , 2010; Lagronova-Churava et al . , 2013; Lochovska et al . , 2015; Peterková et al . , 2005 ) . This signaling center also soon vanishes and by 14 . 0 dpc , a third , even more posterior , signaling center has formed , called the early M1 signaling center . In the lower jaw , the R2 signaling center interacts in a complex manner with the signaling center of the first molar . Both transiently co-exist before fusing to form the mature M1 signaling center known as the pEK ( Lochovska et al . , 2015; Sadier et al . , 2019 ) . This pEK drives ‘cap transition’ , a morphogenetic transition during which the epithelium starts to fold around the mesenchyme to form the tooth crown . From an early cap stage , the R2 primordium becomes thus integrated into the anterior part of the first lower first molar . Although a similar R2 bud is also present in the upper jaw ( Lesot et al . , 1998; Peterková et al . , 1996 ) , its developmental relationship to the upper first molar is less clear . It is not incorporated during the initial stage of molar cap development , as seen for its lower counterpart ( Peterkova et al . , 2006 ) . Recently , we have shown that the distance between R2 and M1 signaling centers is larger in the upper jaw , where the two signaling centers do not fuse at the cap stage ( Sadier et al . , 2019 ) . The anterior position of R2 and the difference between lower and upper R2 , make these structures excellent candidates to explain that the upper molar , and not its lower counterpart , evolves so repeatedly towards anterior elongation ( Renaud et al . , 2011 ) . To get insights into the developmental basis of upper first molar repeated evolution , we looked for two mouse strains reflecting the above-mentioned evolutionary variations: a ‘broad and upper-short’ versus a ‘narrow and upper-elongated’ strain . It was noted long ago that two laboratory mouse strains displayed the elongated upper molar morphology with a small additional anteriormost cusp ( C3H , 101 strain from Harwel , Grüneberg , 1965 ) . In our effort to test the correlation between this morphology and a large body size , we found that the DUHi mice , an inbred strain that was established following artificial selection for increased body size ( Bünger and Herrendörfer , 1994; Bünger et al . , 1982 ) , display narrow molars with elongated first upper molars , and an additional cusp in some individuals ( Figure 1—figure supplement 1 ) . In contrast , the FVB mice , an inbred strain often used to maintain genetic modifications , display wide molars with short first upper molars ( Figure 1—figure supplement 1 ) . After checking that these strains mirror mouse molar phenotypic evolution , we looked for developmental variation in the dynamics of R2 rudimentary buds between strains and jaws , but also within strains . We asked whether variational properties of the upper molar developmental system , whether qualitative or quantitative , may predict the evolutionary variation of the first upper molar .
In order to place these two strains within the context of natural variation in molar tooth shape , we compared the outline of first molars in a number of murine groups , including Mus musculus domesticus from the wild and three lab strains ( two inbred strains: FVB , DUHi; one outbred strain CD1 ) . We performed a principal component analysis ( PCA ) of outline descriptors ( obtained from an outline analysis of the 2D outline , see methods ) as an agnostic method ( i . e . it does not make strong assumptions about the shape of the underlying variation ) to reveal the main direction of variation in the dataset , and a direct comparison of molar Length/Width ratio ( Figure 1 ) . For both the upper and lower molar , the first axis of the PCA contrasts broad with narrow outlines . Hence , this can be considered to be the most important aspect of the outline variation for both teeth , with a morphology ranging from short , compact and rounded teeth to long and narrow teeth . However , this variation is more pronounced for the upper molar ( PCA UM1 = 62% instead of PC1 LM1 = 46% of variance ) because it also involves a change focused at the anterior part of the tooth , and opposing short vs . anteriorly elongated UM1 . This variation corresponds to the evolutionary trend seen repeatedly for the upper molar in the Mus genus: the anterior elongation that can even take the form of a small additional cusp ( Renaud et al . , 2011; Stoetzel et al . , 2013 ) . We note that although the PCA of the outline descriptors could in principle separate these two types of variation , this was not the case , showing that these two types of variation are correlated in the upper molar ( i . e . anterior elongation occurs in otherwise narrow molars , while broad molars have a short anterior part ) . The position of FVB and DUHi strains along the PC1 axis and length/width ratio indicate that the direction of variation between the two strains recapitulates the direction of variation seen in murine rodents as a whole ( e . g . broader FVB molars versus narrower DUHi molars , lower and upper M1 ) . Moreover , it recapitulates the variation specifically seen in the upper first molars of the Mus genus ( short FVB versus anteriorly elongated DUHi upper molars , with additional cusp ) . Indeed , the two strains are representative of extreme wild Mus musculus domesticus samples: DUHi teeth fall with the most elongated upper molars ( e . g . samples from a very small Corsican island: Piana ) , while the FVB upper molars fall with the most ‘short and wide’ upper molars ( e . g . samples from the continent ( Gardouch locality ) or from a large Orkney island: Eday ) . This validates the choice of these strains , as recapitulating the two intermingled main phenotypic variations seen in murine rodents . The next step was to examine the developmental basis of interstrain variation in lower and upper molars . In order to allow the developmental trajectories of the developing molar teeth to be compared , embryonic samples were taken from a wide developmental window in the two strains in question ( 12 . 0 to 15 . 5 dpc ) . For each strain , multiple litters were sampled every half day , ensuring even coverage due to the slight variation in developmental stage within litters . Taking account of the dynamic nature of the system required a numeric estimation of the embryonic age of embryos ( i . e . an age reflecting the progress in embryogenesis ) . Neither the age in days post-coïtum nor the embryonic body weight solely provide a correct estimation ( for further explanation , see Materials and methods and Appendix 1 ) . Therefore , in this study , we devised a Bayesian modeling approach to compute a strain-specific embryonic age for each embryo ( later called cdpc ) by combining these two information ( see Materials and methods and Appendix 1 for further precisions ) . Such a model learns from the data , and could also take into account strain differences without imposing a priori assumptions . Because the model does not comprise information on actual stage differences and we estimate cdpc for each strain , it will not correct for developmental stages differences between strains , but instead reveal them by providing a time framework , the computed age ( e . g . cap transition occurs at earlier cdpc in FVB ) . The detailed model and the script are provided in Appendix 1 . An important stochastic term in the model corresponds to the part of the inter litter variation of the body weight due to pregnancy , for which we explored two values ( Figures of Appendix 1 ) . The first one , which we consider realistic , corresponds to a maximum effect on weight of 20 mg for a 200 mg embryo . The second one , which we consider an extreme upper bound , corresponds to a maximum effect on weight twice as important ( 40 mg for a 200 mg embryo ) . The computed embryonic ages ( next called cdpc , for computed days post coïtum ) presented in the main text have been estimated using the realistic parameter , but the results are qualitatively robust using the upper bound parameter ( Appendix 1—figures 2–8 ) . This demonstrates that our results are robust to noise in embryonic age estimation . We also used embryonic age directly estimated from body weight ( a simplification of the previous model , similar to what we used in our previous studies , Pantalacci et al . , 2009 ) . All the results shown in this study were robust in relation to these estimations , although they differ slightly ( Appendix 1—figures 2–8 ) . In the next paragraphs , the results obtained for upper molar are mostly presented in the main figures and results for lower molar will be found in Appendix 1 . We proceeded to compare the DUHi and FVB lower and upper developmental systems . In both jaws and strains , we see Shh expression at the signaling center of the R2 bud ( Figure 2 and Figure 2—figure supplement 1A–D ) . This expression then fades away ( Figure 2—figure supplements 1 and 2 E-H ) . A second spot of expression appears , which represents the early M1 signaling center ( Figure 2 and Figure 2—figure supplement 1I–L ) . As development proceeds to the cap transition stage , the M1 expression zone increases in size ( Figure 2 and Figure 2—figure supplements 1–2 ) to form a ‘mature M1 signaling center’ ( differences between lower and upper jaw in this process will be emphasized later ) , and the tooth continues to develop ( Figure 2—figure supplements 2–3 ) . A simple examination of embryonic series suggested differences in the dynamics of the signaling centers ( Figure 2 , e . g . see I-J versus K-L ) . In the upper jaw of DUHi mice , we frequently see the co-occurrence of the fading R2 spot with a distinct M1 spot ( Figure 2M–T and Figure 2—figure supplements 1–2 M-T ) . To allow comparisons between jaws and strains , we proceeded to examine development in a quantitative fashion ( Figure 3—figure supplements 1–2 ) . For each sample dated in cdpc , four key characters were scored for two to three possible states ( characters were: the expression of Shh at the premolar R2 signaling center , the expression of Shh at the M1 signaling center , the progression of the bud-cap transition at the level of M1 , and the protrusion of the dental epithelium at the level of R2 , Supplementary file 1 ) . Then , we combined these four characters to compute the developmental state of each sample ( Figure 3 and Figure 3—figure supplement 3 ) . Mathematically speaking , this yields 54 possible total states of the dental epithelium , which can be conceptualized as a theoretical ‘developmental space’ through which each individual will move as it develops . However , not all of this space is occupied; the number of states observed was much smaller than the mathematical maximum theoretically possible ( 11 states for the upper molar , Figure 3; and 9 states for the lower molar , Figure 3—figure supplement 3 ) . This is because the unfolding of normal development imposes rules on characters’ state transitions and combinations . For example , transitioning from a well-developed cap to a bud seems impossible , because normal tooth development proceeds from bud to cap stage . Yet it remains conceivable in mutant situations , where morphogenesis would be stopped and the tissue would start regress . Combining no protuberance in R2 zone with a cap transition is unexpected , because R2 develops before M1 . Combining large M1 Shh expression and no cap transition is also unexpected , because this large expression marks the presence of a PEK and PEK drives cap transition . Yet , again in mutant mice , things may happen differently , because the normal rules of development may be broken ( e . g . defective PEK despite normal Shh expression ) . This emphasizes that the actual occupied proportion of developmental space and the trajectories through this space are properties of a given genotype with a given development . A comparison of the distribution of developmental states seen in DUHi and FVB upper molar samples reveals that certain states are only present in one of these two strains . This is especially true for upper molar ( 3 DUHi-exclusive states , 2 FVB-exclusive states ) as compared with lower molar ( 1 DUHi-exclusive state , evidenced by only two samples ) . Indeed , within a given weight range ( from 14 . 25 to 14 . 75 cdpc; 175 mg to 225 mg ) , all DUHi samples display a ‘DUHi-exclusive’ state , while all FVB samples display an ‘FVB-exclusive’ state . This corresponded to the period between the disappearance of the R2 signaling center and the maturation of the M1 signaling center . Therefore , the developmental differences between these two strains can be conceived of as each of the two strains following distinct trajectories through ‘developmental space’ . Moreover , these differences are especially marked in the upper molar . We then compared the timing of developmental events between the two strains . In upper molars ( Figure 3 and Figure 3—figure supplement 1 ) , the R2 expression zone persists longer in DUHi than in FVB ( logistic regression: presence of R2 signaling center activity in relation to computed embryo age , p<1 . 10E-8 , see Supplementary statistical details 1 ) . The R2 and M1 signaling center frequently co-occur in DUHi , and co-occurence of R2 and a mature ( large ) M1 signaling center was only seen in DUHi upper molars . In contrast , early M1 signaling center was seen without a R2 signaling center only in FVB mice . A logistic regression also revealed a statistically significant difference in the timing of the appearance of the M1 spot , which appears later in DUHi than in FVB embryos ( p<0 . 01 , see Supplementary file 4 ) . The longer persistence of R2 signaling center and later appearance of the early M1 signaling center in the upper jaw of DUHi mice were associated with a more prominent R2 bud , as seen in dissociated epithelium ( Figure 2 , Figure 4-F ) but also in 3D reconstructions of tooth germs ( Figure 4A–D ) . In the lower jaw , we observed a similar tendency , with R2 and M1 signaling centers occasionally coexisting only in DUHi , but never in FVB ( Figure 3—figure supplements 2–3 ) . In both strains , the lower R2 signaling center was ultimately part of the mature M1 signaling center known as the pEK ( Figure 2—figure supplement 1R and P ) , as shown for other mouse strains ( Lochovska et al . , 2015 ) . We recently proposed that activation-inhibition mechanisms acting in a posteriorly growing domain rules the patterning and fate of the R2 signaling center ( Sadier et al . , 2019 ) . The finding that R2 is larger and longer lived in DUHi mice suggests that the mechanism proposed in our previous study differs between the two strains . This could be at different levels , from posterior growth , rate of maturation , influence of mesenchyme or activation-inhibition mechanisms per se . A prediction is that the positioning of the M1 signaling center should differ between the two strains . Our measurements revealed that the M1 signaling center was shifted posteriorly in DUHi mice: the pre-M1 signaling center region was longer in DUHi than in FVB samples ( Figure 4H , t-test , p<0 . 01 ) , while the post-M1 signaling center region was shorter in DUHi upper molars than in FVB ( Figure 4I , t-test , p<0 . 01; see Supplementary file 4 ) . No statistically significant differences in total dental epithelium length was found ( Figure 4J ) . In the case of the lower molars ( Figure 4—figure supplement 1 , see Supplementary file 4 ) , no statistically significant inter-strain differences emerged . This finding suggests that the posterior growth rate is unchanged between the two strains . According to the sensitivity analysis performed in our previous study , two other parameters on top of posterior growth rate ( Figure S3a in Sadier et al . , 2019 Supplementary Material ) could produce more distant signaling centers ( and possibly R2 rescue ) : one modulates activation-inhibition per se ( Figure 5 and S2a in Sadier et al . , 2019 ) and the other modulates the rate of production of the mesenchyme signal that primes the tissue for activation-inhibition ( Figure S4a in Sadier et al . , 2019 ) . Unfortunately , it is very unclear so far which molecular pathways would contribute to each of these two parameters . Despite that , we know many pathways that contribute positively or negatively to molar formation ( inc Wnt , BMP , FGF , Activin , Edar pathways ) . In addition , some genes are known to suppress the potential of R2 to form a premolar-like tooth ( e . g . Klein et al . , 2006; Peterkova et al . , 2009; Sadier et al . , 2019 ) . Next , we looked for gene expression differences between the two strains that would be consistent with a difference in these known pathways . For that , we generated transcriptomes of lower and upper molar germs of FVB versus DUHi mice at the time when R2 and M1 signaling center co-exist in DUHi mice . We found a large number of differentially expressed genes ( see Appendix 2 for detailed results , data available in Supplementary file 2 ) . Among all those DE genes , we identified two genes , Spry2 and Sostdc1 ( also known as Ectodin ) , whose knock-out causes the formation of a premolar-like tooth ( Ahn et al . , 2010; Cho et al . , 2011; Klein et al . , 2006 ) . Where it was specifically investigated , this premolar tooth was demonstrated to arise from R2 revival ( Ahn et al . , 2010; Klein et al . , 2006; Peterkova et al . , 2009 ) . The downregulation of these two genes in DUHi lower and upper molar samples ( Figure 4—figure supplement 2 ) could thus help the partial rescue of R2 bud in this strain . We then focused on the Wnt and BMP4 pathways that have been shown to be key for tooth formation ( O'Connell et al . , 2012 ) . Because tooth formation involves many genes with complex regulatory feedbacks within and between these two key pathways ( and other pathways ) , we did not expect a change in activation-inhibition mechanisms to shift all target genes in a consistent direction . Rather , we expected to find a different equilibrium , with genes changed in both directions , but that may collectively indicate greater or instead weaker activation of these pathways in the DUHi mice . For the BMP4 pathway , there were 20 genes in favor of weaker BMP4 activation in DUHi ( e . g . summing activated targets that are upregulated with repressed targets that are downregulated , see Appendix 2 for more detail and Figure 4N ) versus only 7 genes in favor of greater BMP4 activity ( e . g . summing activated targets that are downregulated with repressed targets that are upregulated ) . This two-fold difference is significant ( p=0 . 01 in a χ2 test ) . For the Wnt pathway , we found no trend with 23 genes in favor of weaker Wnt activity in DUHi and 20 genes in favor of greater Wnt activity ( Figure 4N ) . Intriguingly , this included 4 feedback inhibitors of the Wnt pathway upregulated in DUHi mice ( Axin2 , Kremen1 , Osr2; Sfrp2 ) and 3 feedback inhibitors downregulated in DUHi ( Dkk1 , Wif1 , Sostdc1 ) . Finding these major regulators of Wnt activity in tooth development differentially expressed suggests that Wnt activity differs between FVB and DUHi , although we cannot orient it as for the BMP pathway . This is consistent with recent findings suggesting that both activation and inhibition of the Wnt pathway is required to make teeth , and Wnt activation needs to be carefully controlled by feedback mechanisms ( including a crosstalk with the BMP4 pathway ) to enable the sequential formation of teeth ( Järvinen et al . , 2018 ) . In conclusion , transcriptomic data further suggest a difference in the activation-inhibition balance between the two strains . The balance tends toward weaker BMP4 activity in DUHi mice at the cap transition stage , suggestive of lowered levels of activation in these mice ( either directly , if BMP4 is part of activation-inhibition mechanisms per se , or indirectly , as part of the mesenchymal signal that enables activation-inhibition ) . To further establish that these strains differ in activation-inhibition balance , we tested another prediction , that the two strains should react differently to the same perturbation of activation-inhibition mechanisms . Simply dissecting and culturing teeth ex vivo is known to provide such a perturbation , resulting in incisor germ splitting ( Häärä et al . , 2012 ) or in a partial rescue of the lower R2 rudimentary bud ( Sadier et al . , 2019 ) . Culturing upper molars , we have occasionally observed such a rescue of the upper R2 bud which starts to form an independent tooth cap ( Figure 4M ) . This occurred much more frequently in DUHi than in FVB mice ( Figure 4K–M; Fisher exact test; p=0 . 015 ) . This is consistent with R2 being partially rescued in the DUHi mice . We concluded that the activation-inhibition balance differs between the two strains . Next , we asked if and how these differences could explain an anterior elongation of the adult M1 that is specific to the upper molar . In a previous study , we had put forward differences between the lower and upper molar developmental system in CD1 mice ( Sadier et al . , 2019 ) . In both strains , we found similar lower-upper jaw differences as seen in CD1 mice , namely 1 ) R2 persisted longer in the upper than in the lower jaw , as tested via a Fisher exact test on samples for which data were available for both upper and lower jaws of the same embryo ( p=0 . 04 ) . 2 ) The M1 Shh expression zone increased in size in upper as in lower jaw , but in the lower jaw only it encompassed the zone of the R2 signaling center ( compare Figure 2Q–T and Figure 2—figure supplement 1Q–T ) . Thus , the spatio-temporal dynamics and fate of the R2 signaling center relative to M1 signaling center differs between the two jaws , regardless of difference in activation-inhibition mechanisms between strains . This can be considered a conserved developmental property of the lower and upper developmental systems . This may be the foundation for the lower and upper jaw developmental system reacting non-linearly to a same genetic change between FVB and DUHi mice: the increase in R2 signaling center persistence may be all the stronger as the R2 signaling center is already more persistent in the upper jaw and preserved from an early fusion with the M1 signaling center . But why would this result in anterior elongation in the upper molar only ? Answering this question requires that we make comparisons between the two jaws in order to reveal how the R2 bud may contribute to the first molar . In order to compare developmental relationship between the R2 bud and M1 in the upper and lower jaw of the DUHI and FVB strain , we genetically tracked the fate of R2 signaling center in the developing molar . Using a tamoxifen-inducible Cre/LacZ line , we were able to induce the marking of Shh-expressing cells and their descendants at different timepoints during tooth development . In practice , cells expressing Shh at the given timepoint of tamoxifen injection ( in practice , a time window corresponding to tamoxifen elimination that may be 24–48 hr ) recombine a lacZ transgene , so that these cells and their descendants will then be positive ( blue ) to a X-gal staining . By inducing the tamoxifen at different timepoints ( Figure 5A , B , Supplementary file 3 for a summary of all experiments ) , we could mark cells that descend from either the R2 signaling center and the M1 signaling center population ( Figure 5C , F , treatment at 12 . 5 dpc ) or from the M1 signaling center population only ( Figure 5D–H , treatments at 13 . 5 and 14 . 5 dpc ) . This was done in a CD1 background , where the dynamics of R2 and M1 signaling centers is well known ( Lochovska et al . , 2015; Prochazka et al . , 2010; Sadier et al . , 2019 ) and molars are morphologically intermediate between FVB and DUHi mice ( Figure 1 ) . For upper molars treated during R2 signaling center activity , the anterior part of the tooth is marked at 17 . 5 dpc ( Figure 5C ) . However , this anterior region is unmarked when tamoxifen is applied later , during M1 signaling center activity only ( Figure 5D–E ) . Thus , the fate of the cell populations of R2 signaling center is the anterior region of the first upper molar . In the case of the first lower molar , the same region is stained in all three conditions ( Figure 5F–H ) , consistent with R2 signaling center being overwritten by the mature M1 signaling center ( Sadier et al . , 2019 ) . Moreover , the anterior part of the first molar is unstained . Thus , in the lower molar , the R2 signaling center does not contribute specifically to the most anterior part . From these results , we can deduce that a change in R2 size and R2-M1 centers distance , as seen in DUHi upper jaw , will directly elongate the anterior part of the upper M1 ( Figure 5C ) . In contrast , the more modest change seen in the DUHi lower jaw may elongate the M1 , but not specifically from its anterior part ( compare Figure 5I with Figure 5D–E ) . We conclude that intrinsic differences between lower and upper M1 developmental systems are responsible for their different reactions: marked anterior elongation in the upper molar and a more discrete and isometric elongation of the lower molar ( see Figure 5I ) . A brief examination revealed that DUHi samples display different states for similar embryonic age ( Figure 3 and S7 ) . This suggested that the two strains may exhibit different degree of developmental variability , which we aimed to quantify . This requires disentangling differences in developmental state due to error in embryonic age estimation ( e . g . two embryos are different because they were erroneously attributed the same age ) from real differences due to developmental variation ( e . g . two embryos , effectively of the same age , display different states ) . We developed a method to measure developmental variation , which is described in detail in the Materials and methods section . This method utilises the developmental state scoring system discussed previously . The method works by identifying cases where samples differ greatly in terms of developmental state despite being approximately equally old and allows for statistical comparison of the degree of developmental variation present , via a Wilcoxon rank-sum test . We applied this method to determine whether there is greater developmental variation present in DUHi upper molars than FVB upper molars and in DUHi lower molars than FVB lower molars ( Figure 6A and Figure 6—figure supplement 1A ) . Our method yielded a highly significant result in both cases , ( p<0 . 001 for Wilcoxon tests ) where DUHi was more variable than FVB in both cases . Therefore , not only do the developmental trajectories of these two strains differ from one another , but the degree of variation within each strain is not equivalent . Having established that DUHi has more overall developmental variation than FVB , the next step was to develop a method to examine the temporal profile of developmental variation . This was achieved by tracking the difference in developmental state between each embryo and all other embryos of the same strain that were close in age ( less than 0 . 25 days difference in computed embryo age ) . The local regression line through the developmental state difference present at a given time was then plotted , and is seen in Figure 6B ( upper molars ) and Figure 6—figure supplement 1B ( lower molars ) . A similar pattern is seen in both upper and lower molars , whereby there is a small peak in developmental variation between computed age 14–15 cdpc/weight: 150–250 mg , followed by a decrease in developmental variation . This peak is considerably greater , in both duration and magnitude , in DUHi than FVB samples . Some differences exist between the upper and lower molars , however . In the upper molars , it corresponds to the period when two signaling centers coexist and the M1 signaling center expands ( 14 . 5–14 . 75 cdpc/Weight: 200–220 mg , Figure 6B ) . In the lower molar , it also matches this period in DUHi mice ( around 14 . 5 cpdc/Weight: 200 mg , Figure 6—figure supplement 1B ) , but rather corresponds to an earlier variability in termination of R2 signaling for FVB mice ( around 14 . 1 cpdc/Weight: 160 mg ) . Having established that DUHi embryos display greater variation during development than is seen in their counterparts at equivalent stages in FVB , the next step was to examine whether this developmental variation would be reflected by greater variation in adult morphology . The variation in length and width in DUHi , FVB and CD1 adults was examined for both upper and lower first molars and is shown in Figure 6C and Figure 6—figure supplement 1C . Although they are also an inbred strain , DUHi mice show greater variation in upper molar length than FVB ( p=0 . 095 ) , to an extent comparable to the outbred CD1 mice . The variation in DUHi upper molar length is consistent with the large degree of variation in R2/M1 development in DUHi embryonic upper jaws . Consistent with the developmental data , this variation in length is specific to the upper molars: In the lower molars , DUHi individuals show little variation in length , albeit large variation in width ( Figure 6—figure supplement 1C ) .
Evolutionary trajectories are not random , but tend to take preferred routes . Several factors can explain this , among them the ability of developmental systems to vary in particular directions . Then variation typically follows similar trends at different levels , from inter-individual differences to population differences , up to species differences . This underpins the concepts of evolutionary ‘lines of least resistance’ ( e . g . Schluter , 1996 ) or ‘evolutionary predictability’ inferred from developmental systems rules ( Kavanagh et al . , 2007; Salazar-Ciudad and Jernvall , 2010 ) . Lines of least resistance were primarily thought of as arising from genetic constraints Schluter , 1996 ) , whereas the second concept more directly refers to variational properties of developmental systems . Both relate to the old concept of ‘developmental constraints’ ( e . g . Smith et al . , 1985 ) . These concepts are typically tested by matching the variation at different levels ( e . g . population/macroevolution ) or by simulating it with developmentally realistic models . However , they have rarely been tested by directly examining the variation in developmental systems and its systemic basis , at least for mammalian models . Here , we used a comparative approach , focusing on fine-scale developmental differences that show evolutionary-relevant variation to unravel the generative principles underlying shape variation . We have used inter-strain variation as a proxy for natural variation . We think this choice is relevant for three reasons: 1 ) We show that the morphological variation between FVB and DUHi mice recapitulates the morphological variation in the Mus genus and beyond 2 ) The DUHi mice are a product of artificial selection in the lab for large body size , starting from six different mouse strains ( Bünger and Herrendörfer , 1994; Bünger et al . , 1982 ) . It thus also recapitulates the correlation between anterior elongation and large body size seen in natural populations of Mus musculus domesticus ( Renaud et al . , 2011 ) . 3 ) Lab mice were derived from a mix of several mouse subspecies ( Mus musculus domesticus , Mus musculusmusculus , Mus musculus castaneus ) . This original substantial amount of genetic variation being now split between mouse strains , it is no surprise that different mouse strain could recapitulate trends seen in the Mus genus . We show that the anterior development of the first upper molar varies between DUHi and FVB strains . We show inter-related differences such as , in DUHi mice compared to FVB mice , the larger size of R2 rudimentary bud , the longer persistence of its signaling center , a posteriorisation of the M1 signaling center , a marked tendency for R2 to individualise in culture and differences in gene expression . A similar tendency , although less marked , is seen in the DUHi lower molar . Collectively , these results show that these two strains differ in the settings of the activation-inhibition mechanisms patterning the tooth signaling centers . Finding different settings of activation-inhibition mechanisms between two mouse strains is not surprising since mapping studies have found genetic variation segregating with relative molar proportions ( Navarro and Murat Maga , 2018 ) . We have indications that the balance is shifted for the BMP4 pathway , with the expression levels of BMP4 targets suggesting overall weaker BMP4 activity in DUHi mice , which would mean reduced activation in DUHi mice ( activation is taken here in a broad sense , and may include different mechanisms , as suggested in the results ) . Although R2 partial rescue in a context of reduced activation may appear counter-intuitive at first glance , it is in line with our recent study showing that reduced activation in culture or in the Edar mutant favors the R2 bud in its competition with the M1 signaling center and thereby tends to rescue it Sadier et al . ( 2019 ) . We note that the narrow morphology of DUHi ( lower and upper ) molars as compared with the broad , massive morphology of FVB molars is also consistent with decreased levels of activation in this strain . We have noticed in a study of Orkney house mouse populations that the presence of the anterior cusp was associated with a decrease in overall cusp complexity , again consistent with overall decreased level of activation ( Renaud et al . , 2018 ) . In conclusion , our results thus suggest that the two lines of least resistance seen in murine rodents: the anterior elongation of the upper molar , and the variation in shape seen in both molars are developmentally coupled by a common setting of the activation/inhibition balance . Further work will be necessary to disentangle the different mechanisms that contribute to this balance for example , as proposed in Sadier et al . ( 2019 ) and reveal the basis for the difference between DUHi and FVB . Our study also sheds light on the developmental reasons why these variations in activation-inhibition mechanisms would turn into elongation of the anterior part of the adult molar , specifically and repeatedly in the upper first molar . First , we show that the R2 signaling center is intrinsically stronger in the upper jaw , remains independent of the M1 signaling center and contributes to the anterior part ( cervical loop ) of upper M1 . This is contrasted with the lower jaw , where R2 bud is smaller , included in the mature M1 signaling center ( so-called pEK ) and R2 signaling center cells do not contribute to the anterior cervical loop , but to a more central part of the molar . As a consequence , subtle variations in activation-inhibition mechanisms that would affect R2 signaling center could specifically impact the anterior part of the upper M1 only , whereas this effect could be either buffered in the lower M1 or spread on the whole tooth ( DUHi lower M1 are longer than FVB ) . Secondly , upper molar development appears to be inherently more variable than lower molar development . DUHi and FVB mice are more different in terms of upper molar development than they are in terms of lower molar development; variation in R2 signaling center persistence , in anterior region size and M1 positioning are all stronger for upper than lower molars . This intrinsic variability of the upper molar is also apparent in the phenotypic variability of the DUHi mice . Although these mice are inbred , the upper molar is much more variable in length and this is correlated , again , with greater developmental variability in upper R2 signaling center ( we note , however , that the lower M1 of DUHi mice is highly variable in width , Figure 6—figure supplement 1: this might indicate that the lower M1 also reacts to changes in R2 , albeit very differently ) . This suggests that upper molar development is intrinsically unstable , especially when the tuning of activation-inhibition parameters comes in the ‘DUHi range’ . In summary , we provide evidence for developmental particularities of the first upper molar , explaining why it responds to variations in activation-inhibition parameters by varying its anterior morphology . We thus uncovered the basis for an evolutionary line of evolutionary least resistance in murine rodents , leading to repeated morphological evolution of the first upper molar in mice . A common conception is that developmental variation is minimal at early stages and increases over developmental time . Thus , morphological changes are assumed to result from small changes in development , which result from yet smaller changes in earlier development . This ‘inverted funnel’ bears a certain resemblance to Von Baer’s law of embryonic divergence , and to the ‘hourglass model’ of interspecific developmental similarity ( Abzhanov , 2013; Irie and Kuratani , 2014 ) . Under this view , mild variations in adult phenotypes should result from almost undetectable or at least late-detectable variation in development . Here , our work identifies strong variation in early tooth development between strains ( morphology of the tooth germ and dynamics of signaling centers ) in tandem with relatively mild variation in adult phenotype . Besides influencing our view of developmental variation , this has implications for developmental biology practices , especially in the mouse model in which access to embryo is limited for both ethical and cost reasons . For example , heterozygotes are often used as controls for developmental genetic studies , where heterozygotes do not display an obvious phenotype in adults , because it is assumed that the development underlying that phenotype proceeds normally . However , our work here finds greater variation in early development than in the adult phenotype . In systems like that observed here , heterozygotes may have important differences in early development , making their use as controls unreliable . We also note that discrepancies between observations made in different labs will a priori not be attributed to a difference in the wild type strain used in each lab . In this context , it is interesting that the recognition of R2 vestigial bud presence in mouse molar development was a matter of debate for several years , and R2 is much less transient and discrete in the CD1 strain , where R2 was first identified , than in FVB , in which much heavier sampling is needed to catch the short developmental window when it is present . Therefore , it is well possible that differences between strains worked to obscure the debate . In conclusion , we believe that enhancing the focus on developmental variation will be important to move on from overly simplistic views of developmental variation that more or less consciously influence our practices in biology .
DUHi mice were raised at the PBES; cryopreserved embryos had been obtained from MRC Mary Lyon centre , Oxfordshire , UK . FVB and CD1 mice were purchased from Charles River company . C57BL/6 mice carrying tamoxifen-inducible Cre fused with the Shh allele ( B6 . 129S6-Shh < tm2 ( cre/ERT2 ) Cjt>/J ) and Cre recombinase-sensitive transgenic mice cB6 . 129S4-Gt ( ROSA ) 26Sortm1LacZSor/J ) containing LacZ ( beta-galactosidase ) inserted into the Gt ( ROSA ) 26Sor locus were used for the cell fate tracing study . The breeding pairs were purchased from the Jackson Laboratory ( Maine , USA ) . The mice were genotyped using the Jackson Laboratory’s protocols . This study was performed in a strict accordance with European guidelines ( 2010/63/UE ) . It was approved by the CECCAPP Animal Experimentation Ethics Committee ( Lyon , France; reference ENS_2014_022 ) , by the Professional committee for guarantee of good life-conditions of experimental animals at the Institute of Experimental Medicine IEM CAS , Prague , Czech Republic ) and by the Expert Committee at the Czech Academy of Sciences ( permit number: 027/2011 ) . First upper and lower molars of a set of adult mice from the strains DUHi ( 19 ) , FVB ( 11 ) and CD1 ( 17 ) were pictured using a Leica MZ 9 . 5 stereomicroscope . These teeth were compared with the variation observed within the murine rodents ( Murinae ) . Their morphological diversity was documented by a set of specimens from the Museum National d’Histoire Naturelle ( Paris , France ) covering the main divisions of the group ( Lecompte et al . , 2008 ) : Rattini with Rattus whiteheadi ( 5 ) , Micromys minutus ( 9 ) , Berylmys sp . ( 4 ) , Leopoldamys sabanus ( 10 ) , Bandicota indica and bengalensis ( 11 ) and Sundamys muelleri ( 9 ) ; Arvicanthini with Golunda ellioti ( 6 ) , Lemniscomys barbarus ( 6 ) , Oenomys hypoxanthus ( 5 ) , Rhabdomys pumilio ( 5 ) , Arvicanthis niloticus ( 5 ) and Aethomys chrysophilus and namaquensis ( 10 ) ; Praomyini with Mastomys chrysophilus ( 6 ) and Praomys tullbergi ( 7 ) ; Murini with Mus cervicolor ( 5 ) , Nannomys setulosus ( 6 ) , and Nannomys mattheyi ( 4 ) . Apodemini were represented by Apodemus sylvaticus ( 13; data from Renaud et al . , 2009 ) . The sampling was completed with wild house mouse ( Mus musculus domesticus ) samples , documenting the continental and insular variation: Gardouch , France ( 68 ) , Montpellier , France ( 13 ) ; Eday , Orkney , United Kingdom ( 18 ) , Fango , Corsica , France ( 7 ) , Vaitella , Corsica , France ( 24 ) and the islet Piana off Corsica , France ( 7 ) ( data from Ledevin et al . , 2016; Renaud et al . , 2011 ) . Maximal length and width were automatically extracted , together with 64 points along the outline , using the image analyzing software Optimas 6 . 5 . The morphological variance in the total sample , including laboratory strains and wild species of murine rodents ( 200 specimens , with 198 first upper molars ( UM1 ) and 192 first lower molars ( LM1 ) ) was summarized using principal component analyses ( PCA ) ( one for the upper and one for the lower molars ) on the variance-covariance matrix of the shape coefficients delivered by an outline analysis of the 2D occlusal surface ( see Renaud et al . , 2009; Renaud et al . , 2011 ) . Fourteen variables were considered for both the first upper molar ( UM1 ) and the first lower molar ( LM1 ) ; all were standardized by the size of the respective tooth and corresponded to shape only . Maximum length and width of the outline were also measured and allowed the estimation of the overall elongation of the tooth ( Length/Width ratio ) . Mouse females were mated overnight and the morning detection of a vaginal plug was taken as proof of coitus , noon being taken 0 . 5 days post coïtum ( dpc ) . We used a different day/night regime 12 hr apart to obtain embryos every half day . Pregnant females were sacrificed via cervical dislocation and embryos were harvested on ice and weighted . Embryos were dissected in Hank’s medium to separate upper from lower molars and then treated in Dispase II ( Roche ) 10 mg/mL at 37°C for 1 to 2 hr , depending on embryonic stage . Dental epithelium was then carefully removed and fixed overnight in paraformaldehyde ( PFA ) 4% . Shh probes were transcribed from a plasmid described by Echelard et al . ( 1993 ) , by means of in vitro transcription with the incorporation of digoxigenin-ddUTP , using a premixed DIG RNA labelling mix ( Roche ) . In situ hybridisation was performed with a conventional protocol . The antibody utilised was an anti-DIG antibody coupled with alkaline phosphatase ( Roche ) ; the chromogenic substrate used was BM Purple ready-to-use NBT/BCIP ( Roche ) . Samples were examined with Leica M205 stereomicroscope . Images were taken with the Leica Application Suite 4 . 1 software package . In this study , we needed a numeric estimation of the embryonic age of embryos ( i . e . an age reflecting the progress in embryogenesis ) . In a given strain , harvesting age estimated from the calculation of days-post-coitum provides a very rough estimation , because of important differences between litters ( standard range ±0 to 0 . 5 days variation in embryonic age , notably due to difference in fertilization and implantation time ) and because it does not take into account the slight variation in developmental stage within litters ( standard range from 0 to a quarter day ) . A combined ‘age/weight’ staging has been recommended previously ( Peterka et al . , 2002 ) . Embryonic body weight provides a much better numeric estimation than age in dpc alone ( Pantalacci et al . , 2009; Peterka et al . , 2002 ) . The embryonic body weight alone is especially reliable within litters ( at the stages examined here ) but less reliable between litters , presumably because the nutritional status differs from one pregnancy to the other , causing embryos of similar embryonic age to have smaller or larger body weight . Age in dpc can help to correct for this , since embryos with similar body weight and similar harvesting age will have higher chance to have reached similar developmental stage ( similar embryonic age ) , while embryos with similar body weight but different harvesting age in dpc will have higher chances to be at different embryonic age . On top of these intra-strain differences , there are differences between strains: embryos sampled at a given number of days post-coitum were not of the exact same embryonic age range in each strain . In order to provide a measure of embryonic age , we built a model that could estimate the embryonic age from the body weight and dpc , taking into account intra and inter litter variations and applying a correction effect for the latter . This model and its construction are described in detail in Appendix 1 . The code is available at: https://github . com/msemon/cdpc ( Sémon , 2020a; copy archived at https://github . com/elifesciences-publications/cdpc ) . The developmental state of all samples was assessed by combining four separate developmental criteria . Two are related with Shh expression: 1 ) Shh expression in the R2 ( rudimentary premolar ) zone , 2 ) Shh expression in the M1 ( first molar ) zone , and two are purely morphological criteria: 3 ) the bud-cap transition and 4 ) the appearance of a protrusion , visible in the dissociated epithelium at the site of the R2 signaling center , but only in the latest stages of Shh signaling and after cessation of Shh signaling . For each criterion , the samples were scored as one of 2 or 3 states . This scoring system is summarized in Supplementary file 1 . Two tables with all analyzed embryos , with their weight , age in cdpc , information on litter and scored characteristics , and the code used to analyze these data is provided on github at: https://github . com/luke-hayden/tree/master/dvpap/devstate ( Hayden , 2020; copy archived at https://github . com/elifesciences-publications/dvpap ) . Dissected tooth germs were fixed overnight in 4% PFA and dehydrated through a methanol series . In toto immunolocalization protocol was adapted from Ahnfelt-Rønne et al . ( 2007 ) . Following incubation in methanol added with H202 5% and DMSO 10% for 4 hr at room temperature , they were rehydrated , blocked with serum and incubated successively with an anti-laminin5a antibody ( overnight , 1/800 , kind gift from Jeff Miner , Miner et al . , 1997 ) and a Dylight 549 conjugated Donkey Anti-rabbit antibody ( overnight 1/200 , Jackson immunoresearch ) . Following dehydration , they were clarified and mounted in BABB as described in Yokomizo and Dzierzak ( 2010 ) . They were imaged with a Zeiss LSM710 confocal microscope at the PLATIM ( Lyon , France ) . The basal membrane labeled by the antibody was delineated semi-manually and reconstructed with the AMIRA software . RNA-Seq were made on 12 carefully dissected embryonic lower and upper first molar germs , from DUHi ( embryo weight: 196 , 219 and 239 mg ) and FVB ( 195 , 215 and 233 mg ) strains . Following dissection in culture medium , tooth germs were stored in RNA later at −20°C . RNA was extracted with RNAeasy micro kit ( Qiagen ) , and controlled with Q-bit ( Invitrogen ) and Tapestation ( Agilent technologies ) . RNA-Seq samples were prepared following the TruSeq RNA Sample Preparation v2 Guide , starting from 100 ng of total RNA of top quality ( RINe > 9 . 5 ) . Sequencing was performed with the Illumina HiSeq 4000 system ( single-end 50 bp reads ) . Reads were then mapped to the mouse genome using Kallisto ( Bray et al . , 2016 , version 0 . 44 . 0 , options -l 200 , -s 20 ) . The reference cDNA sequences and annotation files for M . musculus are based on C57B6 strain . They were collected from Ensembl 88 ( 10 5129 cDNAs , [Zerbino et al . , 2018] , GRCm38 ) . Reads were independently mapped to the FVB/NJ strain cDNAs , collected from Ensembl strains 94 , using biomart ( 10 1520 cDNAs , strain FVB_NJ_v1 , accession GCA_001624535 . 1 ) . Tximport was used to import and summarize transcript-level estimates at gene level ( version 1 . 6 , Soneson et al . , 2016 ) . Differentially expressed genes were detected with DESeq2 ( Love et al . , 2014 ) , version 1 . 18 . 1 ) with classical one-factor design , and using FDR significance threshold = 0 . 05 . 19202 genes are in common between FVB and reference strain C57B6 and have a MGI annotation . Out of these genes , 2234 genes ( 11 . 6% ) presented a significant difference of expression between the mapping on the reference strain C57B6 and the mapping on FVB strain ( DESeq2 , adjusted p-value<0 . 05 , considering the mapping effect that is with 12 replicates ) . This is presumably a mapping artifact , due to the sequence divergence between mouse strains . These genes were removed , and the remaining 16 , 968 genes were retained for further analysis . 3619 genes were found to be differentially expressed between the two strains taking into account the jaw of origin ( lower/upper ) ( ~jaw + Strain ) . Processed data with statistics are provided in Supplementary file 2 . Raw and processed data were deposited in NCBI Gene Expression Omnibus ( GEO , accession number GSE135432; https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE135432 ) . The mapping data ( by Kallisto , on each reference strain as discussed in the text ) , R source code and parameters are available on github at: https://github . com/msemon/trDUHi_FVB ( Sémon , 2020b; copy archived at https://github . com/elifesciences-publications/trDUHi_FVB ) . We used a list from supplementary data published by O'Connell et al . ( 2012 ) describing regulatory interactions for BMP4 and Wnt pathways in tooth epithelium and tooth mesenchyme at different developmental stages , by combining data mining with results of their own perturbation experiments . For BMP4 pathway , it describes up or downregulation ( +: upregulated , -: downregulated , o: no change ) of presumptive target genes upon BMP4 treatment ( perturbation + ) or BMP4 knockout ( perturbation - ) . For Wnt pathway , it describes up or downregulation of presumptive target genes upon inhibition of Gsk3b ( i . e . Wnt pathway activation ) , Ctnnb1 overexpression ( i . e . Wnt pathway activation ) , Ctnnb1 knock-out ( i . e . Wnt pathway inhibition ) , Dkk1 overexpression ( i . e . Wnt pathway inhibition ) , Lef1 knock-out ( i . e . Wnt pathway inhibition ) , or treatment with different Wnts . First , we checked if genes of this list were differentially expressed in the above analysis . For those that were DE , we compiled the O’Connell table to determine if the gene was a positive , a negative or not a target of the pathway . A gene that is upregulated upon pathway activation , or downregulated upon pathway inhibition was considered a positive target . A gene that is downregulated upon pathway activation , or upregulated upon pathway inhibition was considered a negative target . When data were conflicting between tissues ( e . g . positive target in epithelium , negative target in mesenchyme; 7 genes , e . g . Dlx1 ) , the gene was excluded from the analysis , because in our analysis the whole tooth germ is examined . When data were conflicting between different sources ( two genes: Egr1 , Ptch1 ) , we kept the result obtained by O'Connell et al . ( 2012 ) because 1 ) these are transcriptomic data and 2 ) most interactions described in the table are from this study only or are confirmed in this study , and thus the result obtained for this gene has more chance to be consistent with results for other genes in the table . The resulting table is shown in Supplementary file 2 . The strain B6 . 129S6-Shh < tm2 ( cre/ERT2 ) Cjt>/J was reciprocally crossed with a reporter strain containing LacZ inserted into the Gt ( ROSA ) 26Sor locus in order to mark the cell population expressing Shh from the time of the tamoxifen injection into pregnant female mice . Pregnant female mice were injected intra-peritoneally with tamoxifen at E12 . 5 ( when Shh is expressed in the R2 expression domain and early M1 expression is not yet apparent ) , E13 . 5 ( when Shh expressing domain in R2 finishes its activity and early M1 signaling center starts to be apparent posteriorly ) or at E14 . 5 ( when only M1 signaling center express Shh ) . Tamoxifen was administrated in a dose of 0 . 225 mg/g of body-weight ( Hayashi and McMahon , 2002 ) . Such a concentration is not hazardous for pregnant mice or embryos and is sufficient for the fast activation of recombination . The embryos were harvested at 17 . 5 dpc , 72 , 96 or 120 hr after tamoxifen application and beta-galactosidase activity was detected on whole embryos or dissociated epithelia of upper and lower cheek region . The X-gal ( Sigma ) concentration in the staining buffer was 3 mM . Samples with positive staining were post-fixed in PFA ( 4% ) overnight . After post-fixation , the samples were washed in PBS and photographed using a Leica MZ6 stereomicroscope equipped with a Leica EC3 digital camera ( Leica Microsystems GmbH , Wetzlar , Germany ) . Data are summarized in Supplementary file 3 . The upper molar region of 13 . 0 dpc FVB or DUHi embryos were dissected and cultured according to standard methods described in Kavanagh et al . ( 2007 ) . Tooth culture was stopped after 40 hr and imaged using a Leica MZ6 stereomicroscope equipped with a Leica EC3 digital camera ( Leica Microsystems GmbH , Wetzlar , Germany ) . Measuring developmental variation is a complex task; it requires that we can measure factors that change with time as development proceeds . If we consider how a developmental system proceeds along its trajectory , we will expect that it changes gradually over time , where a given sample is most similar to those at the closest time-points . However , where developmental variation is present , we expect to find pairs of samples that are of the same embryonic age but differ markedly in form . So , for each strain and for both upper and lower molars separately , we took all of our set of samples , determined their computed embryonic age ( in cdpc ) and their developmental state , then took all possible pairs of samples where both members of the pair were close in age ( less than 0 . 25 cdpc difference ) . For each pair of samples , we can then compute a pairwise developmental distance: the distance between the two samples in terms of developmental state , computed as the sum of the score difference obtained for each four developmental criteria ( Supplementary file 1 ) . For each strain for both upper and lower molars , we could then plot the distribution of these pairwise developmental distances . Finally , in order to compare total developmental variation between strains , we subjected these pairwise developmental state differences ( between pairs of samples at cdpc ) to a Wilcoxon rank-sum test . Using this method , we can measure the degree of developmental variation found in a set of samples and compare between strains . The code is provided on github at: https://github . com/luke-hayden/dvpap/tree/master/devstate ( Hayden , 2020; copy archived at https://github . com/elifesciences-publications/dvpap ) . We first calculated pairwise developmental state distances for each sample in relation to all nearby samples , over computed embryonic time . Then to obtain developmental variation over time , we used locally estimated scatterplot smoothing ( LOESS ) , a non-parametric regression method to plot a complex curve through many data points , weighting the contribution of data points according to their proximity to the point of estimation . The code is provided on github at: https://github . com/luke-hayden/dvpap/tree/master/devstate ( Hayden , 2020; copy archived at https://github . com/elifesciences-publications/dvpap ) . The statistical significance of differences in the timing of developmental events was tested using logistic regression ( embryo weight as a predictor of state , with strain as an additional preictive factor ) , examining changes in the scoring of a developmental criterion ( four criteria scored , see previously ) in relation to computed embryonic age . The statistical significance of differences in the sizes of various morphological features was tested using Student’s t-test . Fisher’s test was used to test differences in the relative rarity of a developmental state within a given window . The code is provided on github: https://github . com/luke-hayden/dvpap/tree/master/devstate ( Hayden , 2020; copy archived at https://github . com/elifesciences-publications/dvpap ) . See also Supplementary file 4 . All statistical analyses were carried out using the R statistical environment ( R Development Core Team , 2014 ) , version 3 . 2 . 3 . Packages used included ggplot2 ( Wickham , 2009 ) , reshape2 ( Wickham , 2007 ) and phytools ( Revell , 2012 ) . | Over time species develop random mutations in their genetic sequence that causes their form to change . If this new form increases the survival of a species it will become favored through natural selection and is more likely to get passed on to future generations . But , the evolution of these new traits also depends on what happens during development . Developmental mechanisms control how an embryo progresses from a single cell to an adult organism made of many cells . Mutations that alter these processes can influence the physical outcome of development , and cause a new trait to form . This means that if many different mutations alter development in a similar way , this can lead to the same physical change , making it ‘easy’ for a new trait to repeatedly occur . Most of the research has focused on finding the mutations that underlie repeated evolution , but rarely on identifying the role of the underlying developmental mechanisms . To bridge this gap , Hayden et al . investigated how changes during development influence the shape and size of molar teeth in mice . In some wild species of mice , the front part of the first upper molar is longer than in other species . This elongation , which is repeatedly found in mice from different islands , likely came from developmental mechanisms . Tooth development in mice has been well-studied in the laboratory , and Hayden et al . started by identifying two strains of laboratory mice that mimic the teeth seen in their wild cousins , one with elongated upper first molars and another with short ones . Comparing how these two strains of mice developed their elongated or short teeth revealed key differences in the embryonic structures that form the upper molar and cause it to elongate . Further work showed that variations in these embryonic structures can even cause mice that are genetically identical to have longer or shorter upper first molars . These findings show how early differences during development can lead to small variations in form between adult species of mice . This study highlights how studying developmental differences as well as genetic sequences can further our understanding of how different species evolved . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"evolutionary",
"biology"
] | 2020 | Developmental variability channels mouse molar evolution |
Ubiquitination by HECT E3 enzymes regulates myriad processes , including tumor suppression , transcription , protein trafficking , and degradation . HECT E3s use a two-step mechanism to ligate ubiquitin to target proteins . The first step is guided by interactions between the catalytic HECT domain and the E2∼ubiquitin intermediate , which promote formation of a transient , thioester-bonded HECT∼ubiquitin intermediate . Here we report that the second step of ligation is mediated by a distinct catalytic architecture established by both the HECT E3 and its covalently linked ubiquitin . The structure of a chemically trapped proxy for an E3∼ubiquitin-substrate intermediate reveals three-way interactions between ubiquitin and the bilobal HECT domain orienting the E3∼ubiquitin thioester bond for ligation , and restricting the location of the substrate-binding domain to prioritize target lysines for ubiquitination . The data allow visualization of an E2-to-E3-to-substrate ubiquitin transfer cascade , and show how HECT-specific ubiquitin interactions driving multiple reactions are repurposed by a major E3 conformational change to promote ligation .
A prevailing mechanism for altering protein function involves post-translational modification by ubiquitin ( Ub ) via E1-E2-E3 trienzyme cascades . First , an E1 activating enzyme catalyzes formation of a transient E2∼Ub intermediate , which is linked by a thioester bond between Ub’s C-terminus and the catalytic cysteine of an E2 conjugating enzyme ( here , ‘∼’ refers to a covalent protein–protein interaction linked by a thioester , oxyester , or isopeptide bond , ‘-’ refers to a noncovalent interaction , and ‘x’ refers to a crosslinked complex ) . E3s subsequently utilize one of two general mechanisms to adjoin the activated Ub C-terminus with targets . E3s in the RING family promote transfer of Ub’s C-terminus from the E2 catalytic cysteine to a substrate ( Deshaies and Joazeiro , 2009 ) . Notably , recent structural studies have revealed how isolated RING domains bind both E2 and Ub to optimally orient and activate the E2∼Ub intermediate for nucleophilic attack ( Dou et al . , 2012; Plechanovová et al . , 2012; Pruneda et al . , 2012 ) . Many ligases utilize a different mechanism possessing an active site cysteine participating directly in catalysis via a two-step mechanism . First , Ub is transferred from an E2∼Ub intermediate to the E3 catalytic cysteine to form a labile , thioester-linked E3∼Ub intermediate . Second , Ub is transferred from the E3 cysteine to a substrate’s primary amino group , which is typically from a lysine side-chain . This catalytic mechanism is common to many classes of E3s , including several effector proteins from bacterial pathogens , the RING-IBR-RING family , and HECT ( homologous to E6AP C-terminus ) ligases ( Scheffner et al . , 1995; Zhang et al . , 2006; Rohde et al . , 2007; Wenzel et al . , 2011 ) . Despite the importance of thioester-forming E3s , the mechanisms by which Ub is ligated from an E3 cysteine remain elusive , because at the time of original manuscript submission there were no structures mimicking an E3∼Ub intermediate , and to date there are none representing an E3∼Ub-substrate complex . HECT ligases were among the first family of E3 enzymes to be cloned , and are the best functionally characterized among the thioester-forming E3s ( Huibregtse et al . , 1995; Scheffner et al . , 1995 ) . Deregulation of HECT E3-mediated ubiquitination is associated with diseases such as cancers , neurological disorders , autoimmunity , and hypertension ( reviewed in Bernassola et al . , 2008; Rotin and Kumar , 2009; Metzger et al . , 2012; Scheffner and Kumar , 2013 ) . Thus , it is important to understand the mechanisms underlying ubiquitination by E3s in the HECT family . The common feature of HECT E3s is the ∼40 kDa C-terminal catalytic HECT domain . Isolated C-terminal HECT catalytic domains are capable of binding E2∼Ub , forming an E3∼Ub intermediate , and ligating the E3-linked ‘donor’ Ub to a specific lysine on an ‘acceptor’ Ub to build a polyubiquitin chain ( reviewed in Kee and Huibregtse , 2007 ) . Accessory domains upstream of the HECT domain recruit specific substrates for Ub ligation . The N-terminal sequences guide the classification of HECT E3s into different sub-families , the largest of which is the NEDD4-family . NEDD4-family members share a common domain organization with an N-terminal membrane binding C2 domain , two to four substrate-binding WW domains , and a C-terminal HECT domain . This modular architecture enables regulation and interactions ( reviewed in Ingham et al . , 2004; Kee and Huibregtse , 2007; Rotin and Kumar , 2009 ) . Substrates often contain recognition sequences that bind directly to one or more WW domains . For example , PPXY ( Pro-Pro-X-Tyr ) is one common targeting sequence . Some HECT substrates are multiprotein complexes , with the PPXY motif and ubiquitination sites located in different polypeptides . Also , many substrates of NEDD4-family E3s are transmembrane proteins in which the PPXY motif and ubiquitination sites reside in separate cytosolic segments . Furthermore , some substrates do not bind directly to NEDD4-family E3s , but are instead recruited by adaptor proteins bearing PPXY-like sequences ( reviewed in Léon and Haguenauer-Tsapis , 2009 ) . In addition , for NEDD4-family E3s , a noncovalent ubiquitin binding site was discovered in the HECT domain , which promotes polyubiquitination through poorly understood mechanisms ( French et al . , 2009; Ogunjimi et al . , 2010; Kim et al . , 2011; Maspero et al . , 2011 ) . The catalytic HECT domain consists of an N-lobe that binds E2 , a flexible linker , and a C-lobe containing the active site cysteine ( Huang et al . , 1999; Verdecia et al . , 2003 ) . A prior structure of an E2∼Ub-HECT intermediate revealed that Ub transfer from an E2 to a NEDD4-family HECT E3 is directed by noncovalent interactions between the HECT domain N-lobe and E2 , and by contacts between the HECT domain C-lobe and the E2-linked Ub ( Kamadurai et al . , 2009 ) . A structure published while the present work was under consideration showed preservation of this HECT domain-Ub conformation immediately upon generation of the E3∼Ub intermediate ( Kamadurai et al . , 2009; Maspero et al . , 2013 ) , but raised the question of how the E3∼Ub thioester bond could be directed to a substrate . Although structural mechanisms underlying Ub ligation from the E3 to a substrate remain poorly understood , a prevailing model is that numerous different relative orientations of the N- and C-lobes , manifested in prior crystal structures of HECT domains , enable Ub ligation at spatially distinct sites associated with polyubiquitination of diverse substrates ( Huang et al . , 1999; Verdecia et al . , 2003; Ogunjimi et al . , 2005 ) . However , to date , this hypothesis has not been experimentally tested , and how Ub and substrates are positioned into the catalytic domain and how the target lysines are prioritized after their initial capture remains unknown . To address the mechanisms of HECT E3-mediated Ub ligation , we studied Rsp5 , the single essential NEDD4-family member from Saccharomyces cerevisiae . Rsp5 is required for proteasomal processing of a membrane-bound transcription factor ( Hoppe et al . , 2000 ) . Rsp5 also functions in a multitude of diverse ubiquitin-dependent pathways including 26S proteasomal degradation of RNA polymerase II , and mediates proteasome-independent regulation of membrane protein trafficking ( reviewed in Ingham et al . , 2004; Kaliszewski and Zoladek , 2008; Rotin and Kumar , 2009 ) . Rsp5 has a typical NEDD4-family domain structure consisting of a C2 domain , three WW domains , and a HECT domain . Although the Rsp5 C2 domain is required for regulation of endocytosis , a construct comprising only the third WW and HECT domains ( hereafter referred to as ‘Rsp5WW3-HECT’ ) is sufficient to support viability under normal growth conditions ( Springael et al . , 1999; Hoppe et al . , 2000 ) . Rsp5 substrates are typically modified either by a single Ub ( ‘monoUb’ ) or by polyubiquitin chains containing isopeptide linkages between Lys63 and the C-terminus of the subsequent Ub in the chain ( Shih et al . , 2003; Kim and Huibregtse , 2009; Wilson et al . , 2012 ) . Several prior studies have suggested that the Rsp5 substrate Sna3 , a small transmembrane protein that undergoes Ub-dependent sorting into the multivesicular body pathway , would be a good model for studying HECT E3-mediated Ub ligation , because both the Rsp5-binding sequence and a major ubiquitination site are encompassed in Sna3’s soluble C-terminal cytoplasmic domain ( McNatt et al . , 2007; Oestreich et al . , 2007; Stawiecka-Mirota et al . , 2007; Watson and Bonifacino , 2007; Macdonald et al . , 2012 ) . Rsp5 binds via its third WW domain to a PPXY motif ( residues 106–109 ) in Sna3 , and ubiquitinates Sna3 at Lys125 ( McNatt et al . , 2007; Oestreich et al . , 2007; Stawiecka-Mirota et al . , 2007; Watson and Bonifacino , 2007; Macdonald et al . , 2012 ) . Here we define mechanisms underlying HECT E3-mediated Ub ligation . We report the crystal structure of a ternary complex in which Rsp5′s active site is simultaneously crosslinked to both Ub’s C-terminus and Sna3 , as well as biochemical , genetic , and molecular modeling experiments . Together , the data reveal interactions between the HECT domain N-lobe , the HECT domain C-lobe , and its thioester-linked Ub , establishing a specific architecture for ligation . This conformation both confines the selection of target lysines and displays a composite active site wherein both HECT domain lobes contribute to catalysis of ubiquitin transfer .
To facilitate biophysical and structural studies , we identified a minimal HECT E3-substrate pair ( Figure 1 ) . For an Rsp5 substrate , we turned to the cytoplasmic domain of Sna3 ( hereafter Sna3C ) , because peptide-like substrates have proven to be very useful for providing insights into the ubiquitin ligation mechanisms of other families of E3 ligases . The corresponding minimal region of Rsp5 mediating Sna3C ubiquitination was identified through deletion mapping using a pulse-chase assay . Briefly , a thioester-bonded E2∼Ub intermediate was enzymatically generated using E1 , the E2 UbcH5B , and a fluorescently labeled version of Ub . After quenching formation of the E2∼Ub intermediate , wild-type or truncation mutant versions of Rsp5 were added alongside a synthetic peptide corresponding to Sna3C . Formation of the Rsp5∼Ub intermediate and the Sna3C∼Ub product were monitored over time ( Figure 1C ) . Thioester-linked intermediates were confirmed by their susceptibility to reduction by DTT ( Figure 1—figure supplement 1 ) . Complexes in which Ub was linked via an isopeptide bond from autoubiquitination of Rsp5 or ligation to a lysine on Sna3C were not susceptible to reduction . As expected , the isolated HECT domain forms a thioester-linked intermediate with Ub , but is incapable of promoting Ub ligation to substrate ( Figure 1D , Figure 1—figure supplement 1 ) . Inclusion of the third WW domain is required for Ub ligation to both Sna3C and also for autoubiquitination of Rsp5 itself ( Figure 1D ) . Notably , little Rsp5WW3-HECT autoubiquitination is observed in the presence of the excess Sna3C substrate in our assays ( Figure 1D ) . 10 . 7554/eLife . 00828 . 003Figure 1 . Rsp5WW3-HECT-Sna3C as a minimal model HECT E3-substrate system to study Ub ligation . ( A ) Schematic view of two-step HECT E3 ubiquitination mechanism . First , Ub is transferred from an E2∼Ub intermediate to the HECT E3 catalytic cysteine ( Cys ) to form a labile , thioester-linked E3∼Ub intermediate . Second , Ub is transferred from the E3 Cys to a substrate’s primary amino group , which is typically from a lysine side-chain . Ub can also be a substrate for the second reaction during polyubiquitination . ( B ) Schematic views of Rsp5 and Sna3 sequences . ( C ) Schematic description of pulse-chase assay . A thioester-bonded E2∼Ub intermediate was enzymatically generated by mixing E1 , the E2 UbcH5B , and a fluorescently labeled version of Ub for 30 min . This ‘pulse’ reaction was quenched by addition of EDTA . In the ‘chase’ , wild-type or deletion mutant versions of Rsp5 were added alongside a synthetic peptide corresponding to Sna3C . Formation of the Rsp5∼Ub intermediate and the Sna3C∼Ub product were monitored at the indicated time-points . ( D ) Imaging of nonreducing SDS-PAGE gels monitoring pulse-chase fluorescent Ub transfer from E2 to Rsp5 to substrate for the indicated versions of Rsp5 , in the absence or presence of Sna3C . Rsp5WW3-HECT is the minimal version mediating Ub ligation to Sna3C . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 00310 . 7554/eLife . 00828 . 004Figure 1—figure supplement 1 . Mutational data defining Rsp5WW3-HECT as a minimal E3 mediating Ub ligation to Sna3C and autoubiquitination . ( A ) DTT-treated controls of the end points of reactions shown in Figure 1B . ( B ) Fluorescent imaging of reducing SDS-PAGE gels monitoring multiple turnover fluorescent Ub ligation by full-length Rsp5 ( Rsp5FL ) , or a construct containing all three WW domains plus the HECT domain ( Rsp5WW1-3-HECT ) , in the presence of wild-type or a P107A PPXY motif mutant version of Sna3C . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 004 Interlobe flexibility observed upon comparing the structures of isolated HECT domains , and mutagenesis of residues linking the N- and C-lobes from a HECT domain , led to the idea that the N- and C-lobes adopt many different relative orientations during polyubiquitination ( Huang et al . , 1999; Verdecia et al . , 2003; Ogunjimi et al . , 2005 ) . In this model , the catalytic requirements for ligation are encompassed exclusively in the C-lobe , its thioester-linked Ub , and the substrate . We considered that alternatively , tethering between the HECT domain N- and C-lobes could specify a limited set of distinct conformations for forming the thioester-linked E3∼Ub intermediate and for ligation . To distinguish between these mechanisms , we performed comprehensive alanine mutagenesis of the HECT domain in Rsp5WW3-HECT . We examined pulse-chase Ub ligation by 65 Rsp5WW3-HECT variants , each with one to four alanine mutations in the HECT domain ( 180 surface residues mutated in total ) ( Figure 2A , Figure 2—figure supplement 1 ) . The rates of Ub transfer to Rsp5WW3-HECT’s cysteine and subsequent ligation to the Sna3C peptide were too rapid to observe the thioester-linked E3∼Ub intermediate at the earliest time-point in this assay . Thus , we considered that mutants displaying persistence of the E2∼Ub intermediate would identify residues important for Ub transfer from E2 to the HECT domain . We anticipated that these mutations would map to the N- and C-lobe surfaces shown to bind E2∼Ub in the prior crystal structure of a trapped E2∼Ub-HECT complex ( Kamadurai et al . , 2009 ) . By contrast , mutants displaying accumulation of the Rsp5WW3-HECT∼Ub intermediate would indicate defective ligation . We hypothesized that if the HECT domain adopts a specific architecture for ligation , then some mutations in the N-lobe may also hinder ligation . Alternatively , if the two HECT domain lobes are flexibly tethered during ligation , we expected that mutations disrupting Ub transfer to Sna3C would map to the C-lobe , where Ub is linked . Results from the Ala scan revealed that mutations defective in the first catalytic step of E2-to-E3 transfer all map to the E2∼Ub binding site and the N-lobe/C-lobe interface identified in the previous E2∼Ub-HECT domain structure ( Figure 2B ) . By contrast , mutants that hindered Ub ligation to Sna3C map to distinctive surfaces on the N- and C-lobes ( Figure 2B ) . Thus , the mutational data suggest a specific HECT domain architecture may be important for Ub ligation to Sna3C , and this conformation would involve both the catalytic C-lobe and the distal N-lobe packing differently from the arrangement promoting ubiquitin transfer from E2-to-E3 . 10 . 7554/eLife . 00828 . 005Figure 2 . Alanine scanning mutagenesis suggests distinct specific HECT domain architectures to receive and ligate ubiquitin . ( A ) Summary of effects of indicated HECT domain Ala mutations on pulse-chase fluorescent Ub transfer from E2 to Rsp5WW3-HECT and then from Rsp5WW3-HECT to Sna3C . The assay scheme is shown in Figure 1C , except a 15 s chase was used . Cyan bars—accumulation of thioester-linked E2∼Ub intermediate , reflecting a defect in Ub transfer from E2 to E3 . Purple bars—accumulation of thioester-linked E3∼Ub intermediate , reflecting a defect in Ub transfer from E3 to substrate . Green bars—ratio of Sna3C∼Ub product formed . Standard deviations are calculated from three independent replications . ( B ) Locations of mutations hindering E2-to-Rsp5WW3-HECT ( blue ) or Rsp5WW3-HECT-to-substrate ( green ) Ub transfer mapped on prior E2∼Ub-HECT domain structure ( Kamadurai et al . , 2009 ) . Magenta surfaces represent residues mutated and found not essential for activity . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 00510 . 7554/eLife . 00828 . 006Figure 2—figure supplement 1 . Representative raw data from Ala scan . Fluorescent imaging of representative nonreducing ( top ) and reducing ( bottom ) gels from assays used for data in Figure 2A , monitoring pulse-chase fluorescent Ub transfer from E2 to the indicated versions of Rsp5WW3-HECT to Sna3C . The chase was 15 s . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 006 To visualize the Rsp5WW3-HECT architecture mediating ligation , we devised a method to bypass the instability of an Rsp5∼Ub thioester bond , allowing generation of a stable proxy for a HECT E3∼Ub-substrate intermediate . A synthetic Sna3C peptide with the acceptor Lys125 substituted with azidohomoalanine was linked by Click chemistry to a homobifunctional maleimide crosslinker ( Figure 3A , B ) . One arm was crosslinked to Rsp5WW3-HECT , and the other to a Cys side-chain substituted for Gly75 of Ub . This enabled us to determine the crystal structure of Rsp5WW3-HECT with the active site cysteine ( Cys777 ) simultaneously crosslinked to the C-terminus of the Ub variant and to Sna3C ( Rsp5WW3-HECTxUbxSna3C ) ( Figure 3C , Figure 3—figure supplement 1 ) . The crystals contained two Rsp5WW3-HECTxUbxSna3C complexes per asymmetric unit , which superimpose with ∼0 . 5 Å RMSD . Accordingly , only one is discussed here . The individual domains in Rsp5WW3-HECTxUbxSna3C overlay well with previous HECT structures . Noncovalent interactions between the HECT domain C-lobe and Ub involved in Ub transfer from E2-to-E3 ( Kamadurai et al . , 2009; Maspero et al . , 2013 ) are preserved in the Rsp5WW3-HECTxUbxSna3C structure . The Sna3C PPXY motif binds Rsp5′s WW3 domain in a typical manner ( Vijay-Kumar et al . , 1987; Kanelis et al . , 2001; Kim et al . , 2011 ) . However , the overall Rsp5WW3-HECTxUbxSna3C assembly is distinctive , with the HECT domain active site and its covalently linked Ub facing the Sna3C substrate ( Figure 3C ) . Importantly , the locations of the HECT domain Ala scan mutations that selectively impaired ligation map to surfaces that anchor the Rsp5 lobes and Ub in the Rsp5WW3-HECTxUbxSna3C structure ( Figure 3D ) . 10 . 7554/eLife . 00828 . 007Figure 3 . Structure of a trapped proxy for an Rsp5∼Ub-Sna3C intermediate . ( A ) Chemical representation of substrate lysine attack of an Rsp5∼Ub thioester bond . ( B ) Chemical representation of the trapped proxy , depicting covalent bonds linking Rsp5 Cys777 , a Ub Cys75 ( in place of Gly75-Gly76 ) and Sna3C Cys125 ( in place of Lys125 ) in Rsp5WW3-HECTxUbxSna3C complex . ( C ) Three views of Rsp5WW3-HECT ( violet ) xUb ( yellow ) xSna3C ( green ) , with catalytic Cys shown as orange sphere . Regions not observed in electron density are indicated with dashed lines . ( D ) Locations of Ala scan mutations hindering E2-to-Rsp5WW3-HECT ( blue ) or Rsp5WW3-HECT-to-substrate ( green ) Ub transfer mapped on Rsp5WW3-HECTxUbxSna3C structure . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 00710 . 7554/eLife . 00828 . 008Figure 3—figure supplement 1 . Electron density for Rsp5WW3-HECTxUbxSna3C structure . ( A ) Final 2Fo−Fc electron density and Rsp5WW3-HECTxUbxSna3C crystal structure , colored by domain: WW3 domain—beige , N-lobe—violet , C-lobe—pink , Ub—yellow , Sna3C—green . The maps are displayed at 1σ contour level , the structures as Cα traces for one copy in the asymmetric unit . Overall , the complex is supported by good electron density , except for the crosslinker , associated C-terminal portion of Sna3C , and the N-lobe catalytic loop , which are not visible or too patchy for building . ( B ) Close-up views highlighting active site region , N-lobe ( violet ) /C-lobe ( pink ) /Ub ( yellow ) interfaces , and PPXY motif ( green ) from Sna3C ( beige ) bound to Rsp5′s WW3 domain . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 008 The HECT domain∼Ub architecture is established by extensive interactions that bury ∼2800 Å of surface area between the N- and C-lobes , and Ub . Below we describe how ( 1 ) the two lobes of the HECT domain cradle Ub’s C-terminus to prime the thioester bond for the ligation reaction , ( 2 ) the HECT domain conformation is established by interactions between the C- and N-lobes , ( 3 ) the HECT domain architecture results in a composite Ub/N-lobe/C-lobe catalytic center , with critical N-lobe residues placed adjacent to Ub’s thioester linkage to the catalytic Cys in the C-lobe , ( 4 ) the catalytic architecture projects the E3∼Ub thioester bond toward the substrate and prioritizes substrate lysines for ubiquitination , and ( 5 ) together with the prior structure of a trapped E2∼Ub-HECT intermediate ( Kamadurai et al . , 2009 ) the new data rationalize the necessity for rotation about the HECT domain N- and C-lobes by enabling visualization of a HECT E3 Ub transfer cascade . The HECT domain architecture in the Rsp5WW3-HECTxUbxSna3C structure results in Ub being sandwiched between the C- and N-lobes ( Figure 4A ) . On one side , the interface between Ub’s globular domain and the HECT C-lobe superimposes with the corresponding region of the E2∼Ub-HECT structure ( 0 . 99 Å RMSD ) ( Figure 4B ) . Here , Ub’s Ile36/Gln40/Leu71/Leu73 hydrophobic patch contacts Leu771 , Leu798 , Ala799 , Glu802 , and Thr803 of Rsp5 ( Figure 4C , D ) . Accordingly , mutating these interface residues of Ub is lethal to yeast ( Sloper-Mould et al . , 2001; Kamadurai et al . , 2009 ) , and Ub transfer from E2 is impaired for L771A/L798A and E802A/T803A/I804A mutant versions of Rsp5WW3-HECT tested in our Ala scan , and for the L73A mutant versions of Ub ( Figures 2A and 4E ) . The E3∼Ub intermediate accumulates and the rate of Ub ligation to Sna3C is also slowed for these mutants , although a relatively lesser effect on Rsp5-mediated Ub ligation to substrate may reflect the larger interface stabilizing E3–Ub interactions for the second reaction step ( Figure 4E ) . Neither Ub transfer from E2 to Rsp5WW3-HECT nor from Rsp5WW3-HECT to Sna3C is substantially affected by an Ala substitution for Ub’s Ile44 , which does not make contacts in either the prior E2∼Ub–NEDD4LHECT or the current Rsp5WW3-HECTxUbxSna3C structures ( Figure 4E ) . Conversely , Ile44 is essential for RING E3-mediated Ub ligation and for E3-independent polyubiquitination by some E2s ( Saha et al . , 2011; Wickliffe et al . , 2011; Dou et al . , 2012; Plechanovova et al . , 2012; Pruneda et al . , 2012 ) . 10 . 7554/eLife . 00828 . 009Figure 4 . HECT domain–Ub interactions for ligation . ( A ) Ub’s C-terminal tail and its covalent linkage to the Rsp5 catalytic Cys are shown sandwiched between the Rsp5 N- and C-lobes in the crystal structure of Rsp5WW3-HECT ( violet ) xUb ( yellow ) xSna3C ( not shown ) . ( B ) C-lobe-Ub portions superimposed for Rsp5WW3-HECT ( violet ) xUb ( yellow ) xSna3C ( not shown ) and E2 ( not shown ) ∼Ub ( orange ) -NEDD4LHECT ( blue ) ( Kamadurai et al . , 2009 ) . ( C ) Close-up view of interactions between Rsp5 C-lobe and the C-terminal tail of covalently linked Ub in the crystal structure of Rsp5WW3-HECT ( violet ) xUb ( yellow ) xSna3C ( not shown ) . ( D ) Close-up view of interactions between NEDD4L C-lobe and the C-terminal tail of the E2-linked Ub in the crystal structure of E2 ( only the Cys-to-Ser mutation at the active site is shown ) ∼Ub ( orange ) -NEDD4LHECT ( blue ) ( Kamadurai et al . , 2009 ) . ( E ) Nonreducing gels from pulse-chase E2-to-Rsp5-to-Sna3C Ub transfer assay , using indicated versions of Rsp5WW3-HECT and of fluorescent or radioactive Ub . Bands corresponding to thioester-linked E2∼Ub and Rsp5WW3-HECT∼Ub intermediates and isopeptide-bonded Sna3C∼Ub product are indicated . ( F ) Yeast complementation assays for the indicated HA-Rsp5 mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 009 The Ub C-terminal tail adopts an extended conformation , secured through an intermolecular β-sheet with the HECT β-strand leading to the covalent linkage with the catalytic Cys777 ( Figure 4C ) . Here , backbone hydrogen bonds are formed between Ub residues 73 and 75 and Rsp5′s Ser774 . Because this three-residue β-sheet is not found when Ub is bound to E2 in the E2∼Ub-HECT complex ( Kamadurai et al . , 2009 ) , this structure apparently results from covalent linkage to the E3 Cys , and is likely specific for ligation ( Figure 4C , D ) . Three-way interactions between the HECT domain N-lobe , C-lobe , and the covalently linked Ub are anchored by Ub’s Arg74 ( Figure 4C ) . The aliphatic portion of Ub’s Arg74 interacts with Phe736 from Rsp5′s C-lobe . On the other face of Ub , the Arg74 guanidino group packs in a region of the N-lobe dominated by acidic residues , and directly contacts Glu502 , which corresponds to a site of mutation in the E6AP HECT E3 in Angelman syndrome ( Huang et al . , 1999; Cooper et al . , 2004 ) . Mutation of Arg74 in Ub or Glu502 in the Rsp5 HECT domain N-lobe impaired Ub ligation to Sna3C in vitro ( Figure 4E ) , and Rsp5′s essential function in vivo ( Figure 4F ) ( Sloper-Mould et al . , 2001 ) . Through rapid quench-flow kinetic analyses , we measured a sevenfold defect in the rate of Ub transfer from Rsp5WW3-HECT to Sna3C for the E502A mutant in our assay ( Figure 5 ) . 10 . 7554/eLife . 00828 . 010Figure 5 . Ala mutations selectively impaired for ligation map to N-lobe/C-lobe interface in Rsp5WW3-HECTxUbxSna3C crystal structure . ( A ) Rsp5WW3-HECTxUbxSna3C crystal structure highlighting locations of Ala mutants ( sticks , colored by mutant as indicated ) selectively impaired for ligation . Dotted line indicates approximate locations of mutations in the N-lobe not visible in electron density . ( B ) Close-up views highlighting N-lobe/C-lobe interface locations of Ala mutants selectively impaired for ligation . Dotted line indicates approximate locations of mutations in the N-lobe not visible in electron density . ( C ) Initial rates of fluorescent Ub transfer from E2 to the indicated versions of Rsp5WW3-HECT measured in rapid quench-flow pulse-chase assays . ( D ) Initial rates of fluorescent Ub transfer from the indicated versions of Rsp5WW3-HECT to Sna3C measured in rapid quench-flow pulse-chase assays . ( E ) Fluorescent images of nonreducing SDS-PAGE gels showing rapid-quench flow pulse-chase fluorescent Ub transfer from E2 to Rsp5WW3-HECT to Sna3C , for the indicated versions of Rsp5WW3-HECT . Bands corresponding to thioester-linked E2∼Ub and Rsp5WW3-HECT∼Ub intermediates and isopeptide-bonded Sna3C∼Ub product are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 010 In the Rsp5WW3-HECTxUbxSna3C structure , almost half the N-lobe nestles and orients the C-lobe ( Figure 5A ) . Notably , the importance of these N-lobe/C-lobe interactions is confirmed by impaired ligation activity for every Ala scan mutant mapping to this interface ( Figures 2A and 3D ) . For example , Arg560 and Phe561 from the N-lobe map to the core of the interface with the C-lobe ( Figure 5B ) . Accordingly , the Arg559 , Arg560 , Phe561 triple Ala mutant was severely impaired in Ub ligation to Sna3C , with no defect in the rate of E2-to-Rsp5 Ub transfer in our assay ( Figure 5C–E ) . Similarly , Arg501 from the N-lobe further anchors the C-lobe , and the combination of R501A and E502A mutations completely abrogates ligation in our assay , with a more severe defect than E502A alone ( Figure 5B–E ) . There are lesser effects of mutating side-chains in the Lys432/Arg433/Asp434/Arg436 and Arg437/Lys438/Ile440/Tyr441 combinations in the N-lobe , and the Val745/Asn746/Lys749/Asp750 combination in the C-lobe , presumably because of their more peripheral locations in the interdomain interface ( Figure 5A ) . Furthermore , the C-lobe’s Phe806 , also known as the ‘-4 phenylalanine’ required for the polyubiquitination cycle ( Salvat et al . , 2004 ) , loosely approaches Phe505 and Leu506 in the N-lobe ( Figure 6A ) . The F806A mutant was severely impaired in the Ala scan , and pulse-chase assays show specific defects in Ub ligation upon mutating either Rsp5WW3-HECT’s Phe806 to Leu , or the Phe505 and Leu506 from the N-lobe simultaneously to aspartates ( Figures 2A and 6B ) . Leu506 is adjacent to Arg461 , and a bulky R461F mutation also hinders ligation ( Figure 6B ) . In agreement with the structural data , these mutations are ligation-specific , and do not impact formation of the thioester-linked Rsp5WW3-HECT∼Ub intermediate . We devised an experiment to dissect the role of the -4 Phe during ligation , by complementing the defect of the C-lobe F806L with a compensatory aromatic residue at position 506 of the N-lobe . Indeed , the defect from the F806L mutation in the C-lobe is largely ameliorated by complementary N-lobe L506W or L506F mutations both in vivo and in vitro ( Figure 6A , C–E ) . Moreover , rapid quench-flow kinetic analyses confirmed wild-type rates of E2-to-E3 Ub transfer for these mutants , and also observed 43-fold rescue in the rate of Ub ligation to Sna3C when the defective F806L mutant was combined with L506F mutation ( Figure 6E ) . This compensation is selective , as L506F does not influence the deleterious effects of mutating the structurally distal residue Phe778 , even after extended reaction times . Overall , these data demonstrate that an aromatic side-chain has a critical function in the ligation reaction by anchoring the two HECT domain lobes . 10 . 7554/eLife . 00828 . 011Figure 6 . HECT domain C-lobe/N-lobe interactions anchoring architecture for ligation . ( A ) Close-up view of a portion of the C-lobe/N-lobe interface in Rsp5WW3-HECTxUbxSna3C structure . ( B , C ) Nonreducing gels from pulse-chase transfer assay of fluorescent Ub from E2-to-Rsp5-to-Sna3C using indicated mutants of Rsp5 . Bands corresponding to thioester-linked E2∼Ub and Rsp5WW3-HECT∼Ub intermediates and isopeptide-bonded Sna3C∼Ub product are indicated . ( D ) Left: HA-tagged WT and mutant rsp5 alleles housed on low copy plasmids were assessed for their ability to complement the essential function of RSP5 in either serially diluted rsp5-1 temperature-sensitive cells grown at restrictive 37°C or in rsp5Δ null cells after eviction of wild-type RSP5 plasmid on 5-FOA . Below: whole cell lysates of rsp5-1 transformants immunoblotted for HA and PGK . ( E ) Rates of pulse-chase fluorescent Ub ligation from the indicated versions of Rsp5WW3-HECT to Sna3C . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 01110 . 7554/eLife . 00828 . 012Figure 6—figure supplement 1 . A role for HECT domain C-terminus in ligation . ( A ) Multiple-turnover assays showing ligation of fluorescent methylated Ub to Sna3C for 2 or 10 min , by the indicated versions of Rsp5WW3-HECT . Δ1 C-ter refers to deletion of the C-terminal residue Glu809 . Reactions are shown in the absence of DTT ( left ) to confirm the abilities of mutant proteins to form E2∼Ub and E3∼Ub intermediates , and with DTT ( right ) to show isopeptide-bonded products . ( B ) Nonreducing gels from pulse-chase transfer assay of fluorescent Ub from E2-to-Rsp5-to-Sna3C using indicated mutants of Rsp5 . Bands corresponding to thioester-linked E2∼Ub and Rsp5WW3-HECT∼Ub intermediates and isopeptide-bonded Sna3C∼Ub product are indicated . ( C ) Surface view of Rsp5WW3-HECTxUbxSna3C structure , with the HECT domain N-lobe in magenta and C-lobe in pink , and Ub in yellow . Ub's residues 72 , 73 , and 74 are shown with nitrogens in blue and oxygens in red to highlight exposed basic patches . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 012 Because Phe806 is near to the C-terminus , we considered whether the Rsp5 C-terminal sequence may play a role in ligation . Deleting Rsp5′s C-terminal residue severely impairs Ub ligation to Sna3C , although not to the same extent as the extremely deleterious D495A mutant identified in our Ala scan ( Figure 6—figure supplement 1A , B ) . Our results are consistent with the recent finding that deleting Rsp5′s C-terminal residue also impairs substrate-independent polyubiquitination by the isolated HECT domain from Rsp5 ( Maspero et al . , 2013 ) . However , although it has been suggested that a C-terminal acidic side-chain may perform a catalytic role in the ligation reaction ( Maspero et al . , 2013 ) , we observed no defect upon simultaneously mutating all three of Rsp5′s C-terminal residues to alanines ( Figure 2A ) , or mutating the Rsp5′s C-terminal residue to Arg ( Figure 6—figure supplement 1A ) . Future studies will be required to definitively determine the role of the C-terminus , which has not been observed in any structure of a HECT E3 , including Rsp5WW3-HECTxUbxSna3C . However , we speculate that for Rsp5 and possibly other HECT E3s , the C-terminus itself may contribute to the ligation reaction . Indeed , C-terminal epitope tags have been shown to hinder activity ( Salvat et al . , 2004 ) . Alternatively , the C-terminus may stabilize the conformation of the HECT∼Ub complex . In this regard , we note that in the Rsp5WW3-HECTxUbxSna3C structure , the backbone amides from Leu73 and Arg74 and other portions of Ub’s extended C-terminal tail are partially exposed and could potentially interact with the acidic C-terminus of the HECT domain ( Figure 6—figure supplement 1C ) . To generate models for the substrate lysine and a functionally important N-lobe loop not visible in the Rsp5WW3-HECTxUbxSna3C electron density , we used the structure prediction program Rosetta . First , we determined the orientation of the thioesterified ubiquitin tail using constraints on the ubiquitin locations derived from the crystal structure . To accommodate the ubiquitin location , the catalytic cysteine must adopt the gauche+ rotamer , which allows formation of the thioester and its packing against His775 , Thr776 , and the noncovalent interactions between Ub and the Rsp5 C-lobe ( Figure 7A ) . Given the geometric requirements for isopeptide bond formation , this rotameric preference directs the acceptor lysine’s path of attack and all models predicted that Lys125 of Sna3 packs against Phe778 ( Figure 7B ) . His775 , Thr776 , and Phe778 correspond to residues that also interact with E2∼Ub during formation of the HECT E3∼Ub intermediate ( Kamadurai et al . , 2009 ) . Their interaction with Ub was recently also observed in the structure of a HECT E3∼Ub intermediate published while our manuscript was under consideration ( Maspero et al . , 2013 ) . Furthermore , H775A and T776A mutations result in modest defects in Ub ligation to Sna3C , and an F778A mutation results in a more severe defect ( Figure 7C ) . While all the models demonstrate a similar approach for Sna3 Lys125 , modeling the additional residues between this Lys125 and the PPXY motif revealed multiple orientations that could accommodate a proper Lys125 active-site approach . Accordingly , neither Ala nor Glu substitutions adjacent to the acceptor lysine influenced substrate selection ( Figure 7—figure supplement 1 ) . 10 . 7554/eLife . 00828 . 013Figure 7 . A distinctive active site for HECT E3-mediated ligation . ( A ) Rosetta-generated models for Sna3C ( different models in different shades of green ) target lysine approaching Rsp5WW3-HECT ( violet ) active site Cys777 thioester-linked to Ub ( yellow ) . ( B ) Model for residues aligning thioester-bound Ub and Sna3C acceptor lysine approaching the active site . ( C ) Nonreducing gels from pulse-chase assay for transfer of fluorescent Ub from E2-to-Rsp5-to-Sna3C using indicated versions of Rsp5WW3-HECT . Bands corresponding to thioester-linked E2∼Ub and Rsp5WW3-HECT∼Ub intermediates and isopeptide-bonded Sna3C∼Ub product are indicated . ( D ) Yeast complementation assays for the indicated HA-Rsp5 mutants . ( E ) Close-up of structural superposition of N-lobes showing catalytic loop from NEDD4LHECT ( Kamadurai et al . , 2009 ) not visible in Rsp5WW3-HECT ( violet ) structure . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 01310 . 7554/eLife . 00828 . 014Figure 7—figure supplement 1 . Residues proximal to the acceptor lysine in Sna3 play insignificant roles in ubiquitination . The residues around K125 ( DNKQQ ) are individually or collectively mutated to Ala or Glu and product formation at 15 s in pulse-chase fluorescent Ub transfer assay with Rsp5WW1-3-HECT is shown . Neither the Ala or Glu mutations significantly affect Sna3C ubiquitination . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 014 Although the Rsp5 loop comprising residues 491–495 is not visible in the structure , our Ala scan results revealed that Glu491 and Asp495 are essential for the ligation reaction , and defective ligation was also observed for the L494A Y496A double mutant ( Figure 2A ) . Accordingly , E491A and D495A Rsp5 mutants are also defective in supporting yeast growth ( Figure 7D ) . To gain initial insights into potential functions of this region , we superimposed the structure of the HECT domain N-lobe from NEDD4L ( Kamadurai et al . , 2009 ) onto Rsp5WW3-HECTxUbxSna3C ( Figure 7E ) . This suggested that the 491–495 loop from the N-lobe is located adjacent to the thioester bond between Ub and the catalytic Cys in the C-lobe , and provides a rationale for the importance of the HECT domain architecture for the ligation reaction . Structural modeling of this critical N-lobe loop using Rosetta suggests that Glu491 could contribute to positioning the loop and may participate in an electrostatic network involving Arg74 of Ub and Arg461 of Rsp5 ( Figure 7B ) . Furthermore , these structural models demonstrate that the acidic loop can adopt several permissive orientations that allow the carboxyl group of Asp495 to approach the epsilon nitrogen of the acceptor lysine ( Figure 7B ) . Asp495 may activate ubiquitin ligation through a variety of mechanisms including guiding the lysine into the active site or by contributing , either directly or indirectly , to the deprotonation of the substrate lysine . Direct deprotonation may be less likely as we were unable to rescue catalytic activity either with a D495H mutant , or chemically by adding high concentrations of formate or acetate ( Figure 7D and data not shown ) . We wished to address whether the catalytic conformation is specific for Sna3C and Rsp5 , or whether there is a general NEDD4-family HECT E3 architecture for ligation . As a first test , we sought to examine the effects of alanine substitutions on another Rsp5 substrate . As NEDD4-family HECT E3s can transfer Ub to a free Ub acceptor , we tested the effects of mutations on di-Ub synthesis . There is a striking similarity in the effects of the Ala mutations in pulse-chase assays for Ub ligation to Sna3C and to free Ub as the substrate ( Figure 8 ) . Some subtle differences appear due to the high concentration of free Ub ameliorating the deleterious effects of a few mutations on E2-to-E3 Ub transfer , potentially due to Ub’s noncovalent association with the HECT domain N-lobe stabilizing the E2 binding site . Also , one mutant ( L494A , Y496A ) showed a greater defect in di-Ub synthesis , perhaps marking the site where Ub binds as a substrate . Nonetheless , the overall agreement between mutational effects on modification of the two distinctive substrates suggests that a specific HECT domain architecture is generally important for Rsp5-mediated Ub ligation . 10 . 7554/eLife . 00828 . 015Figure 8 . Ala scan for HECT domain surfaces required for di-Ub synthesis . ( A ) Summary of the effects of indicated Rsp5WW3-HECT HECT domain Ala mutations on pulse-chase fluorescent Ub transfer from E2 to Rsp5WW3-HECT and then to Ub . Total activity within each gel lane was determined by adding all intensities for all three species ( E2∼Ub , Rsp5WW3-HECT∼Ub , and Ub∼Ub ) and was used to estimate the relative yield of each with respect to wild-type proteins . The error bar shows the standard deviation ( SD ) for three independent replicates . ( B ) Fluorescent imaging of representative nonreducing ( left ) and reducing ( right ) gels from assays used for data in A , monitoring pulse-chase fluorescent Ub transfer from E2 to the indicated versions of Rsp5WW3-HECT to Ub . The chase was 15 s . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 015 As a first step toward testing whether other HECT E3s adopt a similar architecture for ligation , we examined whether other NEDD4-family HECT E3s showed defects upon mutating their N-lobe aspartate corresponding to Rsp5′s Asp495 , which is distal from the catalytic center for Ub transfer from E2 to NEDD4-family E3s ( Kamadurai et al . , 2009 ) , but maps to the composite active site for ligation ( Figure 9 ) . As with the Rsp5WW1-3-HECT D495A mutant , the corresponding D584A and D639A mutant versions of NEDD4WW1-4-HECT and NEDD4LWW1 , 3-4-HECT form an E3∼Ub thioester intermediate . However , the mutations completely eliminate autoubiquitination . Taken together , the results raise the possibility that NEDD4-family HECT E3s use a common , specific catalytic architecture for Ub ligation to various substrates . 10 . 7554/eLife . 00828 . 016Figure 9 . Key residues for ligation are conserved across HECT E3s . ( A ) Alignment of portions of sequence from Rsp5 with corresponding regions of human NEDD4L , NEDD4 , Smurf1 , Smurf2 , WWP1 , WWP2 , ITCH , NEDL1 , NEDL2 , and E6AP showing conservation of key residues establishing the ligation mechanism . Identity to Rsp5 sequence is highlighted in yellow . ( B ) Fluorescent images of nonreducing ( left ) and reducing ( right ) gels of multiple turnover autoubiquitination assays for wild-type and the indicated Asp-to-Ala mutant versions of NEDD4WW1-4-HECT , NEDD4L1-3-4-HECT , and Rsp5WW1-3-HECT . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 016 A fundamental question is how substrate lysines are selected for ubiquitination . A mechanism for prioritizing lysines for modification is suggested by the ligation-primed architecture of the Rsp5WW3-HECTxUbxSna3C complex . In the structure , Ub’s C-terminus and the E3 catalytic Cys face the C-terminus of the substrate’s PPXY motif anchored to the WW3 domain . However , the N-terminus of the bound Sna3C substrate is distal , and points away from the HECT domain active site . We considered that the relative positions of the substrate-binding WW3 domain and the HECT domain N- and C-lobes might influence sites of Ub modification . In the structure , ∼25 Å separates the HECT domain active site and the Sna3C PPXY motif . The distance predicts a requirement for a 10-residue spacer between the PPXY motif that binds the WW3 domain and the acceptor lysine that receives Ub from the HECT cysteine ( Figure 10 ) . By contrast , according to prior models suggesting that the domains are flexibly tethered during ligation , it would be possible to juxtapose the catalytic cysteine in the HECT domain C-lobe with a substrate lysine only two residues C-terminal of the WW3-bound PPXY motif in Sna3C ( Vijay-Kumar et al . , 1987; Kanelis et al . , 2001; Kim et al . , 2011 ) . Thus , as a further test of the relevance of the fixed HECT domain architecture for ligation , we developed an assay for target lysine selectivity , using a TEV-cleavable HisMBP-fusion to Sna3C as a substrate . Fluorescent Ub∼E2 was added to a mixture of Rsp5WW1-3-HECT and HisMBP-TEV-Sna3C , and Ub transfer was terminated by adding DTT to reduce thioester-linked intermediates . After treatment with TEV protease , the levels of fluorescent Ub-ligated Sna3C , HisMBP , and autoubiquitinated Rsp5WW1-3-HECT were compared ( Figure 10B , Figure 10—figure supplement 1 ) . For wild-type HisMBP-TEV-Sna3C , Ub is predominantly ligated to Sna3C . Without HisMBP-TEV-Sna3C , or with a mutant disrupting the HisMBP-TEV-Sna3C PPXY motif , Ub is ligated to Rsp5WW1-3-HECT itself . Conversely , if a HisMBP-TEV-Sna3C with a K125A mutation that eliminates Sna3 ubiquitination but still retains its Rsp5-binding PPXY motif is used , Ub is ligated to both HisMBP and Rsp5WW1-3-HECT . Therefore , the distribution of Ub-ligated Sna3C , HisMBP , and Rsp5WW1-3-HECT products distinguishes preferences among different Ub-accepting lysines on substrates recruited via the same PPXY motif ( Sna3C vs HisMBP ) , and on Rsp5WW1-3-HECT autoubiquitination as an internal control for ligation activity . To test whether the relative position of a lysine influences its ubiquitination , we performed assays with insertion and deletion mutants in the 15-residue linker between the PPXY motif and Ub-acceptor lysine of Sna3C . While substrates were generally tolerant of insertions , reducing the linker to nine residues , irrespective of the sequence location , dramatically decreased Sna3C ubiquitination while increasing ubiquitination of HisMBP and Rsp5WW1-3-HECT . The strong spatial preference for target lysines suggests that the Rsp5 domains are not freely rotating during ligation . Furthermore , the sharp decline in ubiquitination upon reducing the number of residues linking the Ub acceptor lysine below 10 strongly correlates with the distance between the C-terminus of the Sna3C PPXY motif and the Rsp5 active site in the Rsp5WW3-HECTxUbxSna3C structure . Importantly , the N-terminally fused HisMBP was inefficiently ubiquitinated , despite containing 36 lysines . Nor was a lysine placed N-terminal to the PPXY motif . Overall , poor ubiquitination correlated with the structural location of the N-terminus of the PPXY motif , which projects away from the active site , although there may also be some context dependence to substrate selection based on the low activity toward an L121K K125A mutant ( Figure 10C , Figure 10—figure supplement 1 ) . These data are also consistent with in vivo data showing that K125 is the major ubiquitination site of Sna3 and a Sna3-GFP fusion protein , with the GFP moiety undergoing a lower level of ubiquitination via its 20 lysines ( Stawiecka-Mirota et al . , 2007 ) . Notably , the minimum required linker also correlates with the 12-14 residues between the LPXY motif and Ub acceptor lysines in the Rsp5 substrate Mga2p120 ( Bhattacharya et al . , 2009 ) . Overall , the data indicate that the HECT E3 architecture primed for ligation prioritizes potential target lysines by their placement relative to a composite catalytic center for ubiquitination . 10 . 7554/eLife . 00828 . 017Figure 10 . HECT domain architecture for ubiquitin ligation suggests mechanism for substrate lysine prioritization . ( A ) Structure of Rsp5WW3-HECT ( violet ) xUb ( yellow ) xSna3C ( green ) , highlighting 25 Å distance between the alpha carbon of Sna3C Tyr109 in the PPXY motif and the sulfur of Rsp5WW3-HECT Cys777 . ( B ) Schematic view of substrate selection assay . ( C ) Rsp5WW1-3-HECT selection between substrates with relative distributions of non-reducible Sna3C∼Ub ( green bar graph ) , autoubiquitinated Rsp5WW1-3-HECT∼Ub ( violet bar graph ) , and HisMBP∼Ub ( gray bar graph ) products of reactions with substrates bearing the indicated number of residues between the Sna3C PPXY motif and lysine . Effects of Ala mutations in place of the Sna3C acceptor Lys125 and the WW3-binding Pro107 are shown as controls . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 01710 . 7554/eLife . 00828 . 018Figure 10—figure supplement 1 . Data supporting Rsp5WW3-HECT target Lys prioritization . ( A ) Wild-type Sna3C has 15 residues between the PPXY motif and acceptor Lys125 . Deletion mutants were made starting at different positions after the PPXY motif , referred to as Frame1 for deletions immediately following Tyr109 , as Frame2 for deletions immediately following Ala115 , as Frame3 for deletions immediately following Gly116 , and as Frame4 for deletions immediately following Asp113 . The maximum deletion sequence for each frame , including for data shown in B , is represented in gray . Positions of Lys substitution ( tested in C ) in the K125A background are indicated above the sequences . ( B ) Pulse-chase assays performed with fluorescent Ub∼E2 added to a mixture of Rsp5WW1-3-HECT and HisMBP-TEV-Sna3C , and the reaction terminated by adding DTT to reduce any remaining intermediates and identify only isopeptide-bonded reaction products . Shown are reaction products for mutants with the indicated number of linker residues after treatment with TEV protease , allowing comparison of the levels of fluorescent Ub-ligated Sna3C , HisMBP , and autoubiquitinated Rsp5WW1-3-HECT ( see also Figure 10C ) . ( C ) Representative raw data from pulse-chase assays with substrates containing different linker lengths in Frame2 or Lys in different positions in the sequence in the K125A background . All lanes but E2∼Ub alone were treated with DTT to examine ligation products . ( D ) Plots showing the yield for each species from C for the lysine mutants , normalized with respect to total activity with wild-type proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 018
Taken together with a prior crystal structure of an E2∼Ub-HECT domain complex ( Kamadurai et al . , 2009 ) , our data allow visualization of a HECT E3’s multistep Ub transfer cascade . Our systematic Ala scanning mutagenesis ( Figure 2 ) confirms that the previous structure of a complex between oxyester-linked E2∼Ub and a HECT domain ( Kamadurai et al . , 2009 ) explains how Ub is transferred from E2 to NEDD4-family HECT E3s , wherein the HECT domain C-lobe and catalytic cysteine face the E2 . This position is dictated by noncovalent interactions between the HECT domain N-lobe and the E2 , and between HECT domain C-lobe and the E2-linked Ub ( Kamadurai et al . , 2009 ) . Our new data and a study published during manuscript revision ( Maspero et al . , 2013 ) suggest that after the E3∼Ub intermediate is formed , the C-lobe remains associated with its thioester-linked Ub . Here , to our knowledge , we provide the first snapshot of a HECT E3 poised to transfer Ub , enabled through our development of a technology for site-specific three-way crosslinking . The crystal structure of Rsp5 with its active site Cys simultaneously crosslinked to Ub’s C-terminus and to a substrate reveals that the HECT domain C-lobe and Ub together rotate ∼130° to face the substrate bound to an N-terminal WW domain ( Figure 11 ) . The three-way anchored HECT domain N-lobe/C-lobe∼Ub architecture would reduce inherent conformational flexibility ( Verdecia et al . , 2003 ) , and orient the E3∼Ub intermediate toward the substrate for the ligation reaction ( Figure 3 ) . Furthermore , the fixed architecture for the HECT domain and its thioester-bound Ub during ligation would restrict orientations available to the WW domain , thus limiting the positions available for a lysine to access the active site . 10 . 7554/eLife . 00828 . 019Figure 11 . Structural view of the ubiquitin transfer cascade for a HECT E3 . ( A ) Top left: Prior structure of E2 ( pale cyan , with active site as sphere ) ∼Ub ( yellow ) -NEDD4LHECT ( violet , with active site as sphere ) ( Kamadurai et al . , 2009 ) . Right: new structure of Rsp5WW3-HECT ( violet , with active site as sphere ) xUb ( yellow ) xSna3C ( green ) aligned over the N-lobes of the NEDD4L and Rsp5 HECT domains . ( B ) Schematic views of E2-to-E3 Ub transfer and E3-to-substrate Ub ligation . ( C ) The different relative orientations of the HECT domain N-lobe with respect to the HECT domain C-lobe∼Ub are highlighted by superimposing the N-lobes of the E2 ( pale cyan , with active site as sphere ) ∼Ub ( yellow ) -NEDD4LHECT ( pink , with active site as sphere ) ( Kamadurai et al . , 2009 ) and Rsp5WW3-HECT ( violet , with active site as sphere ) xUb ( yellow ) xSna3C ( green ) structures . The HECT domain portions of the two structures are shown on the right . DOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 019 We anticipate that other E3s functioning via a catalytic cysteine , including several effector proteins from bacterial pathogens and members of the RING-IBR-RING family ( Zhang et al . , 2006; Rohde et al . , 2007; Wenzel et al . , 2011 ) , also activate Ub ligation through parallel mechanisms that align their E3∼Ub intermediates . This theme is reflected by how mechanistically unrelated E3s ( e . g . , a SUMO E3 in the ‘neither-HECT-nor-RING’ class , and Ub E3s in the RING family ) bind both E2 and its thioester-linked Ub-like protein ( Ubl ) to reduce conformational heterogeneity and optimally orient the E2∼Ubl nucleophilic attack ( Reverter and Lima , 2005; Pruneda et al . , 2011; Dou et al . , 2012; Plechanovova et al . , 2012; Pruneda et al . , 2012 ) . Nonetheless , there are notable mechanistic differences between HECT E3s . For example , prior to the ligation reaction , there are differences during Ub transfer from the E2 . For HECT E3s , orientation of the E2∼Ub thioester bond involves extensive interactions between Ub and its ‘target’—the C-lobe to which Ub will be transferred . By contrast , RING domains apparently orient the E2∼Ub thioester bond for nucleophilic attack even in the absence of any downstream acceptor . As a result , RING E3s promote discharge of associated E2∼Ub thioester intermediates , whereas E2∼Ub complexes are stable in the presence of catalytic Cys mutant versions of HECT domains . Another mechanistic difference is reflected by exposure of Ub’s Ile44 in HECT E3 intermediates , whereas for RING and related E3s the enzyme bearing the thioester-bound Ub binds the notorious ‘I44 patch’ on Ub to orient the thioester bond . Thus , different E3 families have evolved distinct approaches to activate Ub transfer to appropriate targets . A major question is how lysines are selected for Ub and Ub-like protein modification . To date , this is best understood for E3-independent interactions between E2s and substrates , including Ub during polyubiquitination . These include Ubc9 binding a specific SUMO target sequence ( Bernier-Villamor et al . , 2002 ) , Ube2S binding an acidic portion of Ub that completes its catalytic site ( Wickliffe et al . , 2011 ) , and Ubc13 binding a partner protein that presents a specific Ub target lysine ( Eddins et al . , 2006 ) . Mechanisms underlying RING E3 presentation of substrates are generally less understood . For example , different members of the SCF family of RING E3s appear to utilize distinct selections: context is a major determinant of target lysines during SCFCdc4-dependent ubiquitination of Sic1 in vivo , ( Sadowski et al . , 2010 ) , spatial proximity to the noncovalent binding site is important for in vitro ubiquitination of β-catenin by un-neddylated SCFβ-TRCP ( Wu et al . , 2003 ) , and future studies will be required to determine mechanisms dictating substrate lysine selection by NEDD8-activated SCF E3s . Our data show how a NEDD4-family HECT E3’s architecture prioritizes lysines recruited to a WW domain by their placement relative to a composite catalytic center that involves residues from both HECT domain lobes . Although future studies will be required to reveal how substrates recruited to other domains of HECT E3s are selected for ubiquitination , conservation of key residues indicates that the ligation mechanism used by Rsp5 broadly applies to a range of HECT E3s controlling numerous physiological processes ( Figure 9 ) . Following Ub ligation , the HECT domain must reload with Ub , requiring C-lobe rotation to receive another Ub from E2 . This may be stimulated by discharge of Ub from the HECT domain onto the substrate , which would diminish the number of contacts at the junction between the HECT domain N- and C-lobes , allowing the C-lobe to both face E2 and separate from Ub . Notably , our mutational data also suggest utilization of a common HECT domain catalytic architecture for both substrate ligation and Ub chain formation ( Figures 2A and 8 ) . Although additional investigation will be required to reveal how Ub itself is positioned as an acceptor for polyubiquitination , we note that the HECT C-lobe region we found to interact noncovalently with the donor Ub determines ubiquitin chain lysine specificity in vitro ( Kim and Huibregtse , 2009 ) . We speculate that during polyubiquitination , the sum of weak interactions between HECT E3 domains and various substrates , adaptors , and regulators , sequentially drives the conformational pathway toward the active form at each step in HECT E3 cascades .
Proteins and peptides used in this work are listed in Table 1 . GST-tagged Ub and UbcH5B were expressed and purified as described ( Kamadurai et al . , 2009 ) . His-tagged human E1 ( Berndsen and Wolberger , 2011 ) and Ub were purified by nickel-affinity chromatography and gel filtration . Rsp5WW3-HECT used in structural studies was nickel-affinity-purified , treated with TEV , and purified by ion exchange . NEDD4 , NEDD4L , Rsp5WW3-HECT , and Rsp5HECT proteins used in the assays were pulled down by affinity chromatography and cleaved from their tags on-column with TEV . All other Rsp5 fragments were further purified by ion-exchange and gel filtration . All Sna3C proteins were HisMBP fusions , purified by nickel affinity . For assays , substrates either retained HisMBP or were separated from it by TEV treatment and gel filtration . 10 . 7554/eLife . 00828 . 020Table 1 . Constructs and peptidesDOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 020ConstructsDescriptionRsp5 HisMBP-TEV-Rsp5FLRsp5 residues 1–809 were cloned into pRSF-1b with a His6MBP tag , a poly Asn linker , a TEV site , and a GSGGS linker on the N-terminus GST-TEV-Rsp5WW1-3-HECTRsp5 residues 217–809 cloned into pGEX-4T1 modified to include a TEV site and a GSGGS linker between GST and Rsp5 GST-TEV- Rsp5WW2-3-HECTRsp5 residues 331–809 cloned as above GST-TEV-Rsp5WW3-HECTRsp5 residues 383–809 cloned as above GST-TEV-Rsp5HECTRsp5 residues 432–809 cloned as above HisMBP-TEV-Rsp5WW3-HECT single-CysRsp5WW3-HECT with the His6MBP-polyN-TEV-GSGGS tag in pRSF-1b and C455L , C517G , and C721A mutations; retains C777; used in structure studies NEDD4WW1-4-HECTNEDD4 residues 165–900 cloned into pGEX-4T1 as above NEDD4LWW1 , 3-4-HECTNEDD4L residues 165–955 but lacking 356–399 ( WW2 ) cloned as aboveSna3 HisMBP-TEV-Sna3CSna3 residues 71–133 with His6MBP-polyN-TEV-GSGGS on the N-terminus Sna3C-maliemide peptideSna3 residues 104–128 chemically synthesized with azidohomoalanine at position 125 , N-terminal acetylation , and C-terminal amidation Biotin-Sna3C peptideSame as above but with K125 , N-terminal biotinylation , and C-terminal carboxylationE2 UbcH5BAs described previously ( Kamadurai et al . , 2009 ) E1 Human UBA1Human UBA1 with a non-cleavable His-tag on the N-terminus expressed in pET21dUb His-Ub-G75CHis-Ub terminating with G75C; used for structure His-1Cys UbHis-Ub with Cys N-terminal of Met1 for fluorescent labeling GST-TM-2TK-UbHuman Ub cloned into pGEX2TK ( Kamadurai et al . , 2009 ) Proteins for crosslinking were treated with 10 mM DTT for 30 min and desalted into 25 mM HEPES 7 . 0 , 150 mM NaCl . Rsp5WW3-HECTxUbxSna3C was prepared by reacting single-Cys Rsp5WW3-HECT and Ub G75C with the Sna3C-maleimide peptide in a molar ratio of 1:2:3 , respectively , for 3 min on ice . The crosslinking reaction was terminated with excess DTT and the products were purified using ion exchange and sizing chromatography . Rsp5WW3-HECTxMBPSna3C ( 40 . 8 mg/ml ) crystals were grown by hanging drop vapor diffusion in 0 . 1 M Bis-Tris propane pH 7 . 2 , 21% wt/vol PEG3350 , and 1% wt/vol polypropylene glycol at room temperature , and cryoprotected in 20% glycerol in the mother liquor . Diffraction data were processed using HKL2000 ( Otwinowski and Minor , 1997 ) . PHASER ( McCoy et al . , 2007 ) was used to implement molecular replacement , searching separately for Rsp5 N-lobe , Rsp5 C-lobe , and Ub for the two copies per asymmetric unit . The WW3 domain and Sna3 peptide were built manually and the model was subjected to multiple rounds of refinement and rebuilding using REFMAC ( Murshudov et al . , 1997 ) and Coot ( Table 2; Emsley et al . , 2010 ) . The crosslinker and the Sna3 portion near it were not visible in the map . The final model contains Rsp5 ( 383–806 , chain A; 384–806 , chain B ) , Sna3 ( 105–112 , chain C; 104–117 , chain D ) , and Ub ( 1–75 , chains E and F ) . 10 . 7554/eLife . 00828 . 021Table 2 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 00828 . 021Rsp5WW3-HECTxUbxSna3CData collection* Space groupP21 Cell dimensions a , b , c ( Å ) 83 . 046 , 78 . 923 , 96 . 722 α , β , γ ( ° ) 90 , 101 . 67 , 90 Resolution ( Å ) 30–3 . 1 ( 3 . 21–3 . 1 ) † Rsym or Rmerge7 . 2 ( 36 . 1 ) I/σI15 . 5 ( 2 . 0 ) Completeness ( % ) 94 . 9 ( 92 . 7 ) Redundancy3 . 4 ( 3 . 2 ) Refinement Resolution ( Å ) No . reflections21 , 323 Rwork/Rfree25 . 1/29 . 9 No . atoms Protein7944 Ligand/ion0 Water0 B-factors N-lobe ( chains A , B ) 83 . 1 , 88 . 2 C-lobe ( chains A , B ) 91 . 9 , 109 . 7 WW3 ( chains A , B ) 100 . 0 , 100 . 3 Sna3C ( chains C , D ) 123 . 5 , 139 . 3 Ub ( chains E , F ) 124 . 7 , 122 . 2 LigandsNA WaterNA R . m . s . deviations Bond lengths ( Å ) 0 . 009 Bond angles ( ° ) 1 . 05*Data were collected from single crystals . †Values in parentheses are for highest-resolution shell . Error bars are from triplicates of each experiment . Assays were performed with either radiolabeled ( Kamadurai et al . , 2009 ) or fluorescently labeled Ub and stopped by adding SDS loading buffer , unless specified , and products separated by nonreducing SDS-PAGE and visualized by fluorescent imaging with Typhoon 9200 or autoradiography . DTT-treated controls of either all samples or the final time-point in a time-course assay confirmed thioester-linked intermediates . To produce fluorescently labeled ubiquitin , 10× fluorescein-5-maleimide ( Anaspec ) was reacted with 1Cys Ub for 3 hr at room temperature . Excess label was quenched with DTT and removed through extensive desalting steps involving dialysis and PD-10 columns . Pulse-chase assays to monitor Ub transfer from E2 to Rsp5 to Sna3C or Ub were performed in two steps . First , UbcH5B ( 9 . 1 µM ) was reacted with E1 ( 0 . 42 µM ) and fluorescent Ub ( 16 . 7 µM ) in 50 mM Tris 7 . 6 , 300 mM NaCl , 2 mM ATP , 10 mM MgCl2 , and 0 . 04 mg/ml of ovalbumin for 30 min at room temperature or 1 hr at 18°C for 32P-labeled Ub . These first reactions were quenched by diluting fourfold with 25 mM MES 6 . 5 , 100 mM NaCl , 10 mM EDTA and desalting in the same buffer using Zeba Spin Desalting Columns ( Pierce ) , or diluting fourfold with 25 mM HEPES 7 . 5 , 100 mM NaCl , 25 mM EDTA . Second , E2∼Ub was mixed with excess Rsp5 and Sna3C on ice to initiate a single round of substrate ubiquitination . For Rsp5WW3-HECT mutants and different fragments of Rsp5 , the final concentrations were ∼0 . 4 µM E2∼Ub , 2–2 . 7 µM E3 , and 10–10 . 2 µM Biotin-Sna3C or 200 µM unlabeled Ub . Reactions were quenched at the indicated times and analyzed by fluorescent imaging or autoradiography ( Storm phosphorimager , GE ) of nonreducing SDS-PAGE gels . Thioester-linked intermediates were confirmed by DTT-reduced controls collected by treating half of the last time-point samples with 100 mM DTT . E3∼Ub conjugates that accumulate for the defective mutants are thoroughly reduced with DTT except for the Lys432Ala/Arg433Ala/Asp434Ala/Arg436Ala and Arg437Ala/Lys438Ala/Ile440Ala/Tyr441Ala mutants ( data not shown ) . Total activity within each lane was determined by adding all three intensities and was used to estimate the relative yield of each species ( E2∼Ub , Rsp5WW3-HECT∼Ub , and Sna3C∼Ub or Ub∼Ub ) and scaled with respect to wild-type proteins . For substrate selection assays summarized in Figure 10C and Figure 10—figure supplement 1 , 3 . 6 µM of Rsp5WW1-3-HECT and 6 . 4 µM of HisMBP-TEV-Sna3C were used . In this case , the reaction was terminated at 15 s with 100 mM DTT and products were incubated with TEV for 30 min at room temperature to liberate HisMBP . Following quenching and separating on SDS gels , each product ( Sna3C∼Ub , Rsp5∼Ub , and HisMBP∼Ub ) was quantified using ImageQuant TL . Total activity within each lane was determined by adding all three intensities and was used to estimate the relative yield of each species . These yields were then scaled in proportion to the total activity observed with wild-type proteins to obtain normalized product formation . Rapid-quench flow kinetic studies were performed at 25°C by reacting 2 µM of E2∼Ub with a mixture of 4 µM E3 and 30 µM Biotin-Sna3C , using the KinTeK RQF-3 instrument in 25 mM MES 6 . 5 , 100 mM NaCl , 10 mM EDTA . The reactions were quenched with SDS gel-loading buffer . E2∼Ub was generated as described above but with 0 . 5 µM E1 and 10 µM each of UbcH5B and fluorescent Ub . Product separation by SDS-PAGE , visualization by fluorescent scanning , quantification , and normalization within each lane were performed as described for the substrate selection experiments . The yields were normalized across gels . Error bars are from triplicate measurements . Rates were obtained by fitting the product vs time data to linear functions using DeltaGraph ( Red Rock Software ) . All multiple turnover assays were performed at room temperature in 25 mM HEPES 7 . 5 , 100 mM NaCl , 5–6 mM MgCl2 , 1 mM ATP , 0 . 2 mg/ml ovalbumin , 100 nM E1 , 550 nM E2 . The assay described in Figure 1—figure supplement 1B was performed with Rsp5FL ( 78 nM ) and Rsp5WW1-3-HECT wild-type and WW mutants ( 70 nM ) , 5 µM wild-type or P107A MBP-free Sna3C , and 5 µM fluorescent Ub for 2 min . The substrate ubiquitination assay in Figure 6—figure supplement 1 was performed with 2 . 5 µM methylated fluorescent Ub , 50 µM Biotin-Sna3C , and 1 µM of the indicated Rsp5WW3-HECT mutants . Autoubiquitination of NEDD4 , NEDD4L , and Rsp5 ( Figure 9 ) was assayed with 1 µM E3 and 2 . 5 µM fluorescent Ub for 2 min . Null rsp5Δ mutant ( MATα: leu2-3 , 112 ura3-52 , his3-Δ200 , trp1-Δ901 lys2-801 suc2-Δ9 mel rsp5Δ::HIS3 + pNTAP416-Rsp5-WT ) and rsp5-1 mutant ( LHY23: MATa: rsp5-1 leu2-3 , 112 ura3-52 bar1 ) strains were described previously ( Dunn and Hicke , 2001; Oestreich et al . , 2007 ) . A previously described centromere-containing low copy plasmid expressing N-terminally HA tagged wild-type RSP5 ( Gajewska et al . , 2001 ) was converted to a LEU2 plasmid ( pPL4333 ) and used to create the mutants . For experiments with rsp5Δ null mutants , single colony transformants were grown in minimal media lacking uracil and leucine to select for plasmids expressing both wild-type and mutant versions of RSP5 . Cells ( 1 OD ) grown to OD600 = 1 . 0 were resuspended in water , serially diluted 1:10 , and spotted onto SC-Leu plates with or without 1 mg/ml 5-fluoroorotic acid ( 5-FOA ) and grown at 30°C for 2–5 days . For experiments with rsp5-1 mutants , single colony transformants were similarly serially diluted and spotted onto YPD and grown for 2–5 days at 30°C or 37°C . To immunoblot for HA-Rsp5 , cells were grown to OD600 = 1 . 0 , centrifuged , incubated with 0 . 2 M NaOH for 3 min , and resuspended in Laemmli sample buffer containing 8 M urea . Samples were analyzed by SDS-PAGE followed by immunoblotting using anti-HA and anti-PGK antibodies . We used various modules in the Rosetta3 modeling suite to construct the different features of the catalytic center; modified versions of the UBQ_E2_thioester and FloppyTail algorithms will be available in the next Rosetta3 release ( currently 3 . 4 ) ( Leaver-Fay et al . , 2011 ) . The UBQ_E2_thioester application generates a thioester-bonded complex and samples torsion angles for the thioester and nearby residues , including cysteine chi1 and 2 , Gly76 phi and pseudo-psi , Gly75 phi and psi , Arg74 phi and psi , and Leu73 phi and psi ( Saha et al . , 2011 ) . Additionally , this code was modified to model the location of terminal Ub acceptor , positioning the NZ of the acceptor lysine collinear with the carbonyl bond and 2 . 4 Å from the carbonyl carbon on the C-terminal glycine of Ub . Input structures for Rsp5 , Ub , and Sna3 were obtained from the Rsp5WW3-HECTxUbxSna3C co-crystals . Several constraints were placed on the Ub position derived from the Rsp5WW3-HECTxUbxSna3C three-dimensional structure . We focused on recovering the β-strand pairing between the Ub tail and the C-lobe active site . To precisely position the Ub tail , we introduced two angle constraints between the backbones of Ser774 of Rsp5 and Leu74 of Ub . As a preliminary step , we modeled five-residue Ub tails to determine the range of thioester torsion angles that could accommodate the β-sheet pairing and applied these angles to the starting pose . Additionally , we decreased the size of the moves on the cysteine chi1 and thioester bond to obtain more models with optimal geometry; even with this change models with various cysteine chi1 and thioester torsion angles were recovered . We filtered the subsequent models on constraint energy , RMSD to Ub , and visual inspection , and selected the top five models . For these five models , we used the FloppyTail application to sample possible conformations of the Sna3 substrate ( Kleiger et al . , 2009 ) . Constraints were placed on the location of the PPXY motif . Residues 115–124 , between the PPXY motif and the acceptor lysine , were allowed to sample backbone torsion angles , and the models were filtered by lowest energy and RMSD to the crystal PPXY motif . To ascertain possible N-lobe loop conformations for residues 491–495 , we grafted a homologous segment of NEDD4L ( PDB 3JW0 ) containing residues 630–645 , which correspond to residues 483–505 on Rsp5 . We used a two-step protocol to close and refine the loop . First , we used perturb KIC to close the loop in the absence of the Sna3 peptide and substituted the corrected amino acid identities that varied in the grafted NEDD4L loop . We allowed backbone sampling for residues 488–498 , and selected the lowest energy closed loop; about 5% of the models had successfully closed N-lobe loops . Next , we refined the loop in the presence of the Sna3C peptide , both with and without a 3 Å distance constraint between the acceptor lysine and Asp495 side chain . Although the loop could adopt multiple orientations that could bring these residues in close proximity , without this constraint Rosetta favored orientations that placed Asp495 more distant from the acceptor lysine . We examined the models scoring in the 90th percentile and found that about half of them had a very similar loop orientation and choose a representative structure from this set . | Ubiquitin is a small protein that can be covalently linked to other , ‘target’ , proteins in a cell to influence their behavior . Ubiquitin can be linked to its targets either as single copies or as polyubiquitin chains in which several ubiquitin molecules are bound end-on-end to each other , with one end of the chain attached to the target protein . A multi-step cascade involving enzymes known as E1 , E2 , and E3 adds ubiquitin to its targets . These enzymes function in a manner like runners in a relay , with ubiquitin a baton that is passed from E1 to E2 to E3 to the target . The E3 enzyme is a ligase that catalyzes the formation of a new chemical bond between a ubiquitin and its target . There are approximately 600 different E3 enzymes in human cells that regulate a wide variety of target proteins . A major class of E3 enzymes , called HECT E3s , attaches ubiquitin to its targets in a unique two-step mechanism: the E2 enzymes covalently link a ubiquitin to a HECT E3 to form a complex that subsequently transfers the ubiquitin to its target protein . The ubiquitin is typically added to a particular amino acid , lysine , on the target protein , but the details of how HECT E3s execute this transfer are not well understood . To address this issue , Kamadurai et al . investigate how Rsp5 , a HECT E3 ligase in yeast , attaches ubiquitin to a target protein called Sna3 . All HECT E3s have a domain—the HECT domain—that catalyzes the transfer of ubiquitin to its target protein . This domain consists of two sub-structures: the C-lobe , which can receive ubiquitin from E2 and then itself become linked to ubiquitin , and the N-lobe . These lobes were previously thought to adopt various orientations relative to each other to deliver ubiquitin to sites on different target proteins ( including to multiple lysines on a single target protein ) . Unexpectedly , Kamadurai et al . find that in order to transfer the ubiquitin to Sna3 , Rsp5 adopts a discrete HECT domain architecture that creates an active site in which parts of the C-lobe and the N-lobe , which are normally separated , are brought together with a ubiquitin molecule . This architecture also provides a mechanism that dictates which substrate lysines can be ubiquitinated based on how accessible they are to this active site . The same regions of Rsp5 transfer ubiquitin to targets other than Sna3 , suggesting that a uniform mechanism—which Kamadurai et al . show is conserved in two related human HECT E3 ligases—might transfer ubiquitin to all its targets . These studies therefore represent a significant step toward understanding how a major class of E3 enzymes modulates the functions of their targets . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | Mechanism of ubiquitin ligation and lysine prioritization by a HECT E3 |
Ataxia-telangiectasia mutated ( ATM ) protein kinase regulates the DNA damage response ( DDR ) and is associated with cancer suppression . Here we report a cancer-promoting role for ATM . ATM depletion in metastatic cancer cells reduced cell migration and invasion . Transcription analyses identified a gene network , including the chemokine IL-8 , regulated by ATM . IL-8 expression required ATM and was regulated by oxidative stress . IL-8 was validated as an ATM target by its ability to rescue cell migration and invasion defects in ATM-depleted cells . Finally , ATM-depletion in human breast cancer cells reduced lung tumors in a mouse xenograft model and clinical data validated IL-8 in lung metastasis . These findings provide insights into how ATM activation by oxidative stress regulates IL-8 to sustain cell migration and invasion in cancer cells to promote metastatic potential . Thus , in addition to well-established roles in tumor suppression , these findings identify a role for ATM in tumor progression .
Cancer treatments rely heavily on DNA damaging agents , including radiation and chemotherapeutics , to eliminate cancer cells and decrease tumor burden ( Begg et al . , 2011; Lord and Ashworth , 2012; Cheung-Ong et al . , 2013 ) . These cancer therapies activate complex signaling networks , termed the DNA damage response ( DDR ) , that detect and signal the presence of DNA damage to promote cell cycle arrest and DNA repair ( Jackson and Bartek , 2009; Ciccia and Elledge , 2010 ) . The apical protein kinase ataxia-telangiectasia mutated ( ATM ) initiates a large signaling cascade in response to DNA double-strand breaks ( DSBs ) by phosphorylating many key proteins , including the tumor suppressor p53 , to orchestrate the DDR ( Shiloh and Ziv , 2013; Stracker et al . , 2013 ) . The DDR regulates cell fate decision pathways including apoptosis , senescence and differentiation . Many of these pathways depend on ATM and/or p53 to enforce tumor suppressive anti-cancer barriers to limit proliferation in response to , and in the presence of , DNA damage ( Norbury and Zhivotovsky , 2004; Bartkova et al . , 2005; Gorgoulis et al . , 2005; Bartkova et al . , 2006; Di Micco et al . , 2006; Vousden and Lane , 2007; Sherman et al . , 2011; Roos and Kaina , 2013 ) . However , cancer cells exhibit genome instability ( Hanahan and Weinberg , 2011 ) and often contain endogenous oxidative and replicative damage that can promote genetic alterations to drive malignant transformation ( Klaunig et al . , 2010; Hills and Diffley , 2014 ) . Cancer cells evolve a defective DDR to allow cell proliferation in the presence of DNA damage , furthering genome instability and cancer progression . Defects in the DDR can also profoundly influence DNA damage-dependent therapies both positively and negatively ( Bouwman and Jonkers , 2012 ) . Thus , DDRs can influence both cancer promoting and suppressing mechanisms . This concept is perhaps best exemplified by p53 , which is normally activated by stress signals and promotes tumor suppressor pathways ( Bieging et al . , 2014 ) . However , the tumor-associated gain-of-function p53 mutations that are most commonly observed in human cancers exhibit several tumor-promoting functions ( Muller and Vousden , 2013 ) . Thus , tumor suppressive or promoting activities from the same protein can be dictated by many factors including mutations , cell type , and/or disease context . ATM regulates complex signaling networks that are involved in many biological processes in addition to DSB signaling and repair ( Stracker et al . , 2013 ) . Aside from increased tumorigenesis , ATM-deficiency results in altered metabolism , aberrant immune and inflammatory responses and increased levels of reactive oxygen species ( ROS , Schneider et al . , 2006; Alexander et al . , 2010; Freund et al . , 2011; Kulinski et al . , 2012; Valentin-Vega et al . , 2012 ) . Several of the pathological outcomes in ATM deficient mice have been linked to ROS and many of these pathologies can be reversed by the addition of antioxidants , highlighting the important role of ATM in regulating redox-homeostasis ( Ito et al . , 2004; Reliene and Schiestl , 2006 , 2007; Reliene et al . , 2008; Freund et al . , 2011; Okuno et al . , 2012; D'Souza et al . , 2013 ) . In addition , ATM can be directly activated by ROS , independently from DSB signaling , and has been implicated in mitochondrial quality control , potentially through an ability to localize to mitochondria ( Guo et al . , 2010; Valentin-Vega et al . , 2012 ) . The substrates of ATM following ROS mediated activation , and how distinct they are from those modified following DNA damage , remains unknown . The NF-κB family of transcription factors are similarly activated by multiple stimuli that include DNA damage and ROS . NF-κB signaling regulates inflammation and is involved in cancer where it has been associated , like p53 , with both anti and pro-tumorigenic processes ( Ben-Neriah and Karin , 2011; Oeckinghaus et al . , 2011 ) . NF-κB signaling can be regulated by ATM through the phosphorylation of NEMO , which aids in transmitting nuclear ATM signaling to activate NF-κB in the cytoplasm ( Wu et al . , 2006 ) . Upon activation , NF-κB translocates to the nucleus where it regulates gene expression and influences cell survival pathways . Understanding the crosstalk between the NF-κB and ATM signaling in complex pathologies including inflammation and cancer remains an area of active investigation . Cancer cells acquire several capabilities in addition to uncontrolled proliferation during the multistep process of tumorigenesis , including cell migration and invasion that promote tumor metastasis . ( Friedl and Alexander , 2011; Hanahan and Weinberg , 2011; Valastyan and Weinberg , 2011 ) . Metastasis occurs when a tumor spreads by means of cell migration and invasion to a secondary site , a process linked with the majority of cancer deaths ( Hanahan and Weinberg , 2011 ) . Several pathways suppress this dangerous process as part of tumor suppressive mechanisms ( Mehlen and Puisieux , 2006 ) . Although the role of DNA damage and the DDR in this process has not been extensively explored experimentally , increased genomic instability due to the acquisition of DDR defects has been proposed to play a role in the acquisition of these traits based on human patient data ( Halazonetis et al . , 2008 ) . Here , we studied the contribution of the DDR kinase , ATM , in metastatic programs operating in cancer cells . We found that ATM supports migration and invasion , cellular processes intimately linked with metastasis . Our data is consistent with endogenous oxidative stress triggering these responses independently from DSB signaling . Gene expression analysis identified an ATM and mutant p53 transcriptional program that contained pro-migration and invasion genes , including interleukin-8 ( IL-8 ) . Addition of IL-8 rescued defects in cell migration and invasion observed in ATM-depleted cells thus validating this target . Consistent with the biological relevance of these findings , ATM promoted tumor formation in a xenograft human breast cancer cell model . Based on the data , we propose a non-canonical role for ATM in supporting pro-tumorigenic behavior of cancer cells .
Given the prevalence and success of DNA-damage based cancer treatments , we sought to determine whether DNA damage and the DDR could inhibit pro-metastatic processes as part of its well-established role as an anti-cancer barrier . To this end , we first determined the effects of DNA damaging agents on cell migration , a process intimately linked with metastasis . DNA damage , including ionizing radiation ( IR ) , topoisomerase II poisons etoposide ( Et ) and doxorubicin ( Dox ) , had little effect on cell migration in the highly metastatic human breast cancer cell line MDA-MB-231 as measured by wound healing assays ( Figure 1A , quantified in Figure 1B ) . Induction of the DNA damage marker , phosphorylated histone variant H2AX ( γH2AX ) , confirmed DNA DSBs formations ( Figure 1B ) . These treatments arrested the cell cycle and inhibited proliferation ( Figure 1C , D ) . Thus , cell migration is independent from DSB induction by DNA damaging agents , as well as cell cycle and growth arrest . 10 . 7554/eLife . 07270 . 003Figure 1 . Ataxia-telangiectasia mutated ( ATM ) is required for cell migration and invasion in MDA-MB-231 cells . ( A ) Wound-healing assays of MDA-MB-231 cells untreated ( Unt ) or treated with doxorubicin ( Dox , 100 nM ) , phleomycin ( Phleo , 90 μg/ml ) , etoposide ( Et , 40 μM ) or ionizing radiation ( IR , 20 Gy ) . Drug treatments were for 48 hr . IR treatment was performed at time 0 . All samples were analyzed post-48 hr from wound induction . Images were acquired at 0 and 48 hr . Representative images from three independent experiments are shown . ( B ) Verification of DNA damage induction and quantification of wound healing from ( A ) . Top: Western blot analysis of samples from ( A ) with the DNA damage marker γH2AX . H2AX is a loading control . Bottom: Quantification of wound healing experiments from ( A ) . ( C ) Cell cycle analysis of cells treated in panel A by flow cytometry . Cells were treated as in ( A ) and analyzed by FACS 48 hr post-treatment . ( D ) Proliferation of cells treated as in ( A ) . After 48 hr , cells were trypsinized , counted and normalized to untreated cells at 0 hr . ( E ) ATM promotes cell migration in the absence of induced double-strand breaks ( DSBs ) . Wound-healing assays were performed in siRNA-treated MDA-MB-231 human breast cancercells with siNon-coding ( siNC ) or siATM siRNAs . ( F ) Verification of DNA damage induction and quantification of wound healing from ( E ) . Top: Western blot analysis of samples from ( E ) with the ATM markers KAP1-pS824 and mut-p53-pS15 . Unmodified proteins are loading controls and ATM controls siRNA depletion . Bottom: Quantification of wound healing experiments from ( E ) . ( G ) ATM depletion impairs cell migration and invasion , but not proliferation . Left panel: siNC or siATM cells were analyzed with xCELLigence Real-time cell analyzer ( RTCA ) to measure proliferation , migration and invasion in parallel and real-time . Experiments performed as detailed in ‘Materials and methods’ . Right panel: ATM depletion 96 hr post-transfection . ( mean ± s . e . m . , n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 00310 . 7554/eLife . 07270 . 004Figure 1—figure supplement 1 . Knockdown of ATM impairs cell migration and invasion . ( A and B ) ATM depletion by two independent siRNAs reduces migration and invasion in MDA-MB-231 cells . Proliferation , migration and invasion of cells in siATM-1 ( A ) and siATM-2 ( B ) compared to siNC control cells were analyzed as in Figure 1G . Western blot of ATM depletion for each individual siRNA are shown . H2AX is loading control . ( C ) Treatment with ATM kinase inhibitor reduces cell migration compared to untreated cells . Cells were analyzed as in Figure 1E . ATMi—KU-55933 . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 004 Defects in DDR genes and DNA damage signaling affect cellular responses to chemotherapeutic drugs in some cancer cells ( Bouwman and Jonkers , 2012 ) . As ATM mediates many initial DNA damage signaling events and regulates multiple cellular responses to DNA damage , we tested its role in cell migration . Strikingly , ATM depletion reduced cell migration in the presence or absence of IR-induced damage ( Figure 1E , quantified in Figure 1F ) . Increased DDR signaling as evidenced by phosphorylated H2AX , p53 and KAP1 confirmed DNA damage by IR ( Figure 1F ) . We independently verified the requirement of ATM for cell migration and invasion using an xCelligence system that measures cell migration and invasion , as well as proliferation in real-time and in the same cell populations which provides standardized experimental conditions . Reduced invasion and migration in ATM-depleted cells was independent from proliferation as control and ATM-depleted cells grew similarly ( Figure 1G ) . Experiments with two additional independent ATM siRNAs confirmed these observations thus ruling out possible off-target effects by siRNAs ( Figure 1—figure supplement 1 ) . Thus , ATM promotes cell migration and invasion in MDA-MB-231 cells in the absence of exogenous DNA damage . ATM is activated by at least two distinct mechanisms ( Shiloh and Ziv , 2013 ) . DSBs activate ATM through interactions with the MRE11-RAD50-NBS1 ( MRN ) complex . ATM phosphorylates many targets , including the effector kinase CHK2 , which collectively orchestrate the DDR ( Bakkenist and Kastan , 2003; Lee and Paull , 2005; Matsuoka et al . , 2007; Shiloh and Ziv , 2013 ) . ATM is also activated by oxidative stress independently of DSBs , MRN or CHK2 ( Guo et al . , 2010 ) , although the downstream pathways reliant on this ATM activation are unclear . In contrast to ATM inhibition , depletion of CHK2 , MRE11 or NBS1 did not reduce cell migration , suggesting ATM-mediated cell migration occurred independently of DSB-dependent ATM activation ( Figure 2A , B; depletion analysis shown in Figure 1C ) . Thus , we hypothesized that ATM could promote migration through oxidative stress . In support of this idea , chemical inhibition of oxidative stress in MDA-MB-231 cells using the reducing agent N-acetylcysteine ( NAC ) reduced cell migration similarly as depletion of ATM ( Figure 2D and Figure 2—figure supplement 1 ) . We also showed that NAC treatment inhibits intracellular oxidative stress ( Figure 2E ) . To rule out any influence of proliferation on cell migration in our analyses , we performed live-cell imaging to track the migration of individual cells . We found reduced cell motility in cells treated with NAC compared with control cells ( Figure 2F , G and Videos 1 , 2 ) . These data suggest that ATM , independent from DSB signaling , promotes cell migration . 10 . 7554/eLife . 07270 . 005Figure 2 . ATM promotes cell migration and invasion independently of DNA DSB signaling in MDA-MB-231 cells . ( A ) DSB signaling is not involved in cell migration . Experiments were performed as in Figure 1A with the indicated siRNAs . ( B and C ) Quantification of wound healing ( B ) and siRNA depletions ( C ) in ( A ) . ( D ) Reactive oxygen species ( ROS ) inhibitor N-acetylcysteine ( NAC ) reduces cell migration . Right panel: quantification of wound healing . ( E ) NAC treatment reduces endogenous ROS . Cells treated with 10 mM NAC were analyzed using an intracellular ROS detector as detailed in ‘Materials and methods’ . 4 mM H2O2 treatment serves as a positive control . ( F ) Live-imaging analysis of cells treated with 10 mM NAC or left untreated . Images were acquired every 15 min for 6 hr and cell were tracked using ImageJ . Colored dots and lines represent individual cell paths . Scale bar , 37 . 5 μm . ( G ) Quantification of individual cell speed ( μm/min ) and cell path ( μm ) from ( F ) . Cell parameters were quantified in ImageJ and represent mean data from >100 cells . Error bars = SD . *** p-value <0 . 0001 , unpaired two-tailed t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 00510 . 7554/eLife . 07270 . 006Figure 2—figure supplement 1 . Inhibition of oxidative stress reduces cell migration and invasion . ROS inhibitor NAC reduces cell migration to a similar extent as ATM-depletion . siATM and siNC control MDA-MB-231 cells were either untreated or treated with NAC and analyzed as in Figure 2D for cell migration . Right panel: Western blot of ATM depletion and quantification of cell migration for each sample . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 00610 . 7554/eLife . 07270 . 007Video 1 . Live cell imaging and tracking of untreated MDA-MB-231 cells for Figure 2 . Images were taken every 15 min for 6 hr and tracking was performed in ImageJ . Still images and quantifications are provided in Figure 2E , F . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 00710 . 7554/eLife . 07270 . 008Video 2 . Live cell imaging and tracking of MDA-MB-231 cells treated with 10 mM NAC for Figure 2 . Images were taken every 15 min for 6 hr and tracking was performed in ImageJ . Still images and quantifications are provided in Figure 2E , F . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 008 While DSBs trigger ATM-dependent phosphorylation of more than 1000 proteins ( Matsuoka et al . , 2007; Bennetzen et al . , 2010; Bensimon et al . , 2010 ) , very few targets of ATM in response to other types of damage , including oxidative stress , are known ( Guo et al . , 2010 ) . However , p53 is phosphorylated by ATM after oxidative stress and mutant p53 is involved in TGFβ-dependent cell migration in MDA-MB-231 cells ( Adorno et al . , 2009; Guo et al . , 2010 ) . We therefore speculated that mutant p53 could act in concert with ATM to promote cell migration . To test this hypothesis , we depleted ATM , mutant p53 or both and analyzed cell migration . As expected , inhibition of mutant p53 resulted in reduced cell migration ( Figure 3A , quantified in Figure 3B ) . Interestingly , co-depletion of ATM and mutant p53 resulted in a similar reduction of cell migration as either single gene knockdown alone ( Figure 3A , quantified in Figure 3B ) . We confirmed and extended these results using real-time , simultaneous analyses of proliferation , migration and invasion . Although mutant p53 depletion mildly reduced proliferation , co-depletion of ATM resulted in an epistatic reduction of cell migration and invasion ( Figure 3C ) . Live-cell imaging of migrating cells revealed reduced speed and migration path length in ATM and mutant p53 depleted cells ( Figure 3D , E and Videos 3–5 ) . These results further corroborate the role of ATM and mutant p53 in promoting cell migration , independently from cell proliferation . These results are consistent with our analyses of DNA damaging agents , which inhibit proliferation without altering cell migration ( Figure 1A–D ) . We repeated these experiments in the highly metastatic BT-549 breast cancer cells that have mutant p53 . Consistent with results from MDA-MB-231 cells , depletion of ATM reduced cell migration by ∼ 60% in BT-549 ( Figure 3—figure supplement 1A ) . ATM and/or mutant p53-depleted BT-549 cells exhibited similar levels of migration ( Figure 3—figure supplement 1B ) , which is in accord with data obtained in MDA-MB-231 cells ( Figure 3A–C ) . Taken together , these results from multiple human cancer cell lines suggest that ATM and mutant p53 are required for the cell migration and invasion phenotypes observed in these highly invasive cancer cell lines . 10 . 7554/eLife . 07270 . 009Figure 3 . ATM-mutant p53 axis of the DNA damage response ( DDR ) promotes cell migration and invasion in MDA-MB-231 cells . ( A ) ATM or mutant p53 depletion , as well as co-depletion , impairs cell motility similarly . Wound-healing assays were performed with the indicated siRNAs as in Figure 2A . ( B ) siRNA depletions and quantification of wound healing for ( A ) . ( C ) Real-time analysis of cell dynamics in siATM , simutant-p53 and co-depleted cells . Experiments performed as in Figure 1G with indicated siRNAs . Right: ATM and mutant p53 levels in cell samples . ( D ) Live cell imaging of cell migration defects in ATM and mutant p53 depleted MDA-MB-231 cells . Experiments were performed and analyzed as in Figure 2F . Scale bar , 20 μm . ( E ) Quantification of individual cell speed ( μm/min ) and cell path ( μm ) from ( D ) . Cell parameters were quantified as in Figure 2G . Error bars = SD . *** p-value <0 . 0001 , unpaired two-tailed t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 00910 . 7554/eLife . 07270 . 010Figure 3—figure supplement 1 . Analysis of ATM and mutant p53 functions in human breast cancer cells . ( A ) ATM depletion impairs cell migration . Cells transfected with indicated siRNA were analyzed by wound-healing assays to determine cell motility . Right: Quantification of wound healing and ATM siRNA knockdown efficiency . ( B ) Reduction of ATM and/or p53 reduces cell migration . Cells were treated and analyzed with the indicated siRNAs as Figure 3A . Quantification of wound healing and depletion analysis is in right . ( C ) Treatment of MDA-MB-231 cells with ATM kinase inhibitor reduces mutant p53-pS35 compared to untreated cells . Experiments were performed as in Figure 1F . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01010 . 7554/eLife . 07270 . 011Video 3 . Live cell imaging and tracking of siNC MDA-MB-231 cells for Figure 2 . Images were taken every 15 min for 6 hr and tracking was performed in ImageJ . Still images and quantifications are provided in Figure 3D , E . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01110 . 7554/eLife . 07270 . 012Video 4 . Live cell imaging and tracking of siATM MDA-MB-231 cells for Figure 3 . Images were taken every 15 min for 6 hr and tracking was performed in ImageJ . Still images and quantifications are provided in Figure 3D , E . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01210 . 7554/eLife . 07270 . 013Video 5 . Live cell imaging and tracking of si mutant p53 MDA-MB-231 cells for Figure 3 . Images were taken every 15 min for 6 hr and tracking was performed in ImageJ . Still images and quantifications are provided in Figure 3D , E . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 013 We next investigated the molecular mechanism by which ATM promotes cell migration and invasion . The identification of an epistatic relationship between ATM and mutant p53 in promoting cell migration pointed towards a common molecular pathway . Given the role of mutant p53 in transcriptional regulation , we hypothesized that ATM could similarly regulate the expression of genes involved in cell migration ( Riley et al . , 2008; Shiloh and Ziv , 2013 ) . Therefore , we performed a microarray-based comparative gene expression analysis in control , ATM and mutant p53 siRNA-treated cells . Out of the hundreds of genes in ATM-depleted cells whose mRNA levels were differentially regulated more than 1 . 5-fold compared to control siNon-coding ( siNC ) cells , only ∼40 genes showed equivalent regulation between siATM and si mutant p53 cells ( Figure 4A and Supplementary file 1A , B ) . We observed comparable numbers of co-regulated genes that were either up-regulated or down-regulated similarly in both siATM and si mutant p53 cells ( Figure 4A and Supplementary file 1C ) . Gene ontology ( GO ) analysis indicated that genes involved in the response to wound healing , including cell migration genes , were in the top 10 GO categories for both ATM and mutant p53 depleted samples ( Figure 4—figure supplement 1 ) . GO analysis of genes co-regulated by ATM and mutant p53 identified almost exclusively pathways involving cell migration ( Figure 4—figure supplement 1C and Supplementary file 1C ) in agreement with our data showing reduced migration in ATM or mutant p53 deficient cells . Collectively these data support a role for mutant p53 and ATM in the co-regulation of a gene network regulating cell migration under these conditions . 10 . 7554/eLife . 07270 . 014Figure 4 . ATM-mutant p53 regulates cytokine interleukin-8 . ( A ) Differential transcriptome expression analysis in siATM- and simut-p53- depleted cells identifies reduced IL-8 expression in both samples . Upper: Venn diagram of differentially expressed genes in simut-p53 , siATM or both . Numbers indicate genes differentially expressed 1 . 5-fold or greater compared to siNC . Heatmap represents the 38 genes co-regulated similarly in siATM and simut-p53 cells . Expression data was normalized to control siNC cells . Cut-off = 1 . 5-fold normalized to siNC control cells . ( B and C ) qRT-PCR analysis of IL-8 mRNA levels in MDA-MB-231 ( B ) and BT-549 ( C ) siATM or sipmut-53 depleted cells . ( D ) IL-8 promoter activity by luciferase reporter assay in siATM and simut-p53 cells . Depletion of ( E ) NF-κB or ( F ) NEMO impairs IL-8 expression . ( G ) ATM or mutant p53 depletion abrogates NF-κB p65 nuclear localization . Cells treated with indicated siRNA were harvested to obtain cytoplasmic extract ( CE ) and nuclear extract ( NE ) to analyze NF-κB p65 localization . ( H ) ATM or mutant p53 deletion impairs NF-κB p65 binding to IL-8 promoter using chromatin immunoprecipitation ( ChIP ) analysis . Actin promoter serves as a negative control . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01410 . 7554/eLife . 07270 . 015Figure 4—figure supplement 1 . Gene ontology ( GO ) analysis of ATM and mutant p53 regulated genes in MDA-MB-231 cells . ( A–C ) Functional classification of regulated genes in MDA-MB-231 from siATM ( A ) , simut-p53 ( B ) and genes regulated similarly in both siATM and simut-p53 by more than 1 . 5-fold ( C ) . GO analysis was analyzed using the ‘Functional Annotation Tool’ in DAVID ( http://david . abcc . ncifcrf . gov/home . jsp ) and biological process terms are shown . GO pathways are shown for the upregulated genes ( top ) and downregulated genes ( bottom ) . X-axis plots—log10 p-values obtained by the GO analysis . The top 12 pathways are shown for each data set . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01510 . 7554/eLife . 07270 . 016Figure 4—figure supplement 2 . Regulation of IL-8 and NF-kB by ATM and p53 . ATM promotes IL-8 mRNA expression in the context of mutant p53 ( A and B ) Depletion of ATM in WT p53-containing cancer cell lines ( A ) U2OS or ( B ) MCF7 did not reduce IL-8 mRNA levels . ( C and D ) Depletion of ATM or mutantp53 does not reduce the mRNA levels of ( C ) p65 or ( D ) NEMO . ( E ) p65 protein levels are not reduced upon depletion of ATM , mutant p53 or NEMO . sip65 acts as a control . ( F ) Depletion of ATM or mutant p53 results in reduced phosphorylation of NEMO at Ser-85 . All experiments were performed in MDA-MB-231 cells except A and B . Experiments were performed as in Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 016 We focused our analysis on the down-regulated genes identified in both ATM and mutant p53 data sets since GO analysis only identified the wound healing pathway in this gene set ( Figure 4—figure supplement 1 ) . Interestingly , the inflammatory cytokine interleukin-8 ( IL-8 ) was the most down-regulated gene in both siATM and si mutant p53 cells ( Figure 4A ) . We confirmed IL-8 down-regulation in mutant p53-containing cell lines MDA-MB-231 and BT-549 upon ATM depletion ( Figure 4B , C ) . Conversely , depletion of ATM in cancer cell lines containing WT p53 resulted in increased IL-8 mRNA levels ( Figure 4—figure supplement 2A , B ) . These results suggest that ATM promotes IL-8 levels in the context of mutant p53 . IL-8 is upregulated in several cancers , including breast cancer , where it mediates several cancer promoting pathways including cell migration ( Campbell et al . , 2013; Singh et al . , 2013 ) . The IL-8 promoter contains many transcription factor binding sites , including NF-κB , which regulates IL-8 expression and is linked to the DDR through ATM activation by DSBs ( Mukaida et al . , 1990; Biton and Ashkenazi , 2011; McCool and Miyamoto , 2012 ) . We confirmed IL-8 promoter regulation by NF-κB as ∼90% of IL-8 promoter activity was lost by mutating the NF-κB binding site ( mut IL-8 , Figure 4D ) . Interestingly , depletion of ATM or mutant p53 reduced IL-8 promoter activity similarly as mut IL-8 , showing ATM regulation of IL-8 occurs at the transcriptional level ( Figure 4D ) . As expected , we observed that depletion of NF-κB p65 , a subunit of NF-κB dimer , or NEMO abrogated IL-8 expression in MDA-MB-231 ( Figure 4E , Freund et al . , 2004 ) . Both ATM and p53 are known to be required for NF-κB localization and activation in the nucleus upon various stimuli including cellular stress ( Wuerzberger-Davis et al . , 2007; Hoesel and Schmid , 2013 ) . To determine whether NF-κB function required ATM or mutant p53 in our cell system , we investigated the nuclear localization of the NF-κB subunit p65 in MDA-MB-231 cells under normal growth conditions . Nuclear localization of the p50/p65 NF-κB dimer enables transcriptional activation of this complex so we analyzed p65 nuclear accumulation as a readout of NF-κB localization ( Hayden and Ghosh , 2012 ) . We observed reduced p65 nuclear localization and NEMO phosphorylation in ATM- and mutant p53-depleted cells compared to control cells , which is inline with the reduced IL-8 expression that occurs under these conditions ( Figure 4G , Figure 4—figure supplement 2F ) . We next performed chromatin immunoprecipitation ( ChIP ) of NF-κB on the IL-8 promoter to analyze directly the involvement of NF-κB in regulating IL-8 transcription and how this is affected by ATM and mutant p53 . ChIP analyses revealed that reduced levels of ATM or mutant p53 impaired NF-κB accumulation on the IL-8 promoter ( Figure 4H ) . Collectively , our results strongly suggest that ATM and mutant p53 are required for NF-κB activity , which is necessary to regulate IL-8 expression . Further analyses supported the notion of IL-8 as the gene responsible for reduced migration in ATM-depleted MDA-MB-231 cells as ( 1 ) IL-8 depletion reduced cell migration and invasion , ( 2 ) NAC treatment reduced IL-8 mRNA levels and ( 3 ) oxidative stress induction by H2O2 increased IL-8 levels and ( 4 ) H2O2-induced IL-8 expression was dependent on ATM ( Figure 5A–E ) . Taken together , these results suggest that ATM regulates a transcriptional network that includes the NF-κB-regulated gene IL-8 . Our data suggests that this ATM pathway promotes cell migration and invasion in MDA-MB-231 cells through a cell intrinsic mechanism that is reliant on endogenous oxidative stress . 10 . 7554/eLife . 07270 . 017Figure 5 . ATM promotes pro-metastatic IL-8-dependent cellular processes . ( A ) IL-8 depletion reduces cell migration and invasion . Experiments performed as in Figure 1G . Error bars = SEM . * p-value <0 . 05 , *** p-value <0 . 001 , unpaired two-tailed t-test . ( B ) IL-8 qRT-PCR analysis from samples in ( A ) . ( C ) ROS inhibitor NAC reduces IL-8 expression . ( D ) H2O2-induced oxidative stress increases IL-8 expression . ( E ) H2O2-induced IL-8 expression is dependent on ATM . Cells treated with indicated siRNAs were incubated with 0 . 025 mM H2O2 or mock treated and analyzed by qPCR . ( F and G ) IL-8 addition restores impaired migration and invasion for ATM-depleted ( F ) or mutant p53-depleted ( G ) cells . Experiments performed as in Figure 1G with or without recombinant IL-8 . For all graphs , mean ± s . e . m . , n = 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01710 . 7554/eLife . 07270 . 018Figure 5—figure supplement 1 . Exogenous addition of IL-8 rescues migration and invasion defects in ATM-depleted cells . ( A and B ) IL-8 addition restores impaired migration and invasion for cells treated with independent siATM-1 ( A ) and siATM-2 ( B ) compared to siNC control cells . Experiments performed as in Figure 5F , G ( mean ± s . e . m . , n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 018 The importance of IL-8 in promoting cell migration in MDA-MB-231 cells , and its reduction upon ATM inhibition , prompted us to test whether reduced IL-8 expression in ATM-depleted cells was responsible for reduced migration and invasion in these cells . Supporting this hypothesis , the addition of recombinant IL-8 rescued both migration and invasion properties in both mutant p53 and ATM-depleted cells ( Figure 5F , G ) . These results were confirmed with two individual siRNAs targeting ATM to ensure that these results were not due to any siRNA off-target effects ( Figure 5—figure supplement 1 ) . These data identify IL-8 as an ATM regulated gene target that strongly influences the reduced migration and invasion of ATM and mutant p53 deficient MDA-MB-231 breast cancer cells . ATM is considered a tumor suppressor as its deletion in mice results in tumors , patients with mutations in ATM in the human disorder Ataxia telangiectasia have increased cancer risks , and several hematological cancers have been reported to express defective ATM ( Cremona and Behrens , 2013; Shiloh and Ziv , 2013; Stracker et al . , 2013 ) . However , our identification for a requirement of ATM to promote IL-8 expression , cell migration and invasion in vitro , as well as the association of increased levels of IL-8 with the metastatic potential of several cancer types , including colon cancer and melanoma ( Lee et al . , 2012; Wu et al . , 2012 ) , prompted us to evaluate if ATM could influence tumorigenesis and promote pro-metastatic processes . To address this , we generated a MDA-MB-231 cell line with stable ATM depletion via shRNA . shATM and shNC cells exhibited similar cellular morphology and growth rates in vitro ( Figure 6—figure supplement 1A , B ) and shATM cells exhibited reduced cell migration and invasion , as well as reduced IL8 expression ( Figure 6—figure supplement 1A , C ) . In addition , both shATM and shNC cells formed colonies with similar efficiency in soft agar ( Figure 6A , B ) . Thus , ATM depletion by shRNA did not overtly alter the growth properties of MDA-MB0231 cells . 10 . 7554/eLife . 07270 . 019Figure 6 . ATM promotes tumor progression in vivo . ( A ) ATM is not required for colony growth in soft agar . Representative images of shNC and shATM are shown . ( B ) Quantification of soft agar assays . Colonies were counted from 10 fields of view . Differences between shNC and shATM are not significantly different ( NS ) by unpaired two-tailed t-test . ( C ) MDA-MB-231 shATM cells increase cancer cell proliferation compared to shNC cells by measuring orthotopic tumor growth . ( D ) ATM-depletion reduces lung tumor formation . Representative India ink stained lungs at necropsy from mice ( N = 12 for each group ) following tail-vein injection xenografting of shNC or shATM MDA-MB-231 cells . Lung tumors indicated by arrows and dotted lines . ( E ) Quantification of lung tumors from D . ( F ) Kaplan-Meier plot of probability of lung metastasis free survival in 560 breast cancer patients based on IL8 expression levels . ( G ) Model for ATM pathway in tumor progression . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 01910 . 7554/eLife . 07270 . 020Figure 6—figure supplement 1 . ATM is required for cell motility and IL-8 expression . ( A ) shATM cells exhibit defects in cell migration and invasion but not proliferation compared to shNC cells . Experiments were performed as in Figure 1G . Western blot showed ATM depletion in shATM cells compared to shNC cells . ( B ) Morphology of shNC and shATM MDA-MB-231 cells . ( C ) IL-8 expression is reduced in shATM cells compared to shNC cells . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 02010 . 7554/eLife . 07270 . 021Figure 6—figure supplement 2 . Analysis of gene expression data from breast cancer patients . ( A ) Boxplot of IL8 ( left panel ) and ATM ( right panel ) expression in patients grouped according to IL8 expression ( see Figure 6F ) . ( B ) Pearson analysis of correlation between ATM and IL8 in the 560 patient samples used in the analysis . ( C ) Histogram of p-values of a Univariate Cox proportional hazard regression for the expression of all genes in the microarray . In color we show the p-values obtained with IL8 expression ( 0 . 0033 percentile ) and ATM expression ( 0 . 4676 percentile ) . ( D ) Kaplan-Meier plots of probability of brain and bone metastasis free survival in 560 breast cancer patients based on IL8 expression levels . DOI: http://dx . doi . org/10 . 7554/eLife . 07270 . 021 To examine the influence of ATM on tumor growth in vivo , we then orthotopically injected these cells into the mammary fat pad . There , we observed an increase in primary tumor growth , consistent with the well-established role of ATM in tumor suppression ( Figure 6C ) . To address whether ATM depletion could influence the ability of these cells to colonize secondary sites , we performed tail vein injections with shATM and shNC MDA-MB 231 cells and assessed lung colonization after 9 weeks . In contrast to the pro-tumorigenic effects of ATM depletion on primary growth , lung colonization was strongly suppressed in cells with reduced ATM levels compared to control cells ( Figure 6D , E ) , consistent with the ability of ATM activity to regulate IL-8 mediated cell migration and invasion capacity . To examine the relationship between ATM and IL-8 expression and metastasis in human breast cancer , we analyzed a set of expression arrays from 560 breast cancer patients with clinical annotation ( Morales et al . , 2014 ) . The expression of IL-8 was significantly associated with metastasis to the lung but not bone or brain ( Figure 6F , Figure 6—figure supplement 2 and Supplementary file 2 ) . These data suggest that IL-8 , and potentially ATM , do not affect all metastatic processes similarly in all tissues . The mRNA expression levels of ATM were not similarly associated with the probability of metastasis , however , ATM is regulated primarily at the post-transcriptional level and these analyses do not reflect its activity ( Figure 6F ) . Collectively , these results demonstrate that ATM can play dual functions in breast cancer tumor progression . On the one hand , ATM serves as a tumor suppressor to limit the proliferation or survival of cancer cells during primary tumor formation . However , based on our findings , ATM can also promote pro-metastatic processes such as lung colonization . These functions of ATM are likely through its ability to control the transcription of key regulators , including IL-8 , that promote cell migration and invasion to influence metastatic colonization .
In summary , we have identified a tumor promoting activity for ATM . Our data is consistent with this pathways being triggered by intrinsic oxidative stress that occurs in highly metastatic cancer cells . This ATM-dependent pathway regulates the transcription of pro-metastatic genes including IL-8 through the activation of the transcriptional regulator NF-κB . The expression of these genes , including IL-8 , results in increased cell migration and invasion that aid in tumor progression ( Figure 6G ) . ATM is known to regulate several pro-inflammatory cytokines as part of a senescence-associated secretory phenotype ( SASP ) in senescent cells ( Rodier et al . , 2009; Coppe et al . , 2010 ) . The SASP can be mediated by DNA damage induced ATM signaling through MRN and CHK2 , but not p53 . In addition , a role for p38 MAPK in the SASP has been defined that is independent of the initial ATM-dependent damage response ( Freund et al . , 2011 ) . Senescent cells are thought to promote tumor progression in the microenvironment through paracrine signaling by secreted pro-inflammatory factors including cytokines ( Coppe et al . , 2008 , 2010 ) . ATM has also been shown to regulate IL-8 secretion in response to DSB formation by etoposide , which is dependent on NF-κB and NEMO but independent from p53 ( Biton and Ashkenazi . , 2011 ) . From our data , we conclude that ATM , through a pathway requiring mutant p53 , can support a transcriptional program that includes several pro-metastatic genes including IL-8 that maintain cell migration and invasion . These processes are operational in cancer cells and have the ability to support cellular activities contributing to tumorigenesis . Interestingly , a recent study identified a fundamentally different type of anti-cancer pathway mediated by ATM in a MLL-AF9 driven acute myeloid leukemia model ( Santos et al . , 2014 ) . In this instance , ATM was required to block the differentiation capacity of these cells , thereby promoting their tumorigenic properties . p38 MAPK plays both pro and anti-tumorigenic roles and its downregulation has been demonstrated to reduce tumor growth and survival but also to promote the metastasis of colorectal cancer to the lung through the modulation of inflammatory responses ( Gupta et al . , 2014; Urosevic et al . , 2014 ) . Thus , emerging evidence indicates that ATM represents another DDR factor that , like p53 or p38 , can differentially influence pathways relevant to diverse aspects of tumorigenesis in a context dependent manner . Our study emphasizes the ability of ATM to regulate the transcription of a sub-set of genes involved in tumor promoting pathways . Comparative expression analysis revealed wound healing as a major pathway regulated by both ATM and mutant p53 . Several pro-metastatic genes from this analysis in addition to IL-8 , including CXCL11 , CCL2 , MMP1 and SAA2 , were identified . By identifying IL-8 as a target of oxidative stress and ATM , we provide a link between activated ATM by oxidative stress and pro-metastatic pathways operating in cancer cells . Our results provide additional insights into other disease pathways involving ATM including angiogenesis , a pathway triggered by oxidative stress and dependent on IL-8 ( Koch et al . , 1992; Okuno et al . , 2012 ) . Inhibitors of the IL-8 pathway are actively being developed and evaluated in pre-clinical studies and clinical trials for cancer and inflammatory diseases ( Waugh and Wilson , 2008; Singh et al . , 2013 ) . Our work provides a rationale for evaluating the clinical efficacy of IL-8 inhibition in the context of lung metastasis , including the use of ATM inhibitors beyond their ability to radio-sensitize cancer cells . It is intriguing that high IL-8 expression levels in breast cancers were specifically correlated with lung metastasis , a tissue location with high oxygen burden ( Rosanna and Salvatore , 2012 ) . Thus , regulation of IL-8 by ATM could promote a metastatic program that is important in an organ-specific context where oxidative stress and cytokine secretion occur . We hypothesize based on our findings that targeting ATM could inhibit IL-8 dependent processes involving tumor progression and metastasis in certain cancers . In addition , based on our study and those of others , ATM inhibition could also be efficacious in the setting of other diseases involving oxidative stress and pro-inflammatory cytokines including inflammation , angiogenesis and aging ( Osorio et al . , 2012; Okuno et al . , 2012 ) .
MDA-MB-231 , MCF-7 , and U2OS cells were grown in Dulbecco's Modified Eagle's Medium ( DMEM ) supplemented with 10% fetal bovine serum ( FBS ) , 100 U/ml penicillin , 100 mg/ml streptomycin and 2 mM L-glutamine and BT-549 cells were maintained in RPMI-1640 medium with the same supplements . DNA-damaging treatments used in wound-healing assay were as follows: ATM inhibitor KU-55933 ( ATMi , 10 μM ) , doxorubicin ( Dox , 100 nM ) , phleomycin ( Phleo , 90 μg/ml ) , etoposide ( Et , 40 μM ) , and IR ( 20 Gy ) by a Faxitron X-ray unit , ( Faxitron X-ray Corporation , Tucson , AZ ) . Cells were incubated with these agents for 48 hr except IR , which is imposed on cells at 0 hr followed by incubating the cells with regular medium . Cells were washed once with PBS ( phosphate buffered saline ) , and lysed in Laemmli buffer ( 4% SDS , 20% glycerol and 120 mM Tris , pH 6 . 8 ) to obtain whole cellular lysate . The cell lysates were sonicated in a Diagenode Bioruptor 300 for 10 min , boiled for 5 min at 95°C followed by centrifugation before loading . Samples were separated by SDS-PAGE and detected using standard chemiluminescence ( GE Healthcare Amersham ECL prime , Piscataway , NJ ) using a Bio-Rad Molecular Imager ChemiDoc XRS+ system . For ChIP , cells were crosslinked in 1% formaldehyde for 10 min followed by 125 mM glycine treatment to stop the reaction . Cells were then lysed in RIPA buffer ( 50 mM Tris-HCl , pH 8 . 0 , 150 mM NaCl , 2 mM EDTA , pH 8 . 0 , 1% NP-40 , 0 . 5% Sodium Deoxycholate , 0 . 1% SDS and a proteinase inhibitor ) and sonicated in a Biorupter ( Diagenode , Denville , NJ ) . Samples were immunoprecipitatd with NF-κB p65 antibody overnight at 4°C . After incubation of protein A agarose for 4 hr at 4°C , beads were washed once with the following solutions: TSE-150 ( 1% Triton X-100 , 0 . 1% SDS , 2 mM EDTA , 20 mM Tris-HCl , pH 8 . 0 , 150 mM NaCl ) , TSE-500 ( 1% Triton X-100 , 0 . 1% SDS , 2 mM EDTA , 20 mM Tris-HCl , pH 8 . 0 , 500 mM NaCl ) , LiCl buffer ( 0 . 25 M LiCl , 1% NP-40 , 1% SOC , 1 mM EDTA , 10 mM Tris , pH 8 . 0 ) , and TE buffer ( 10 mM Tris pH 8 . 0 , 0 . 1 mM EDTA pH 8 . 0 ) . After reverse crosslinking , DNA was purified using a PCR purification kit ( Qiagen , Netherlands and Germany ) . qRT-PCR analysis was performed using SYBR green ( Applied Biosystems , Foster City , CA ) on an Applied Biosystems StepOnePlus system . For cytoplasmic and nuclear extract , 3 × 106 cells were suspended in 150 μl solution containing 10 mM HEPES , pH7 . 9 , 10 mM KCl , 1 . 5 mM MgCl2 , 0 . 34 M sucrose , 10% glycerol , 1 mM DTT , 0 . 1% TritonX-100 and a proteinase inhibitor . Followed by incubating on ice for 5 min , cells were harvested by centrifugation at 1300×g for 4 min to acquire cytoplasmic ( supernatant ) and nuclear ( pellet ) fractions . Then the pellet was added 200 μl of Laemmli buffer and sonicated for 10 min . These extracts were analyzed as above described . Antibodies used for western blotting were as follows: H2AX ( #2595; Cell Signaling , Beverly , MA ) , γH2AX ( #NB100-384; Novus Biologicals , Littleton , CO ) , Phospho-p53 ( Ser15 ) ( #9284; Cell Signaling ) , p53 ( #39553; Active Motif , Carlsbad , CA ) , ATM ( #sc-135663; Santa Cruz Biotechnology , Santa Cruz , CA ) , CHK2 ( #2662; Cell Signaling ) , KAP1 ( #sc-33186; Santa Cruz Biotechnology ) , phosphor-KAP1 Ser824 ( #A100-767A; Bethyl Laboratories , Montgomery , TX ) , MRE11 ( #ab214; Abcam , United Kingdom ) , NBS1 ( #NB100-143; Novus Biologicals ) , NF-κB p65 ( #sc-109; Santa Cruz Biotechnology ) , NF-κB p65 Ser536 ( #3033; Cell Signaling ) , NF-κB p65 Ser468 ( #3039; Cell Signaling ) , NF-κB p65 Ser276 ( #sc-101749; Santa Cruz Biotechnology ) , NEMO ( #sc-8330; Santa Cruz Biotechnology ) , NEMO Ser85 ( #ab63551; Abcam ) , and β-tubulin ( #ab6046; Abcam ) . Secondary antibodies conjugated to horseradish peroxidase ( Cell Signaling ) were used for enhanced chemiluminescence . Small interfering RNA ( siRNA ) SMARTpools were obtained from Dharmacon: siNC ( Non-targeting pool ) , siATM , siCHK2 , siMRE11 , siNBS1 and sip53 . The sequences for independent siATM-1 and siATM-2 , which are part of siATM SMARTpools , are GCAAAGCCCUAGUAACAUA and GGGCAUUACGGGUGUUGAA ( Dharmacon , Lafayette , CO ) . The sequence for siIL-8 is GCCAAGGAGUGCUAAAGAA ( Dharmacon ) . The sequence for sip65 is GAUCAAUGGCUACACAGGAUU ( Dai et al . , 2005 ) . The sequence for siNEMO is GGAAGAGCCAACUGUGUGAUU ( Hinz et al . , 2010 ) . Lipofectamine RNAiMAX ( Invitrogen , Carlsbad , CA ) was used to transfect the indicated siRNA into cells following the manufacturer's instructions . The promoter region of IL-8 corresponding to −1423/+96 bp was cloned from MDA-MB-231 cells by PCR using 5′-CTCGAGGTAACCCAGGCATTATTTTA-3′ and 5′-AGATCTAGCTTGTGTGCTCTGCTGTC-3′ primers . The PCR fragment was then subcloned into pCR-Blunt II-TOPO vector ( Invitrogen ) . The IL-8 promoter was cut with XhoI and BglII and cloned into Xho1/BglII treated pNANOG-Luc ( Addgene No . 25900 ) . After sequencing , this IL-8 promoter driven luciferase construct was named as pIL8-Luc . As a negative control , we made another IL-8 reporter with mutated NF-κB binding ( named as mut pIL8-Luc ) by using pIL8-Luc as a template and the following primers: 5′-ATGGGCCATCAGTTGCAAATCGTTAACTTTCCTCTGACATAATGAAAAGA-3′ 5′-TCTTTTCATTATGTCAGAGGAAAGTTAACGATTTGCAACTGATGGCCCAT-3′ The luciferase assay was performed by transfecting with indicated siRNA on day 1 . After 24 hr , pIL8-Luc or mut pIL8-Luc and a trace amount of pRL-SV40P as an internal control ( Addgene No . 27163 ) were transfected into cells with FugeneHD ( Promega , Madison , WI ) . On day 3 or 4 , cells were lysed and both firefly and renilla luciferase were measured by using Dual-Glo Luciferase assay system ( Promega ) on a microplate reader ( TECAN , Switzerland ) . All firefly luciferase measurements were corrected for renilla luciferase values . The ratios were normalized to control groups and expressed as relative luciferase unit . MDA-MB-231 and BT-549 cells were seeded into 6-well plates and cultured to confluence . Cells were then wounded with sterile pipette tips and washed with PBS . Then cells were treated or untreated with DNA-damaging agents as described . For wound-healing assay with siRNA treatments , cells were transfected with indicated siRNA and after 48 hr , wounds were made as described above . Pictures were acquired at 0 hr and 48 hr using an EVOS fl fluorescence microscope . Representative pictures were from three independent experiments and quantified by using TScratch software ( Geback et al . , 2009 ) . Soft agar assays were performed on shNC and shATM MDA-MB-231 cells essentially as described ( Xhemalce et al . , 2012 ) . For live analysis of cell motility , MDA-MB-231 cells were plated in glass round petri dish ( TED PELLA , INC , Redding , CA ) and treated as indicated . Images were taken every 15 min using Olympus FluoView FV1000 confocal microscope and analyzed with ImageJ manual tracking plugin . Greater than 100 cells were analyzed for each experiment and sample and results provided are from three independent experiments . Cellular proliferation of MDA-MB-231 was counted by trypan blue staining . Briefly , cells were plated in 6-well plates and treated with indicated DNA-damaging agents . Then , the cells were trypsinized and counted by trypan blue staining . For cell-cycle analysis , MDA-MB-231 cells untreated or treated with indicated DNA-damaging agents for 48 hr , except that cells were exposed to IR at 0 hr followed by incubated with regular medium , were trypsinized and fixed in 70% ice-cold ethanol overnight at 4°C . Cells were centrifuged , suspended and incubated in PBS containing propidium iodide ( 50 μg/ml ) and RNase A ( 100 μg/ml ) for 30 min at room temperature . DNA content was analyzed by using BD Accuri C6 flow cytometer , and FlowJo software was used to analyze flow cytometry data . Intracellular ROS was determined by CM-H2DCFDA ( Invitrogen ) . Briefly , following indicated treatments , cells were trypsinized and incubated with 5 μM CM-H2DCFDA in PBS for 30 min in the 37°C incubator . Cells were then returned to phenol red-free medium for 15 min recovery period and immediately analyzed by BD Accuri C6 flow cytometer . 4 mM H2O2 treatment served as a positive control and 10 mM NAC was used as a reducing agent . Data was analyzed using FlowJo and mean fluorescence intensity was used as a measure of ROS . xCELLigence Real-Time Cell Analyzer ( RTCA; Roche Diagnostics , Switzerland ) was used to monitor cell proliferation , migration and invasion independently in a label-free , real-time setting . When cells contact and adhere to electrical sensors , this leads to increasing electrical impedance . The impedance correlates with an increase in proliferating , migrated or invaded cells , derived as a parameter called cell index . For proliferation experiment , cells were treated as indicated and seeded in quadruplicates on E-plate . For migration and invasion assays , following indicated treatments , cells were starved in serum-free medium overnight and seeded in quadruplicates on CIM-plate . Matrigel ( BD Biosciences ) was diluted in serum-free medium at a ratio of 1:40 and coated on CIM-plate only for invasion . Regular growth medium was added in the lower chamber as chemoattractant . The cell index was measured every 15 min , and the results were represented as normalized cell index . For exogenous addition of interleukin-8 ( IL-8 ) ( PEPROTECH ) , 150 ng/ml of IL-8 was added to cells transfected with siATM and 200 ng/ml of IL-8 was added to cells depleted for p53 . Total RNA form cells transfected with siNC , siATM or sip53 was extracted using RNeasy Mini kit ( Qiagen ) following manufacturer's instructions . All RNA was DNase treated before sendind to Microarray core facility at DANA-Farber Cancer Institute to perform Affymetrix GeneChip Human Gene 2 . 0 ST array ( n = 2 in each group ) . Affymetrix Expression Console ( EC ) was used to generate CHP files from CEL files . Then CHP files were loaded into Affymetrix Transcriptome Analysis Concole ( TAC ) software . Normalizaton and gene expression analysis were performed in TAC software . GO analysis was analyzed using the ‘Functional Annotation Tool’ in DAVID ( http://david . abcc . ncifcrf . gov/home . jsp ) and biological process terms are shown ( Huang da et al . , 2009 ) . Total RNA from cells treated as indicated was purified using RNeasy Mini kit ( Qiagen ) and treated with DNase according to manufacturer's instructions . 1 μg of total RNA was used for cDNA systhesis with SuperScript III first-strand synthesis system . To analyze IL-8 mRNA expression , we designed gene-specific qPCR primers: 5′-AAGAAACCACCGGAAGGAAC-3′ and 5′-ACTCCTTGGCAAAACTGCAC-3′ for IL8 . GAPDH ( glyceraldehyde-3-phosphate dehydrogenase ) ( Quantitect primer assay , Qiagen ) was used for normalization . For analyzing NEMO and p65 expression , we designed the following qPCR primers: 5′-AGAGTCTCCTCTGGGGAAGC-3′ and 5′-GCTTGGAAATGCAGAAGCTC-3′ for NEMO; 5′-ACAACCCCTTCCAAGTTCCT-3′ and 5′-ATCTTGAGCTCGGCAGTGTT-3′ for p65 . To perform ChIP-qPCR , we used the following primers: ( 1 ) 5′-GTGTGATGACTCAGGTTTGC-3′ and 5′-GTTTGTGCCTTATGGAGTGC for IL-8 promoter and ( 2 ) 5′-CGGAAAGATCGCCATATATGGAC-3′ and 5′-ACCGGCAGAGAAACGCGA-3′ for actin promoter . Quantitative real-time PCR analysis was performed using SYBR green ( Applied Biosystems ) on an Applied Biosystems StepOne Plus system . Stable MDA-MB-231 shNC and shATM cells were made by using shATM ( sequence: GAAGTAGAAGGAACCAGTTACCATGAATC ) and shNC ( control ) plasmids from Origene . These cells were maintained in regular DMEM with 0 . 2 μg/ml puromycin . Balb/c mice were maintained according to the University of Texas at Austin Institutional Animal Care and Use Committee guidelines . 0 . 5 × 106 shNC or shATM MDA MB 231 breast cancer cells were injected into the mouse tail vein . Mice were then monitored daily for symptoms of distress , including difficulty breathing , ruffled appearance , and limited mobility . Surviving mice were euthanized 9 weeks after cell injections . Mouse lungs were perfused with India ink to quantify metastatic lesions as previously reported ( Williams et al . , 2004 ) . For orthotopic tumor experiment , 2 × 106 shATM or shNC MDA MB 231 cells were injected into the #4 mammary fat pad of 6 week-old Balb/c mice . Mice were palpated three times weekly until tumors reached 750 mm3 . Tumor size was expressed as tumor volume ( mm3 ) and calculated by the formula: volume = ( smaller dimension2 × larger dimension ) /2 . Publically available breast cancer patient gene expression data from GEO was pooled ( GSE2603 , GSE2034 , GSE5327 , and GSE12276 ) as previously described ( Minn et al . , 2005; Barrett et al . , 2007; Morales et al . , 2014 ) . In order to remove systematic biases , expression measurements were converted to z‐scores for all genes before merging . Multiple probes from the same gene were converted to a mean , patients were grouped based on expression levels and Kaplan–Meier survival was plotted . The hazard ratio ( HR ) and p value for each gene was calculated using a Cox proportional hazards model and performing likelihood ratio tests . The HR was checked for consistency over time , fulfilling assumptions of the Cox model . All significance measurements were done using expression as a continuous variable . | Damaged DNA threatens the normal activity of living cells , so cells use a number of mechanisms to ensure that this damage is repaired . When DNA is damaged , an enzyme called ATM activates several other proteins that ultimately lead to the DNA being repaired . Problems with detecting and repairing damaged DNA have been linked to cancer . Thus , these pathways , including the activity of ATM , were previously thought to only be involved in cancer suppression . Now , Chen et al . report a new cancer-promoting role for ATM . The experiments reveal that reducing the amount of ATM in cancer cells actually made them less able to migrate and less invasive . Likewise , human breast cancer cells in which the levels of ATM had been depleted formed fewer lung tumors than normal breast cancer cells when they were transplated into mice . Oxidative stress—a build-up of harmful chemicals inside cells—is a signature feature of cancer cells and is known to be another signal that activates ATM . Chen et al . found that activating ATM through oxidative stress , but not by DNA damage , encouraged the cancer cells to migrate and become invasive . Further analysis of cellular responses following ATM activation by oxidative stress revealed that this enzyme regulates the production of a small protein called interleukin-8 . This protein is an important pro-inflammatory molecule that has been implicated in cancer , in particular , in helping cancer cells to migrate to other tissues . When interleukin-8 was added to ATM-depleted cancer cells , it rescued their defects in spreading and invasiveness , thereby providing strong evidence that interleukin-8 is a biologically important target of ATM . Clinical data confirmed that breast cancer cells that had also spread to the patient's lungs often produced high levels of interleukin-8 . Together , these findings suggest that ATM could be a potential target for anti-cancer therapies , as inhibiting this enzyme would inhibit interleukin-8 , and in turn slow the progression and spread of cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"cell",
"biology"
] | 2015 | ATM regulation of IL-8 links oxidative stress to cancer cell migration and invasion |
Actin-based thin filament arrays constitute a fundamental core component of muscle sarcomeres . We have used formation of the Drosophila indirect flight musculature for studying the assembly and maturation of thin-filament arrays in a skeletal muscle model system . Employing GFP-tagged actin monomer incorporation , we identify several distinct phases in the dynamic construction of thin-filament arrays . This sequence includes assembly of nascent arrays after an initial period of intensive microfilament synthesis , followed by array elongation , primarily from filament pointed-ends , radial growth of the arrays via recruitment of peripheral filaments and continuous barbed-end turnover . Using genetic approaches we have identified Fhos , the single Drosophila homolog of the FHOD sub-family of formins , as a primary and versatile mediator of IFM thin-filament organization . Localization of Fhos to the barbed-ends of the arrays , achieved via a novel N-terminal domain , appears to be a critical aspect of its sarcomeric roles .
Sarcomeres constitute the basic functional units of muscle fibers , endowing these large and specialized cells with their contractile capacity . Central to sarcomere function is the lattice-like organization of two filament systems: an actin-based thin-filament array , which provides a stiff backbone along which thick filaments , composed of myosin motor proteins , 'slide' in order to produce force and contractile motion ( Squire , 1997 ) . The spatial organization and efficient operation of this remarkable cellular machinery relies on a host of dedicated proteins and protein complexes , which act to regulate sarcomere size and streamline its activity , and to coordinate between the multiple sarcomeric units that comprise individual myofibrils ( Clark et al . , 2002; Ehler and Gautel , 2008; Gautel and Djinovic-Carugo , 2016 ) . Despite their fundamental significance , elucidation of the molecular mechanisms underlying assembly , maturation and maintenance of thin-filament arrays remains one of the major open issues in the study of sarcomere structure and function . While mechanisms relating to size definition and stability of the arrays have been extensively investigated ( Fernandes and Schock , 2014; Meyer and Wright , 2013 ) , other key aspects of microfilament array formation and dynamics , including determination of distinct phases of array maturation , the identity and regulation of elements mediating filament nucleation/elongation , and the processes governing incorporation of additional filaments into nascent arrays are not resolved ( Ono , 2010 ) . Here we address these matters in the context of formation and development of the Drosophila indirect flight muscles ( IFMs ) . These are the largest muscles of the adult fly , which power flight by regulated contraction of the thorax ( Dickinson , 2006 ) . A major subset of the IFMs , the dorso-longitudinal muscles ( DLMs ) , closely resemble vertebrate skeletal muscles in both their developmental program and in their mature myofibrillar structure ( Dutta and VijayRaghavan , 2006; Roy and VijayRaghavan , 1999 ) , making them a particularly attractive model system , in which the powerful molecular genetic approaches available to study Drosophila development can be harnessed to investigate and elucidate general principles of myogenesis . DLM formation initiates by fusion of hundreds of individual myoblasts to a set of larval muscles during the first 24-30 hours of pupal development ( Fernandes et al . , 1991 ) . The subsequent ~80 hrs of myogenesis leading up to eclosion of the adult fly include formation and maturation of a parallel arrangement of myofibrils and assembly and growth of sarcomeric units within them ( Reedy and Beall , 1993; Weitkunat et al . , 2014 ) . This sequence of events takes place over a wide time window , providing an opportunity to temporally dissected and manipulate the coordinated processes giving rise to thin-filament array assembly and maturation . IFM sarcomeres initiate as small , nascent structures , that grow considerably over the course of pupal development ( Sparrow and Schock , 2009 ) , reaching a final , uniform size of 3 . 4 µm in length and 1 . 5 µm in diameter . Spatial organization of the mature , evenly-spaced IFM sarcomeres closely mirrors that of striated vertebrate skeletal muscle ( Reedy and Beall , 1993 ) . Individual sarcomeric units are defined by Z-disc borders , which serve as anchoring sites for the barbed ends of the thin-filament arrays , and by a central , microfilament-free H-zone , bordered by the pointed-ends of neighboring thin-filament arrays . We utilized temporally-controlled expression of GFP-tagged actin monomers ( Roper et al . , 2005 ) to follow thin-filament array dynamics , and recognize key phases and distinct transitions of the arrays throughout sarcomerogenesis . In parallel we identified Fhos , the single Drosophila member of the conserved FHOD family of formin proteins ( Schonichen and Geyer , 2010 ) , as a major contributor to thin-filament array assembly and growth . An important aspect of Fhos involvement is its critical role in radial growth of the arrays , by mediating incorporation of new peripheral filaments to the nascent core structure . The elongation of thin filament arrays , shown to occur primarily from their pointed ends ( Mardahl-Dumesnil and Fowler , 2001 ) , is mediated by the WH2-domain actin regulator Sals protein . While the combined activities of Fhos and Sals can account for most aspects of thin-filament array growth and maturation during pupal stages , other elements are likely involved in additional processes that shape and maintain the arrays , such as continuous monomer exchange at the barbed-ends .
Assembly of sarcomeric thin-filament arrays has been traditionally studied by monitoring global changes in sarcomere structure in fixed samples of muscle tissue . However , in order to decipher the underlying regulatory mechanisms and uncover the machineries that drive this process , it is essential to monitor the dynamic patterns of actin monomer incorporation into the growing sarcomere . We therefore followed the incorporation of GFP-tagged actin monomers into IFM sarcomeres , throughout the pupal stages of development , as a means of revealing the assembly and maturation of IFM sarcomeric thin-filament arrays in developing flies . Inducible UAS-based GFP-actin transgenes ( Roper et al . , 2005; Verkhusha et al . , 1999 ) have proven to be reliable tools and have been used extensively to study microfilament localization and dynamics during Drosophila development ( Fulga et al . , 2007; Jacinto et al . , 2000; Kaltschmidt et al . , 2002; Perkins and Tanentzapf , 2014; Schottenfeld-Roames and Ghabrial , 2012 ) . The actin isoform at chromosomal position 88F , one of six Drosophila actin genes , is specifically expressed during the formation of the pupal muscles , and represents the major actin isoform used in this tissue ( Beall et al . , 1989; Fyrberg et al . , 1983 ) . We therefore employed temporally-restricted induction protocols of UAS-GFP-actin88F ( Roper et al . , 2005 ) , and compared the resulting GFP patterns to the outlines of phalloidin-stained sarcomeres , as a primary tool for following the dynamics of IFM thin-filament array development ( Figure 1A ) . 10 . 7554/eLife . 16540 . 003Figure 1 . Four distinct modes of GFP-actin monomer incorporation contribute to formation of IFM thin-filament arrays . ( A ) Scheme of IFM development intervals used for unrestricted and temporally restricted expression of GFP-actin88F ( B–E” ) or GFP-actin5C ( G–I ) . ( B–B” ) Induction of GFP-actin88F expression ( green , gray ) with mef2-Gal4 throughout fly development results in full monomer incorporation into the thin-filament arrays ( phalloidin- blue , gray ) , as monitored in IFMs of young ( 1–3 days old ) adults . Z-discs are indicated by anti-Zasp52 ( red ) . The designations 'Z' and M' are used throughout to mark the Z-disc ( array barbed-end ) and H-zone/M line ( array pointed end ) regions of the sarcomere . ( C–E’’ ) Incorporation patterns of GFP-actin88F ( green , gray ) following temporally restricted expression pulses using the mef2-GAL4 driver and the GAL80ts/TARGET system . Microfilaments are visualized with phalloidin ( blue , gray ) . Z-discs are indicated by anti-Zasp52 ( red ) . ( C–C” ) 0–30 hrs APF . Initial uniform incorporation . ( D–D” ) 30–45 hrs APF . 'Patched' incorporation of monomers . This mode occurs mainly at array ends ( insets in D and D’ ) , in proximity to the future Z-disc ( Z , white arrow ) or towards the opposite boundary of the nascent arrays ( M , red arrow ) . ( E–E’’ ) 50–90 hr APF . Monomer incorporation into a 'frame' generated by peripheral 'thickening' ( white arrowhead in panel E inset ) and pointed-end growth ( M and red asterisk in panel E inset ) . Red arrowhead ( panel E’ inset ) points to an absence of incorporated monomers at the barbed-end boundary ( Z ) of the arrays . ( F ) Schematic representations of the incorporation process . Blue filaments denote previously incorporated ( 'old' ) actin , while green ( 88F ) and orange ( 5C ) marks monomers newly incorporated during the indicated pulse . An initial period ( 0–30 hr APF ) of extensive actin polymerization and the establishment of nascent thin-filament arrays are followed by an interim period ( 30–45 hr APF ) of 'patchy' incorporation during which individual , uniform sarcomeric units are defined . The second half of pupal development is devoted to array growth via pointed-end elongation and recruitment of circumferential filaments , as well as turnover at array barbed ends . ( G–I ) Incorporation patterns of GFP-actin5C ( green ) following temporally restricted expression pulses using the GAL80ts/TARGET system . Z-discs are indicated by anti-Zasp52 ( red ) . Restricted expression 'windows' corresponded to 40–90 ( G ) , 50–90 ( H ) and 70–90 ( I ) hrs APF . Insets show the GFP-actin5C incorporation patterns in single sarcomeres , in which the prominent Z-disc associated stripe is outlined ( Z , Z-disc region; M , M-line region ) . ( J ) Quantification reveals a constant width of the GFP-actin5C incorporation stripe overlying the Z-disc region despite the different pulse durations , n = 200 ( 50 sarcomeres each from 4 different flies ) . Scale bars correspond to 5 μm in all main panels , 2 μm in the insets . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 00310 . 7554/eLife . 16540 . 004Figure 1—figure supplement 1 . Peripheral growth of the nascent arrays is continuous . ( A ) Scheme of IFM development intervals used for temporally restricted expression of GFP-actin88F . ( B , D ) Incorporation patterns of GFP-actin88F ( green ) following temporally restricted expression pulses at 30–45 hrs APF . Thin filament array ends are marked by Zasp52 ( red ) for the barbed ends and Obscurin ( red ) for the pointed end; microfilaments are visualized with phalloidin ( blue ) . ( C , E ) Representative intensity profiles along the white dashed line in ( B , D ) , shows the positions of 'patch' incorporation events in nascent sarcomeres . ( F ) The distribution of 'patch' incorporation events across the thin filament arrays , broken up into four categories . Data obtained from 50 sarcomeres in 5 different pupae ( n=250 ) . ( G–J ) Incorporation patterns of GFP-actin88F ( green , gray ) following temporally restricted expression pulses at 40–90 hrs APF ( G–G’’ ) and 60–90 hrs APF ( H–H’’ ) , using the mef2-GAL4 driver and the GAL80ts/TARGET system . Z-discs are indicated by anti-Zasp52 ( red ) and microfilaments are visualized with phalloidin ( blue ) . Incorporation “frames” are outlined and schematized ( G” , H” ) to show relative frame sizes following the two expression pulses and to distinguish between the array 'core' ( blue ) , elongation from pointed-ends ( green ) and radial thickening ( orange ) . Scale bars in all panels correspond to 5 μm . ( I , J ) Quantifications show significant ( ~40% , p<0 . 0001 ) differences between the long and short expression pulses in both the elongation ( I ) and radial ( J ) incorporation areas , indicative of continuous growth in both modes , n=200 ( 50 sarcomeres each from 4 different flies ) . P values determined by Mann-Whitney test . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 004 Following continuous expression of GFP-actin88F , a close correspondence between the sarcomeric GFP-actin and phalloidin patterns is observed in IFMs isolated from adult flies ( Figure 1B–B” ) . The normal size and appearance of the IFM sarcomeres indicates that the GFP-tagged actin does not block or interfere with the structuring and organization of the thin filament arrays throughout the process , and the overlap between the phalloidin and GFP-actin patterns demonstrates the reliability of this tool for monitoring thin-filament array assembly . We now attempted to break down into stages the processes of IFM thin-filament array organization and growth , by restricting actin-GFP expression to defined 'time windows' during pupal development . Temporal control of expression was achieved by using the general muscle driver mef2-GAL4 ( Ranganayakulu et al . , 1996 ) in combination with the GAL80ts/TARGET system ( McGuire et al . , 2004 ) . While the timing of induction of GFP-actin is readily controlled using this system , effective 'chase' is not possible due to the stability of the GFP-actin monomers . We first induced expression of GFP-actin88F at the onset of pupariation , and examined the nascent IFMs at 30 hr APF ( after puparium formation ) ( Figure 1C–C” ) . Individual myofibrils are already apparent at this early stage , displaying a nearly homogeneous distribution of microfilaments and an irregular pattern of nascent Z-disc structures , revealed using the early Z-disc marker Zasp52 ( Jani and Schock , 2007; Katzemich et al . , 2013 ) . A full correspondence between the GFP-actin88F and phalloidin patterns is observed , implying that the bulk of early IFM microfilaments are formed by extensive de novo filament polymerization . A very different profile of actin monomer incorporation is observed when a 15 hr pulse of actin88F-GFP expression is provided at 30–45 hr APF , immediately following the initial period of extensive polymerization ( Figure 1D–D” ) . The GFP-actin and microfilament distributions become markedly distinct from each other during this phase , in which the patterns of Zasp52 , the M-line marker Obscurin ( Burkart et al . , 2007 ) and phalloidin are now indicative of repeating sarcomeric units ( Figure 1D–D” and Figure 1—figure supplement 1B–E ) . In contrast to the early 'smeared' pattern that filled the myofibrils , actin-GFP is now restricted to discrete , isolated spots , the great majority of which ( ~80% ) positioned at either or both the ends of the nascent arrays ( insets Figure 1D , D’ and Figure 1—figure supplement 1B–F ) . This pattern suggests that the initial establishment and structuring of individual sarcomere scaffolds , achieved during this interim phase of pupal development , relies on organization of the existing microfilaments , produced earlier ( between 0–30 hr APF ) . Utilization of newly produced monomers during this interim phase of assembly is limited , and primarily involves 'patchy' incorporation along and ( mostly ) at the ends of nascent thin-filament arrays , thereby contributing to the generation of uniformly-sized sarcomeres . A subsequent pulse , between 50 and 90 hr APF , revealed yet a third pattern of actin incorporation ( Figure 1E–E” and Figure 1—figure supplement 1G–H” ) . During this ( final ) phase of pupal development , sarcomere units grow noticeably in both length and width , and the characteristic striated pattern of alternating Z-discs and filament-free H zones becomes clearly evident . Newly added actin is observed to form a distinct 'frame' that surrounds a dark rectangular core ( insets Figure 1E and E’ ) . The core presumably corresponds to the initial thin-filament array assembled during earlier phases . Conversely , the frame-like structure is likely composed of two separate contributions of newly synthesized actin to the nascent core array: extension of the initial fibers at their 'pointed' ( M-line associated ) ends , and addition of complete new fibers at the circumference of the sarcomere . This picture coincides well with previous studies of global IFM sarcomere growth ( Mardahl-Dumesnil and Fowler , 2001; Reedy and Beall , 1993 ) . Subdividing this relatively large interval further , by initiating GFP-actin88F production at 60 hrs APF , generated similarly shaped 'frames' of monomer incorporation , but of smaller size ( Figure 1—figure supplement 1H–J ) . These observations demonstrate that both aspects of actin incorporation- pointed end growth and peripheral thickening- continue throughout the entire period . Monitoring of actin monomer incorporation patterns reveals therefore a dynamic , multi-faceted timeline of IFM thin-filament array assembly during pupal development ( Figure 1F ) . An additional , prominent feature of the GFP-actin pattern following the 50–90 hr APF pulse was a dark stripe overlying the entire Z-disc ( insets Figure 1E’ ) , implying that actin88F-GFP was not incorporated at this position , in neither the core array nor in newly added filaments at the periphery ( Figure 1E , E’’ ) . This observation suggests that elongation and thickening of the array during the final 50–90 hr APF interval are accompanied by a third mode of monomer incorporation , at the barbed end of the arrays . Absence of GFP-actin88F incorporation at the Z-disc was previously noted by Roper et al ( 2005 ) , who also demonstrated preferential localization of other GFP-actin isoforms at this site , implying a specific , possibly steric hindrance of GFP-actin88F incorporation . We therefore chose to use the GFP-tagged version of actin5C ( Roper et al . , 2005 ) , a ubiquitous isoform , as a tool for monitoring monomer incorporation dynamics at barbed ends of the thin-filament array . Expression of GFP-actin5C during the 40–90 hr APF time-window resulted in a nearly complementary incorporation profile to the one generated by actin88F-GFP: a prominent stripe of GFP-actin adjacent to the Z-disc and relatively limited incorporation in the peripheral actin strands and pointed ends ( Figure 1G ) . Initiation of the GFP-actin5C pulse at different times within this 50 hr interval , generated a thin bright incorporation stripe of constant width adjacent to the Z-disc in all cases ( Figure 1H–J ) . This result implies continuous exchange and turnover of actin , rather than actual growth at the barbed ends of the arrays , and is consistent with the notion that lateral growth of array filaments occurs primarily at their pointed ends ( Littlefield et al . , 2001; Mardahl-Dumesnil and Fowler , 2001; Molnar et al . , 2014 ) . The correspondence between the regular length of the sarcomeres and the size of the actin monomers can provide a rough estimate of the number of actin monomers that build the entire structure and the proportion undergoing turnover . In a mature sarcomere , the length of thin-filament actin fibers ( as measured from the Z-disc to the H-zone ) is ~1 . 70 µm . It is difficult to accurately measure the width of the domain of dynamic actin-monomer exchange due to the limitations of light microscopy resolution , but it is roughly 0 . 15–0 . 3 µm on each side of the Z-disc . Given an estimated actin subunit size of 2 . 7 nm ( Sept et al . , 1999 ) , we can say that a complete thin filament is comprised of ~650 monomers , whereas the zone of continuous exchange at the Z-disc encompasses 50–100 monomers . To identify actin regulators that are involved in the different phases of IFM thin-filament array organization and growth , we focused on members of the formin protein family , which are major mediators of nucleation and elongation of linear microfilament arrays ( Campellone and Welch , 2010 ) . The Drosophila genome harbors six members of this protein family , each representing a distinct formin sub-family ( Liu et al . , 2010; Mi-Mi et al . , 2012 ) . To assess their involvement in the IFM sarcomere formation , we used the muscle-specific driver mef2-Gal4 to induce expression of RNAi directed against each of the six formins throughout development via UAS-based transgenic constructs , and examined IFM morphology following their isolation from newly eclosed or pharate adults . In most instances , IFM development was only mildly affected , if at all , following individual knockdown of the different Drosophila formins . A severe IFM phenotype was obtained , however , following an expression of RNAi constructs directed against Fhos , the single Drosophila FHOD sub-family homolog ( Schonichen and Geyer , 2010 ) . A normally-sized set of six DLM fibers formed in Fhos knockdown flies ( Figure 2—figure supplement 1A , B ) , indicating that the IFM developmental program is properly initiated . However , the internal organization of these fibers was severely disrupted . This was made apparent by staining the DLMs for key structural components , including α-actinin as a marker for sarcomeric Z-discs , microfilaments and muscle myosin ( Figure 2A–B” ) . Fhos knockdown DLMs appear to contain myofibril-like elements , but these are thin and randomly oriented ( Figure 2B ) . Furthermore , in contrast to the highly regular division of wildtype myofibrils into repetitive sarcomeric units ( Figure 2A , A’ ) , the abnormally thin Fhos knockdown myofibrils display only sporadic α-actinin -stained structures , and a 'smeared' , uneven distribution of microfilaments ( Figure 2B , B’ ) . In addition , muscle-specific myosin is disorganized , and to a large extent lacks an obvious association with microfilaments ( Figure 2B , B” ) . 10 . 7554/eLife . 16540 . 005Figure 2 . The formin Fhos is essential for organization and growth of thin filament arrays . ( A–C” ) Confocal images of IFMs dissected from 1 day old flies or pharate adults and stained with anti- α-actinin ( red ) to mark Z-disc structures , phalloidin ( blue , gray ) to visualize microfilaments and anti-MHC ( green , gray ) to visualize myosin . ( A–A” ) mef2-GAL4 control . Z and M mark the Z-disc and M-line of a single sarcomere . ( B–B” ) mef2-GAL4>UAS-fhos RNAi ( knockdown of all fhos isoforms ) . Myofibril and sarcomere structure and organization are defective , but sporadic , undersized sarcomeric units can be observed ( white arrowhead in B ) . ( C–C’’ ) fhosΔ1/Df ( 3L ) BSC612 ( fhos null ) . Deletion of the fhos locus results in full impairment of myofibril and sarcomeric organization . ( D–E ) TEM micrographs of longitudinal sections of IFMs dissected from control 1 day old flies ( D ) and fhos null ( fhosΔ1/Df ( 3L ) BSC612 ) pharate adults ( E ) . Distinction in the overall myofibril organization is readily apparent , with fhos null IFMs lacking typical myofibril and sarcomeric individualization . The insets contrast the stereotypic , highly-ordered structure of the control sarcomeric units ( inset D ) with the poor organization of arrays within fhos null myofibrils and their failure to form individual sarcomeres ( inset E ) . Red arrowheads in ( E ) point to dispersed , rudimentary Z-discs . ( F–G’ ) TEM micrographs of transverse sections of IFMs dissected from control 1 day old flies ( F , F’ ) and fhos null ( fhosΔ1/Df ( 3L ) BSC612 ) pharate adults ( G , G’ ) . Primed panels are magnifications of the dashed squares in panels ( F ) and ( G ) . In contrast to the highly ordered hexagonal lattice of thick ( orange ) and thin filaments ( blue ) in control myofibrils ( F’ ) , fhos null myobrils lack a defined spatial organization ( G’ ) . Scale bars: 5 μm ( A-E ) , 500 nm ( insets in D , E , and F , G ) , 100 nm ( F’ , G’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 00510 . 7554/eLife . 16540 . 006Figure 2—figure supplement 1 . Fhos function is required during the early stages of sarcomerogenesis . ( A–D ) Low magnification images of hemithoraces from 1 day old or pharate adults , dissected from a control ( mef2-GAL4 ) fly ( A ) , a fhos knockdown fly ( all isoforms ) ( B ) , a fhos null fly ( fhosΔ1/Df ( 3L ) BSC612 ) ( C ) and a fhos knockdown fly ( long isoforms only ) ( D ) , and stained with anti-MHC ( green ) and phalloidin ( blue ) . No significant differences in muscle fiber size , number or overall morphology are observed . ( E–F’ ) Higher magnification views of IFMs dissected from control ( E , E” ) or fhos null ( F , F’ ) pupae at 50 hr APF , and stained with the Z-disc marker anti-α-actinin ( red ) and phalloidin ( blue , gray ) . The fhos null IFMs display a severely defective organization of myofibrils and sarcomeres . ( G–J ) TEM views of longitudinal ( G , I ) or cross ( H , J ) sectioned material from IFMs dissected from control ( G , H ) or fhos null ( I , J ) pupae at 50 hr APF . Control myofibrils run parallel to each other ( G ) and contain regularly-spaced Z-discs ( red arrowheads in G ) and nascent filament arrays , displaying an ordered , hexagonal lattice organization ( H , dashed circles and inset ( thick [orange] and thin filaments [blue] ) . Only sporadic myofibrils bearing short , weakly-organized filament arrays are found within the fhos null muscle fibers ( I , J , dashed squares and insets ) . Scale bars: 50 μm ( A–D ) , 5 μm ( E–F’ ) , 500 nm ( G and J ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 00610 . 7554/eLife . 16540 . 007Figure 2—figure supplement 2 . Fhos is required for proper sizing of thin-filament arrays . ( A–B’ ) Confocal images of IFMs dissected from young adult flies and stained with anti-α-actinin ( red ) to mark Z-disc structures and phalloidin ( blue , gray ) to visualize microfilaments . ( A–A’ ) act88F-GAL4 control . ( B–B’ ) act88F-GAL4>UAS-fhos RNAi ( knockdown of fhos initiating at mid-pupal stages ) . Z and M mark the Z-disc and M-line of a single sarcomere . Vertical bars ( A , B ) mark the width of a single sarcomere , while horizontal bars ( A’ , B’ ) mark the length of a single thin-filament array . ( C–D ) Quantification of the average width and length of thin-filament arrays in control ( act88F-GAL4 ) and fhos knockdown IFMs dissected from young ( 1 day old ) flies . The data represent the measurements of 50 sarcomeres each from 7 flies ( n = 350 ) for each genotype . Following fhos knockdown arrays are significantly thinner ( ~15% , p<0 . 0001 ) ( C ) and shorter ( ~30% , p<0 . 0001 ) P values determined by Mann-Whitney test for width and length measurements . Scale bars in all panels correspond to 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 007 We sought to complement and enhance the analysis of Fhos-knockdown IFMs by studying mutant alleles in the Fhos locus . A recent study ( Lammel et al . , 2014 ) described several such alleles , including FhosΔ1 , a small deficiency that completely removes the coding regions of most Fhos isoforms , and thus represents a severe , possibly null , gene disruption . FhosΔ1 homozygotes die as pharate adults , allowing to assess the effects of Fhos gene knockout on IFM development . Immuno-fluorescent staining with informative markers revealed that in FhosΔ1 hemizygous flies that reach the pharate adult stage , DLMs are highly disorganized , lacking even the trace appearance of sarcomeric units observed in Fhos knockdown DLMs ( Figure 2C–C” ) . We extended this study by subjecting the mutant DLMs to transmission electron microscopy ( TEM ) analysis . Longitudinal TEM sections underscored the disorganized nature of the Fhos mutant DLMs , which appear to be composed of irregularly shaped myofibrils , lacking a defined spatial orientation ( Figure 2D , E ) . While myofilament arrays can be found within these structures , they fail to exhibit any of the features of regularly spaced sarcomeric units characteristic of wildtype DLMs ( Figure 2D ) , and display only a few sporadic electron-dense spots that may represent rudimentary Z bands ( Figure 2E ) . In contrast to the highly-ordered hexagonal lattice of thick and thin filaments within wildtype myofibrils , revealed by TEM cross-sectional views ( Figure 2F , F’ ) , the myofilament arrays in Fhos mutant DLMs are small , irregularly-spaced , and lack a defined spatial organization ( Figure 2G , G’ ) . Highly defective myofibril and sarcomere organizations were already clearly apparent in FhosΔ1 hemizygotes at 50 hrs APF via both light microscopy ( Figure 2—figure supplement 1E–F’ ) and TEM Figure 2—figure supplement 1G–J ) analyses , demonstrating that the mutant phenotypes are a consequence of developmental abnormalities initiating at the onset of IFM sarcomere formation , rather than deterioration of normally formed structures . The severe phenotypes of Fhos knockdown and null mutant pupae demonstrate an essential role for Fhos in the assembly and organization of sarcomeric units within IFMs . The nearly complete lack of sarcomeric organization within myofibrils in the absence of Fhos activity implies a critical requirement for Fhos already at early stages of sarcomere assembly . Such early arrest in sarcomerogenesis may mask potential requirements for Fhos at later stages of the process . To address this issue , we induced RNAi directed at Fhos using the IFM-specific act88F-Gal4 driver ( Gajewski and Schulz , 2010 ) , thereby delaying onset of Fhos knockdown to a more advanced phase of IFM development . While such flies emerged from the pupal case , they were flightless . A more detailed examination following isolation of IFMs revealed the establishment of an ordered array of intact , regularly-spaced sarcomeric units ( Figure 2—figure supplement 2A–B’ ) . However , these Fhos knockdown sarcomeres exhibited significantly shorter widths and lengths than sarcomeres from age-matched controls ( Figure 2—figure supplement 2C , D ) , implying a requirement for Fhos in the elongation and peripheral thickening mechanisms underlying thin-filament array maturation . Taken together , the range of phenotypic abnormalities associated with the various forms of disruption to Fhos function suggest that Fhos is a major mediator of sarcomere formation , contributing throughout pupariation to different aspects of thin-filament array assembly and maturation . The Fhos locus is composed of two classes of transcripts , utilizing different promoters ( Figure 3A ) . Both transcript classes share a set of 3’ exons , but differ in their 5’ regions , which include distinct non-coding and coding exons . As a result , the locus generates two main Fhos protein isoforms , in which a conventional FHOD-family formin , containing all of the canonical formin regulatory and actin-related functional domains , is appended to different N-terminal segments ( Figure 3B ) . The isoform encoded by transcripts RA-RG features a short ( 63 residue ) N-terminal domain , while transcripts RH-RJ encode an isoform bearing distinct and substantially larger N-terminal region that is conserved among Neopteran winged-insects ( Bechtold et al . , 2014 ) . 10 . 7554/eLife . 16540 . 008Figure 3 . Localization of Fhos at the Z-disc is essential for its function . ( A ) Map of the Fhos genomic locus and transcripts ( after Flybase , ( Attrill et al . , 2016 ) . Shown are the nine known fhos transcripts ( designated Fhos-RA--Fhos-RI ) , which are divided into two groups ( RA-RG and RH-RI ) . The two groups share a nearly identical set of 3’ exons ( red dashed rectangle ) , which encode a conventional FHOD-family formin , but are expressed via distinct regulatory regions , and possess distinct sets of 5’ exons , including 5’ coding exons ( blue and purple dashed rectangles ) that encode different N-terminal domains . The two transcript variants , RH and RA , used to generate , respectively , the long and short transgenic UAS-Fhos constructs are indicated by orange arrows . The insertion positions of three MiMIC elements , MI04231 ( inserted downstream of long isoform initiation sites ) , MI01421 ( inserted downstream of all transcript initiation sites ) , and MI09324 ( used to produce the Fhos-GFP 'protein trap' ) are indicated by inverted triangles . Positions of two dsRNA target sequences used , one common to all fhos isoforms ( red bar ) and the other specific to the long forms ( blue bar ) are shown above the transcript map . The CRISPR/Cas9-generated deletion of the guanine residue at position 99 of the short isoform transcript and its adjacent sequence are indicated . ( B ) Schematic representation of three representative Fhos protein isoforms . The canonical formin domains common to all forms are colored red , while the alternative N-terminal domains are in blue ( long forms ) and purple ( short form ) . Canonical domains indicated include the GTPase binding domain ( GBD ) , formin homology ( FH ) domains 1/2 and 3 , and the diaphanous autoregulatory domain ( DAD ) . The positions of the I966A point mutation in the FH2 domain and the premature stop codon , generate by the frameshift mutation ΔG99 in the Fhos-PA N-terminal domain are indicated . ( C–D’ ) Zasp ( red ) and phalloidin ( blue and gray ) stainings demonstrate the severe , null-like disruption of myofibril and sarcomere microfilament organization in hemizygous FhosMI01421/Df ( 3L ) BSC612 pharate adult flies ( C , C’ ) , similar to that observed in fhosΔ1 hemizygotes . No rescue is observed following expression of UAS-GFP-Fhos-PA ( green ) driven by arm-Gal4 in this background ( D , D’ ) . ( E–F ) α-actinin ( red ) and phalloidin ( blue and gray ) stainings demonstrate the severe , null-like phenotypes following specific RNAi mediated knockdown of the Fhos long-isoforms ( E , E’ ) and in hemizygous FhosMI04231/Df ( 3L ) BSC612 ( F ) pharate adult flies . ( G ) Zasp ( red ) and phalloidin ( blue ) stainings demonstrate normal myofibril and sarcomeric structure of Fhos ΔG99/Df ( 3L ) BSC612 hemizygotes , in which the short Fhos isoforms are not expressed . ( H–L’’ ) Fhos localization in myofibrils , as monitored at two distinct pupal developmental time points , 45 hr APF ( H–I’ ) , and 65 hr APF ( J–J’’ ) , via a GFP 'exon trap' engineered at the insertion site of the MiMIC transposon MI09324 ( green triangle in A ) . The GFP-tagged Fhos proteins ( all isoforms ) generated in this manner are visualized with anti-GFP ( green or gray ) , Z-discs are visualized with anti-α-actinin or anti-Zasp ( red ) , thin filament pointed ends visualized by anti-Tmod ( blue ) and microfilaments with phalloidin ( blue ) . The diffuse/punctate initial localization of Fhos-GFP overlying broad portions of the growing myofibrils ( H ) , in some cases shows an adjacent localization to the nascent Z-disc ( I white arrowhead ) or to array pointed ends ( I’ red arrowhead ) . The initial punctate localization gives way to a striated pattern restricted to the vicinities of both the barbed ( Z ) and pointed ( M ) ends of the thin-filament arrays ( J–J’’ ) . ( K–K” ) Localization of the short isoform of Fhos in IFMs from a young adult fly , visualized by expression of UAS-GFP-Fhos-PA using the mef2-GAL4 driver ( anti-GFP , green or gray ) . Z-discs are visualized with anti-α-actinin ( red ) , and microfilaments with phalloidin ( blue ) . GFP-Fhos-PA localizes to the vicinity of the pointed ends of the arrays ( M ) . ( L–L” ) Localization of the long isoforms of Fhos in IFMs from a young adult fly , visualized with anti-HA ( green and gray ) , following expression of UAS-HA-Fhos-PH using the mef2-GAL4 driver . Fhos-PH-PA localizes to the vicinity of the barbed ends of the arrays ( Z ) , where it overlaps with the general Fhos distribution to both the barbed and pointed ends ( M ) of the arrays ( visualized with anti-Fhos [red] ) . Microfilaments visualized with phalloidin ( blue ) . Scale bars in all panels correspond to 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 00810 . 7554/eLife . 16540 . 009Figure 3—figure supplement 1 . Fhos localization at the Z-disc is essential for its function . ( A–A’’’ ) Fhos localization in IFM myofibrils from young adult flies . Shown are myofibrils from a fly in which GFP was inserted into the Fhos locus ( see text and Figure 3A ) . Fhos localization was monitored by anti-GFP ( green or gray ) and anti-Fhos ( red or gray ) and microfilaments are visualized with phalloidin ( blue ) . Both methods demonstrate enrichment of Fhos in the vicinities of the pointed ( M ) and barbed ( Z ) ends of the thin-filament arrays . ( B–B’’’ ) Sporadic rescue of the disrupted myofibril and sarcomere organization of null hemizygous FhosMI01421/Df ( 3L ) BSC612 young ( 1 day old ) flies following mef2-Gal4 driven expression of UAS-GFP-Fhos-PA ( green ) . Z-discs are marked by anti-Zasp52 ( red ) and microfilaments are visualized with phalloidin ( blue or gray ) . Formation of sarcomeres is always accompanied by localization of GFP-Fhos-PA not only to the pointed ( M ) but also to the barbed ( Z ) ends of the thin-filament arrays . ( C–C’’’ ) Partial rescue of null hemizygous FhosMI01421/Df ( 3L ) BSC612 young ( 1 day old ) flies following mef2-Gal4 driven expression of UAS-3XHA-Fhos-PH ( green ) , a long isoform that localizes strictly to the Z-disc region . Z-discs are marked by anti-Zasp52 ( red ) and microfilaments are visualized with phalloidin ( blue or gray ) . ( D ) The properly ordered arrays from a hemizygous Fhos ΔG99/Df ( 3L ) BSC612 fly- expressing only the long isoforms- show Fhos localization ( visualized with anti-Fhos , green ) near both the pointed ( M ) and barbed ( Z ) ends of the thin-filament arrays; microfilaments are visualized with phalloidin ( blue ) . The dashed rectangle in ( A , C , D ) corresponds to the magnified inset . Scale bars in all panels correspond to 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 009 The short Fhos variant , represented by form RA , has been shown to rescue Fhos null mutant flies to adult viability , and to restore normal function in affected tissues ( e . g . macrophage motility and wing inflation ) , when expressed ubiquitously via armadillo-GAL4 ( Lammel et al . , 2014 ) . We were therefore surprised to discover that the flight muscles of such EGFP-Fhos-PA-rescued flies continued to exhibit severe Fhos mutant phenotypes ( Figure 3C–D’ ) , implying that the short form of Fhos , which mediates most Fhos developmental functions , is insufficient in this context . These observations raised the possibility that the larger isoforms , which have not been extensively studied , provide Fhos activities necessary for IFM sarcomere formation . To address this issue directly , we utilized a transgenic RNAi construct specifically targeting the large 5’ coding exon ( Figure 3A ) . Expression of this RNAi construct in muscle cells , which should eliminate only the long isoforms in this tissue , led to a strong disruption of sarcomere organization , closely resembling null mutations ( Figure 3E , E’; see also [Schnorrer et al . , 2010] ) . Furthermore , we observed similar deleterious effects on the IFM organization in flies hemizygous for MI04231 ( Figure 3F ) , a MiMIC insertion allele that is predicted to specifically disrupt the long Fhos isoforms and leave the short isoforms intact ( Figure 3A ) . These results lead us to conclude that the long N-terminal protein domain is critical for IFM function of Fhos . To determine whether the long Fhos isoforms are sufficient , we used a CRISPR/Cas9 approach to generate small deletions in the exon encoding the N-terminal region of the short Fhos isoforms . One of these , FhosΔG99 , results in a single nucleotide deletion , generating a translational frameshift and a predicted termination of translation of the short Fhos isoform after only 58 residues ( Figure 3A , B ) . FhosΔG99 hemizygous flies , which do not express functional short Fhos isoforms , are fully viable and fertile , and do not exhibit any obvious morphological defects . Importantly , the IFMs of these flies display normal myofibril and sarcomere organization ( Figure 3G ) . These observations imply that the long Fhos isoforms are sufficient for proper IFM myogenesis , and furthermore , that they can provide most , if not all , functional requirements for Fhos . Identifiable protein domains are not found within the 1198 residue long protein sequence encoded by the large 5’ exon , which raised the possibility that this extension to the canonical FHOD-like formin provides a localization cue , rather than an additional functional moiety . To pursue this notion , we set out to determine the localization patterns of the different Fhos isoforms . We first made use of a MiMIC transposable element insertion in the Fhos gene locus ( MI09324 ) and the RMCE technique ( Venken et al . , 2011 ) to generate a GFP 'protein trap' ( Fhos-GFP ) , so that all isoforms of endogenous Fhos would also harbor a GFP tag ( Figure 3A ) . Monitoring the Fhos-GFP signal at 45 hr APF revealed an initial diffuse localization to myofibrils , with some enrichment over nascent sarcomeric units ( Figure 3H–I’ ) . At later stages ( 65 hr APF ) , the localization of Fhos refines to discrete stripes overlying the barbed and pointed ends of the microfilament arrays within the sarcomere ( Figure 3J–J’’ ) , a pattern that persists throughout pupal stages and is still observed in young adult flies ( Figure 3—figure supplement 1A–A”’ ) . Staining with an antibody we raised to the Fhos C-terminal domains shared by all isoforms of the protein , displayed a similar pattern ( Figure 3—figure supplement 1A–A”’ ) . We next examined the localization patterns of representative tagged versions of the short and long isoforms of Fhos , following their expression in IFMs . EGFP-Fhos-PA , representing the shorter isoform , was found to localize exclusively to the vicinity of sarcomere M-lines , corresponding to the pointed-ends of the thin-filament arrays ( Figure 3K–K” ) . This observation raised the possibility that the short isoform is not functional on its own in IFMs due to its restricted localization pattern , which does not include the Z-disc associated barbed-ends of the thin-filament arrays . Notably , instances of ectopic localization of EGFP-Fhos-PA to Z-discs , which were sporadically observed when this construct was over-expressed in a null Fhos mutant background , were associated with markedly improved organization of the affected sarcomeres ( Figure 3—figure supplement 1B–B”’ ) . This observation supports the notion that localization of Fhos to the Z-disc region is critical for its function in the growing sarcomere . To monitor localization of the long Fhos isoform , we generated an HA-tagged version of Fhos-PH , which contains 410 residues of the novel N-terminal domain ( Figure 3B ) . Remarkably , HA-Fhos-PH was found to localize exclusively to the Z-disc region of IFM sarcomeres following expression via mef2-GAL4 ( Figure 3L–L” ) , in complementary fashion to EGFP-Fhos-PA . Taken together with the genetic analysis , which identified the long isoform as the functional Fhos variant in IFMs , we conclude that localization of Fhos to the Z-disc/array barbed-end region , mediated by the long , novel N-terminal domain , is critical for its sarcomeric function . Interestingly , despite the localization of the HA-Fhos-PH construct to the Z-disc region , it only partially rescued the sarcomere organization defects of Fhos null flies ( Figure 3—figure supplement 1C–C”’ ) . In addition , we observe that the long Fhos isoforms , which provide full sarcomeric functionality , localize to both the Z-disc and M-line regions in FhosΔG99 hemizygotes ( Figure 3—figure supplement 1D ) . Thus , while Z-disc localization of Fhos is an essential requirement for proper sarcomere assembly , we cannot rule out that Fhos performs functional roles at additional sites within maturing IFM sarcomeres . Having established that Fhos is a major contributor to IFM sarcomere organization , we now sought to elucidate its specific roles , by monitoring actin monomer incorporation patterns in the absence of Fhos function . Towards this end , we examined IFMs from Fhos knockdown pupae , following a restricted expression of GFP-actin88F during early ( 0–30 hr APF ) , interim ( 30–60 hr APF ) and late ( 60–90 hr APF ) phases of pupal development ( Figure 4A ) . During the initial stages of IFM development , incorporation of GFP-actin88F into unstructured microfilament arrays within nascent myofibrils proceeded normally in Fhos knockdown IFMs ( Figure 4B , C ) . This finding implies that Fhos is not essential for the initial 'burst' of strong polymerization activity characteristic of this phase ( Figure 1C–C” ) , and is consistent with the establishment of properly sized but internally disorganized DLM myofibers in Fhos knockdown and mutant flies ( Figure 2 , Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 16540 . 010Figure 4 . Fhos is required for the 'patchy' actin monomer incorporation and radial expansion aspects of thin-filament array growth . ( A ) Scheme of IFM development intervals used for temporally restricted expression of GFP-actin88F ( B–J ) or GFP-actin5C ( K–M’ ) in wildtype ( B–B” , D , D’ , H , H’ , K , K’ ) or Fhos knockdown ( C–C” , F , F’ , I , I’ , L , L’ ) IFMs . ( B-I’ ) GFP-actin88F ( green , gray ) expression between 0–30 ( B–C’ ) , 30–60 ( D-G ) and 50–90 ( H-I’ ) hr APF . Z-discs are visualized with anti-Zasp52 ( red ) and microfilaments with phalloidin ( blue , gray ) . ( B–C’ ) The general and uniform incorporation of monomers into microfilaments characteristic of the initial phase of sarcomere formation ( B–B” ) is not affected in fhos knockdown myofibrils ( C–C” ) . ( D–G ) The dispersed and 'spotty' wildtype incorporation pattern during the interim ( 30–60 hr APF ) phase ( D ) is replaced by a pointed-end centered pattern in fhos knockdown myofibrils ( F ) . These pattern distinctions are further demonstrated by heat maps of the GFP-actin88F distribution ( D’ , F’ ) and quantification of GFP intensity ( E , G ) derived from 10 μm profiles covering approximately four sarcomeric units ( white lines in D and F; data were acquired for 15 profiles from 7 different pupae for each genotype [n = 105] ) . ( H–I’ ) The incorporation 'frames'normally generated by late GFP-actin88F expression pulses ( H , H’ ) lack peripheral incorporation ( arrowheads ) following fhos knockdown ( I , I’ ) , but these abnormally thin myofibrils retain proper incorporation at array 'pointed' ends ( red asterisks ) . The lack of incorporation 'frames' is also demonstrated by heat maps ( H’ , I’ ) , and the quantification of GFP intensity ( J; data were acquired for 100 profiles from 4 different flies for each genotype [n = 400] ) along vertical profiles ( white dashed line H’ , I’ ) , which show the loss of peripheral incorporation and thinner sarcomeres ( red dashed lines in J ) . ( K-M ) GFP-actin5C ( green , gray ) expression between 50–90 hr APF . Z-discs are visualized with anti-Zasp52 ( red ) and microfilaments with phalloidin ( blue ) . GFP-actin5C induction in parallel to fhos knockdown showed normal Z-disc associated turnover ( arrowheads ) . ( M ) Quantification of Z-disc incorporation band width ( data were acquired for 50 Z-discs from 4 different flies for each genotype [n = 200] ) . Scale bars in all panels correspond to 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 010 In contrast to these observations , a marked effect of Fhos knockdown on the actin monomer incorporation pattern can be discerned during the interim ( 30–60 hr APF ) period of pupal development ( Figure 4D–G ) . While wildtype IFMs display an irregular 'patchy' pattern of incorporation spread out over the nascent sarcomeric arrays ( Figure 4D–E ) , Fhos knockdown IFMs exhibited a repetitive , undulating pattern , with peaks of incorporation centered at the pointed-ends of the arrays ( Figure 4F–G ) . Fhos therefore appears to mediate monomer incorporation into filament patches , but is not involved in the emerging process of array elongation from pointed ends . This feature of Fhos knockdown IFMs persists during the final phase of pupal development , when wildtype sarcomeres display a frame-like pattern of incorporation ( Figure 4H–J ) . While such 50–90 hr old Fhos knockdown IFMs retain pointed-end incorporation , they appear to be thinner and lack the circumferential accumulation of incorporated GFP-actin88F , which represents radial growth of the arrays through addition of peripheral microfilaments ( Figure 4I , J ) , implying a requirement for Fhos in the array 'thickening' process . Finally , a GFP-actin5C incorporation band of normal width was readily detected adjacent to sarcomere Z-discs in Fhos knockdown IFMs ( Figure 4K–M ) , implying that actin monomer exchange at barbed ends of the arrays is Fhos independent . Analysis of actin monomer incorporation patterns thus provides a higher resolution and reveals specific roles for Fhos in mediating thin-filament array assembly and growth . These include the organization of microfilaments into nascent structures , shaping sarcomeres into uniformly sized and regularly-spaced units and , finally , radial growth of the arrays via peripheral thickening . On the other hand , the synthesis of the initial pool of microfilaments , array elongation from pointed ends and actin exchange at the Z-disc do not require Fhos , and thus rely on the activity of other actin regulators . As the incorporation data suggests that filament extension from the pointed-ends of the arrays does not require Fhos , we sought to identify alternative elements that may mediate this key aspect of sarcomere growth . Such factors are unlikely to include formin family members , since formins are primarily thought to extend actin filaments from their barbed ends ( Campellone and Welch , 2010; Goode and Eck , 2007 ) . The WH2-domain protein Sarcomere length short ( Sals ) is an attractive candidate , since it was shown to contribute to pointed-end filament elongation in Drosophila larval muscles , and to be localized to the pointed ends of thin-filament arrays in IFM sarcomeres ( Bai et al . , 2007 ) . We first assessed the role of Sals in IFM sarcomerogenesis by using the mef2-GAL4 driver and the GAL80ts/TARGET system to express an RNAi construct targeted against sals from the onset of pupariation , and examined the effect on IFMs of pharate adult flies . The overall length of the sals-knockdown thin-filament arrays is significantly shorter than control arrays , and their pointed end borders are abnormally shaped and discontinuous ( Figure 5A–C’ , Figure 5—figure supplement 1D ) , consistent with a pointed-end extension function for Sals . 10 . 7554/eLife . 16540 . 011Figure 5 . Sals is required for 'pointed-end' thin-filament growth . ( A ) Scheme of IFM development intervals used for temporally restricted expression of sals RNAi and GFP-actin88F . ( B–C’ ) IFMs dissected from control ( mef2-GAL4 ) 1 day old flies ( B , B’ ) and sals knockdown pharate adults ( C , C’ ) , in which RNAi expression was initiated at 0 hr APF . Z-discs are visualized with anti-Zasp52 ( red ) and microfilaments with phalloidin ( blue or gray ) . sals knockdown results in sarcomere shortening and 'pointed' end abnormalities ( insets in B’ and C’; for quantification see Figure 5—figure supplement 1D ) . ( D–E’ ) IFMs dissected from young ( 1–2 day old ) flies in which GFP-actin88F expression ( anti-GFP , green , gray ) was initiated at 50 hr APF on its own ( D , D’ ) or together with sals RNAi ( E , E’ ) . M lines are visualized with anti-Obscurin ( red ) and microfilaments with phalloidin ( blue ) . ( F ) The GFP-actin88F incorporation band at the 'pointed' ends is significantly decreased following sals knockdown , as shown by the M/Z intensity ratio ( p<0 . 0001 , P values determined by Mann-Whitney test ) , while the addition of peripheral microfilaments is unaffected . ( G ) Quantification of phalloidin intensities derived from 13 μm profiles ( white line in D ) . sals RNAi myofibrils ( red line ) exhibit shorter sarcomeric units compared to control ( blue line ) . Scale bars in all panels correspond to 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 01110 . 7554/eLife . 16540 . 012Figure 5—figure supplement 1 . Involvement of Tmod in nascent thin filament array elongation . ( A ) Scheme of IFM developmental intervals used for temporally restricted expression of tmod RNAi and GFP-actin88F . ( B–C’ ) IFMs dissected from control ( mef2-GAL4 ) ( B , B’ ) and tmod knockdown ( C , C’ ) 1 day old flies , in which RNAi expression was initiated at 0 hr APF . Anti-Tmod ( red ) visualized array pointed ends , Z-discs are visualized with anti-Zasp52 ( green ) and microfilaments with phalloidin ( blue or gray ) . tmod knockdown results in mild significant sarcomere shortening ( ~10% , p<0 . 01 ) , which is , however , considerably milder than shortening following sals knockdown ( ~40% , p<0 . 0001 ) . The data for sals and tmod knockdown represent measurements of 50 sarcomeres each from 5 flies ( n = 250 ) . The P values determined by one-way ANOVA followed by Dunnett’s multiple comparison test for array length measurements . ( E , E’ ) IFMs dissected from young ( 1–2 day old ) flies in which GFP-actin88F expression ( anti-GFP , green , gray ) was initiated at 50 hr APF together with tmod RNAi . Pointed ends are visualized with anti-Tmod ( red ) and microfilaments with phalloidin ( blue ) . The GFP-actin88F incorporation pattern remains intact following tmod knockdown: ( G ) The M/Z ratio of incorporation band intensity is not significantly changed ( p>0 . 01 , P values determined by Mann-Whitney test ) and quantification of phalloidin intensity ( H ) derived from 13 μm profiles ( white line in E ) , indicates only slight shortening of the arrays . ( F ) Interestingly , IFMs dissected from a 90 hr APF pupa exhibit a complete absence of Tmod staining ( red ) at the vicinity of pointed ends . Scale bars in all panels correspond to 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 012 We next utilized the GFP-actin88F incorporation assay to further examine Sals activity in developing IFMs . Specifically , sals function was disrupted by knockdown beginning at 50 hr APF , when 'core' thin-filament arrays have already formed , monomer incorporation was monitored in young adults . The resulting incorporation pattern is a near 'mirror-image' of Fhos knockdown during this period: shortened sarcomeres displaying normal peripheral thickening of the core arrays ( Figure 5D–E’ , G ) , coupled with a significant decrease of actin incorporation at their pointed ends ( Figure 5F ) . The conserved pointed-end capping protein Tropomodulin ( Tmod ) ( Gregorio and Fowler , 1996 ) is a second factor that could contribute to filament pointed-end elongation . Induction of RNAi targeting tmod at 0 hr APF indeed results in a significant shortening of IFM thin-filament arrays , but to a considerably lesser extent than the shortening observed following knockdown of sals ( Figure 5—figure supplement 1A–D ) . Furthermore , tmod knockdown during the second half of pupal development only weakly affects the array length and pointed-end incorporation of actin monomers ( Figure 5—figure supplement 1E–H ) . Taken together , these results identify Sals as a major actin regulator , specifically mediating pointed-end growth of thin-filament arrays throughout IFM maturation , while the contribution of Tmod appears to be restricted to early stages of IFM development . Formins employ a variety of molecular mechanisms for regulating microfilament organization and dynamics , including microfilament nucleation , elongation , bundling and capping activities ( Goode and Eck , 2007; Harris and Higgs , 2006; Schonichen and Geyer , 2010 ) . To begin to address this issue in the context of Fhos IFM function , we used CRISPR/Cas9 technology to insert a point mutation into the endogenous Fhos locus , thereby generating a single amino acid substitution ( I966A according to residue numbering of the short form of Fhos , Figure 3B ) . Mutating this highly conserved residue in a variety of formins , including the mammalian Fhos homolog FHOD3 , consistently abolished activities requiring barbed-end association ( actin nucleation , elongation and capping ) ( Harris et al . , 2006; Taniguchi et al . , 2009; Xu et al . , 2004 ) . FhosI966A hemizygous flies are viable and exhibit an externally normal morphology , implying that the nucleation activity of Fhos is generally dispensable . However , the IFMs of these flies contain abnormally thin myofibrils . Most of these myofibrils display , nevertheless , an organized structure of repeated sarcomeric units , with clear demarcation of the Z-discs ( Figure 6A–A” ) , suggesting an arrest in sarcomere growth following proper initial assembly and organization . This notion was further borne out following TEM-level visualization , which showed that the thin FhosI966A myofibrils house properly structured and evenly-spaced sarcomeres ( Figure 6B ) , harboring well-ordered lattices of thick and thin filaments ( Figure 6C ) . 10 . 7554/eLife . 16540 . 013Figure 6 . Early Fhos function does not require barbed-end activities . ( A–A’’ ) IFM myofibrils from a FhosI966A/Df ( 3L ) BSC612 1 day old adult fly . Z-discs are marked by anti-α-actinin ( red or gray ) and microfilaments are visualized with phalloidin ( blue or gray ) . The thin myofibrils display organized arrays of repeated sarcomeric units ( red outlines in A’’ ) . ( B–F ) TEM analysis of IFM myofibrils from FhosI966A/Df ( 3L ) BSC612 flies . ( B , C ) One day old adult flies . A longitudinal section ( B ) shows a stereotypic sarcomeric unit displaying clear Z-disc ( Z ) and M line ( M ) structures . A cross section ( C ) shows an individual sarcomeric unit ( red dashed circle ) harboring a well-formed lattice of thick and thin filaments . Arrows point to accumulations of nearby filaments , which could serve as a source for radial growth , but have not been recruited . The degree of lattice organization within and outside the sarcomere can be appreciated from the spatial arrangement of representative thick ( orange ) and thin ( blue ) filaments . ( D ) Quantification of sarcomere size in wildtype and FhosI966A/Df ( 3L ) BSC612 pupae and adult flies at the indicated ages , based on the number of thick filament units in TEM cross-sections ( n = 20 for each background ) . ( E , F ) 50 hr APF pupae . Longitudinal ( E ) and cross ( F ) sections show myofibril individualization ( red dashed lines in E ) and formation of nascent sarcomeric units ( red dashed circles in F ) with defined Z-disc borders ( red arrowheads in E; see also Figure 2—figure supplement 1G , H ) . Scale bars correspond to 5 μm ( A ) , 100 nm ( B ) , 500 nm ( C , F ) and 200 nm ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 013 The actual radial size of FhosI966A sarcomeres was assessed by determining the number of thick filament units in myofibril TEM cross sections . This analysis revealed that the sarcomeres of 1 day old Fhos-I966A mutants are ~ 8 times smaller than wildtype sarcomeres from similarly aged flies , and correspond in size to normal sarcomeres between 50–60 hr APF ( Figure 6D ) . This observation is in keeping with the notion of an arrest in sarcomere growth during intermediate pupal stages . Furthermore , the appearance and organization of sarcomeres within myofibrils of FhosI966A hemizygotes at 50 hr APF ( Figure 6E , F ) closely matches those of sarcomeres from age-matched wildtype pupae ( Figure 2—figure supplement 1G , H ) . These light and electron microscopy analyses suggest therefore that the normal involvement of Fhos at the initial stages of IFM sarcomere assembly does not require a barbed end-associated activity , consistent with the early diffuse localization pattern of Fhos in myofibrils . However , the barbed-end localization and associated activities of Fhos appear to be essential for the maturation of nascent arrays , implying a multi-faceted involvement of this formin in sarcomere formation .
Formation of the adult Drosophila IFMs during pupariation provides an established model system to study formation of skeletal muscles , and in particular the generation of the repeated sarcomere structure , the core functional unit underlying muscle contractility . The IFM system possesses several key features that make it amenable for detailed analysis . These include an extended developmental time window of ~60 hr , the availability of genetic methods for investigation ( e . g . , RNAi-based knockdown ) and the highly ordered and repetitive organization of sarcomeric units within IFM myofibrils , which allows for the detection and analysis of both major and subtle alterations and defects . Our study focused on the actin based thin-filament array component of sarcomeres . Towards this end , we utilized an additional , highly useful feature of the IFM system: the capacity to monitor the incorporation pattern of new actin monomers at discrete stages of the process , using transgenic GFP-actin constructs whose expression can be readily manipulated . The analyses performed using the various approaches and tools available for the study of the IFMs allow us to chart the principles , timeline and molecular basis for assembly , organization and maturation of thin-filament arrays within this model of sarcomerogenesis . While it is likely that regulators of linear actin belonging to the formin family play a central role in the formation and maintenance of actin filament arrays in the sarcomere , redundancy among them appears to be commonplace in this context ( Mi-Mi et al . , 2012; Rosado et al . , 2014 ) , complicating the elucidation of distinct functional roles . Our detection of severe sarcomeric phenotypes following disruption of Fhos activity on its own , identifies Fhos as a key contributor to the processes governing IFM thin-filament array assembly , and sets the stage for studying the involvement of formins in this major aspect of microfilament organization , via utilization of genetic approaches . FHOD-family formins have been previously associated with sarcomere organization in cardiac muscle ( Iskratsch et al . , 2010; Kan et al . , 2012; Taniguchi et al . , 2009; Wooten et al . , 2013 ) , but the roles these proteins play during skeletal myogenesis have been difficult to ascertain . The ability to dissect the program of IFM sarcomere formation and to disrupt the activity of Fhos to different extents and at distinct phases , provides a comprehensive view of the role of this formin-family protein in the process , and of skeletal muscle thin-filament array assembly at large ( Figure 7 ) . 10 . 7554/eLife . 16540 . 014Figure 7 . Model of IFM thin-filament array assembly and the roles of Fhos and Sals . Four distinct stages of thin-filament array assembly and maturation are represented in correlation to pupal developmental stages . ( A ) Extensive filament polymerization ( 0–30 hr APF ) , which takes place in a Fhos-independent manner . ( B ) Organization of nascent sarcomeres and patched actin incorporation ( 30–45 hr APF ) . Fhos localizes to the nascent arrays , and is required for their organization into discrete structural units . ( C , D ) Growth and maturation of nascent sarcomeres ( 45 hr APF- eclosion ) . Fhos localizes to the vicinity of the Z-discs , and is essential for radial growth of the thin-filament arrays , possibly via peripheral filament recruitment . Two additional actin incorporation modes are executed in a Fhos independent manner . Z-disc associated monomer turnover , and 'pointed' end elongation , which is mediated primarily by Sals . DOI: http://dx . doi . org/10 . 7554/eLife . 16540 . 014 1 . The onset of pupal development is accompanied by a prominent burst of actin polymerization , generating a store of microfilaments for use during subsequent stages , when polymerization activity declines considerably . The identity of actin nucleators and in particular formins involved in the extensive initial polymerization is not known , and it is certainly possible that several formins act redundantly , given our failure to disrupt this process by single formin knockdowns , including that of Fhos . An initial sign of internal myofiber organization is seen in the segregation of the abundant microfilaments produced during the early phase of pupal development , into elongated myofibrils that align in parallel to the fiber . We do not know the nature of the signal governing this alignment , but it is noteworthy that previous studies of the dynamic organization of sarcomeres in cultured muscle cells have indicated that microtubules which are aligned along the fiber may provide an initial cue for the recruitment and orientation of myosin heavy chain and possibly microfilaments as well ( Pizon et al . , 2005 ) . 2 . The initial , widespread polymerization gives way to a more limited mode of incorporation , characterized by 'patches' of actin monomers that are added to a nascent microfilament array , suggesting that this period is devoted to proper structuring of the arrays within individual , uniformly-sized and regularly separated sarcomeric units . It is during this interim period that disruption of Fhos activity first leads to alteration of the monomer incorporation pattern , in what we interpret to be a telling fashion , as it retains an underlying , highly regular pattern of monomer incorporation at the pointed-ends of the thin filament arrays ( Figure 4D–G ) . This observation constitutes a direct demonstration that IFM arrays elongate from their pointed-ends , contrary to the conventional barbed end-biased growth of microfilaments , and in keeping with previous findings based on studies of the pointed-end capping protein Tropomodulin in both IFMs and vertebrate cardiomyocytes ( Littlefield et al . , 2001; Mardahl-Dumesnil and Fowler , 2001 ) . Furthermore , our investigation identifies the WH2-domain protein Sals as a major mediator of the pointed-end growth of IFM thin-filament arrays , similar to its function in Drosophila larval muscle sarcomeres ( Bai et al . , 2007 ) . Interestingly , Leiomodin acts to mediate pointed-end elongation in vertebrate cardiomyocytes ( Chereau et al . , 2008; Tsukada et al . , 2010 ) , implying a conserved function for WH2-domain actin regulators in sarcomerogenesis . Interfering with Fhos activity results in severe impairment of myofibril and sarcomere organization ( Figure 2 ) , raising the question of how to reconcile the strong mutant phenotypes with the limited degree of Fhos-dependent actin incorporation and array growth during the early and interim periods of pupal development . We suggest that , rather than participating directly in the process of actin polymerization , Fhos plays a key organizational role during this period , which leads to the initial structuring of thin-filament arrays , using microfilaments generated by other formins and actin regulators . FHOD-family formins are considered to be poor microfilament nucleators ( Schonichen et al . , 2013; Taniguchi et al . , 2009 ) , and are thought to act via alternative modes- such as microfilament bundling ( Kutscheidt et al . , 2014; Schonichen et al . , 2013 ) - to provide shape and structure to microfilament arrays . Our analysis is consistent with this notion , as we have demonstrated that FhosI966A , a Fhos variant presumably lacking barbed end associated activities , supports formation of small but properly organized sarcomeres , similar to the sarcomeres normally formed during the early and interim stages of pupal development . We suggest therefore , that Fhos plays a critical early role , coupling organization of pre-existing filaments into ordered arrays , with a secondary but important capacity to coordinate array size and structure through 'patchy' monomer incorporation . 3 . Once the rudimentary sarcomere 'core' is assembled and defined Z-discs with fixed spacing are established , further extension and addition of actin filaments is dictated by the initial organization of the sarcomere . It is at this stage that Fhos assumes a striated localization pattern corresponding to thin-filament array ends , and where localization to the barbed end region becomes critical for Fhos function . We identified three different modes of actin incorporation which contribute to the nascent sarcomere maturation during the later stages of pupal development: The contribution of barbed-end turnover to thin-filament array organization is not readily apparent . Turnover may be part of a mechanism that maintains the filament integrity by a continuous process of elongation and disassembly within the dynamic environment of the maturing Z-disc . Turnover does not require Fhos , and its molecular basis is currently unknown . It will be interesting to examine if a similar process is operating in adult muscles , to maintain their integrity in the face of extensive contraction activity during flight ( Perkins and Tanentzapf , 2014 ) . In conclusion , our analysis of IFM sarcomeres implies that the assembly of the highly structured thin-filament array is an elaborate , stepwise process involving diverse aspects and machineries of microfilament nucleation , growth and organization . The formin protein Fhos plays a central role , contributing to array assembly at several stages of the process . Fhos acts initially to mediate the assembly of thin-filament arrays within discrete sarcomeric units . Fhos localization to the Z-disc region , an apparently conserved feature among FHOD-family formins ( Iskratsch et al . , 2010; Mi-Mi et al . , 2012; Rosado et al . , 2014 ) , becomes essential for function during the later stages of IFM development , where Fhos plays an essential role in sarcomere radial growth . Interestingly , it has been suggested that localization of FHOD3 to the Z-lines of murine cardiomyocytes is a regulated process , relying on phosphorylation of a short domain encoded by an alternatively-spliced exon ( Iskratsch and Ehler , 2011; Iskratsch et al . , 2010 ) , echoing the importance of Fhos Z-disc localization described here . While this study has elucidated specific roles for a FHOD-family formin in a model system of skeletal muscle sarcomerogenesis , key issues remain open . The precise molecular nature of the microfilament-associated activities of Fhos , the mechanistic significance of its spatial localization patterns , regulation of its sarcomeric activities and their coordination with other functional elements of the actin-based cytoskeleton , all await further investigation .
GAL4 drivers included mef2-GAL4 ( Ranganayakulu et al . , 1996 ) and act88F-GAL4 Gajewski and Schulz , 2010] , BDSC 38461 ) . UAS-Dicer2 elements were included for enhancement of RNAi activity ( Dietzl et al . , 2007 ) . GFP-actin lines ( Roper et al . , 2005 ) included UAS-GFP-act5C ( BDSC # 9258 ) and UAS-GFP-act88F ( BDSC # 9253 and # 9254 ) . UAS-dsRNA lines used: fhos ( VDRC GD2374 [knockdown of long forms] and VDRC KK108388 [general Fhos knockdown] ) ; sals ( VDRC KK112869 and TRiP JF01110 ) ; tmod ( VDRC GD32602 and TRiP JF01094 ) Crosses were commonly kept at 25°C . Temporally-controlled expression protocols utilized the GAL80ts/TARGET system ( McGuire et al . , 2004 ) . F1 progeny were raised at 18°C and shifted to 29°C at 0 hr APF ( white pupae ) . Pupae were then grown at 29°C until the desired developmental time , taking into account the accelerated pupal development ( 1 hr at 29°C equals approximately 1 hr and 20 min at 25°C ) and the time for a complete substitution to inactive form of GAL80 ( approximately 5 hr ) . All indicated time windows are equivalent to the developmental periods for flies grown at 25°C . Generation of the different fly lines described in the study was achieved as follows: Fhos-GFP protein trap line: obtained by injection of the protein trap plasmid Splice phase 0 EGFP-FIAsH-StrepII-TEV-3xFlag , into the MiMIC insertion line FhosMI09324 , as described ( Venken et al . , 2011 ) . UAS-PH-Fhos transgenic line: A full length Fhos-RH cDNA was assembled from clones IP17223 and SD08909 using restriction free cloning as described ( Unger et al . , 2010 ) . The resulting construct was amplified by PCR , cloned into the NotI–KpnI sites of the pUAST–attB vector containing N-terminal 3XHA and injected following sequence verification into an attP40 line to produce transgenic flies . Short isoform mutant alleles: A CRISPR/Cas9-based approach ( Gratz et al . , 2013 ) was used to target the 1st coding exon of the short Fhos isoforms . A guide RNA template complementary to sequences within the exon ( 5’CTTCGCCGCCTTCCCGATCCCGGTG3’ ) was synthesized and cloned into the pU6-BbsI-chiRNA plasmid ( Addgene ) , and the plasmid was injected into vasa-Cas9 embryos ( BestGene ) . Lines were established from the progeny of the injected flies . Mutations were identified by sequencing PCR-amplified genomic DNA encompassing the relevant exon from flies bearing candidate mutant 3rd chromosomes . Two deletion events were identified in this manner , a single guanine nucleotide deletion giving rise to FhosΔG99 ( Figure 3A ) and a four nucleotide deletion ( positions 101–104 ) . Both deletions result in translational frameshifts and premature translational arrest at amino-acid position 58 of the protein sequence , and display identical phenotype and localization features . Primers used included 5’ GCGTGGCGTGCCAACAATTTG3’ and 5’ GATCGCGATAATGCGATCCACC ) for genomic DNA amplification and 5’GTATCTCGTAAATGCGCAG3’ for sequencing . FhosI966A substitution allele: CRISPR-based genome editing was used to generate the point mutant allele FhosI966A as described ( Gratz et al . , 2013 ) . Briefly , three 1000 bp fragments , which cover the Fhos FH2 domain genomic region , including exon #20 , were synthesized , and cloned into the pHD-DsRed-attB vector ( Addgene ) . The two flanking fragments served as homology arms , while the middle fragment harbored a point mutation leading to substitution of Isoleucine 966 to Alanine . Two different gRNAs 5’CCTCATATAACACCCAATTGGTC and 5’CCATGTAAGAATTAACTTTTGTA ) homologous to sites 5’ and 3’ of the replaced genomic area were synthesized and cloned into pU6-BbsI-chiRNA . The plasmid mixture was injected into vasa-Cas9 flies ( BDSC#55821 ) ( BestGene ) . All plasmid constructs were verified by sequencing . Antibodies were raised to the 93 C-terminal residues of the Fhos protein . The relevant sequence was amplified by PCR from the SD08909 cDNA clone and cloned into the pDEST17-6xHis vector ( Invitrogen ) . The recombinant protein expressed in BL21 cells , purified and injected into Wistar rats to raise the polyclonal antisera . A modified protocol from ( Weitkunat and Schnorrer , 2014 ) was used for both adult and pupal IFM tissue preparation . Briefly , staged pupae were removed from the pupal case , pinned down on Sylgard plates and dissected in cold relaxation buffer ( 20 mM phosphate buffer , pH 7 . 0; 5 mM MgCl2; 5 mM EGTA , 5 mM ATP ) . For adult IFMs , thoraces of young adults ( not older than 48 hr post enclosure ) were bisected on the longitudinal axis and collected in the cold relaxation buffer . In both cases fixation was carried out with 4% paraformaldehyde for 20 ( pupa ) or 30 ( adults ) minutes at room temp . Following washes and permabilization with PBS+0 . 3% Triton-X ( pupa ) and PBS+0 . 5% Triton-X ( adults ) , the samples were incubated in blocking solution containing 0 . 1% bovine serum albumin ( BSA ) + 5% Normal goat serum ( NGS ) . Staining that involved anti-α-actinin and/or anti-MHC required an additional 30 min blocking step with Image-IT FX signal enhancer reagent ( Thermo ) prior to the standard blocking step . All primary antibodies were diluted in standard blocking solution ( 0 . 1% BSA + 5% NGS ) and were added for overnight incubation at 4°C . Following washes , secondary antibodies were added for 2 hr at room temp . Adult hemi-thoraces were cleared in 80% glycerol at 4°C overnight prior to mounting . All samples were mounted in Immu-Mount ( Thermo ) . Primary antibodies and dilutions used included: anti-GFP ( chicken , 1:1000 , Abcam ) ; anti-MHC ( rabbit , 1:1000 , kindly provided by P . Fisher , Stony Brook ) ; anti-α-actinin ( rat , 1:50 , Babraham institute , UK ) ; anti-Zasp52 ( rabbit 1:500 , [Katzemich et al . , 2013] ) ; anti-Obscurin ( rabbit , 1:500 , [Burkart et al . , 2007] ) ; anti-Fhos ( rat , 1:200 ) was generated as described above; anti-Tmod ( Rat , 1:500 , kindly provided by Velia Fowler ) . Secondary antibodies used included Alexa Fluor 405 , Alexa Flour 488 , Alexa Fluor 555 , Alexa Fluor 568 and Alexa Flour 647 conjugated to anti-rabbit , mouse , rat , or chick antibodies ( Molecular Probes ) and applied at a dilution of 1:1000 . Atto647N-Phalloidin ( Fluka ) was used at 5 μg/ml . Immunofluorescent images of fixed samples were acquired using Zeiss LSM 710 or Zeiss LSM 780 confocal scanning systems , equipped with a Zeiss Axiovert microscope , and using a ×20 0 . 8 N . A or ×63 oil immersion 1 . 4 N . A lenses . The initial image acquisition was performed using the imaging system Zen software . Thoraces of young adults or isolated IFMs were collected in the ice cold relaxation buffer ( 20 mM phosphate buffer , pH 7 . 0; 5 mM MgCl2; 5 mM EGTA , 5 mM ATP ) . Following 15 min incubation the samples were transferred into 1 mM sodium cacodylate buffer ( pH 7 . 4 ) containing fixative ( 4% paraformaldehyde and 2 . 5% glutaraldehyde ) . Samples were fixed for 1 hr ( IFMs ) or 2 hr ( adults ) at room temp and transferred to 4°C overnight . Samples were washed x3 with sodium cacodylate buffer , post fixed in 1% OSO4 solution for 1 hr at room temp , washed x3 with sodium cacodylate buffer , incubated in 2% aqueous uranyl acetate for 1 hr and washed x3 with distilled water . Samples were taken through an ethanol dehydration series and incubated in propylene oxide ( x3 , 10 min each ) . Infiltration was performed with a series of propylene oxide: Epon mixtures , culminating in incubation in 100% Epon ( x3 , 12 hr each ) . Infiltrated samples were embedded in plastic moulds ( EMS ) and polymerized for 48 hr at 60° . Ultra thin sections were cut using diamond knife 35° ( Diatome , Switzerland ) on a Leica Reichert ultra cut UCT . Sections were post stained with 1% lead citrate and 2% uranyl acetate . Images were recorded using an FEI T12 spirit BioTWIN transmission electron microscope ( TEM ) operating at 120KV and equipped with an Eagle 2Kx2K CCD camera ( FEI ) . Measurements of various geometric properties of the sarcomere and tagged actin monomers incorporation were performed using Fiji image analysis software . For sarcomere length and width , the Fiji measurement tool was used to draw a vertical line ( width ) or horizontal line ( length ) across the Z-disc of a single sarcomere from pointed end to pointed end , using phalloidin staining as a guide . 50 sarcomeres were measured form 7 different flies for each genotype ( 350 sarcomeres in total ) . Horizontal lines or polygons were drawn to measure the Z-disc associated incorporation band length ( act5C control and following Fhos knockdown ) or the incorporation frame area ( act88F ) respectively . 50 sarcomeres were measured form 4 different flies for each genotype ( 200 sarcomeres in total ) . The intensity distribution in Figure 4E and G was measured along 10 μm horizontal profiles starting from a Z-disc and drawn at the middle of the myofibril . The data represent an average of normalized values collected from 15 profiles in 7 different flies ( n = 105 ) for each genotype . Vertical profiles drawn across half sarcomeres was used to measure the GFP intensity distribution in Figure 4J . The data represents an average of normalized values collected from 100 half sarcomeres in 4 different flies ( 400 sarcomeres in total ) . To measure the M/Z incorporation intensity ratio in Figure 5 and S5 polygons were drawn around the Z-disc or M-line area . The date represents an intensity measurement from 50 sarcomeres in 5 different flies ( 250 sarcomeres in total ) . The F-actin intensity distribution in Figure 5 and S5 was measured along a 13 μm horizontal profile . The data represents an average of normalized values collected from 15 profiles in 5 different flies ( n = 75 ) . Distribution of actin incorporation events was obtained from 50 sarcomeres in 5 different flies ( 250 sarcomeres in total ) . The incorporation events were visualized by a GFP intensity profile drawn along the contour of the nascent arrays . Z-disc and M-line vicinity markers ( Zasp52 and Obscurin , respectively ) determined the location of the incorporation event . For counting thick filaments in TEM cross-sections , a threshold base segmentation was applied and the number of filaments determined by using Fiji Analyze Particles tool . The data represent an average thick filament number per myofibril from 20 myofibrils from 3 different samples . All graphs and statistic tests were done using GraphPad Prism software . The figures were assembled and organized using Adobe Photoshop CS6 . | Muscles owe their ability to contract to structural units called sarcomeres , and a single muscle fiber can contain many thousands of these structures , aligned one next to the other . Each mature sarcomere is made up of precisely arranged and intertwined thin filaments of actin and thicker bundles of motor proteins , surrounded by other proteins . Sliding the motors along the filaments provides the force needed to contract the muscle . However , it was far from clear how sarcomeres , especially the arrays of thin-filaments , are assembled from scratch in developing muscles . When the fruit fly Drosophila transforms from a larva into an adult , it needs to build muscles to move its newly forming wings . While smaller in size , these flight muscles closely resemble the skeletal muscles of animals with backbones , and therefore serve as a good model for muscle formation in general . New muscles require new sarcomeres too , and now Shwartz et al . have observed and monitored sarcomeres assembling in developing flight muscles of fruit flies , a process that takes about three days . The analysis made use of genetically engineered flies in which the gene for a fluorescently labeled version of actin , the building block of the thin filaments , could be switched on at specific points in time . Looking at how these green-glowing proteins become incorporated into the growing sarcomere revealed that the assembly process involves four different phases . First , a large store of unorganized and newly-made thin filaments is generated for future use . These filaments are then assembled into rudimentary structures in which the filaments are roughly aligned . Once these core structures are formed , the existing filaments are elongated , while additional filaments are brought in to expand the structure further . Finally , actin proteins are continuously added and removed at the part of the sarcomere where the thin filaments are anchored . Shwartz et al . went on to identify a protein termed Fhos as the chief player in the process . Fhos is a member of a family of proteins that are known to elongate and organize actin filaments in many different settings . Without Fhos , the thin-filament arrays cannot properly begin to assemble , and the subsequent steps of growth and expansion are blocked as well . The next challenges will be to understand what guides the initial stages in the assembly of the thin-filament array , and how the coordination between assembly of actin filament arrays and motor proteins is executed . It will also be important to determine how sarcomeres are maintained throughout the life of the organism when defective actin filaments are replaced , and which proteins are responsible for carrying out this process . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2016 | The Drosophila formin Fhos is a primary mediator of sarcomeric thin-filament array assembly |
In the developing mammalian brain , differentiating neurons mature morphologically via neuronal polarity programs . Despite discovery of polarity pathways acting concurrently with differentiation , it's unclear how neurons traverse complex polarity transitions or how neuronal progenitors delay polarization during development . We report that zinc finger and homeobox transcription factor-1 ( Zeb1 ) , a master regulator of epithelial polarity , controls neuronal differentiation by transcriptionally repressing polarity genes in neuronal progenitors . Necessity-sufficiency testing and functional target screening in cerebellar granule neuron progenitors ( GNPs ) reveal that Zeb1 inhibits polarization and retains progenitors in their germinal zone ( GZ ) . Zeb1 expression is elevated in the Sonic Hedgehog ( SHH ) medulloblastoma subgroup originating from GNPs with persistent SHH activation . Restored polarity signaling promotes differentiation and rescues GZ exit , suggesting a model for future differentiative therapies . These results reveal unexpected parallels between neuronal differentiation and mesenchymal-to-epithelial transition and suggest that active polarity inhibition contributes to altered GZ exit in pediatric brain cancers .
Construction of the central nervous system’s circuitry requires that newborn neurons exit their germinal zone ( GZ ) , elaborate axons and dendrites , migrate to a final position and synaptically engage other neurons . Emerging evidence suggests that classic cell polarity signaling molecules , including the Numb endocytic adaptor , the Partitioning defective ( Pard ) polarity complex and LKB1/SAD kinases , create the cellular asymmetry required for neuronal development and circuit assembly ( Shi et al . , 2003; Solecki et al . , 2004; Kishi et al . , 2005; Cappello et al . , 2006; Rasin et al . , 2007; Shelly et al . , 2007; Barnes et al . , 2007; Bultje et al . , 2009; Hengst et al . , 2009; Zhou et al . , 2011; Chen et al . , 2013; Famulski et al . , 2010 ) . Indeed , defective neuronal polarization is proposed to underlie the pathology of some neurodevelopmental or neurodegenerative diseases , and restored polarity has been suggested as a potential therapeutic approach for syndromes involving perturbed polarity-linked mechanisms ( May-Simera and Liu , 2013 ) . Given the importance of polarity for neuronal maturation events , great efforts have been made to define mechanisms that cell-extrinsically or -intrinsically control polarity during neuronal differentiation . Most current models suggest the activation of signaling cascades ( Barnes and Polleux , 2009; Lewis et al . , 2013; Funahashi et al . , 2014 ) , transcriptional networks ( de la Torre-Ubieta and Bonni , 2011 ) , or chromatin states ( Hirabayashi and Gotoh , 2010; Yamada et al . , 2014 ) promotes or maintains cell polarity in differentiated neurons . However , it remains unclear how developing neurons undergo discrete transitions during which polarity is delayed or promoted ( Cooper , 2014; Singh and Solecki , 2015 ) . As an example , maturing cortical neurons undergo enhanced polarization via a multipolar to bipolar transition , while cerebellar granule neuron progenitors ( GNPs ) remain unpolarized for an extended period while their progenitor pool expands during cerebellar development . We discovered that the transcription factor Zeb1 , a critical regulator of epithelial polarity ( Vandewalle et al . , 2009 ) , is highly expressed in unpolarized GNPs and that its expression diminishes as these cells become polarized cerebellar granule neurons ( CGNs ) . Developing CGNs provide an excellent model of the mechanisms regulating neurogenesis , neuronal differentiation , polarization linked to morphological maturation , and GZ exit ( Hatten and Roussel , 2011; Hatten and Roussel , 2011 ) . They also provide a model of migration mechanisms , since they undergo two migration phases: morphologically unpolarized GNPs and newly postmitotic CGNs migrate tangentially near the cerebellar surface in the external granule layer ( EGL ) while polarized CGNs migrate radially away from their GZ and cross the molecular layer ( ML ) to reside within the internal granule layer ( IGL ) ( Hatten , 2002; Chédotal , 2010; Legué et al . , 2015 ) . In cerebellar medulloblastoma ( MB ) , excessive or constitutive mitogenic signaling in GNPs disrupts the intricate balance of GZ exit and radial migration via unknown motility mechanisms ( Goodrich et al . , 1997; Kim et al . , 2003; Yang et al . , 2008; Ayrault et al . , 2010 ) . Zeb1 functions in many organ systems , including muscle , lymphocytes , and nervous system ( Takagi et al . , 1998; Liu et al . , 2008 ) . Proliferating progenitors express Zeb1 in GZs in the developing mouse brain ( Darling et al . , 2003 ) . While loss of Zeb1 function in the developing neocortex reduces proliferation in the VZ and SVZ ( Liu et al . , 2008 ) , it remains unknown how Zeb1 regulates neural progenitor populations . Studies examining Zeb1 regulation of epithelial cell polarity provide insights . Zeb1 activates stemness pathways in immature , unpolarized epithelial cells and their transformed counterparts ( Spaderna et al . , 2008 ) . It also controls transitions in epithelial differentiation and polarity plasticity: high Zeb1 expression inhibits epithelial differentiation and drives cells toward epithelial-to-mesenchymal transition ( EMT ) , while low expression allows mesenchymal-to-epithelial transition ( MET ) . During EMT , Zeb1 acts as a transcriptional repressor that silences adherens junction ( AJ ) and apical-basal polarity genes ( Aigner et al . , 2007 ) . Thus , Zeb1 simultaneously blocks differentiation , apical-basal polarity , and junction formation of epithelial cells , locking them into the mesenchymal state . In the developing nervous system , EMT-like events have been observed in the transition of polarized radial glia to their delaminating progeny ( Rousso et al . , 2012; Itoh et al . , 2013 ) . Given that neuronal progeny undergo multiple polarity transitions after delamination from a radial glial cell , it is open to question how polarity is re-acquired after delamination as nascent neurons mature ( Cooper , 2014; Singh and Solecki , 2015; Barnes et al . , 2008 ) . Do nascent neurons that undergo an EMT-like process also then transition through an MET-like process , like epithelial cells ? We have known for more than a decade that persistent Sonic hedgehog ( SHH ) signaling blocks GNP GZ exit , but the mechanism has remained a mystery . Here we hypothesized that MET-like events control the onset of neuronal differentiation and GZ exit , which involve cell polarity and cell-cell adhesion transitions . By using gain- and loss-of-function approaches , we found that Zeb1 is necessary and sufficient to maintain GNPs in an undifferentiated , unpolarized , transiently amplifying state within the EGL and to control the onset of their GZ exit . Zeb1 represses transcription of polarity and cell adhesion genes , such as Pard6a , Pard3a and close homolog of L1 ( Chl1 ) . By using a functional screen , we found that restored expression of these genes rescues GNP differentiation , neurite extension , and GZ exit . Finally , we examined the link between morphogens and Zeb1 in controlling this process . We found that SHH , a potent GNP mitogen , maintains Zeb1 expression . Moreover , Zeb1 expression persists in MB tumor cells , the transformed GNP counterpart in which SHH signaling is persistently activated . Zeb1 loss-of-function or restored Zeb1 target expression rescued the GZ exit phenotype in Patched1 ( Ptch1 ) -deficient GNPs , the progenitors of SHH-subgroup MB . Our findings show that CGN differentiation bears a remarkable similarity to mesenchymal-to-epithelial transition . The balance of EMT-like vs . MET-like processes and of proliferative vs . maturation processes may be a key developmental mechanism that , when disrupted , contributes to the pathological alteration of GZ exit in neurodevelopmental disorders and pediatric cancers .
To test the hypothesis that MET-like events control the onset of GNP differentiation and GZ exit , we first surveyed expression of the canonical EMT regulators Snail1 , Snail2 , Twist , and Zeb1 in GNPs . Quantitative RT-PCR revealed that Zeb1 is the primary EMT factor expressed in GNPs during the early postnatal ( P ) peak of neurogenesis , and that expression diminishes as GNPs exit the cell cycle to differentiate into CGNs: at P7 Zeb1 mRNA was 28-fold higher than the next most abundant transcription factor , Snail1 ( Figure 1a ) . Zeb1 protein expression confirmed our RNA analysis where it is expressed primarily in the EGL at P7 and greatly reduced at P15 ( Figure 1b ) . At P7 , Zeb1 is co-expressed with the proliferation marker Ki67 and two markers of GNP identity Siah2 , and Meis1/2 , and is greatly reduced in cyclin-dependent kinase inhibitory protein p27Kip1/Cdkn1b ( referred as p27 thereafter ) -positive postmitotic CGNs in the inner EGL . We noted a subpopulation of Zeb1 positive cells in deeper layers of the cerebellum at P7 . These cells represent a mixture of white matter interneuron or oligodendrocyte precursors as these cells also express Pax2 ( Maricich and Herrup , 1999 ) or Olig2 ( Chung et al . , 2013 ) ( Figure 1—figure supplement 1 ) . In GNPs , Zeb1 mRNA expression was inversely correlated with the expression of the apical-basal polarity genes Pard6a and Prkcz ( Figure 1c ) . Not only did Pard6a mRNA increase as CGN differentiation proceeded , but the promoter of this gene was active in individual GNPs at the border of the GZ , prior to their entry into the inner EGL ( Figure 1d ) . Taken together , these results indicate that GNPs are mesenchymal-like , as they express a high level of Zeb1 and low levels of polarity genes . 10 . 7554/eLife . 12717 . 003Figure 1 . Zeb1 is the primary EMT regulator expressed in the developing cerebellum . ( a ) qRT-PCR shows that Zeb1 mRNA is more abundant than other EMT factors ( Twist , Snail1 , Snail2 ) in GNPs . Zeb1 mRNA diminishes in GNPs at P10 and P15 ( Zeb mRNA was significantly different at all times , t-test p<0 . 01 ) . ( b ) Immunohistochemistry in P7 and P15 cerebellum shows Zeb1 ( red ) GNP expression at P7 coincident with that of Ki67 , Meis1/2 and Siah2 ( green ) but complementary to the p27Kip marker ( green ) . Zeb1 protein diminishes at P15 . ( c ) qRT-PCR shows increasing Pard6a and Prkcz mRNA as GNPs at P10 and P15 . ( d ) Immunohistochemistry in the P7 cerebellum of Pard6a-EGFP BAC transgenic mice shows little Pard6a promoter activity ( green ) in the outer EGL but elevated activity in the inner EGL with TAG1-positive CGNs ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 00310 . 7554/eLife . 12717 . 004Figure 1—figure supplement 1 . Zeb1 is expressed in Pax2 and Olig2 positive progenitors in the developing cerebellar white matter . Immunohistochemistry in P7 cerebellum shows Zeb1 ( green ) expression at P7 partially overlaps with ( a ) Pax2 and ( b ) Olig2 in cerebellar white matter . This indicates that Zeb1 positive cells located in deeper cerebellar layers are interneuronal- or oligodendrocyte/glial-progenitors , not IGL resident CGNs . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 004 Given the Zeb1 expression profile , we reasoned that this transcription factor might regulate GNP differentiation . We used a gain-of-function approach to examine Zeb1’s role in this process , as this method maintained Zeb1 expression in GNPs and because diminished Zeb1 expression coincides with differentiation to CGNs . Purified P7 GNPs were nucleofected with an expression vector that encodes mouse Zeb1 . After 1 day in vitro , control GNPs displayed features of differentiated CGNs: they extended neurites , expressed p27 and no longer expressed Ki67 and Atoh1 , a marker of proliferating GNPs ( Figure 2a , b ) ( Ayrault et al . , 2010; Flora et al . , 2009 ) . In contrast , Zeb1-expressing cells had short , multipolar extensions ( x¯ length=140 ± 13 μm vs 60 ± 3 μm ) , expressed reduced p27 and sustained levels of Ki67 and Atoh1 , indicating arrested maturation and proliferating , GNP-like state . While Zeb1-expressing GNPs were motile on time-lapse microscopy in dissociated cultures , they did not display the typical two-stroke nucleokinesis cycle used by differentiated CGNs and had an apolar , isotropic f-actin distribution reminiscent of GNP morphology in vivo ( Videos 1 and 2 ) . At the moment , it’s unclear whether this mesenchymal-like morphology and random migration direction is due to a disturbed intrinsic polarity program or perturbed glial binding . 10 . 7554/eLife . 12717 . 005Figure 2 . Zeb1 gain- or loss-of-function determines GNP differentiation . ( a ) Micrographs of purified CGNs nucleofected with Centrin2-Venus alone ( green ) or Myc-Zeb1 ( magenta ) . After 24 hr in culture , control cells extend long neurites ( x¯ = 139 . 8 ± 13 . 3 μm . n = 1045 cells ) , while Zeb1-expressing cells have short neurites ( x¯ = 59 . 6 ± 3 . 0 μm , n = 1164 cells , χ2 test , p<0 . 01 ) . ( b ) Micrographs of purified CGNs nucleofected with Centrin2-Venus ( green ) cytoplasmic marker and Myc-Zeb1 . After 24 hr , levels of p27 labeling decreased , while that of Ki67 and Atoh1 increased ( t-test all conditions p<0 . 05 ) . C , D . P7 EGL was co-electroporated with indicated vector and H2B-mCherry . After 24 ( C ) or 48 ( D ) hr of ex vivo culture , the migration distance of labeled CGN from the pial layer ( dashed line ) was analyzed in 3 experiments . Histograms show migration distributions . Zeb1-silenced cells incorporated EdU at lower rates than control cells . ( c ) Most control shRNA-expressing cells ( black ) remain within the EGL ( dashed lines , x¯ = 34 . 2 ± 10 . 5 µm ) at 24 hr , while Zeb1-silenced cells pre-maturely enter the ML and IGL ( x¯ = 67 . 5 ± 18 . 1 µm ) . ( d ) Control cells ( black ) entered the ML and IGL by 48h ( x¯ = 75 . 2 ± 3 . 5 µm ) , while Zeb1-expressing cells remain within the EGL ( x¯ = 40 . 2 ± 6 . 0 µm ) . T-tests and χ2 test showed significant differences in both conditions ( p<0 . 01 , n = 4500 to 9700 cells ) . ( e ) Immunohistochemistry in E18 . 5 cerebellum of wild type and Zeb1 mutant embryos shows the expected absence of Zeb1 expression in mutant embryos . Moreover , increased expression of Tag1 and NeuN differentiation markers is observed in the absence of Zeb1 . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 00510 . 7554/eLife . 12717 . 006Figure 2—figure supplement 1 . In depth quantitation of slice migration assays from Figure 2 . P7 EGL was co-electroporated with the indicated expression constructs and H2B-mCherry . After 24 ( a–d ) or 48 ( e–f ) hours of ex vivo culture , CGN migration distance was analyzed in 3 imaging experiments . Red overlay indicates the average migration distribution of control cells ( error bar , SD ) . ( a , b ) Most control ( n = 7 , 358 cells ) migrated 34 . 2 ± 10 . 1 µm [x¯ ± sd] at 24 hr , while Zeb1-silenced ( n = 4 , 693 cells ) migrated an Av . distance of 67 . 5 ± 18 . 1 µm . χ2 analysis showed distribution of data to be significantly different ( p<0 . 01 ) . ( c ) Binning distribution of the second mir30 based shRNA 846 used to confirm the precocious migration associated with Zeb1 silencing displayed in Figure 2C ( n = 4879 cells ) . ( d ) Representative image of the mir30 based shRNA 846 illustrating Zeb1 silencing with a mirRNA based shRNA also spurs GZ exit . ( e , f ) While control ( n = 9 , 744 cells ) cells entered the ML and IGL after 48 hr ( av distance = 75 . 2 ± 3 . 5 µm ) , However , Zeb1 ( n = 5 , 359 cells ) over-expressing cells remained in the EGL ( Av . distance migrated = 40 . 2 ± 6 . 0 µm ) , which was shown to be significantly different by both χ2 analysis and t-test ( p<0 . 03 ) . ( g ) Summary of average distance migrated for the Zeb1-silencing and over-expression . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 00610 . 7554/eLife . 12717 . 007Figure 2—figure supplement 2 . shRNA knockdown of Zeb1 . ( a ) Immunoblots of lysates of NS5 cells transduced with shLuc or shZeb1 . Zeb1 levels are lower than control NS5 cell or cells expression a luciferase control shRNA . shZeb1 was used in Figure 2 . ( b ) Immunoblots of lysates of HEK293 cells expressing Myc-Zeb1 with or without the corresponding shmiRNA constructs . After 48 hr Myc-Zeb1 protein levels are substantially less than those in controls , actin was loading control . The Zeb1 shmiRNA was used in Figure 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 00710 . 7554/eLife . 12717 . 008Video 1 . Representative time lapse imaging sequence of a CGN migrating in a dissociated culture labeled with Centrin2-Venus ( green , centrosome ) and RFP-UTRCH ABD ( f-actin ) . The focused cell undergoes typical two-stroke nucleokinesis with f-actin accumulation in the leading process . Time stamp= hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 00810 . 7554/eLife . 12717 . 009Video 2 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs migrating in a dissociated culture labeled with Centrin2-Venus ( green , centrosome ) and RFP-UTRCH ABD ( f-actin ) . The featured cells undergo random amoeboid movements with isotropic f-actin decorating the cell periphery . Note the centrosome does not adopt a polarized configuration as in Video 1 . Time stamp= hours: minutes: seconds . Scale bar =10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 009 We next assessed the effect of Zeb1 function on GNP differentiation , GZ exit and migration to the IGL with the ex vivo cerebellar slice assay developed in our laboratory that specifically label GNPs ( Figure 2c and d , see Figure 2—figure supplement 1 for detailed analysis ) . We used two independent shRNA vectors to silence Zeb1 in P7 EGL ( see Figure 2—figure supplements 1 and 2 for second shRNA migration data and validation ) . After 24 hr ex vivo , control EGL cells resided in the GZ and incorporated EdU , not having differentiated into CGNs or begun migrating to the IGL . In contrast , Zeb1 silencing increased migration toward the IGL ( x¯ distance=34 ± 10 μm vs 68 ± 18 μm ) and reduced EdU incorporation ( 22 . 6 ± 1 . 0% vs 7 . 6 ± 1 . 8% EdU positive ) , showing that Zeb1 loss-of-function promotes differentiation and migration toward the IGL . We next confirmed that Zeb1 activity inhibited GZ exit , using a gain-of-function approach . P7 EGL was electroporated with an expression vector for Zeb1 . After 2 days ex vivo , control CGNs entered the molecular layer and IGL , while Zeb1-expressing CGNs remained within the EGL ( x¯ distance=75 ± 3 μm vs 40 ± 6 μm , Figure 2d ) and continued to incorporate EdU ( 3 . 3 ± 0 . 4% vs 10 . 9 ± 0 . 1% EdU positive ) . To further examine the role of Zeb1 in GNP differentiation in vivo we scored Tag1 and NeuN expression in the EGL of E18 . 5 Zeb1 null embryos ( a time-point prior perinatal lethality observed in Zeb1 null embryos ) . Consistent with our ex vivo gene silencing results , loss of Zeb1 function in vivo leads to an increase in Tag1 and NeuN differentiation marker gene expression , indicating an increase of neuronal differentiation in the absence of Zeb1 ( Figure 2e ) . These observations indicate that Zeb1 inhibits differentiation of GNPs to CGNs and is necessary and sufficient to restrict GNPs to their GZ niche . They also suggest that Zeb1 inhibits GNP polarization , as neurite extension , two-stroke nucleokinesis and GZ exit depend on polarity signaling complexes in CGNs . Having learned that Zeb1 inhibits GNP differentiation and potentially the downstream events associated with CGN polarization , we next sought to identify Zeb1 targets to determine how sustained Zeb1 expression maintains GNPs . We reasoned that as Zeb1 gain-of-function strongly inhibits GNP differentiation , it would provide a basis to identify potential Zeb1 targets . We prepared RNA from P0 , P7 and P15 GNPs and used Affymetrix DNA arrays to compare their transcriptomes of these cells with those of pure , FACS-sorted GNP populations nucleofected with control , Zeb1- or HES1 expression vectors ( Figure 3a , ArrayExpress accession number: E-MTAB-3557 ) . We included the transcription factor HES1 because it is a known repressor of GNP differentiation downstream of the Notch2 receptor ( Solecki et al . , 2001 ) . GNPs were selected for our developmental expression analyses as it is well established this transiently amplifying progenitor population expresses early CGN differentiation markers; such as TAG1 , L1 , NRCAM , NeuroD1 or TIS21 prior to their final cell cycle ( Miyata et al . , 1999; Iacopetti et al . , 1999; Xenaki et al . , 2011 ) . Zeb1 gain-of-function suppressed a group of genes increasingly expressed between P0 and P15 ( Figure 3b , Figure 3—figure supplement 1 ) , consistent with previous observations that Zeb1 acts as a transcriptional repressor . Gene ontology analysis revealed this group of genes to be associated with tissue morphogenesis , epithelial polarization , cell adhesion and control of cell motility . Key members of the apical or basolateral polarity pathways ( Pard6a Pard3a , Dlg2 and Lin7a ) and Cdh1 AJ adhesion molecule were among the Zeb1-repressed genes . In parallel , we analyzed the EMT/MET signature upon Zeb1 gain-of-function in GNPs , using a pathway-focused PCR array . Various genes previously shown to be induced during EMT were enriched in these GNPs , while a class of MET-related genes were repressed ( see Tables in Supplementary file 1B ) . For further validation we selected a group of genes that included polarity complex genes ( Pard6a , Pard3a , Dlg2 and Lin7a ) , cell adhesion genes ( Cdh1 and Chl1 ) , transcription factors associated with cell differentiation ( Bhlhe40 and Nfib ) , and three randomly selected genes ( Sorl1 , Flt1 , and Cdk5r1 ) , most of which were not significantly repressed by HES1 . Not only were many of these genes increasingly expressed in the normal developmental time course ( Figure 3—figure supplement 2 ) and validated as suppressed in Zeb1-expressing GNPs ( Figure 3c ) , the protein expression of many of them was mutually exclusive with Zeb1 or Ki67 in vivo ( Figure 3d ) . A previous study in our laboratory demonstrated a similar expression profile for Pard3a ( Famulski et al . , 2010 ) . These results suggest that many of the putative targets identified are bona fide CGN differentiation markers expressed at low levels in early postnatal GNPs . Increased polarity gene expression in differentiated CGNs , their mutually exclusive expression with GNP markers , and their suppression by Zeb1 , further suggest a parallel between GNP differentiation and MET . 10 . 7554/eLife . 12717 . 010Figure 3 . Zeb1 transcriptionally represses neuronal differentiation , cell polarity , and cell adhesion genes . ( a ) Schematic of procedure used to produce pure populations of CGNs for array studies . ( b ) Heat map of the transcriptomes of GNPs and CGNs purified from P0 , P7 , and P15 compared to pure populations of control ( e . g . H2B-mCherry vector alone ) , Zeb1-expressing ( e . g . H2B-mCherry and Zeb1 vector ) and HES1-expressing ( e . g . H2B-mCherry and HES1 vector ) GNPs cultured for 24 hr in vitro . Yellow rectangle highlights genes whose expression increases with development and are repressed by Zeb1 . ( c ) qRT-PCR shows that ectopic Zeb1 expression inhibits transcription of most of the panel of CGN differentiation markers examined . ( d ) Immunohistochemistry in P7 cerebellum shows Zeb1 ( red ) and Ki67 ( green ) expression complementary with expression of the Lin7a , Sorl1 , Cdk5r1 , Chl1 , Dlg2 and Pard6a ( red ) CGN markers . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01010 . 7554/eLife . 12717 . 011Figure 3—figure supplement 1 . PCA analysis of array experiments shown in Figure 3 . Principal component analysis ( PCA ) of Zeb1 , Hes1-overexpression GNPs , controls and cerebellum granule neuron ( CGN ) cells purified at P0 , P7 and P15 . Total of 40 . 9% of variation among these samples can be explained by the first three component ( PCA1 = 20 . 2% , PCA2 = 10 . 6% , PCA3 = 10 . 1% ) . Hes1 samples are well separated from the rest along the first component and Zeb1 are separated from the controls along PCA3 . Comparing with the Hes/Zeb controls , if the distance between the centroids of P7 CGNs and the control samples is 1 , the distance for P0 CGNs , P15 CGNs , Zeb1 samples and Hes1 samples are 1 . 3 , 1 . 2 , 1 . 2 and 2 . 4 indicating that the P7 CGNs are most similar to the control samples . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01110 . 7554/eLife . 12717 . 012Figure 3—figure supplement 2 . PCA demonstrating purity of GNPs/CGNs prepared at different developmental stages from Figure 3 . Principal component analysis ( PCA ) of GNP/CGN cells purified at P0 , P7 and P15 , compared to purified cerebellar glial cells . Variation among these samples can be explained by the first three components ( PCA1 = 35 . 3% , PCA2 = 15 . 4% , PCA3 = 9 . 84% ) . The purified glial cells are well separated indicating low levels of this most common contaminating cell population . The purified cells from each developmental stage are well clustered statically verifying the consistency of the purification procedure at the level of the whole transcriptome . P0 and P7 are most similar . The separation of P15 cells from the earlier developmental stages is on a principle component axis that unique from the purified glial cell population , indicating the differences are related to developmental changes in the cell population than contamination with a non-GNP/CGN cell population . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01210 . 7554/eLife . 12717 . 013Figure 3—figure supplement 3 . qRT-PCR analysis of Zeb1 target mRNA expression in GNPs or whole cerebellum . RNA from purified GNPs ( blue bars ) or whole cerebellum ( red bars ) was extracted at p0 , p7 , and p15 . qRT-PCR shows validate Zeb1 target mRNA expression increases as GNPs mature . Cdk5r1 expression declines as GNPs mature . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 013 We next sought to investigate whether Zeb1 directly regulates genes differentially expressed in our array . As a prelude , we first assessed global Zeb1 binding sites in a ChIP-seq data set from NS5 mouse neural stem cells , which , like GNPs , express high levels of Zeb1 ( Figure 4—figure supplement 1 , ArrayExpress accession number: E-MTAB-3560 ) . Many of the proximal promoters of key apical-basal polarity genes ( Pard6a , Pard6b , Pard3a , Pard6g ) and other Zeb1-regulated genes identified in our screen ( Chl1 , Limk2 ) showed clear Zeb1 binding peaks , suggesting that they are direct targets ( Figure 4—figure supplement 2 ) . Computational analyses comparing the genome-wide Zeb1 binding profile to the Zeb1-regulated genes identified by expression profiling showed Zeb1 binding events are highly associated with downregulated genes ( Figure 4a–c ) , further pointing to Zeb1 as a transcriptional repressor in neural stem/progenitor cells . We next validated Zeb1 binding to key genes in purified P7 GNPs by ChIP PCR ( Figure 4d ) . No binding was detected at non-functional regions of the genome in the androgen receptor and GAPDH genes . Weak but consistent binding was detected in the proximal upstream regions of Limk2 and Lin7a . Strong Zeb1 binding was observed at positive control regions in the Zeb1 gene and Cdh1 gene , at the proximal upstream sequences of Pard6 and Pard3a genes and at an intronic site in the Chl1 gene . Overall , these results indicate that Zeb1 directly regulates genes expected to play a role in cell adhesion and apical-basal polarity . 10 . 7554/eLife . 12717 . 014Figure 4 . Zeb1 binds to the genomic loci of target genes identified in the expression screen . ( a ) Zeb1 binding events are significantly associated with down-regulated genes ( right ) but not with up-regulated genes ( left ) between the NS5 CHIP-Seq and CGN expression array data . Red bars: total number of binding events associated with each group of genes; boxplots: distribution of binding events associations with 1000 random sets of genes . Test data are represented as a boxplot showing the test median and 1st and 3rd quartiles; whiskers are ± 1 . 5 the interquartile range . ( b ) Biological processes representing clusters of gene ontology terms enriched among genes directly targeted by Zeb1 . Parentheses show number of genes associated with each term . ( c ) Heat-map displaying the cumulative fraction of deregulated genes that are directly regulated by Zeb1 ( up-top left panel; down-bottom left panel ) . Transcripts are divided in equal bins of decreasing expression fold change and plotted against Zeb1 binding events with increasing p-value . Control: 100 sets of random binding events ( right panels , the mean value shown ) . ( d ) CHIP PCR Validation of Zeb1 binding in P7 GNPs . The schematic on the left displays gene structure . Exons are pink rectangles , Zeb1 binding unoccupied motifs are colored light green and validated Zeb1 binding sites are colored dark green . The graph on the right shows fold enrichment at the listed genes . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01410 . 7554/eLife . 12717 . 015Figure 4—figure supplement 1 . Overview of Zeb1 ChIP-Seq dataset in NS5 neural stem cells . ( a ) Location of Zeb1 binding events respective to the closest annotated TSS . ( b ) Locations of Zeb1 binding events respective to various genomic features . ( c ) Density plot of Zeb1 ChIP-seq reads mapping to the 4 Kb genomic regions surrounding peak summits . Signal intensity represents the ChIP-seq normalized tag count ( left ) . Total bound sites were divided in 10 bins and the top motif found enriched at vicinity of summits is shown for each bin , with respective fold enrichment over genomic background ( middle ) . The frequency of E-box motif is shown , centered on peak summits ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01510 . 7554/eLife . 12717 . 016Figure 4—figure supplement 2 . Annotated ChIP peaks in polarity genes and putative Zeb1 targets identified in NS5 data set . Visual representation of Zeb1 ChIP-seq enrichment in the vicinity of various putative Zeb1 targets . UTRs are represented as red rectangles , translated exons as black rectangles and the direction of transcription of a locus is represented by an arrow . The graph below each gene indicate the relative Zeb1 ChIP-Seq reads per million at each genomic position near the displayed genes . ( a ) Shows Zeb1 binding enrichment at core PAR complex genes . Each core member of the PAR complex contains a region of enriched Zeb1 binding near Exon 1 . ( b ) Shows Zeb1 , Chl1 and Limk2 genes . A scale bar indicates that size of each locus . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 016 Given the expression profile of Zeb1 targets and the mutually exclusive expression of these genes and Zeb1 , we postulated that some of these targets may facilitate CGN differentiation , neurite extension and GZ exit downstream of Zeb1 . We individually expressed validated targets in the context of our in vitro ( neurite extension , Ki67 or p27 expression status ) and ex vivo ( GZ exit and EdU incorporation status ) Zeb1 gain-of-function assays ( Figure 2a , b and d ) to determine whether restoring individual target expression would rescue the Zeb1 phenotypes and thus functionally prioritize these targets . For this small functional screen we selected key polarity molecules ( Pard6a , Pard3a , Lin7a , Dlg2 ) , adhesion receptors ( Cdh1 , Chl1 , constitutively active JAM-C ) , genes associated with cell differentiation ( Sorl1 , Bhlhe40 , Nfib ) and randomly selected genes ( Flt1 VEGF receptor , Cdk5r1 ) . Our laboratory has previously shown that Pard6a and Pard3a are required for CGN migration and GZ exit ( Solecki et al . , 2004; Famulski et al . , 2010 ) . Chl1 regulates neurite initiation , neuronal migration and neuronal dendrite orientation in the developing neocortex ( Demyanenko et al . , 2004; Demyanenko et al . , 2010 ) . Lin7 and Dlg homologs are components of the apical or basolateral polarity complexes in epithelial cells where Dlg recruits Lin7 to distinct membrane domains ( Bachmann et al . , 2004 ) . Nfib regulates CGN differentiation ( Wang et al . , 2007 ) , and Cdk5r1 regulates Cdk5 activity during neuronal migration ( Gupta et al . , 2003 ) . JAM-C is not a Zeb1 target but was included because reduction of Pard3a activity reduces JAM-C adhesion and this constitutively active receptor complements CGN adhesion in the absence of Pard3a function ( Famulski et al . , 2010 ) . Prior to the screen , we carefully titrated the quantity of expression vector needed to roughly double each target’s expression in control CGNs to complement Zeb1-mediated target repression ( data not shown ) . We observed diversity in the way individual targets modified the Zeb1 gain-of-function phenotypes in our in vitro and ex vivo assays ( Supplemental Figure 5 and 6; Figure 6—figure supplement 1 , Supplementary file 2 ) . Restored expression of Pard6a , Pard3a , and Chl1 rescued all measured phenotypes to normal levels in CGNs . Individual introduction of each of these downstream Zeb1 targets allowed GNPs to acquire mature CGN status , characterize long neurites , expression of the p27 cell cycle inhibitor , absence of Ki67 labeling or EdU incorporation and GZ exit with subsequent migration to the IGL , even with Zeb1 gain-of-function . Constitutively active Jam-C and the basolateral polarity protein Lin7a did not influence maturation parameters in vitro ( Figure 5 ) ; however , both stimulated cell cycle exit , GZ exit and migration ex vivo ( Figure 6 ) , suggesting that they act non-cell–autonomously in the complex ex vivo environment . Dlg2 stimulated cell cycle exit and p27 expression in all conditions tested but was unable to rescue neurite extension or migration ex vivo . Four genes , Sorl1 , Bhlhe40 , Nfib , and Cdk5r1 , reestablished p27 expression . p27 is known for its cell cycle inhibitory and cytoskeletal regulatory properties; however , p27 expression alone in Sorl1- , Bhlhe40- , Nfib- or Cdk5r1-expressing cells was insufficient to rescue the other features of mature CGNs , such as neurite extension , loss of Ki67 labeling/EdU incorporation , or GZ exit and migration to the IGL . Restored expression of Nfib and Flt1 enhanced neurite extension but failed to rescue the full spectrum of mature CGN features , much like the genes that stimulated p27 expression . Additionally , longer term ex vivo incubations revealed that Cdh1 , Cdk5r1 and Sorl1 were not sufficient to rescue IGL-directed migration of Zeb1 over-expressing cells ( Figure 6—figure supplement 2 ) . Interestingly , Bhlhe40 expression , a negative regulator of EMT could rescue with a 72 hr ex vivo incubation . Time-lapse imaging revealed of cultured neurons revealed that Pard6a- , Pard3a- , and Chl1-rescue also restored two-stroke nucleokinesis and JAM-C adhesion levels , two cell biological outputs of the PAR complex function in maturing CGNs ( Solecki et al . , 2004; Famulski et al . , 2010 ) ( see Videos 3–12 ) . 10 . 7554/eLife . 12717 . 017Figure 5 . Restored expression of Zeb1-Target genes rescues neurite extension and CGN differentiation status in vitro . The rectangular images show representative morphological information and myc-Zeb1 expression; box plot below each quantifies neurite lengths in each experimental condition . On average control cells extended neurites 115 . 4 ± 17 . 7 µm [x¯ ± sd] compared to 55 . 2 ± 2 . 6 µm . Asterisks indicate conditions significantly different to the Zeb1 data as determined by t-test ( p<0 . 01 ) . Images on right show representative Ki67 or p27 labeling , quantified below . Asterisks indicate statistically significant rescue of the Zeb1 phenotype by target expression determined by t-test ( p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01710 . 7554/eLife . 12717 . 018Figure 6 . Restored expression of Zeb1 target genes rescues GNP proliferation , GZ exit and IGL-directed migration in ex vivo cerebellar slices . ( a ) Rectangles show representative P7 EGL slice images assessing GZ exit and IGL-directed migration . Labeled cell ( black ) migrate from the lateral surface ( dashed line ) to the IGL ( to the right ) . Below each image is a cumulative distribution plot of all cells relative to a 450 μm scale . Arrowhead indicates the 99th percentile of the total population . Control cells migrated 74 . 0 ± 8 . 3 μm ( x¯ ± sd ) while Zeb1 migrated 42 . 4 ± 7 . 6 μm . Images at right show representative EdU labeling with% labeling index . A statistically significant rescue of a Zeb1 phenotype in the slice migration assay is indicated by the presence of p>0 . 01 ( t-test mean migration distance vs . control ) . Zeb1 and additional target expression conditions had a p-value < 0 . 01 vs . control indicating GZ was not rescued . Asterix indicates a statistical difference of EdU incorporation between Zeb1 and target expression condition by t-test [both p<0 . 01] ) . Reduced EdU labeling indicates a rescue of elevated proliferation in the Zeb1 gain-of-function condition . b Average migration distance shown in accompanying graph , a Student’s t-test shows rescue conditions ( Pard6a , Pard3a , Chl1 , Jam/Nec , and Lin7a ) with a p value>0 . 01 indicating no statistical difference from the control . Zeb1 alone and Zeb1 plus Dlg2 , Sorl1 , Bhlhe40 , Cdh1 , Nfib , Flt1 or Cdk5r1 migration differences were statistically lower than the control ( t-test p<0 . 01 ) , indicating GZ was not rescued with these targets . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01810 . 7554/eLife . 12717 . 019Figure 6—figure supplement 1 . In depth quantitation of slice migration assays from Figure 6 . P7 EGL was co-electroporated with the indicated expression constructs and H2B-mCherry . After 48 hr of ex vivo culture , CGN migration distance was analyzed in 3 imaging experiments . Red overlay indicates the average migration distribution of control cells ( error bar , SD ) . While control ( 13 , 064 cells , 74 . 0 ± 8 . 3 µm [n , x¯ ± sd] ) cells entered the ML and IGL after 48 hr and Zeb1 over-expressing cells ( 13 , 424 cells , 42 . 4 ± 7 . 6 µm ) remained in the EGL , Addition of Pard6a ( 3 , 886 cells , 79 . 8 ± 5 . 2 µm ) , Pard3a ( 8 , 622 cells , 73 . 9 ± 4 . 5 µm ) , Jam/Nectin ( 11 , 333 cells , 71 . 7 ± 5 . 5 µm ) ) , Chl1 ( 3 . 006 calls , 79 . 4 ± 6 . 9 µm ) to Zeb1-expressing CGNs restored migration , suggesting that the Zeb1 migration phenotype are dependent on key polarity or cell adhesion molecule repression . Note: the criteria for rescue was set if the condition resulted in cell distribution that was 80% similar to the Control distribution ( χ2-test p>0 . 8 ) and the average migration distance was less than 3% similar than the Zeb1 condition . Control vs Zeb1 , [χ2 test] p ( χ2 ) = 1 . 9 x 10–7 , [t-test] p ( t ) = 6 . 44 x 10–5 . Control vs Pard6a p ( χ2 ) = 0 . 96 , p ( t ) = 3 . 68 x 10–6 . Control vs Pard3a p ( χ2 ) = 0 . 99 , p ( t ) = 5 . 67 x 10–5 . Control vs Jam/Nec p ( χ2 ) = 0 . 98 , p ( t ) = 3 . 67 x 10–4 . Control vs Chl1 p ( χ2 ) = 0 . 98 , p ( t ) = 3 . 68 x 10–6 . Control vs Lin7a p ( χ2 ) = 0 . 84 , p ( t ) = 0 . 02 . Control vs Dlg1 p ( χ2 ) = 0 . 25 , p ( t ) = 0 . 06 . Control vs Sorl1 p ( χ2 ) = 0 . 04 , p ( t ) = 0 . 01 . Control vs Bhlhe40 p ( χ2 ) = 0 . 13 , p ( t ) = 0 . 06 . Control vs Cdh1 p ( χ2 ) = 5 . 57 x 10–13 , p ( t ) = 0 . 44 . Control vs Flt1 p ( χ2 ) = 1 . 70 x 10–4 , p ( t ) = 0 . 27 . Control vs Cdk5r1 p ( χ2 ) = 0 . 31 , p ( t ) = 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 01910 . 7554/eLife . 12717 . 020Figure 6—figure supplement 2 . Longer term ex vivo epistasis analysis . P7 EGL was co-electroporated with the indicated expression constructs and H2B-mCherry . After 72 hr of ex vivo culture , CGN migration distance was analyzed for a minimum of 4000 nucleofected cells in each experimental condition . Red overlay indicates the average migration distribution of control cells ( error bar , SD ) . While control cells entered the ML and IGL after 72 hr , Zeb1 over-expressing cells remained in the EGL even with longer-term incubation . Bhlhe40 expression , but not Cdh1 , Cdk5r1 and Sorl1 , significantly restores IGL-directed migration of the context of Zeb1 gain-of-function ( determined by Student t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02010 . 7554/eLife . 12717 . 021Video 3 . Representative time lapse imaging sequence of control CGNs labeled with Centrin2-Venus ( green , centrosome ) and H2B-mCherry ( nucleus ) in a dissociated culture . The migrating cells in the field undergo typical two-stroke nucleokinesis with centrosome entering the leading process prior to somal translocation . Note: even stationary cells extend long neurites . Time stamp= hours: minutes: seconds . Scale bar= 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02110 . 7554/eLife . 12717 . 022Video 4 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs labeled with Centrin2-Venus ( green , centrosome ) and H2B-mCherry ( nucleus ) in a dissociated culture . The migrating cells in the field undergo random amoeboid movements where the centrosome adopts an unpolarized position in the cell body . Note: even stationary cells extend do not extend long neurites . Time stamp = hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02210 . 7554/eLife . 12717 . 023Video 5 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs with restored Pard6a expression labeled with Centrin2-Venus ( green , centrosome ) and H2B-mCherry ( nucleus ) in a dissociated culture . The migrating cells in the field undergo typical two-stroke nucleokinesis with centrosome entering the leading process prior to somal translocation . Note: even stationary cell extend long neurites . Time stamp= hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02310 . 7554/eLife . 12717 . 024Video 6 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs with restored Pard3a expression labeled with Centrin2-Venus ( green , centrosome ) and H2B-mCherry ( nucleus ) in a dissociated culture . The migrating cells in the field undergo typical two-stroke nucleokinesis with centrosome entering the leading process prior to somal translocation . Note: even stationary cell extend long neurites . Time stamp = hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02410 . 7554/eLife . 12717 . 025Video 7 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs with restored Chl1 expression labeled with Centrin2-Venus ( green , centrosome ) and H2B-mCherry ( nucleus ) in a dissociated culture . The migrating cells in the field undergo typical two-stroke nucleokinesis with centrosome entering the leading process prior to somal translocation . Note: even stationary cell extend long neurites . Time stamp= hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02510 . 7554/eLife . 12717 . 026Video 8 . Representative time lapse imaging sequence of control CGNs labeled with JAM-C-pHluorin ( green , adhesions ) and H2B-mCherry ( nucleus ) in a dissociated culture . Note: exuberant cell contacts are observed among most cells . Time stamp= hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02610 . 7554/eLife . 12717 . 027Video 9 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs labeled with JAM-C-pHluorin ( green , adhesions ) and H2B-mCherry ( nucleus ) in a dissociated culture . Sparse cell contacts are observed among most cells . Time stamp = hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02710 . 7554/eLife . 12717 . 028Video 10 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs with restored Pard6a expression labeled with JAM-C-pHluorin ( green , adhesions ) and H2B-mCherry ( nucleus ) in a dissociated culture . Note: note cell contacts are observed among most cells . Time stamp = hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02810 . 7554/eLife . 12717 . 029Video 11 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs with restored Pard3a expression labeled with JAM-C-pHluorin ( green , adhesions ) and H2B-mCherry ( nucleus ) in a dissociated culture . Note: restored cell contacts are observed among most cells . Time stamp= hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 02910 . 7554/eLife . 12717 . 030Video 12 . Representative time lapse imaging sequence of Zeb1 over-expressing CGNs with restored Chl1 expression labeled with JAM-C-pHluorin ( green , adhesions ) and H2B-mCherry ( nucleus ) in a dissociated culture . Note: restored cell contacts are observed among most cells . Time stamp= hours: minutes: seconds . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 030 Given that Pard6a and Chl1 were among the targets whose restoration most potently rescued Zeb1 gain-of-function phenotypes , we sought mechanistic insight into this rescue by further characterizing expression of key factors in proliferating GNPs , D type Cyclins and Atoh1 . Restored Pard6a and Chl1 expression did not affect the levels at which Zeb1 suppressed its target genes , indicating that Pard6a and Chl1 did not counteract Zeb1 at the transcriptional level or non-specifically reduce Zeb1 target repression in our assay system ( Figure 7a ) . Moreover in the case of Chl1 protein , restored Pard6a and Pard3 did not rescue Chl1 expression as assayed by immunocytochemistry in dissociated CGNs ( Figure 7—figure supplement 1 ) . Restored Pard6a or Chl1 expression reduced Zeb1-mediated activation of CyclinD1 and CylinD2 mRNA and Atoh1 protein levels , all of which are required to maintain GNPs in the undifferentiated state ( Figure 7b and c ) ( Ayrault et al . , 2010; Flora et al . , 2009; Huard et al . , 1999 ) . The broad rescue of Zeb1 gain-of-function phenotypes by the Pard6a and Pard3a polarity proteins and the Chl1 adhesion molecule demonstrates that these Zeb1-supressed targets are prerequisites for mature CGN characteristics . These findings also reinforce the parallel between CGN differentiation and polarity regulation in cells of epithelial origin . Not only do Zeb1 and polarity proteins show mutually exclusive expression in GNPs and CGNs , but the functional screen also shows that their functional antagonism regulates the balance between the GNP and CGN states . 10 . 7554/eLife . 12717 . 031Figure 7 . Pard6a and Chl1 rescue neuronal differentiation in the Zeb1 gain-of-function context . Cultured CGNs were nucleofected with a marker plasmid encoding H2B mCherry ( or Centrin2-Venus in Panel c ) alone or in combination with plasmids encoding Myc-Zeb1 plus single plasmids encoding Pard6a or Chl1 in our array expression screen . After 24 hr in culture , nucleofected cells were FACS sorted to isolate mRNA ( a , b , c ) or stained with antibodies to highlight morphology/Atoh1 expression ( c ) . a . qRT-PCR analyses shows that: 1 ) Pard6a and Pard3a expression continues to be suppressed in Chl1 rescued GNPs and 2 ) Pard3 and Chl1 expression continues to be suppressed by Zeb1 Pard6a rescued GNPs . NS = not shown . ( b ) qRT-PCR analyses shows that Zeb1 gain-of-function induced CyclinD1 and CyclinD2 mRNA expression and that both restored expression of Chl1 and Pard6a reduces D-type cyclin expression . ( c ) qRT-PCR analyses shows that Zeb1 gain-of-function mildly induced Atoh1 mRNA expression . While Chl1 and Pard6a rescue have little affect on Atoh1 mRNA expression , restored expression of both these genes strongly reduce Atoh1 protein expression detected by immunocytochemistry . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 03110 . 7554/eLife . 12717 . 032Figure 7—figure supplement 1 . Immunocytochemical analysis of Chl1 expression in Control , Zeb1-expressing or Pard6a and Pard3 rescued CGNs . Dissociated CGNs were prepared and nucleofected with the indicated expression constructs . 18 hr post-nucleofection , cultures were fixed and stained with antibodies recognizing EGFP and Chl1 . Control neurons express robust levels of Chl1 protein in their somas and proximal leading process . In contrast , Zeb1 expressing as well as Pard6a or Pard3 rescued cells expressed lower amounts of Chl1 immunoreactivity . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 032 Having found that Zeb1 controls GNP differentiation and GZ exit by regulating neuronal polarity and adhesion , we next sought to identify factors that can regulate Zeb1 in GNPs . We reasoned that SHH , the required mitogen for GNP proliferation , may regulate Zeb1 expression given that it not only stimulates progenitor proliferation but also blocks CGN differentiation ( Wechsler-Reya and Scott , 1999 ) . GNP cultures treated with SAG , a potent small-molecule SHH agonist , displayed not only elevated Zeb1 but also decreased Pard6a and Chl1 proteins ( Figure 8a , b ) . These results suggest that Zeb1 and some of its targets act downstream of the SHH signaling cascade . 10 . 7554/eLife . 12717 . 033Figure 8 . Zeb1 expression is linked to SHH signaling , and restoring polarity of Ptch1-deficient GNPs rescues GZ exit . ( a ) GNPs were cultured in the presence or absence of SAG , a small-molecule agonist of SHH , fixed and stained for DAPI ( blue ) , Zeb1 ( red ) or the Zeb1 targets Pard6a , Chl1 and Dlg2 . Zeb1 expression was maintained , but Zeb1 target expression diminished . ( b ) Western blotting with anti-Zeb1 confirmed that Zeb1 expression was maintained in the presence of SAG . Fibrillarin was loading control ( t-test , p<0 . 01 ) . ( c ) Immunohistochemistry shows maintained expression of Zeb1 ( red ) in a Ptch1+/- , Cdkn2c-/- SHH-type mouse MB; Zeb1 expression is complementary to Tuji1 staining ( green ) . ( d ) qRT-PCR of mRNA from Ptch1+/- , Cdkn2c-/- mouse MBs shows much higher Zeb1 mRNA expression in MB cells than in P7 GNPs . Most of the targets in our screen are expressed at a lower level in SHH MB than in P7 GNPs . ( e ) Zeb1 mRNA expression in 4 MB subgroups . Data set includes 74 MBs ( WNT n = 8; SHH n = 11; G3 n = 17; G4 n = 38 ) profiled on the Affymetrix U133plus2 array . ( f ) The migration distance of CGNs ( black dots ) from the pial layer ( dashed line ) was analyzed ( n = 8 , 800 to 11 , 300 cells ) . Control cells expressing catalytically inactive Cre enter the ML and IGL ( 71 . 2 ± 7 . 8 µm [x¯ ± sd] ) , while Ptch1-deficient GNPs expressing wild-type Cre remain within the EGL ( 41 . 4 ± 5 . 8 µm ) . Zeb1 silencing and restored expression of Pard6a , Chl1 and Lin7a rescued the defective GZ exit ( Asterisks indicate conditions where rescue observed ( χ2 test vs Cre mutant , p>0 . 8; t-test vs Cre WT , p<0 . 01 ) . Below each image is a cumulative distribution plot showing the area relative to a 450 μm scale . Arrowhead indicates 99th population percentile . ( g ) Average migration distance shown in accompanying graph , a Student’s t-test shows rescue conditions with a p value <0 . 01 vs Cre wild type . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 03310 . 7554/eLife . 12717 . 034Figure 8—figure supplement 1 . In depth quantitation of slice migration assays from Figure 8 . P7 EGL of Ptch1 flox/flox animals was co-electroporated with the indicated expression constructs and H2B-mCherry . After 48 hr of ex vivo culture , CGN migration distance was analyzed in 3 imaging experiments . Red overlay indicates the average migration distribution of control cells ( error bar , SD ) . While Cre Mutant ( n = 9 , 471 cells ) cells entered the ML and IGL after 48 hr and Cre wild type ( n = 9 , 872 ) over-expressing cells remained in the EGL , Zeb1 silencing ( n = 11 , 383 ) or addition of Pard6a ( n = 8 , 839 ) , Lin7a ( n = 10 , 543 ) , or Chl1 ( n = 11 , 348 ) to Zeb1-expressing CGNs restored migration , suggesting that the GZ exit phenotype Ptch1 deficient GNPs is dependent on Zeb1 repression of its targets . Average migration distance shown in accompanying graph . Note: the criteria for rescue was set if the condition resulted in cell distribution that was 80% similar to the Cre Mut distribution ( χ2- test p>0 . 8 ) and the average migration distance was less than 3% similar than the Cre WT condition ( t-test p<0 . 03 ) . Cre Mut vs Cre WT , [χ2 test] p ( χ2 ) = 2 . 86 x 10–3 , [t-test] p ( t ) = 3 . 03 x 10–3 . Controls vs Zeb1 shRNA p ( χ2 ) = 0 . 48 , p ( t ) = 0 . 03 . Controls vs Pard6a p ( χ2 ) = 1 . 00 , p ( t ) = 0 . 01 . Controls vs Chl1 p ( χ2 ) = 0 . 86 , p ( t ) = 4 . 78 x 10–3 . Controls vs Lin7a p ( χ2 ) = 1 . 00 , p ( t ) = 1 . 99 x 10–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 12717 . 034 The SHH pathway is activated in both mouse and human MBs derived from GNPs ( Pomeroy et al . , 2002; Lee et al . , 2003; Robinson et al . , 2012; Northcott et al . , 2012 ) . As SHH activation led to elevated Zeb1 in normal GNPs , we next examined the expression levels of Zeb1 and its targets in a mouse SHH MB model from Ptch1 +/- , Cdkn2c -/- mice in which SHH signaling is constitutively activated ( Uziel et al . , 2005 ) . Unlike normal P15 cerebellum ( Figure 1b ) , MBs from adult Ptch1+/- , Cdkn2c-/- mice displayed high levels of Zeb1 expression ( Figure 8c ) . MBs contain subpopulations of cells that can proceed with neuronal differentiation . Zeb1 expression was complementary with that of class III beta-tubulin/Tuj1 , an early neuronal differentiation marker , indicating that Zeb1 expression is extinguished in both normal and tumor-derived cells proceeding toward the differentiated phenotype . We also quantified the RNA expression of Zeb1 and its targets by qRT-PCR in normal GNPs and in mouse SHH MBs . Mouse MBs contained higher levels of Zeb1 RNA than GNPs purified at P7 , the time of peak Zeb1 expression ( Figure 8d ) . Moreover , most Zeb1 targets identified in our Affymetrix Gene Chip array were expressed at lower levels in mouse MBs than in P7 GNPs , with the sole exception of Pard3a ( Figure 8d ) . To broaden our analysis outside of mouse MB , we quantified ZEB1 RNA in human MB samples ( Robinson et al . , 2012 ) . ZEB1 RNA was about four times higher in the human SHH MB subgroup compared to WNT , Group3 and Group4 MBs ( Figure 8e ) . These results indicate that in mouse and human MB , Zeb1 expression is elevated when the SHH pathway is activated , supporting the link we observed between SHH and Zeb1 in normal GNPs . Elevated Zeb1 expression paralleled reduced expression of the targets identified in our Zeb1 gain-of-function expression profiling , validating our findings in primary GNPs . Pre-neoplastic GNPs show a greatly delayed GZ exit , the first overt phenotype observed in mouse MB models with chronic SHH activation ( Ptch1+/-; Ptch1+/- , Cdkn2c-/-; and Ptch1Floxed mice ) ( Goodrich , 1997; Yang et al . , 2008; Uziel et al . , 2005 ) . While there is a firm link between proliferation and delayed differentiation in pre-neoplastic GNPs , it is unknown how deregulated SHH signaling delays GZ exit . Given that Zeb1 controls GNP differentiation and GZ exit and that its expression is linked with elevated SHH signaling , we postulated that Zeb1 function , and its transcriptional repression of polarity genes , may be related to the GZ exit phenotypes of GNPs with an activated SHH pathway . We developed an ex vivo model to examine the GZ exit status of GNPs exposed to chronic SHH stimulation: we electroporated vectors encoding codon-optimized Cre recombinase or its inactive mutant into P7 cerebellar EGL from mice homozygous for Ptch1 harboring loxP sites flanking exons 8–9 ( Ptch1 flox/flox mice ) ( Ellis et al . , 2003 ) . As the Ptch1 receptor is a negative regulator of SHH signaling , conditional Ptch1 deletion leads to potent constitutive activation of the pathway and , over a longer time , GNP malignant transformation . GNPs expressing Cre recombinase remained largely within the EGL , but migration was unaltered by a catalytically inactive mutant ( x¯ distance = 41 . 4 ± 5 . 8 µm vs . 71 . 1 ± 7 . 8 µm; Figure 8f , Figure 8—figure supplement 1 ) . To examine Zeb1 and Zeb1-target function in the GZ exit phenotype of Ptch1-deficient GNPs , we co-electroporated P7 EGL from Ptch1 flox/flox mice with Cre recombinase and an shRNA silencing Zeb1 or vectors encoding Pard6a , Chl1 and Lin7a , which were expressed at low levels in the mouse Zeb1-expressing MB cells . Zeb1 silencing or increased Pard6a , Chl1 and Lin7a expression restored GZ exit and migration to the IGL to near wild-type levels . Taken together , these results show that Zeb1 is functionally required downstream of SHH signaling to control GZ exit . Moreover , the MET-like transition that occurs in CGN differentiation is evident not only during normal development but also in an ex vivo model of pathological GZ exit implicated in cerebellar tumorigenesis .
Throughout the developing brain , newborn neurons are similarly challenged to depart their GZ niche and integrate into a functional circuit ( Hatten , 2002; Itoh et al . , 2013 ) , and at each stage of their differentiation these cells must undergo reorganization of their polarity ( de la Torre-Ubieta and Bonni , 2011; Barnes et al . , 2008 ) . While radial glial cells , migrating neurons and neurons elaborating axons or dendrites display a polarized morphology , transiently amplifying progenitors and newly delaminated neurons are temporarily less polarized . Conceptual parallels have been made between epithelial and neuronal polarity ( Colman , 1999 ) . Recently , Foxp- or Scratch-mediated inhibition of classical cadherins was shown to spur neuronal AJ loss , transition away from radial glial polarity , and delamination from the VZs of the spinal cord and cortex ( Rousso et al . , 2012; Itoh et al . , 2013 ) . The parallel between Foxp- and Scratch-mediated delamination of neurons and EMTs in epithelial cells is incomplete , as both delamination events may occur in postmitotic neuronal progeny . Also , we still have no clear idea how immature neurons or their progenitors transition out of their low polarity states during terminal differentiation . Our work demonstrates that transiently amplifying cerebellar progenitors display mesenchymal characteristics , expressing high levels of Zeb1 and low levels of polarity proteins and adhesion molecules needed for maturation to CGNs . How similar are CGN differentiation and MET ? As illustrated in our model ( Figure 9 ) , METs are associated with acquisition of a mature , polarized morphology . Zeb1 locks GNPs into an immature morphology , just as it blocks apical-basal polarization in epithelia . Second , a common MET pattern is extensive migration followed by a final integrative positioning event ( Thiery and Sleeman , 2006 ) . At the population level , GNPs migrate to cover the cerebellar anlage , migrate within the EGL , and finally undergo differentiative migration to the IGL . Our results show that Zeb1 is necessary and sufficient to confine GNPs to their GZ niche , where migration is restricted to the cerebellar surface . Finally , METs involve a changing balance of cell-matrix and cell-cell contacts , in which mesenchymal cells engage in extracellular matrix adhesions and differentiated epithelial cells engage in cell-cell adhesions ( Nelson , 2009 ) . Similarly , early electron microscopy studies showed that GNPs remain largely contiguous with the matrix-rich pial basal lamina until they differentiate ( Hausmann and Sievers , 1985 ) and develop extensive cell-cell contacts ( Rakic , 1971; Del Cerro and Snider , 1972 ) . Interestingly , we found that promotion of cell-cell contact with constitutively active JAM-C and restored Chl-1 expression rescues Zeb1 gain-of-function phenotypes . One key difference between GNP differentiation and epithelial polarity is the mir200 class of micro RNAs that inhibits Zeb1 expression in epithelial cells are not expressed in CGNs ( Uziel et al . , 2009 ) . Overall , CGN differentiation , which is accompanied by downregulation of Zeb1 , enhanced Zeb1 target expression , morphological maturation and GZ exit , bears remarkable similarity to the METs of epithelial cells as they incorporate into epithelial tissues . At the moment , it is unclear if additional EMT regulatory transcription factors behave similarly in GNP differentiation . While Zeb1 was clearly the highest expressed EMT regulatory factor relative to 18S RNA , necessity and sufficiency testing was not performed on low abundance genes like Snai1 or Snai2 . We anticipate that MET associated with Zeb1 downregulation is also relevant to other brain regions . Both GNPs and cortical intermediate progenitors have delaminated from a parental radial glia , amplify transiently in a displaced GZ ( EGL vs SVZ ) , express some similar markers ( Tbr2 , Id proteins , Tis21 , Zeb1 ) , and assume a simple morphologic form before differentiation . Our ChIP-seq studies show that Zeb1 occupies the promoters of polarity genes in mouse neural stem cells with telencephalic features , raising the possibility that Zeb1 may regulate the polarity of telencephalon cells . We observed that Zeb1 inhibits GNP expression of the GTPases Rnd1 and Rnd3 ( data not shown ) , which promote VZ delamination , inhibit intermediate progenitor proliferation and enhance multipolar to bipolar transition in the neocortex , much as Zeb1 targets function in GNPs ( Heng et al . , 2008; Pacary et al . , 2011 ) . Neuronal polarity regulation by Zeb1 differs from the mechanisms described in forebrain and cerebellar neurons . Neuronal polarization in the hippocampus and cortex depends on the balance of cues and signaling from extracellular , intracellular and cytoskeletal sources that shape forming axons or dendrites ( Lewis et al . , 2013 ) . Transcriptional control mechanisms involving FOXO , SnoN1/2 , NeuroD1 and NeuroD2 have been found to promote discrete stages of morphological CGN maturation , illustrating the partial dependence of axon-dendrite morphogenesis on competence that develops during differentiation ( de la Torre-Ubieta and Bonni , 2011 ) . Our findings show a new level of regulation of the onset of neuronal polarity in which active gene expression programs in neuronal progenitors cells can delay their competence to polarize . Thus , transiently amplifying progenitors are unpolarized not only because they do not yet express intrinsic maturation components but also because they express factors , like Zeb1 , that restrain their polarization . CGNs offer not only a model of neural development but also an excellent system to study the dysregulation of signaling pathways in disease . The best example is the link between SHH signaling , GNP proliferation , and MB tumorigenesis . Humans with activating mutations in the SHH pathway are genetically predisposed to MBs that bear many similarities to GNPs ( Raffel et al . , 1997; Lam et al . , 1999; Taylor et al . , 2002 ) . Available mouse models can recapitulate SHH-associated MB ( Goodrich , 1997; Yang et al . , 2008; Uziel et al . , 2005 ) . During cerebellar development , GNPs stream from the rhombic lip to cover the cerebellar anlage , expand clonally in the EGL in response to Purkinje cell-derived SHH , then exit mitosis and their GZ niche and migrate inward to the IGL ( Hatten et al . , 1997 ) . When the SHH signaling pathway is deregulated in vivo , cohorts of GNPs fail to exit their GZ niche and continue to proliferate on the cerebellar surface well past the normal interval ( Goodrich , 1997 ) . Although migration from the mitotic niche is proposed to be linked to GNP cell cycle exit ( Choi et al . , 2005 ) , the specific downstream GZ exit or migration mechanisms are unknown . Our finding that SHH maintains Zeb1 expression and that Zeb1 target expression is reduced in MB reveals an antagonism between the main GNP mitogen and the polarity required for GZ exit . This antagonism suggests that SHH inhibits the MET-like event we showed to control GNP GZ exit and that pre-neoplastic GNPs or MB cells are inherently polarity-deficient . The possibility that Zeb1 controls an active program to block polarization is particularly relevant to MB . These tumor cells express high levels of the FOXO and NeuroD transcription factors that promote CGN polarization , but they are insufficient to induce polarization of transformed GNPs . Thus , Zeb1 is a candidate factor that may act downstream of SHH in MB to counteract the polarization program . Finally , our results suggest future studies to determine whether restoring the polarity balance in MB will yield therapeutic benefit as a complement to existing first line- or targeted therapies . In Ptch1-deficient , Zeb1-overexpressing GNPs , restored expression of selected Zeb1 targets rescues CGN differentiation , GZ exit and migration to the IGL . How do the targets , such as the PAR complex and Chl1 , promote these events ? In the context of Zeb1 gain-of-function , Pard6a and Chl1 expression reduced Zeb1 activation of CyclinD1 , CyclinD2 , and Atoh1 , each of which is essential to maintain GNP proliferation ( Ayrault et al . , 2010; Flora et al . , 2009; Huard et al . , 1999 ) . Thus , Pard6a and Chl1 appear to cell-intrinsically promote CGN differentiation . Consistent with this hypothesis , Pard6a and Chl1 gain-of-function in normal GNPs spurs precocious germinal zone exit ( data not shown ) . In preliminary time-lapse imaging studies , Pard6a , Pard3a and Chl1 also rescued two-stroke motility and JAM-C adhesion levels ( see Videos 3–12 ) . While it is intriguing that PAR complex and Chl1 behave similarly in our functional genomics screen , further studies are necessary to clarify their potential functional interactions . Finally , an additional area of further investigation is the cooperation between transcriptional and post-transcriptional mechanisms for polarity regulation . While Pard3a is clearly transcriptionally repressed by Zeb1 , it’s mRNA does not display the same elevation displayed by other targets after Zeb1 expression diminishes . Interestingly , Pard3a protein expression levels is controlled by the Siah2 E3 ubiquitin ligase , thus regulation of Pard3a expression may be due to a complex interplay between transcriptional and post-translation mechanisms . In conclusion , further examination of Zeb1 function in neural progenitors and its relation to other GZ exit pathways and the MET-like conceptual model may be useful not only in understanding how normal GNPs transition to the CGN state , but also in understanding the pathogenesis of pediatric cancers linked to defective GZ exit .
All mouse lines were maintained in standard conditions in accordance with guidelines established and approved by Institutional Animal Care and Use Committee at St . Jude Children’s Research Hospital ( protocol number=483 ) . B6N . 129-Ptch1tm1Hahn/J strain mice were obtained from Jackson labs . All cDNAs encoding protein of interest were commercially synthesized and subcloned into pCIG2 by Genscript ( Piscataway , NJ , USA ) . Expression plasmid for Pard3a , Pard6a , Jam-C-Nectin3 and Fluorescent fusion proteins such as pCIG2 H2B-mCherry , pCIG2 RFP-UTRCH , pCIG2 Centin2-Venus and pCIG2 JAM-C-pHluorin were subcloned as previously described ( Solecki et al . , 2009 ) . CGNs were prepared as described ( Hatten , 1985 ) . Briefly , cerebella were dissected from the brains of P7 mice and pial layer removed; the tissue was treated with trypsin/DNase and triturated into a single-cell suspension using fine-bore Pasteur pipettes . The suspension was layered onto a discontinuous Percoll gradient and separated by centrifugation . The small-cell fraction was then isolated . The resulting cultures routinely contain 95% CGNs and 5% glia . For imaging experiments , expression vectors encoding fluorescently labeled cytoskeletal proteins and pCIG2 expressing protein of interests were introduced into granule neurons via Amaxa nucleofection , using the Amaxa mouse neuron nucleofector kit per the manufacturer's instructions and program A030 . The concentration pCIG2 expression vectors used was determined such that increase in protein expression was at least two fold . After cells recovered for 10 min from the nucleofection , they were plated in either plated in 16 well slides for IHC or in movie dishes ( Mattek ) coated with low concentrations of poly-L-ornithine to facilitate the attachment of neurons to glial processes ( according to methods established by ( Edmondson and Hatten , 1987 ) Chromatin immunoprecipitation ( ChIP ) was performed by using EZ ChIP reagents ( Millipore ) in the presence of phosphatase and protease inhibitors according to the manufacturer's instructions . Briefly , chromatin from CGNs ( ≥ 1 × 106 ) was cross-linked for 10 min at RT with 1% formaldehyde , sonically disrupted , diluted and precleared before immunoprecipitation with either 5 µg of Zeb1 antibody or rabbit IgG as control at 4°C overnight . Protein G-agarose beads ( 60 μL/sample ) were added and incubated for a further 1 hr at 4°C . After washing with salt gradient stringent buffers , LiCl and TE buffers , immunoprecipitated protein-DNA complexes were eluted in 200 μL of elution buffer ( 50 mmol/L NaHCO3 , 1% SDS ) . Formaldehyde crosslinking was then reversed by adding 8 μL of 5 mol/L NaCl and incubating at 65ºC overnight . RNA and protein were removed by sequential treatment with RNase for 30 min at 37°C and proteinase K at 45ºC for 2 hr , respectively . Purified DNA fragments were then analysed with qRT-PCR using specific primer for the promoter region see Table is Supplementary file 3B and SYBR Green PCR Master Mix ( Applied Biosystems ) . The results were normalised against the input control . Normalised data of three independent experiments were averaged and are presented using fold change/enrichment of each promoter region expressed as a ratio of PCR signal of samples to that of input . For example , fold increase of promoter binding is defined as the ratio of Zeb1 binding DNA compared to DNA precipitated with the IgG control antibody ( set as a fixed value of 1 . 0 ) . NS5 cells ( Pollard et al . , 2006 ) were fixed sequentially with di ( N-succimidyl ) glutarate and 1% formaldehyde in phosphate buffered saline ( PBS ) and then lysed , sonicated and immunoprecipitated with anti-Zeb1 antibody ( HPA027524 , Sigma ) , as previously described ( Castro et al . , 2011 ) . DNA libraries were prepared from 10 ng of immunoprecipitated DNA according to the standard Illumina ChIP-seq protocol and sequenced with Illumina GAIIx . Sequenced reads were processed after mapping with SAMTools for format conversion and removal of PCR duplicates ( Li et al . , 2009 ) and mapped to the mouse genome ( NCBI37/mm9 ) with Bowtie 0 . 12 . 7 ( Langmead et al . , 2009 ) , resulting in 25 million uniquely mapped reads ( ArrayExpress accession number: E-MTAB-3560 ) . Peak calling was performed with MACS 1 . 4 . 1 Zhang et al . , 2008 ) ( default parameters ) . Profiles of genomic regions were generated using D-peaks source code ( Brohée et al . , 2012 ) . A de novo search for motifs enriched at peak summits was done with Cisfinder ( Sharov and Ko , 2009 ) using default parameters and a background control set of 100 bp genomic regions located 3Kp upstream input regions . Calculation of P-values for the association between binding events and deregulated genes was performed by sampling the number of genes represented in the microarray 1000 times and assuming a normal distribution . Annotation of binding events and association with genomic features was performed with PeakAnalyzer ( Salmon-Divon et al . , 2010 ) and the R/Bioconductor package ChIPpeakanno ( Zhu et al . , 2010 ) . CGN cultures were imaged with a Marianas Spinning Disk Confocal Microscope ( Intelligent Imaging Innovations ) comprising a Zeiss Axio Observer microscope equipped with 40×/1 . 0 NA ( oil immersion ) and 63×/1 . 4 NA ( oil immersion ) PlanApochromat objectives . An Ultraview CSUX1 confocal head with 440 to 514 nm or 488/561/642 nm excitation filters and ImageEM-intensified CCD camera ( Hamamatsu ) were used for high-resolution imaging . Neurite length measurements were performed using the ruler function of SlideBook software ( Intelligent Imaging Innovations ) by measuring the longest neurite from one end to the longest neurite on the opposite end . At least three independent biological replicates were done for each target gene . While measuring neurite length in the rescue experiments , only CGNs that showed at least two fold increase in Zeb1 expression were included for neurite measurement . Data was statistically analyzed using Microsoft Excel and graphed using Kaleidagraph v4 . 03 . P7 cerebella were dissected , soaked in endotoxin-free plasmid DNA suspended in Hanks balanced salt solution ( 1–5 μg/μL of each DNA was generally used , pCIG2-mCherryH2B was electroporated as a nuclear marker for migrating CGNs ) , transferred to a CUY520-P5 platinum block petri dish electrode ( Protech International ) and electroporated with a CUY21EDIT ( Protech International ) square wave electroporator ( 80 V , 5 pulses , 50 ms pulse , 500 ms interval ) . Electroporated cerebella were embedded in 4% low melting point agarose and 250 μm sagittal cerebellar slices were prepared using a VT1200 Vibratome ( Leica Microsystems ) . Slices were transferred to Millicell tissue culture inserts ( Millipore ) and cultured in basal Eagle medium supplemented with 2 mM L-glutamine , 0 . 5% glucose , 50 U/ml penicillin-streptomycin , 1x B27 and 1x N2 supplements ( Invitrogen ) at the air-media interface for the times indicated in the Figures . In experiments that assayed proliferation , 25 μM EdU was added to culture medium and EdU incorporation was assayed by using the Click-iT assay as per manufacturer instructions ( Invitrogen ) . Previous characterization of this method show that greater than 97% of cell manipulated by this method are Pax6 positive CGNs in outer EGL ( Famulski et al . , 2010 ) . For analysis of fixed specimens , slices were fixed 4% paraformaldehyde after 24 or 48 hr of culture and mounted on slides by using ProLong Gold ( Invitrogen ) . Migration distance was measured in fixed slices by measuring the distance between the cerebellar surface and center of individual cell nuclei marked by mCherry-H2B . Central coordinates were exported from SlideBook ( Intelligent Imaging Innovations ) into IGOR Pro ( WaveMetrics Inc . ) , where the distance of cells from the nearest cerebellar surface was measured and logged . Statistical analysis used Microsoft Excel and was graphed by using Kaleidagraph v4 . 03 . For live-imaging analysis of the migration of H2B-mCherry labeled CGNs , slice cultures were transferred at 28 hr to the humidified chamber of the spinning disk confocal microscope described above . Z-stacks ( 60–80 μm width , ~20 sections per stack ) were collected at multiple x , y stage positions every 15 min for 24–48 hr . | During the formation of the brain , developing neurons are faced with a logistical problem . After newborn neurons form they must change in shape and move to their final location in the brain . Despite much speculation , little is known about these processes . Neurons mature via the activity of several pathways that control the activity , or expression , of the neuron’s genes . One way of controlling such gene expression is through proteins called transcription factors . At the same time , the developing neurons go through a process called polarization , where different regions of the cell develop different characteristics . However , it was not known how the maturation and polarization processes are linked , or how the developing neurons actively regulate polarization . By studying the developing mouse brain , Singh et al . found that a transcription factor called Zeb1 keeps neurons in a immature state , stopping them from becoming polarized . Further investigation revealed that Zeb1 does this by preventing the production of a group of proteins that helps to polarize the cells . The most common type of malignant brain tumour in children is called a medulloblastoma . Singh et al . analyzed the genes expressed in mice that have a type of medulloblastoma that results from the constant activity of a gene called Sonic Hedgehog in developing neurons . This revealed that these tumour cells contain abnormally high levels of Zeb1 , and so do not take on a polarized form . However , artificially restoring other factors that encourage the cells to polarize caused the neurons to mature normally . Further investigation is now needed to find out whether the activity of the Sonic Hedgehog gene regulates Zeb1 activity , and to discover whether inhibiting Zeb1 could prevent brain tumours from developing . | [
"Abstract",
"Introduction",
"Result",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Zeb1 controls neuron differentiation and germinal zone exit by a mesenchymal-epithelial-like transition |
No evidence has shown whether insect-borne viruses manipulate the c-Jun N-terminal kinase ( JNK ) signaling pathway of vector insects . Using a system comprising the plant virus Rice stripe virus ( RSV ) and its vector insect , the small brown planthopper , we have studied the response of the vector insect’s JNK pathway to plant virus infection . We found that RSV increased the level of Tumor Necrosis Factor-α and decreased the level of G protein Pathway Suppressor 2 ( GPS2 ) in the insect vector . The virus capsid protein competitively bound GPS2 to release it from inhibiting the JNK activation machinery . We confirmed that JNK activation promoted RSV replication in the vector , whereas JNK inhibition caused a significant reduction in virus production and thus delayed the disease incidence of plants . These findings suggest that inhibition of insect vector JNK may be a useful strategy for controling the transmission of plant viruses .
Most plant viruses depend on sap-feeding insects from the order Hemiptera for their transmission ( Hogenhout et al . , 2008 ) . The success and efficiency of virus transmission , especially for the persistent-propagative viruses , depend on specific interactions between the virus and proteins of the vector insects , allowing transport of the virus in and out of insect tissues and overcoming insect immune reactions for successful replication ( Blanc et al . , 2014 ) . A detailed knowledge of the vector proteins that participate in viral transmission and replication could provide important clues for the development of control strategies to plant viruses . Rice stripe virus ( RSV ) , a member of the genus Tenuivirus , causes rice stripe disease , one of the most notorious rice diseases in temperate and subtropical East Asia ( Toriyama , 1986 ) . RSV is transmitted mainly by the small brown planthopper , Laodelphax striatellus , in a persistent , circulative-propagative manner ( Toriyama , 1986 ) . As with most persistent-propagative viruses , RSV cannot be transmitted directly between plants in the field , and there is a large difference in pathogenicity of the virus from the insect vectors and from the viruliferous plants ( Zhao et al . , 2016b ) . The viral genome consists of four single-stranded RNA segments and encodes seven proteins: RNA-dependent RNA polymerase , NS2 ( RNA silencing suppressor ) , NSvc2 ( putative membrane glycoprotein ) , NS3 ( gene-silencing suppressor ) , CP ( capsid protein ) , SP ( nonstructural disease-specific protein ) , and NSvc4 ( movement protein ) ( Cui et al . , 2016; Du et al . , 2011 ) . CP is the most abundant viral protein and plays a crucial role in the retention and movement of the virus in the insect vector ( Blanc et al . , 2014 ) . Direct interaction between insect vector proteins and RSV proteins has been established , but knowledge about the functions of these vector proteins is limited . Numerous proteins from small brown planthopper interact with CP of RSV , such as atlasin , cuticular protein CPR1 , jagunal , NAC domain protein , ribosomal proteins RPL5 , RPL7a and RPL8 , and vitellogenin ( Huo et al . , 2014; Li et al . , 2011; Liu et al . , 2015 ) . There is evidence that cuticular protein CPR1 and vitellogenin , but not other vector proteins , affect RSV accumulation and transmission by binding with viral CP ( Huo et al . , 2014; Liu et al . , 2015 ) . RSV NS3 protein is able to attenuate the 26S proteasome- mediated host defense response by interacting directly with the planthopper RPN3 protein ( Xu et al . , 2015 ) . G protein pathway suppressor 2 ( GPS2 ) is a small protein that was initially found to inhibit G protein ( Ras ) -activated MAPK ( mitogen-activated protein kinase ) signaling , especially JNK ( c-Jun N-terminal kinase ) , in yeast and mammalian cells ( Spain et al . , 1996 ) . The JNK signaling pathway can be activated by environmental stress and is implicated in multiple physiological processes , such as cell apoptosis/survival signaling , tumor development , diabetes , metabolism , embryonic development and aging ( Weston and Davis , 2002 ) . GPS2 in mammals is bifunctional . In the cell nucleus , GPS2 is a transcriptional cofactor that mediates both gene repression and activation as an intrinsic component of a major transcriptional repressor complex , the NcoR/SMRT nuclear receptor corepressor complex ( Zhang et al . , 2002 ) . In the cytoplasm , GPS2 has a function in the regulation of JNK activation by inhibiting TRAF2/Ubc13 enzymatic activity in response to a proinflammatory cytokine , Tumor Necrosis Factor-α ( TNF-α ) ( Cardamone et al . , 2012 ) . JNK is activated through phosphorylation of threonine and tyrosine residues within a Thr-Pro-Tyr motif located in kinase subdomain VIII . The activation of JNK requires the presence of the E2 ubiquitin-conjugating enzyme Ubc13 , the assembly of K63-ubiquitin chains , and the integrity of the TRAF2 RING domain . GPS2 interacts with the TRAF2/Ubc13 K63 ubiquitylation machinery , translating into a significant decrease of JNK activation ( Cardamone et al . , 2012 ) . The functions of GPS2 in invertebrates have not yet been determined . Because viruses are obligatory cellular parasites , viral proteins can regulate host cellular signaling pathways . In mammals , hepatitis B virus ( Doria et al . , 1995 ) , herpesvirus ( Jung and Desrosiers , 1995 ) , and human T-cell lymphotrophic virus ( Jin et al . , 1997 ) have been shown to affect Ras-dependent kinase cascades . In a compatible vector–virus relationship , such as vector insects and their transmitted plant or mammal viruses , how the viruses regulate the insect's Ras-dependent kinase cascadesand what benefits the viruses would get from this manipulation remain open questions . It has been reported that the JNK-like protein regulates phagocytosis and endocytosis in the Aedes albopictus mosquito cell line , C6/36 , and West Nile virus infection may depend on JNK-mediated endocytosis ( Mizutani et al . , 2003b ) . In this study , we explored the effect of RSV on the JNK signaling pathway in the virus's vector insect . We found that RSV activates the JNK signaling pathway in several ways , especially through the interaction of CP and GPS2 , to secure a benefit in viral replication and disease outbreak in plants .
To detect virus interactive proteins in the small brown planthoppers , we used RSV CP as bait to screen the insect cDNA library in a yeast two-hybrid system . A 1056 bp fragment encoding partial GPS2 was isolated on the SD/–Leu/–Trp/–His medium , but not on the SD/–Leu/–Trp/–His/–Ade medium . The interaction between CP and partial GPS2 was confirmed by further analysis with the yeast two-hybrid system . Yeast was co-transformed with the plasmids pDBLeu-CP and pDEST22-GPS2 , and positive clones were obtained only on the SD/–Leu/–Trp/–His medium ( Figure 1A ) , indicating a moderate strength of the interaction between CP and the partial GPS2 . 10 . 7554/eLife . 26591 . 003Figure 1 . Small brown planthopper GPS2 binds Rice Stripe Virus capsid protein . ( A ) GPS2 binds CP in yeast two-hybrid assays . ( 1 ) pDBLeu-CP and pDEST22-GPS2; ( 2 ) pDBLeu-CP and pDEST22 , self activation; ( 3 ) pDBLeu and pDEST22-GPS2 , self activation; ( 4 ) pDBLeu and pDEST22 , negative control; ( 5 ) pGBKT7-53 and pGADT7 , positive control; ( 6 ) pGBKT7-Lam and pGADT7 , negative control . Interaction between pDBLeu-CP and pDEST22-GPS2 was observed only on the SD/–Leu/–Trp/–His medium . ( B ) His-tag pull-down assay . Recombinantly expressed GPS2-His was bound to Ni Sepharose as a bait to hook the prey protein , recombinantly expressed CP-GST . The products from His vector ( pET28a ) and GST vector ( pGEX-3X ) were applied as negative controls . Anti-His , anti-GST , or anti-CP antibody was used to detect proteins . ( C ) Co-immunoprecipitation ( CoIP ) for recombinantly expressed CP-His and GPS2 from non-viruliferous planthopper total proteins . For ( B ) and ( C ) , three independent biological replicates were carried out for each experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 00310 . 7554/eLife . 26591 . 004Figure 1—figure supplement 1 . Phylogenetic tree of planthopper GPS2 and GPS1 proteins and those from other insect species . The neighbor-joining method ( pairwise deletion and p-distance model ) was used in Mega 6 . 06 software . Bootstrap analysis ( 1000 replicates ) was applied to evaluate the internal support for the tree topology . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 004 The full-length GPS2 and RSV CP were recombinantly expressed in Escherichia coli , and the interaction between the two proteins was tested in vitro using a pull-down assay . The results showed that His-tagged GPS2 bound GST-fused CP , while no binding appeared in negative controls ( Figure 1B ) . A co-immunoprecipitation assay using CP monoclonal antibody and recombinantly expressed His-CP pulled down GPS2 from the planthopper in vivo , consistent with an interaction between CP and GPS2 ( Figure 1C ) . After a BLAST search of the small brown planthopper transcriptome ( Zhao et al . , 2016a ) with the partial GPS2 sequence , the full-length open reading frame ( ORF ) of the planthopper GPS2 was identified and cloned . Its 1299-bp nucleotides encode a 432 amino acid residue protein ( GenBank accession KY435901 ) . The GPS2 partial sequence identified from the yeast two-hybrid was from nucleotide 277 to 1186 of the full-length ORF . The predicted molecular mass of the GPS2 protein was 47 . 8 kD , and the protein included an N-terminal coiled-coil region ( nucleotides 73–336 ) . Cellular localization of GPS2 was predicted to be both nuclear and cytoplasmic . Phylogenetic analysis of protein sequences supported the identity of the cloned GPS2 gene because it clustered with mammalian and other insects’ GPS2 genes . Planthopper GPS2 was most similar to that of the termite Zootermopsis nevadensis ( Figure 1—figure supplement 1 ) . A paralogous gene , GPS1 ( GenBank accession KY435902 ) , which clustered with mammalian GPS1 gene ( Figure 1—figure supplement 1 ) , was also identified in the small brown planthopper transcriptome ( Zhao et al . , 2016a ) . Planthopper GPS1 and GPS2 had 17% amino acid sequence identity and GPS1 had no coiled-coil domain . In order to clarify the influence of RSV infection on the expression of GPS2 , the temporal and spatial expressions of GPS2 were checked in non-viruliferous and viruliferous planthoppers using quantitative real-time PCR . In the non-viruliferous planthoppers , the transcript level of GPS2 was highest in the eggs , was lowest in the middle nymph stages , and then increased until adult stage ( Figure 2A ) . In adults , the GPS2 transcript levels were highest in reproductive organs , testicle and ovary , and lowest in gut and fatbody ( Figure 2B ) . The subcellular location of GPS2 was examined with immunohistochemistry in planthopper salivary gland cells ( taking advantage of their large size ) using anti-human GPS2 polyclonal antibody . This antibody had a high specificity for planthopper GPS2 ( Figure 2—figure supplement 1 ) . Immunofluorescence showed that GPS2 was located in both the nucleus and the cytoplasm ( Figure 2C ) . When total protein was isolated from the nuclei and the cytoplasm of insect whole bodies , more GPS2 protein was detected in the cytoplasm than in the nucleus ( Figure 2D ) . 10 . 7554/eLife . 26591 . 005Figure 2 . Temporal , spatial and subcellular expression of GPS2 in the planthopper . ( A ) Relative transcript levels of GPS2 throughout the developmental stages of non-viruliferous planthoppers . ( B ) Relative transcript levels of GPS2 in various tissues of non-viruliferous planthoppers . The transcript levels of GPS2 are normalized to that of the ef2 transcript . Different letters indicate significant differences in Tukey’s multiple comparison test . ( C ) Subcellular localization of GPS2 in the salivary gland cells of non-viruliferous planthoppers revealed by an anti-human GPS2 polyclonal antibody in immunohistochemistry . Red is the positive signal . Nuclei are stained blue . ( D ) Western blotting showing GPS2 in the nuclei and cytoplasm of cells in insect whole-body samples . Reference proteins for nuclear and cytoplasmic proteins were histone H3 and tubulin , respectively , which were displayed using an anti-human H3 monoclonal antibody and an anti-human tubulin monoclonal antibody . ( E ) Comparison of relative transcript levels of GPS2 between viruliferous and non-viruliferous fourth instar nymphs , females and males . ( F ) Western blotting showing GPS2 protein in fourth instar nymphs of viruliferous and non-viruliferous planthoppers . ( G ) Comparison of relative transcript levels of GPS2 between the viruliferous and non-viruliferous various tissues . ( H ) Relative transcript levels of GPS2 when RSV is incubated within the insects for five days . *p<0 . 05; **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 00510 . 7554/eLife . 26591 . 006Figure 2—source data 1 . Numerical data that are represented as graphs in Figure 2A , B , E , G , H . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 00610 . 7554/eLife . 26591 . 007Figure 2—figure supplement 1 . Western blotting to show the specificity of the anti-human GPS2 polyclonal antibody for recognizing planthopper GPS2 . Total proteins were extracted from non-viruliferous fourth instar planthoppers using T-PER Tissue Protein Extraction Reagent , containing a protease inhibitor cocktail . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 00710 . 7554/eLife . 26591 . 008Figure 2—figure supplement 2 . Densitometry analysis for the GPS2 image bands from Figure 2F . The relative densities of GPS2 were normalized with those of tubulin and presented as mean ± SE . **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 008 Compared to those in non-viruliferous planthoppers , the transcript level and protein level of GPS2 in viruliferous insects was decreased in nymphs and adults ( Figure 2E , F , Figure 2—figure supplement 2 ) . The transcript levels of GPS2 were lower in the salivary glands , ovaries , and testicles of viruliferous insects than of non-viruliferous insects ( Figure 2G ) . When non-viruliferous insects were fed a diet containing RSV for 8 hr and the virus was allowed to incubate in insects for five days , the transcript level of GPS2 decreased significantly on the fifth day ( Figure 2H ) . To clarify whether GPS2 functions as a repressor of JNK activation in the planthoppers ( as is the case in yeast and mammals [Spain et al . , 1996] ) , JNK activation was investigated through the phosphorylation status of JNK when GPS2 was knocked down by injection of double-strand RNAs . Two JNK genes were identified in the small brown planthopper transcriptome ( Zhao et al . , 2016a ) using three human JNKs ( GenBank accessions P45983 , P45984 , and P53779 ) as queries in BLASTp searches . The ORFs of the two putative JNKs were 1287 bp and 1125 bp , encoding a 49 kD protein ( JNK1 , GenBank accession KY435903 ) and a 43 kD protein ( JNK2 , GenBank accession KY435904 ) , respectively . The amino acid identities of the two JNKs to the three JNKs of human were from 56% to 65% ( Figure 3—figure supplement 1 ) . The recombinantly expressed planthopper JNK1 and JNK2 were recognized by the anti-human JNK2 polyclonal antibody ( Figure 3—figure supplement 2 ) , and two bands corresponding to endogenous JNK1 and JNK2 appeared in planthopper samples using this antibody in western blotting ( Figure 3A ) . When using an anti-phospho-human JNK2 antibody to show the phosphorylated JNK , only one band , apparently phosphorylated JNK1 , was detected . The immune reaction decreased after the sample was treated with λ-phosphatase ( Figure 3A ) . 10 . 7554/eLife . 26591 . 009Figure 3 . GPS2 represses JNK activation in the planthopper . ( A ) Western blot of the phosphorylated JNK ( P-JNK ) and total JNKs in non-viruliferous fourth instar planthoppers before and after the treatment with λ-phosphatase . Total protein was incubated with λ-phosphatase for 1 hr at 30°C . Three independent biological replicates were carried out . Here we show one representative result . ( B ) Western blotting showing P-JNK in viruliferous or non-viruliferous fourth instar nymphs when dsGPS2-RNA was injected . Levels were determined 3 d after injections . dsGFP–RNA injection was used as control . The P-JNK , total JNKs , and tubulin were detected using an anti-phospho-human JNK2 antibody , an anti-human JNK2 polyclonal antibody , and an anti-human tubulin monoclonal antibody , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 00910 . 7554/eLife . 26591 . 010Figure 3—figure supplement 1 . Amino acid sequence alignments of planthopper JNKs ( LsJNK1 and LsJNK2 ) and human JNKs ( HsJNK1 , HsJNK2 , HsJNK3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01010 . 7554/eLife . 26591 . 011Figure 3—figure supplement 2 . Recombinant expression of planthopper JNK1 and JNK2 for specificity verification of the anti-human JNK2 polyclonal antibody . ( A ) SDS-PAGE showed the recombinant expression of JNK1 and JNK2 in E . coli . Arrows indicate the target proteins . ( B ) and ( C ) Western blot analysis of JNK1 and JNK2 using the anti-human JNK2 polyclonal antibody . JNK1 inclusions and supernatant JNK2 were applied . The 51-kD recombinant JNK1 and 44-kD recombinant JNK2 were revealed . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01110 . 7554/eLife . 26591 . 012Figure 3—figure supplement 3 . Densitometry analysis for the phosphorylated JNK ( P-JNK ) image bands from Figure 3B . The relative densities of P-JNK were normalized with those of tubulin and presented as mean ± SE . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 012 A 219-bp double-strand RNA of GPS2 was injected into viruliferous or non-viruliferous fourth instar planthoppers for transcript knockdown . The transcript knockdown for GPS2 at the third day after injection was 87% and 91% in viruliferous and non-viruliferous insects , respectively . In both viruliferous and non-viruliferous insects , the phosphorylation of JNK was higher in the GPS2 knockdown groups compared to the control groups , as revealed by the anti-phospho-human JNK2 antibody ( Figure 3B , Figure 3—figure supplement 3 ) , indicating that JNK was more significantly activated with less GPS2 and that GPS2 can repress JNK activation in the planthoppers . Considering that GPS2 represses JNK activation and that RSV’s CP can bind GPS2 , we examined whether RSV infection activated JNK signaling in the planthoppers . The phosphorylation level of JNK was compared between the viruliferous and non-viruliferous adult female , adult male , or fourth instar insects using the anti-phospho-human JNK2 antibody . A higher level of JNK phosphorylation was observed in the viruliferous instars ( Figure 4A , Figure 4—figure supplement 1A ) , but not in viruliferous female or male adults ( Figure 4—figure supplement 2 ) . The most obvious increases in JNK phosphorylation level appeared in the gut , salivary glands , ovary , and testicle of viruliferous insects compared to the non-viruliferous insects ( Figure 4B , Figure 4—figure supplement 1B–E ) . 10 . 7554/eLife . 26591 . 013Figure 4 . Rice stripe virus activates the JNK signaling pathway . ( A ) Western blotting showing the phosphorylated JNK ( P-JNK ) in the fourth instar nymphs of viruliferous and non-viruliferous planthoppers . ( B ) Western blotting showing the P-JNK in salivary gland , gut , ovary , and testis of viruliferous and non-viruliferous planthoppers . ( C ) The TNF-α levels in the fourth instar nymphs of viruliferous and non-viruliferous planthoppers , or in CP-His-protein-injected non-viruliferous insects and negative control groups , determined using a human TNF-α ELISA Kit . ( D ) Western blotting showing the P-JNK in the fourth instar nymphs when dsTNF-α-RNA was injected and RSV was incubated in insects for five days . ( E ) Western blotting showing the P-JNK and GPS2 in non-viruliferous fourth instar nymphs at 24 hr after CP-His protein injection . The control insects were injected with purified products from pET28a vector . ( F ) Relative transcript levels of GPS2 after CP-His protein injection . The P-JNK , total JNK , GPS2 , and tubulin were detected using an anti-phospho-human JNK2 polyclonal antibody , an anti-human JNK2 polyclonal antibody , an anti-human GPS2 polyclonal antibody , and an anti-human tubulin monoclonal antibody , respectively . *p<0 . 05; **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01310 . 7554/eLife . 26591 . 014Figure 4—source data 1 . Numerical data that are represented as graphs in Figure 4C , F . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01410 . 7554/eLife . 26591 . 015Figure 4—figure supplement 1 . Densitometry analysis for the phosphorylated JNK ( P-JNK ) and GPS2 image bands from Figures 4A , B , D and E . The relative densities of P-JNK and GPS2 were normalized with those of tubulin and presented as mean ± SE . *p<0 . 05; **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01510 . 7554/eLife . 26591 . 016Figure 4—figure supplement 2 . Western blot of the phosphorylated JNK ( P-JNK ) in female and male viruliferous and non-viruliferous planthoppers . ( A ) and ( B ) are Western blots . The P-JNK , total JNK , and tubulin were detected using an anti-phospho-human JNK2 antibody , an anti-human JNK2 polyclonal antibody , and an anti-human tubulin monoclonal antibody , respectively . ( C ) and ( D ) show the relative densities of P-JNK , which were normalized with those of tubulin and are presented as mean ± SE . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01610 . 7554/eLife . 26591 . 017Figure 4—figure supplement 3 . Protein characteristics of TNF-α from small brown planthopper , Drosophila fruit fly , and human . Domains were predicted at http://smart . embl-heidelberg . de/ . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 017 In many systems , TNF-α is known to be an upstream proinflammatory signaling molecule for JNK activation ( Grivennikov et al . , 2010 ) . A putative TNF-α transcript ( GenBank accession KY435905 ) , encoding 389 amino acid residues , was identified from our small brown planthopper transcriptome ( Zhao et al . , 2016a ) . The encoded protein contains a conserved TNF domain , as do the human homolog ( NP_000585 ) and the Drosophila homolog ( also called Eiger ) ( NP_724878 ) ( Figure 4—figure supplement 3 ) . An enzyme linked immunosorbent assay using human TNF-α monoclonal antibody showed that the level of the TNF-α protein increased two-fold when insects were infected by RSV ( Figure 4C ) . However , when the expression of TNF-α was lowered ( 70% knockdown of its transcript level ) , RSV did not upregulate the phosphorylation of JNK ( Figure 4D , Figure 4—figure supplement 1F ) . We conclude that RSV infection stimulated an inflammatory response and activated the JNK signaling pathway in the planthopper . To determine whether JNK activation was mediated by the capsid protein of RSV , recombinantly expressed and purified CP was injected into non-viruliferous fourth instar planthoppers . A higher level of JNK phosphorylation was observed in western blotting in insects at 24 hr after CP administration than in the control group , indicating that CP activated JNK ( Figure 4E , Figure 4—figure supplement 1G ) . At the same time , the transcript level ( Figure 4F ) and the protein level of GPS2 ( Figure 4E , Figure 4—figure supplement 1H ) were significantly lower in the presence of CP . The TNF-α protein level increased 2 . 2-fold after CP administration ( Figure 4C ) . Thus , regulation of the JNK pathway by RSV in the planthopper is caused by the virus's capsid protein . Because the E2 ubiquitin-conjugating enzyme Ubc13 is part of the enzymatic machinery for JNK activation in mammals ( Cardamone et al . , 2012 ) , we studied the function of Ubc13 in the JNK pathway in the planthopper . First , we found a putative Ubc13 ORF from the small brown planthopper transcriptome . The planthopper Ubc13 encoded a 17-kD protein ( GenBank accession KY435906 ) and had 82% amino acid identity with the human homolog ( NP_003339 ) . The recombinantly expressed Ubc13 with a GST-tag was a 44 . 2-kD protein . Next , we knocked down the Ubc13 transcript and tested the JNK activation using the anti-phospho-human JNK2 antibody and treatment withmouse TNF-α . The mouse TNF-α was able to stimulate JNK phosphorylation in the insects ( Figure 5—figure supplement 1 ) . The knockdown of the Ubc13 transcript was 91% , and the extent of JNK phosphorylation was less than in the control group ( Figure 5A , Figure 5—figure supplement 2 ) . This means that Ubc13 takes part in the JNK activation in planthoppers . 10 . 7554/eLife . 26591 . 018Figure 5 . The Rice stripe virus capsid protein competes with Ubc13 in binding GPS2 . ( A ) Western blotting showing the phosphorylated JNK ( P-JNK ) when dsUbc13-RNA and mouse TNF-α were injected into the non-viruliferous fourth instar nymphs of the planthopper . Injections of dsGFP–RNA or 20 mM Tris-HCl containing 0 . 1% Tween 20 were used as controls . Relative transcript levels of Ubc13 were compared after three days of treatment . **p<0 . 01 . ( B ) and ( C ) His-tag pull-down assay for the competitive binding of GPS2 by CP and Ubc13 using recombinantly expressed proteins . Recombinantly expressed GPS2-His was bound to Ni Sepharose as a bait for hooking Ubc13-GST at a series of dilutions of CP-GST . ( D ) His-tag pull-down assay for the competitive binding of GPS2 by CP and in vivo Ubc13 . In vitro expressed GPS2-His was bound to Ni Sepharose . CP-GST or GST was applied to the GPS2-His bound Sepharose . Then , the total proteins from non-viruliferous planthoppers were applied to the Sepharose . ( E ) His-tag pull-down assay for the competitive binding of GPS2 by in vivo CP and in vivo Ubc13 . Recombinantly expressed GPS2-His was bound to Ni Sepharose . Then the total proteins from viruliferous or non-viruliferous planthoppers were applied to the Sepharose . ( F ) His-tag pull-down assay for the interaction between CP-GST and the N1-His ( nucleotides 1–465 ) , N2-His ( 1-657 ) , or N3-His ( 1-903 ) fragment of GPS2 . The expression products from His vector ( pET28a ) and GST vector ( pGEX-3X ) were applied as negative controls . ( G ) His-tag pull-down assay for the interaction between the N1-His , N2-His , or N3-His fragment of GPS2 and the CP from the total protein of viruliferous planthoppers . The expression product from His vector ( pET28a ) was applied as negative control . ( H ) His-tag pull-down assay for the interaction between the N1-His , N2-His , N3-His , or C-His ( nucleotides 886–1296 ) fragment of GPS2 and Ubc13-GST . The expression product from His vector ( pET28a ) was applied as negative control . Anti-phospho-human JNK2 polyclonal antibody , anti-human JNK2 polyclonal antibody , anti-human tubulin monoclonal antibody , anti-human Ubc13 monoclonal antibody , anti-CP monoclonal , anti-His monoclonal , and anti-GST polyclonal antibody were used to detect proteins . From ( B ) to ( H ) , three independent replicates were carried out for each experiment . We show one representative result for each experiment here . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01810 . 7554/eLife . 26591 . 019Figure 5—source data 1 . Numerical data that are represented as a graph in Figure 5A . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 01910 . 7554/eLife . 26591 . 020Figure 5—figure supplement 1 . Western blot of the phosphorylated JNK ( P-JNK ) in response to TNF-α treatment . ( A ) Western blotting . The non-viruliferous fourth instar planthoppers were injected with 23 nL of 50 μM mouse TNF-α . The control groups were injected with 23 nL of 20 mM Tris-HCl containing 0 . 1% Tween 20 . Activation or inhibition of JNK were checked after 3 d of treatment . P-JNK , total JNK , and tubulin were detected using an anti-phospho-human JNK2 antibody , an anti-human JNK2 polyclonal antibody , and an anti-human tubulin monoclonal antibody , respectively . ( B ) The relative densities of P-JNKwere normalized with those of tubulin and are presented as mean ± SE . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 02010 . 7554/eLife . 26591 . 021Figure 5—figure supplement 2 . Densitometry analysis for the phosphorylated JNK ( P-JNK ) image bands from Figure 5A . The relative densities of P-JNK were normalized with those of tubulin and are presented as mean ± SE . **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 02110 . 7554/eLife . 26591 . 022Figure 5—figure supplement 3 . Positions of the five fragments , N1 , N2 , N3 , N4 , and C , within GPS2 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 02210 . 7554/eLife . 26591 . 023Figure 5—figure supplement 4 . His-tag pull down assay for the interactions of GPS2 fragments with CP or Ubc13 . ( A ) C-His fragment ( nucleotides 886–1296 ) of GPS2 and CP-GST . ( B ) N4-GST fragment ( 466-885 ) of GPS2 and CP-His . ( C ) N4-GST and Ubc13-His . The expression products from His vector ( pET28a ) and GST vector ( pGEX-3X ) were used as negative controls . Anti-His and anti-GST antibodies were used to detect the proteins with corresponding tags . Three independent replicates were carried out for each experiment . Here we show one representative result . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 023 A pull-down experiment using three recombinantly expressed proteins showed that GPS2 was able to pull Ubc13 down ( Figure 5B ) . However , when CP was present in different amounts , the binding between GPS2 and Ubc13 was affected , the more CP , the less Ubc13 binding to GPS2 occurred ( Figure 5B and C ) . Recombinantly expressed GPS2 can pull down Ubc13 from an extract of non-viruliferous planthoppers . When the recombinantly expressed CP was loaded , less Ubc13 was pulled down from non-viruliferous planthoppers ( Figure 5D ) . When an extract from non-viruliferous or viruliferous insects was passed over a Ni Sepharose column to which recombinantly expressed GPS2 was bound , much more Ubc13 was recovered from non-viruliferous than from viruliferous insects ( Figure 5E ) . These results indicate that RSV’s CP competes with Ubc13 in binding GPS2 , thus releasing GPS2 from inhibiting the JNK pathway . To verify whether the competition between CP and Ubc13 happened at a specific region of GPS2 , five fragments ( Figure 5—figure supplement 3 ) — N1 ( nucleotides 1–465 ) , N2 ( 1–657 ) , N3 ( 1–903 ) , N4 ( 466–885 ) , and C ( 886–1296 ) of GPS2 — were expressed and tested for potential interactions with CP or Ubc13 using pull-down assays . The results showed that N1 , N2 , and N3 were all able to pull down recombinantly expressed CP ( Figure 5F ) as well as the CP from viruliferous insects ( Figure 5G ) . The same fragments were also able to pull down recombinantly expressed Ubc13 ( Figure 5H ) . By contrast , neither fragment N4 nor fragment C bound either CP ( Figure 5—figure supplement 4A , B ) or Ubc13 ( Figure 5H , Figure 5—figure supplement 4C ) . N1 , N2 , and N3 all contain the coiled-coil domain , whereas N4 does not , suggesting that the N-terminal coiled-coil region of GPS2 is the binding region for CP and Ubc13 . To explore the effect of JNK activation on RSV transmission , we assessed virus proliferation in insects and the virus transmission efficiency when JNK was activated or inhibited . First , we knocked down the GPS2 transcript by injection of dsGPS2-RNA to activate JNK . Compared to the control group ( dsGFP-RNA injected ) , the transcript level of GPS2 in the treatment group was reduced by 49% while the RNA level of CP increased by 18-fold ( Figure 6A and B ) , demonstrating that RSV replicated more in insects with lower GPS2 amounts . Second , we lowered the TNF-α transcript level by injection of dsTNF-α-RNA to inhibit JNK . When the transcript level of TNF-α was knocked down by 70% , the RNA level of CP decreased by 80% ( Figure 6C ) . 10 . 7554/eLife . 26591 . 024Figure 6 . JNK activation facilitates the proliferation of Rice stripe virus in insects . ( A ) Relative transcript levels of GPS2 7 d after dsGPS2-RNA or dsGFP-RNA injection . ( B ) Relative RNA levels of RSV CP in dsGPS2-RNA-injected and dsGFP-RNA-injected planthoppers after 5 d of virus incubation . ( C ) Relative RNA levels of RSV CP in dsTNF-α-RNA-injected and dsGFP-RNA-injected planthoppers after 5 d of virus incubation . ( D ) Relative transcript levels of JNK1 and JNK2 3 d after injection of the mixture of dsJNK1-RNA and dsJNK2-RNA . ( E ) Relative RNA levels of RSV CP in dsJNK1/2-RNA-injected and dsGFP-RNA-injected planthoppers after 5 d of virus incubation . ( F ) Relative RNA level of RSV CP in JNK-specific chemical inhibitor SP600125-injected planthoppers after 5 d of virus incubation . The control group was injected with an equal amount of 2% DMSO . ( G ) Relative RNA level of RSV CP in mouse TNF-α-injected planthoppers after 5 d of virus incubation . The control group was injected with an equal amount of 20 mM Tris-HCl containing 0 . 1% Tween 20 . ( H ) The disease incidence rate of the rice plants fed upon by JNK1 and JNK2 knockdown planthoppers . The plants were placed at 26°C with 16 hr of light per day to observe disease symptoms . Six groups of plants per replicate and four replicates were used to calculate the disease incidence rate on each day . Plants fed upon by dsGFP-RNA injected insects were used as controls . *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 02410 . 7554/eLife . 26591 . 025Figure 6—source data 1 . Numerical data that are represented as graphs in Figure 6A–G . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 02510 . 7554/eLife . 26591 . 026Figure 6—figure supplement 1 . Western blot of phosphorylated JNK ( P-JNK ) after SP600125 treatment . ( A ) The non-viruliferous fourth instar planthoppers were injected with 23 nL of 0 . 9 μM SP600125 . The control groups were injected with 23 nL of 2% DMSO . Activation or inhibition of JNK were checked after 3 d of treatment . The P-JNK , total JNK , and tubulin were detected using an anti-phospho-human JNK2 antibody , an anti-human JNK2 polyclonal antibody , and an anti-human tubulin monoclonal antibody , respectively . ( B ) The relative densities of P-JNK , which were normalized with those of tubulin and are presented as mean ± SE . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 02610 . 7554/eLife . 26591 . 027Figure 6—figure supplement 2 . The disease incidence rate of the rice plants fed upon by GPS2 knockdown planthoppers . The plants were placed at 20°C with 16 hr of light per day to observe disease symptoms . Six groups of plants per replicate and five replicates were used to calculate the disease incidence rate on each day . Plants fed upon by dsGFP-RNA injected insects were used as controls . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 027 We also manipulated JNK directly to test the effect of JNK signaling on RSV proliferation in the planthoppers . When dsJNK1-RNA and dsJNK2-RNA were injected simultaneously into insects , the transcript levels of JNK1 and JNK2 were reduced by 65% and 82% , respectively ( Figure 6D ) . There was 71% less CP RNA in the dsJNKs-RNA-injected insects than in the dsGFP-RNA-injected insects ( Figure 6E ) . In pharmacological experiments , the JNK-specific chemical inhibitor SP600125 or the JNK agonist TNF-α from mouse was injected into the insects . The JNK phosphorylation level reduced by 62% after treatment with SP600125 in the small brown planthoppers ( Figure 6—figure supplement 1 ) , and the CP RNA level reduced by 28% ( Figure 6F ) . By contrast , treatment with TNF-α increased the CP RNA level around two-fold with 50% increase of the JNK phosphorylation level ( Figure 6G , Figure 5—figure supplement 1 ) . Taken together , these results show that JNK activation facilitated the proliferation of RSV in the planthoppers . To show the effect of manipulation of the JNK pathway of vector insects in rice plants , the disease incidence of rice plants was determined when the plants were fed upon by viruliferous insects in which the expressions of GPS2 or JNKs were lowered . In the dsGPS2-RNA-injected insects , RSV replication increased ( Figure 6B ) , but the disease incidence rates of the rice plants fed upon by the dsGPS2-RNA-injected insects were similar to those fed upon by the dsGFP-RNA-injected insects ( Figure 6—figure supplement 2 ) . When the levels of the two JNK transcripts were lowered in insects in which RSV replication was inhibited ( Figure 6E ) , the disease incidence of plants was delayed ( Figure 6H ) . Within 15 d , 50% of the plants in the control group showed disease symptoms whereas only 8% of plants in the JNKs-knocked-down group showed disease symptoms . However , 75% plants in the JNKs-knocked-down group showed disease symptom by 21 d , close to the 95% disease incidence rate in the control group . This suggests that inhibition of the JNK pathway retards the incidence of disease in plants by restricting RSV replication in the planthoppers .
Even in the very well-studied mosquito systems , little is known about how viruses that cause diseases in humans , animals or plantsmanipulate the cellular JNK signaling pathway of their vector insects . Our study clarifies for the first time that the virus enhances its replication in a vector insect by activating the vector’s JNK pathway , and thatJNK inhibition results in a delayed disease incidence in plants . We provide a model to summarize how RSV manipulates the upstream signal molecules and the downstream suppressor to activate the JNK pathway in the planthoppers ( Figure 7 ) . 10 . 7554/eLife . 26591 . 028Figure 7 . Model of RSV regulation of the JNK signaling pathway of its vector insect causing increased replication . The virus activates the JNK signaling pathway of the vector insects in three ways: i ) upregulating TNF-α; ii ) decreasing the expression of GPS2 , which is a repressor of JNK activation; and iii ) binding of the virus capsid protein to GPS2 to prevent it from inhibiting the JNK activation machinery . JNK activation is beneficial to the ability of the virus to replicate in its vector insects . DOI: http://dx . doi . org/10 . 7554/eLife . 26591 . 028 Our finding that JNK activation by RSV facilitates virus replication in the small brown planthopper reflects a conservative evolution scenario in which the virus–vector and virus–host interactions both utilize the JNK pathway . The JNK signaling pathway in mammal or invertebrate hosts has been reported to be involved in or affected by infections by various viruses . For instance , infection with rhesus rotavirus , herpes simplex virus type 1 , or varicella-zoster virus results in the activation of JNK in mammalian cells . Inhibition of JNK by SP600125 reduced virus replication ( Holloway and Coulson , 2006; McLean and Bachenheimer , 1999; Zapata et al . , 2007 ) . In Bombyx mori , the magnitude and pattern of JNK activation were dependent on the multiplicity of nucleopolyhedrovirus infection , and inhibition of JNK reduced occlusion body formation and budded virus production ( Katsuma et al . , 2007 ) . In a crustacean , Litopenaeus vannamei , JNK was activated in response to white spot syndrome virus infection and the virus proliferation benefited from JNK activation ( Shi et al . , 2012 ) . Although this phenotype is common in virus–vector and virus–host systems , there should be some mechanisms in vector insects to limit the virus replication to a sustainable level . Our study shows that the capsid protein of RSV not only induces a decrease of GPS2 expression , but also directly binds GPS2 so as to release GPS2 from repressing the JNK pathway . The N-terminal coiled-coil region of GPS2 is evidently key for the binding of CP . This explains the moderate strength of the interaction in the yeast two-hybrid assay between CP and the partial GPS2 that lacks the intact coiled-coil region . The N-terminal coiled-coil region of GPS2 is also the binding site of planthopper Ubc13 . We proved that the planthopper Ubc13 is indispensable for JNK activation , as is the mammal homolog . Binding of GPS2 to Ubc13 leads to the inhibition of JNK activation . Thus the presence of CP that competitively binds GPS2 activates the JNK pathway . On the other hand , we do not exclude the possibility that other viral proteins could also be involved in the JNK regulation . A study on human T-cell lymphotrophic virus ( Jin et al . , 1997 ) is of particular interest in light of our results . Jin et al . ( 1997 ) found that this virus’ protein Tax , a Type-I oncoprotein , binds to GPS2 . Given our results on the binding of the RSV capsid protein to GPS2 , it is possible that Tax’s binding to GPS2 activates JNK . They also showed that induction of Tax expression reduced the level of GPS2 in Jurkat cells . In the present study , we found that JNK activation by RSV is caused by increased levels of the proinflammatory signaling molecule , TNF-α . When the expression of TNF-α was lowered , RSV did not upregulate the phosphorylation of JNK . In mammals , TNF-α is an upstream JNK-stimulating signal that regulates a wide range of physiological activities , such as immune responses and inflammation ( Grivennikov et al . , 2010 ) . RSV capsid protein , as a surface recognition factor , induces this immune response , as does the entire virus . In mammals , excessive activation of proinflammatory signaling pathways is harmful , promoting the development of human diseases such as autoimmune disorder , neurodegeneration , and cancer ( Amor et al . , 2010; Grivennikov et al . , 2010 ) . However , the inflammation in the planthopper caused by RSV seems sustainable . Perhaps this is due to the rather short lifespan of the planthopper ( about 40 days under optimal conditions ) . Activation of the JNK signaling pathway in the planthoppers by RSV might promote stress resistance in the insects . The JNK signaling pathway serves as a molecular sensor for various stresses . The protective functions of JNK activation have been shown in oxidative stress tolerance at the cellular level and in the aging of the organism . For example , JNK signaling activity alleviates the toxic effects of reactive oxygen species and increases the lifespan of Drosophila melanogaster ( Wang et al . , 2003 ) . JNK overexpression and phosphorylation increase the lifespan of Caenorhabditis elegans as well as its resistance to oxidative stress and heat stress ( Oh et al . , 2005 ) . JNK signaling also protects a host against bacterial infections by promoting apoptosis or phagocytosis ( Mizutani et al . , 2003a; Wandler and Guillemin , 2012 ) . Only a few studies have shown effects of RSV infection on the physiology of the planthoppers , such as reduced fecundity and progeny hatchability , accelerated development , greater body weight , and increased abundance of a yeast-like symbiont ( Li et al . , 2015; Wan et al . , 2015 ) . Possible influences of RSV on the stress tolerance of the planthoppers , such as their ability to tolerate cold stress or heat stress , deserve to be further explored . Given the effects of JNK inhibition in postponing the incidence of disease in rice plants infected with RSV , it seems reasonable to suggest that inhibition of the JNK pathway is a potential means to benefit rice agriculture . Perhaps we can generalize from our results on the knockdown of the JNK transcripts in this study to suggest that such inhibition of the JNK pathway —by lowering JNK transcript levels , by strengthening interactions with GPS2 or by weakening the effects of TNF-α or Ubc13 — could be beneficial agriculturally . Such alterations could possibly be achieved through breeding or other means of genetic modification or , especially in the case of weakening effects , by administering appropriate chemical compounds . A successful case has been reported in which genetically engineered rice plants that expressed the RSV CP showed resistance to RSV infection ( Hayakawa et al . , 1992 ) . In conclusion , our study has uncovered strategies through which the RSV manipulates the JNK signaling pathway of its vector insect , the small brown planthopper , for its own benefit . Inhibition of the JNK pathway of insect vectors may , more generally , provide an important means of intervening in the transmission of persistent plant viruses .
The viruliferous and non-viruliferous small brown planthopper strains used in this study were established from a field population collected in Hai’an , Jiangsu Province , China . They were reared separately on 2 cm to 3 cm seedlings of rice Oryza sativa L . spp . japonica var . nippobare in glass incubators in the laboratory , as described by Zhao et al . ( 2016a ) . The RSV-carrying frequency of the viruliferous strain was maintained at no less than 90% through a purification selection every three months utsing dot-ELISA with the monoclonal anti-CP antibody ( Zhao et al . , 2016b ) . Yeast two-hybrid screening was performed with the ProQuest two-hybrid system ( Invitrogen , Carlsbad , CA , USA ) according to the manufacturer’s protocol . A whole-body cDNA library for small brown planthopper was constructed in the GAL4 activation domain of the vector pDEST22 to create prey plasmids using a CloneMiner II kit ( Invitrogen ) . Full-length RSV CP was cloned in the GAL4 DNA binding domain of the vector pDBLeu to create pDBLeu-CP bait plasmid . MaV203 yeast cells were co-transformed with pDBLeu-CP and pDEST22-library . Positive clones were selected on the triple dropout medium ( SD/–Leu/–Trp/–His ) and quadruple dropout medium ( SD/–Leu/–Trp/–His/–Ade ) . Prey plasmids were isolated from these clones for sequencing . To confirm the interaction of bait and prey proteins , we co-transformed the two plasmids into yeast strain AH109 and repeated the selection on the triple and quadruple dropout media . The prey sequences were used in a BLAST search of the small brown planthopper transcriptome ( Zhao et al . , 2016a ) . Highly viruliferous or non-viruliferous planthoppers in various developmental stages and six tissues ( brain , salivary gland , gut , fatbody , ovary , and testicle ) were ground in liquid nitrogen , and total RNA was isolated following the standard TRIzol reagent protocol ( Invitrogen ) . The concentration and quality of total RNA were determined using a NanoDrop spectrophotometer ( Thermo Scientific , Waltham , MA , USA ) and by gel electrophoresis . RNA was treated using the TURBO DNA-free kit ( Ambion , Austin , TX , USA ) to remove genomic DNA contamination before being used for cDNA synthesis . RNA ( 1 µg ) was reverse-transcribed to cDNA using the Superscript III First-Strand Synthesis System ( Invitrogen ) and random primers ( Promega , Madison , WI , USA ) following the manufacturer’s instructions . Based on the small brown planthopper transcriptome ( Zhao et al . , 2016a ) , full-length open reading frames ( ORF ) of the planthopper GPS2 gene and Ubc13 gene were amplified with the primer pairs gps2-F/gps2-R and ubc13-F/ubc13-R ( Supplementary file 1 ) , respectively , from a viruliferous planthopper cDNA library and sequenced . The molecular weight of encoded GPS2 protein was predicted in ExPASy ( http://web . expasy . org/compute_pi/ ) . The cellular localization of GPS2 was predicted using PSORT II Prediction ( http://psort . hgc . jp/form2 . html ) . The phylogenetic relation of planthopper GPS2 and its paralog , GPS1 ( identified from the small brown planthopper transcriptome ) , to those from other insect species was analyzed with the neighbor-joining method ( pairwise deletion and p-distance model ) using Mega 6 . 06 software ( RRID:SCR_000667 ) . Bootstrap analysis ( 1000 replicates ) was applied to evaluate the internal support for the tree topology . The full-length ORF of the RSV CP gene ( DQ299151 ) was amplified with the primer pair cp-F/cp-R ( Supplementary file 1 ) from the viruliferous planthopper cDNA library and sequenced . The ORF of GPS2 was constructed in the pET28a vector between the restriction sites NdeI-BamHI with the primer pair gps2-his-F/gps2-his-R ( Supplementary file 1 ) to generate GPS2-His plasmid . Four fragments of GPS2 ( N1 [1–465] , N2 [1–657] , N3 [1–903] , and C [886–1296] ) , JNK1 , and JNK2 were constructed in the pET28a vector between the restriction sites NcoI-SalI with the corresponding primers , gps2N-his-F/gps2N1-his-R , gps2N-his-F/gps2N2-his-R , gps2N-his-F/gps2N3-his-R , gps2C-his-F/gps2C-his-R , jnk1-his-F/jnk1-his-R , and jnk2-his-F/jnk2-his-R , respectively ( Supplementary file 1 ) . An N4 fragment of GPS2 ( 466–885 ) was constructed in the pGEX-3X vector at the SmaI site with the primers gps2N4-gst-F/gps2N4-gst-R ( Supplementary file 1 ) to produce N4-GST plasmid . The ORF of CP was constructed in the pET28a vector between the restriction sites BamHI-XhoI with the primer pair cp-his-F/cp-his-R , and also in the pGEX-3X vector between the restriction sites BamHI-SmaI with the primers cp-gst-F/cp-gst-R to generate CP-His and CP-GST plasmids , respectively ( Supplementary file 1 ) . The ORF of Ubc13 was constructed in the pET28a vector between the restriction sites NcoI-SalI with primers ubc13-his-F/ubc13-his-R , and in the pGEX-3X vector at the SmaI site with the primers ubc13-gst-F/ubc13-gst-R , to generate Ubc13-His and Ubc13-GST plasmids , respectively ( Supplementary file 1 ) . The recombinant plasmids of CP , Ubc13 , and JNK1 were used to transform Escherichia coli strain BL21 ( DE3 ) for expression . GPS2 , various GPS2 fragments , and JNK2 were expressed in E . coli Transetta cells . After 4 hr induction with 0 . 4 mM isopropyl β-D-thiogalactoside at 37°C , cells were pelleted by centrifugation and sonicated for 30 min in ice water . The supernatant from the sonicated cells was used for pull-down assay or protein purification . The expressed recombinant protein CP-His was purified using Ni Sepharose ( GE Healthcare , Buckinghamshire , UK ) following the manufacturer’s instructions and served as antigen to produce mouse anti-CP monoclonal antibody ( Beijing Protein Institute Co . , Ltd . , Beijing , China ) . Escherichia coli-expressed His-tagged recombinant proteins were bound to Ni Sepharose ( GE Healthcare ) for 1 hr at 4°C . Then the GST-tagged recombinant proteins were added and incubated for 2 hr at 4°C . After washing with lysis buffer ( 20 mM sodium phosphate containing 50 mM imidazole , pH 7 . 4 ) , proteins were released with elution buffer ( 20 mM sodium phosphate containing 250 mM imidazole , pH 7 . 4 ) , and separated by SDS-PAGE gel electrophoresis . The presence of target proteins was verified by western blotting with an anti-His monoclonal antibody ( CWBiotech , Beijing , China ) , an anti-GST polyclonal antibody ( CWBiotech ) , or an anti-CP monoclonal antibody . Expression products from pET28a vector and pGEX-3X vector were used as negative controls . For experiments with His-tag pull-down from insects , recombinantly expressed GPS2-His was bound to Ni Sepharose . CP-GST or GST was added and incubated for 2 hr at 4°C . Then the total proteins extracted from non-viruliferous planthoppers using PBS buffer ( pH 7 . 4 ) were added and incubated for another 2 hr at 4°C . In another experiment , after the recombinantly expressed GPS2-His was bound to Ni Sepharose , the total proteins extracted from non-viruliferous or viruliferous planthoppers using PBS buffer ( pH 7 . 4 ) were added and incubated for 2 hr at 4°C . After washing with lysis buffer ( 20 mM sodium phosphate containing 50 mM imidazole , pH 7 . 4 ) , target proteins were collected with elution buffer ( 20 mM sodium phosphate containing 250 mM imidazole , pH 7 . 4 ) . For co-immunoprecipitation , total protein was extracted from about 20 mg of the fourth instar non-viruliferous planthoppers using T-PER Tissue Protein Extraction Reagent , containing a protease inhibitor cocktail ( Thermo Fisher Scientific ) . Around 10% of total extracted protein was reserved as input for further western blot analyses . The remaining protein was cleared with 10 μL of Dynabeads Protein G ( Novex by Thermo Fisher Scientific ) . 40 μL of Protein G beads were prepared with 5 μg of CP monoclonal antibody bound , and then incubated with 200 μL of CP-His protein for 30 min . The mouse IgG ( Merck Millpore , Billerica , MA , USA ) was used as negative control . 200 μL of cleared total protein of planthoppers were then immunoprecipitated with the bead–antibody–CP complex for 30 min . The antibody–CP-GPS2 complex was dissociated from the beads with elution buffer ( Novex by Thermo Fisher Scientific ) for western blot analysis . Planthopper GPS2 , Ubc13 , and CP proteins in insects were recognized by an anti-human GPS2 goat polyclonal antibody ( Santa Cruz Biotechnology , Dallas , TX , USA; RRID:AB_10841240 ) , an anti-human Ubc13 mouse monoclonal antibody ( Santa Cruz Biotechnology; RRID:AB_11150503 ) , and an anti-CP monoclonal antibody , respectively , in western blotting . 10 mg of insects were used in all western blotting analyses . For JNK detection , total protein was extracted using RIPA lysis buffer , containing protease inhibitor cocktail and phosphatase inhibitor cocktail ( CWBiotech ) . Total JNK , phosphorylated JNK , and tubulin in various planthopper samples were detected using an anti-human JNK2 polyclonal antibody ( Cell Signaling Technology , Danvers , MA , USA; RRID:AB_2250373 ) , an anti-phospho-human polyclonal JNK2 antibody ( Cell Signaling Technology; RRID:AB_331659 ) , and an anti-human tubulin monoclonal antibody ( CWBiotech ) , respectively , in western blotting . For another group of experiments , total protein was extracted using RIPA lysis buffer and incubated with λ-phosphatase ( New England Biolabs , Ipswich , MA , USA ) for 1 hr at 30°C . After the treatment , total JNK and phosphorylated JNK were detected by western blotting . The density of phosphorylated JNK was quantified with image analysis software ImageJ and normalized to that of tubulin . Differences were statistically evaluated using Student’s t-test in SPSS 17 . 0 ( RRID:SCR_002865 ) . Proteins were isolated from nuclei and cytoplasm of insect whole body using the Nuclear and Cytoplasmic Protein Extraction Kits ( Beyotime , Jiangsu , China ) according to manufacturer’s instructions . The reference proteins for nuclear and cytoplasmic proteins were histone H3 and tubulin , respectively , which were displayed using an anti-human H3 monoclonal antibody ( Beijing Biodragon Immunotechnologies , Beijing , China ) and an anti-human tubulin monoclonal antibody ( CWBiotech ) . Salivary glands were dissected from non-viruliferous four-instar planthopper nymphs in cold distilled water on a glass plate , and fixed in 4% paraformaldehyde for 2 hr at room temperature . After being permeabilized with osmotic buffer ( 0 . 01 M phosphate-buffered saline containing 2% Triton X-100 , pH 7 . 4 ) for 4 hr , the salivary glands were blocked with 1% bovine serum albumin for 30 min at room temperature . The samples were incubated with the primary antibody , anti-human GPS2 goat polyclonal antibody ( Santa Cruz Biotechnology ) , overnight at 4°C . After washing with 0 . 01 M phosphate-buffered saline containing 1% Tween-20 ( pH 7 . 4 ) , the secondary antibody , Alexa Fluor 594 ( red ) affinipure donkey anti-goat IgG ( YEASEN , Shanghai , China ) , was added . The nuclei were counterstained with Hoechst ( blue ) in accordance with the manufacturer’s instructions ( Invitrogen ) . Negative control was without the primary antibody . The images were viewed under a Leica TCS SP5 confocal microscope ( Leica Microsystems , Solms , Germany ) . Twenty salivary glands were tested . Non-viruliferous four-instar nymphs were fed on an artificial diet containing RSV crude preparations from RSV-infected rice seedlings for 8 hr as previously described ( Zhao et al . , 2016b ) and then transferred to healthy rice seedlings . PCR primers with T7 promoter sequences , gps2-dsRNA-F/gps2-dsRNA-R , jnk1-dsRNA-F/jnk1-dsRNA-R , jnk2-dsRNA-F/jnk2-dsRNA-R , ubc13-dsRNA-F/ubc13-dsRNA-R , or TNF-α-dsRNA-F/ TNF-α-dsRNA-R were used to prepare 219-bp double-stranded RNA ( dsRNA ) of GPS2 , 124-bp dsRNA of JNK1 , 167-bp dsRNA of JNK2 , 142-bp dsRNA of Ubc13 , or 107-bp dsRNA of TNF-α ( Supplementary file 1 ) . A 420-bp dsRNA for green fluorescent protein ( GFP ) was amplified using primers gfp-dsRNA-F and gfp-dsRNA-R as negative controls ( Supplementary file 1 ) . dsRNA was generated using the T7 RiboMAX Express RNAi System ( Promega , Madison , Wisconsin , USA ) and purified using Wizard SV Gel and the PCR Clean-Up System ( Promega ) following the manufacturers’ protocols . Injection of 23 nL of dsRNAs at 6 μg/μL was performed on the fourth instar nymphs . The dsRNAs were delivered into hemolymph in the ventral thorax by microinjection through a glass needle using Nanoliter 2000 ( World Precision Instruments , Sarasota , Florida , USA ) . Quantitative real-time PCR ( qRT-PCR ) was used to quantify the relative RNA levels of RSV CP and the transcript levels of GPS2 , JNKs , TNF-α , and Ubc13 in extracts of whole body or various tissues of planthoppers . A 172-bp fragment of CP , a 148-bp fragment of GPS2 , a 118-bp fragment of JNK1 , a 110-bp fragment of JNK2 , a 100-bp fragment of Ubc13 , and a 100-bp fragment of TNF-α were amplified using the primer pairs cp-q-F/cp-q-R , gps2-q-F/gps2-q-R , jnk1-q-F/jnk1-q-R , jnk2-q-F/jnk2-q-R , ubc13-q-F/ubc13-q-R , and TNF-α-q-F/TNF-α-q-R , respectively ( Supplementary file 1 ) . qRT-PCR was carried out in 20 μl of reaction agent composed of 2 . 5 μl of template cDNA , 10 μl of 2×SYBR Green PCR Master Mix ( Fermentas , Waltham , MA , USA ) , and 0 . 25 μM each primer on Light Cycler 480 II ( Roche , Basel , Switzerland ) . The thermal cycling conditions were 95°C for 2 min , followed by 40 cycles of 95°C for 30 s , 60°C for 30 s and 68°C for 40 s . The transcript level of planthopper translation elongation factor 2 ( ef2 ) was quantified with primer pair ef2-q-F/ef2-q-R to normalize the cDNA templates of planthoppers ( Supplementary file 1 ) . The relative transcript level of each gene was reported as mean ± SE . Differences were statistically evaluated using SPSS 17 . 0 . Student’s t-test was performed to compare two means , whereas one-way ANOVA followed by a Tukey’s test was applied for multiple comparisons . Six tissues ( brain , salivary glands , digestive gut , fat body , ovary , and testicle ) were collected from 20 to 50 viruliferous and non-viruliferous adult planthoppers for RNA extraction . Six replicates for each tissue were prepared . RNA was also isolated from the eggs , the first to fifth instars , and female and male adults of non-viruliferous planthoppers , and from the fourth instars , and female and male adults of viruliferous planthoppers . Six replicates and six to ten instars or adults , and 30 eggs per replicate were prepared for each developmental stage . The non-viruliferous fourth instars were fed on the artificial diet containing RSV crude preparations for 8 hr and then raised on healthy rice seedlings for from one to five days . The insects whose food did not contain RSV were used as controls . Five replicates and six insects per replicate from each day were prepared for RNA isolation . The expression levels of GPS2 were quantified using qRT-PCR . Recombinantly expressed and purified CP-His was micro-injected into the hemolymph of non-viruliferous fourth instars in the ventral thorax using Nanoliter 2000 ( World Precision Instruments ) . For each insect from the treatment group , 23 nL of CP-His at 330 μg/mL was injected . An equal volume of the expression and purified products from the pET28a vector was injected into the insects from the control group . Activation of JNK , accumulation of TNF-α , and GPS2 transcript levels were checked at 24 hr after CP-His administration . Three to six biological replicates and eight insects per replicate were used . Non-viruliferous fourth instar planthoppers were injected with 23 nL of dsRNA of Ubc13 at 6 μg/μL and with 23 nL of 50 μM mouse TNF-α ( Sigma-Aldrich , Santa Clara , CA , USA ) using Nanoliter 2000 ( World Precision Instruments ) . The insects from the control group were injected with 23 nL of dsRNA of GFP at 6 μg/μL and 23 nL of 20 mM Tris-HCl containing 0 . 1% Tween 20 . Ubc13 transcript levels and activation of JNK were checked after 3 d of treatment . At least three biological replicates and eight insects per replicate were used . Eight four-instar nymphs from viruliferous or non-viruliferous planthopper strains , or from CP-His injected non-viruliferous planthoppers or negative control insects , were ground in 200 μL of 0 . 01M PBS buffer ( pH 7 . 2 ) containing protease inhibitor cocktail ( CWBIO , Beijing , China ) . The supernatant was collected after centrifugation at 12 , 000 g for 15 min at 4°C and the protein concentration was determined using the Bradford method . The TNF-α concentration from each sample of planthoppers was determined using a human TNF-α ElISA Kit ( Liuhe , Wuhan , China ) , in which the monoclonal antibody of human TNF-α and a standard curve of human TNF-α with a two-fold dilution from 1000 pg/ml to 15 . 6 pg/ml were supplied . The quantity of TNF-α from each sample was designated as pg of TNF-α per mg of total protein . Six ( viruliferous versus non-viruliferous planthoppers ) or eight ( CP-His injected experiment ) biological replicates and eight fourth instars per replicate were used for each group . Non-viruliferous fourth instar planthoppers were fed an artificial diet containing RSV crude preparations for 8 hr , and then raised on healthy rice seedlings . After 2 d , the insects were injected with 23 nL of dsRNA mixture of JNK1 and JNK2 at 6 μg/μL , or dsRNA of TNF-α at 6 μg/μL , or 0 . 9 μM JNK specific chemical inhibitor SP600125 ( Sigma-Aldrich ) , or 50 μM mouse TNF-α ( Sigma-Aldrich ) , and then raised on healthy rice seedlings for 3 d . For GPS2 knockdown , insects were injected with 23 nL of dsRNA of GPS2 at 6 μg/μL . Two days later the insects were fed artificial diet containing RSV crude preparations for 8 hr and then raised on healthy rice seedlings for 5 d . The transcript knockdown of JNKs , TNF-α , GPS2 and the RNA level of CP were measured using qRT-PCR . Insects injected with 23 nL of dsGFP-RNA at 6 μg/μL were used as controls . Control groups for SP600125 and TNF-α were injected with 23 nL of 2% DMSO or 20 mM Tris-HCl containing 0 . 1% Tween 20 , respectively . Six biological replicates and four insects per replicate were used for the dsRNA injection experiment and the pharmacological treatment experiment . After RSV was incubated in the GPS2 or JNKs knocked down planthoppers for 5 d , five insects were transferred to two new healthy rice seedlings for 24 hr and then removed from the plants . The plants were then placed in a greenhouse at 20°C or 26°C with 16 hr of light per day to allow observation of disease symptoms . Six groups of plants per replicate and four or five replicates were used to calculate the disease incidence rates on each day . Insects injected with 23 nL of dsGFP-RNA at 6 μg/μL were used as controls . | There are over a thousand different viruses that infect plants . Many plant viruses are transmitted by insects that feed on the plants , much as mosquitoes spread diseases between people when feeding on blood . Often the plant virus can replicate inside the cells of the insect . However , unlike in the plant hosts , the viruses do not seem to cause disease in the insects that carry them . Rice stripe disease is a major viral disease of rice that can reduce the crop’s yield by more than 50% in some areas . An insect called the small brown planthopper spreads the rice stripe virus between plants . Like other animals , insects have an immune system that protects them against viral infections . This means that the rice stripe virus must manipulate the planthopper’s immune system in order to replicate inside the insect’s cells . It was not clear how the virus did this , but answering this question could provide important clues to help scientists develop new ways to protect crops against plant viruses . Wang , Zhao , Li et al . now show that rice stripe virus manipulates its insect host to produce more of a protein called TNF-α and less of a protein called GPS2 . Moreover , a protein that makes up part of the virus also binds to GPS2 . This stops GPS2 from inhibiting a conserved signaling pathway that involves an enzyme known as JNK . When the JNK signaling pathway becomes active , replication of the rice stripe virus inside the insect is accelerated . Further experiments showed that inhibiting JNK made it harder for the virus to replicate , which meant that it took longer for the disease to develop in rice plants . These findings uncover a host of proteins that could be manipulated in insects to benefit rice agriculture . Such alterations could possibly be achieved through breeding or otherwise genetically modifying the insects to make them less able to carry viruses and then releasing them into wild populations . Alternatively , if further studies can identify chemicals that cause insect cells to alter the levels of the proteins , such chemicals could be administered to farmland to reduce the spread of viruses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | The c-Jun N-terminal kinase pathway of a vector insect is activated by virus capsid protein and promotes viral replication |
Chromatin remodeling processes are among the most important regulatory mechanisms in controlling cell proliferation and regeneration . Drosophila intestinal stem cells ( ISCs ) exhibit self-renewal potentials , maintain tissue homeostasis , and serve as an excellent model for studying cell growth and regeneration . In this study , we show that Brahma ( Brm ) chromatin-remodeling complex is required for ISC proliferation and damage-induced midgut regeneration in a lineage-specific manner . ISCs and enteroblasts exhibit high levels of Brm proteins; and without Brm , ISC proliferation and differentiation are impaired . Importantly , the Brm complex participates in ISC proliferation induced by the Scalloped–Yorkie transcriptional complex and that the Hippo ( Hpo ) signaling pathway directly restricted ISC proliferation by regulating Brm protein levels by inducing caspase-dependent cleavage of Brm . The cleavage resistant form of Brm protein promoted ISC proliferation . Our findings highlighted the importance of Hpo signaling in regulating epigenetic components such as Brm to control downstream transcription and hence ISC proliferation .
Central to the animal development is how chromatin assembly and regulation orchestrate cell-fate determination . Four epigenetic factors , DNA methylation , histone modifications , ATP-dependent chromatin remodeling , and the recently discovered non-coding RNAs play major roles in epigenetic regulation at the chromatin level . The SWI/SNF family is one of the most-studied families of ATP-dependent chromatin remodeling complexes , which regulate gene expression by destabilizing nucleosome structures to alter the DNA accessibility for transcription factors ( Cairns , 2007; Hargreaves and Crabtree , 2011 ) . Studies have implicated diverse roles for the mammalian SWI/SNF complexes in embryonic stem cell proliferation and differentiation . SWI/SNF complexes also function in neural , heart , and muscle development ( Bultman et al . , 2000; Lickert et al . , 2004; Ho et al . , 2009; Ho and Crabtree , 2010; Zhan et al . , 2011 ) . In Drosophila , there are two SWI/SNF complexes , the Brahma ( Brm ) -associated proteins ( BAP ) complex and the polybromo-containing BAP ( PBAP ) complex . The BAP complex has a signature subunit Osa , while PBAP complex is defined by BAP170 , Polybromo , and Syap ( Elfring et al . , 1998; Chalkley et al . , 2008 ) . Brm is a unique DNA-stimulated ATPase and common subunit for both BAP and PBAP complexes . Progress has been made in understanding the function of the Brm complex during Drosophila development ( Treisman et al . , 1997; Collins and Treisman , 2000; Janody et al . , 2004; Moshkin et al . , 2007; Carrera et al . , 2008; Terriente-Felix and de Celis , 2009; Neumuller et al . , 2011 ) , yet little is known about Brm complex functions in maintaining stem cell pluripotency of the epithelial tissues . The simplicity of the structure and the multipotency of Drosophila posterior midgut make it an excellent model to study adult epithelial tissue homeostasis and regeneration ( Micchelli and Perrimon , 2006; Ohlstein and Spradling , 2006 ) . The midgut is composed of four cell types: enterocytes ( ECs ) , enteroendocrine ( ee ) cells , intestinal stem cells ( ISCs ) , and enteroblasts ( EBs ) . The mature ECs are large polyploid cells of absorptive function and frame the midgut lining; ee and ISCs are the two types of diploid cells in the midgut that are less abundant . ISCs evenly locate at basal position underneath the ECs with a wedge-like morphology ( Ohlstein and Spradling , 2006 , 2007 ) and are the only known cell type in the posterior midgut that proliferates . On cell division , ISCs undergo self-renewal or proliferation to become EBs , quiescent progenitor cells that ultimately differentiate to ECs or ee cells with the ratio 9:1 under the control of Delta ( Dl ) and Notch ( Micchelli and Perrimon , 2006; Ohlstein and Spradling , 2006 ) . Since the active Dl expression is retained in self-renewed ISCs and is lost in the newly generated EBs , antibody against the active Dl is used as the specific and the only known marker for Drosophila ISCs ( Ohlstein and Spradling , 2007 ) . It has been demonstrated that the proliferation and differentiation of ISCs are tightly controlled by Notch , Janus kinase/signal transducer and activator of transcription ( JAK/STAT ) , epidermal growth factor receptor/mitogen-activated protein kinase ( EGFR ) , Hippo ( Hpo ) , and Wingless signaling pathways ( Jiang and Edgar , 2011 ) . The evolutionarily conserved Hpo pathway controls organ size by regulating cell proliferation and apoptosis ( Pan , 2010; Halder and Johnson , 2011; Yin and Zhang , 2011; Irvine , 2012 ) . Hpo is a serine/threonine Ste20-like kinase ( Harvey et al . , 2003; Jia et al . , 2003; Pantalacci et al . , 2003; Udan et al . , 2003; Wu et al . , 2003 ) that directly phosphorylates and activates downstream nuclear Dbf2-related ( NDR ) family protein kinase Warts ( Wts ) . Wts activation mediated by Hpo requires scaffold proteins Salvador ( Sav ) ( Kango-Singh et al . , 2002; Tapon et al . , 2002 ) and mob as tumor suppressor ( Mats ) ( Lai et al . , 2005 ) . Together , these proteins inhibit Yorkie ( Yki ) nuclear translocation . In the absence of Wts-mediated suppression , Yki forms a complex with transcription factor ( s ) such as Scalloped ( Sd ) ( Goulev et al . , 2008; Wu et al . , 2008; Zhang et al . , 2008 ) in the nucleus to regulate the expression of a plethora of genes involved in cell proliferation , cell cycle progression , and apoptosis ( Halder and Johnson , 2011; Yin and Zhang , 2011; Irvine , 2012 ) . In addition , the Hpo pathway maintains tissue homeostasis by regulating the balance between diap1 expression and basal levels of activated caspases via the control of Dronc ( Drosophila Nedd-2-like caspase orthologous to human Caspase 9 ) ( Verghese et al . , 2012 ) . We present evidence that Brm is required for ISC proliferation in both normal and regenerating midguts , and it is required in ISCs for EC differentiation in normal midguts . In addition , we show that the Brm complex is physically associated with the Sd–Yki transcriptional complex in the nucleus and functions downstream of the Hpo pathway to regulate ISC proliferation . We also demonstrate that Brm is regulated by the Hpo pathway at the protein level through Hpo kinase-induced , caspase-dependent , cleavage of Brm at its D718 site . Altogether , as exemplified in the Drosophila ISCs , our study unravels a novel mechanism of the chromatin remodeling Brm complex in maintaining adult stem cell pluripotency of epithelial tissues .
To gain insights on homeostasis and proliferation of Drosophila midguts , a small-scale screen searching for candidates that genetically alters the midgut regeneration and homeostasis was carried out . During the screen , escargot-Gal4 ( esg-Gal4 ) was used to drive RNAi expressions of different genes in ISCs and EBs in the presence of a temperature-sensitive Gal4 repressor , tubGal80 ( henceforth esg80ts ) . Adult esg80ts flies grown at the permissive temperature do not express GFP or RNAi in ISCs and EBs . Once shifted to the non-permissive temperature , RNAi expressions in ISCs and EBs are induced and simultaneously marked by esg-Gal4-driven GFP signals ( Micchelli and Perrimon , 2006 ) . Interestingly , among the RNAi lines , VDRC ( 37720 ) and Bloomington ( 31712 ) abolished the expression of Brm , the energy providing subunit in Drosophila Brm complex ( Neumuller et al . , 2011; Waldholm et al . , 2011 ) . On Brm RNAi expression , the number of GFP positive ( GFP+ ) cells in the adult posterior midgut decreased . Concomitantly , the number of phospho-histone3 positive ( PH3+ ) cells also reduced , suggesting that ISC proliferation is affected ( compare Figure 1B , B′ with Figure 1A , A′ , also Figure 1E ) . Immunostaining using an antibody against Brm 505–775 aa ( Elfring et al . , 1998 ) confirmed that endogenous Brm protein can be efficiently knocked down in the cells of both wing imaginal discs and midguts that express Brm RNAi transgenes ( compare Figure 1—figure supplement 1B–B′′ , C–C′ with Figure 1—figure supplement 1A; and compare Figure 1—figure supplement 1E–E′′′ with Figure 1—figure supplement 1D–D′′′ ) . In addition , GFP+ cells exhibited a spherical shape in the absence of Brm compared with the angular shaped control cells ( compare Figure 1B , B′ with Figure 1A , A′ ) , suggesting that the attachment of GFP+ cells to surrounding cells is disrupted . We further tested whether knockdown of Brm in ISCs/EBs affects the division of ISCs . On Brm RNAi expression , EBs in the intestinal epithelium labeled with the expression of the Suppressor of Hairless reporter ( Su ( H ) -LacZ , a specific marker for EBs ) ( Micchelli and Perrimon , 2006 ) were detected ( compare Figure 1—figure supplement 1G–G′′ with Figure 1—figure supplement 1F–F′′ ) . This piece of evidence suggests that EBs are still formed even when Brm expression is inhibited and ISC proliferation is blocked . Expression of BrmK804R , a dominant negative form of Brm defective for ATP hydrolysis activity without affecting the complex assembly ( Elfring et al . , 1998 ) , results in similar effects compared to Brm RNAi ( Figure 1C , C′ and Figure 1E ) . Of note , we observed a mild increase in the ISC/EB numbers on Brm overexpression , and the PH3+ cell number was slightly increased ( compare Figure 1D , D′ with Figure 1A , A′ , E ) . 10 . 7554/eLife . 00999 . 003Figure 1 . Brm is required for ISC proliferation in midguts . ( A–D′ ) Adult fly midguts expressing esg80ts-Gal4/UAS-GFP ( esg80ts ) ( A and A′ ) , Brm RNAi ( esg80ts-Brm RNAi ) ( B and B′ ) , esg80ts-Gal4/UAS-GFP-BrmK804R ( esg80ts-BrmK804R ) ( C and C′ ) or esg80ts-Gal4/UAS-GFP-Brm ( esg80ts-Brm ) ( D and D′ ) were immunostained with DAPI ( blue ) . ISCs and EBs were marked by esgGal4-driven GFP expression . ( E ) Quantification of PH3+ cells of adult midguts of the indicated genotypes . The results represent the mean ± SEM , n = 10 for each genotype . ( F–G′ ) Adult midguts containing nuclear localized GFP-labeled control MARCM clones ( F and F′ ) or brm null allele brm2 clones ( G and G′ ) were immunostained for DAPI ( blue ) . Guts were dissected from the adult flies 72 hr after clone induction . ( H ) Quantification of the cell numbers of the control or mutant clones of the indicated genotypes . The results represent the mean ± SEM , n = 10 for each genotype . See also Figure 1—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 00310 . 7554/eLife . 00999 . 004Figure 1—figure supplement 1 . Brm is required for ISC proliferation . ( A–C′ ) The efficiency of Brm antibody was tested by immunostaining of the endogenous Brm in wild-type wing discs ( A ) or in the discs expressing Brm RNAi V37720 ( B–B′′ ) and B31712 ( C and C′ ) in the posterior compartment using hhGal4 driver . Discs were immunostained for Brm ( green ) and Ci ( blue ) . ( D–E′′′ ) Adult flies expressing esg80ts ( D–D′′′ ) , esg80ts-Brm RNAi ( E–E′′′ ) were cultured at 29°C for 7 days . Midguts were dissected and immunostained for Brm ( red ) and DAPI ( blue ) . ( F–G′′ ) Flies of Su ( H ) Z controls ( F–F′′ ) or flies expressing Brm RNAi in the ISCs/EBs ( G–G′′ ) were cultured at 29°C for 7 days . Dl is detected by immunostaining ( red ) . Su ( H ) -lacZ staining identifies the EBs with elevated Notch signaling ( green ) . Cells that retain ISC identity ( small nuclei , Dl positive and lacZ-negative ) are indicated by yellow arrows , and EBs are indicated by white arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 00410 . 7554/eLife . 00999 . 005Figure 1—figure supplement 2 . Brm complex is required for ISC proliferation . ( A–B′ ) Adult midguts containing nuclear localized GFP-labeled control MARCM clones ( A and A′ ) or brm null allele brm2 clones ( B and B′ ) were immunostained with DAPI ( blue ) . Guts were dissected from the adult flies 8 days after clone induction . ( C ) Quantification of the cell numbers of the indicated control clones or mutant clones . The results represent the mean ± SEM , n > 10 for each genotype . ( D–J′ ) Subunits of Brm complex function in ISC proliferation . Adult midguts expressing Bap60 ( E and E′ ) , Bap60 RNAi ( NIG 4303R-1 , F and F′ ) , Mor ( G and G′ ) , Mor RNAi ( VDRC 6969 , H and H′ ) , Osa ( I and I′ ) and Osa RNAi ( VDRC 7810 , J and J′ ) with esg80ts driver were immunostained with DAPI ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 005 Interestingly , ISC/EB reduction induced by the loss of Brm might be due to an alternation in the rate of proliferation and differentiation . We hypothesized that the loss of Brm might result in an inhibition of ISC proliferation , precocious ISC differentiation , or a blockage of ISC differentiation . To test these possible mechanisms , the Mosaic analysis with a repressible cell marker ( MARCM ) approach ( Lee and Luo , 2001 ) was used to generate brm null allele ( brm2 ) clones , and its impact on midgut proliferation was analyzed ( Micchelli and Perrimon , 2006; Ohlstein and Spradling , 2006 ) . The wild-type MARCM stem cell clones divided indefinitely , their sizes increased linearly , and contained several or all midgut cell types ( Micchelli and Perrimon , 2006; Ohlstein and Spradling , 2006 ) . If Brm is important for ISC proliferation , the brm2 clones will be retained in a limited size; if Brm is necessary for EC differentiation , the brm2 clones should mostly contain the small nuclear ISCs/EBs . Compared with the control clones that contain an average of five cells including both large nuclear cells and small nuclear cells within each clone , 3-day brm2 clones contain only one or two cells , which are all small nuclear cells ( Figure 1H , and compare Figure 1G , G′ with Figure 1F , F′ ) . In addition , 8-day brm2 clones contain only one or two cells ( Figure 1—figure supplement 2A , A′ , B , B′ , C ) . These results suggest that both proliferation of these clones and the EC differentiation are affected , suggesting that Brm is indispensable for ISC proliferation and EC differentiation in midguts . We further tested the function of other subunits of the Brm complex in ISC proliferation . We found that the knockdown of other components in the Brm complex , including Bap60 , Mor , and Osa by RNAi respectively under the control of esg80ts inhibited ISC proliferation to different extents and the GFP signal intensities were reduced simultaneously ( compare Figure 1—figure supplement 2F , F′ , H , H′ , J , J′ with Figure 1—figure supplement 2D , D′ ) . Similar to Brm overexpression , overexpression of other Brm complex components induced only a mild enhancement on midgut ISC proliferation ( compare Figure 1—figure supplement 2E , E′ , G , G′ , I , I′ with Figure 1—figure supplement 2D , D′ ) . In toto , these findings indicate that the maintenance of ISCs and EBs requires the presence of Brm complex . Our results indicated that brm2 clones only contained small nuclear cells ( Figure 1G , G′ ) , suggesting that Brm plays a role during ISC differentiation in addition to ISC proliferation . We first analyzed the expression pattern of Brm during ISC cell maturation using Myo1AGal4-GFP ( Morgan et al . , 1994 ) . Myo1AGal4 is an enhancer trap in the gut-specific brush border myosin 1A gene that combined tubGal80ts with the Myo1AGal4 driver and UAS-GFP ( together referred to as Myo1A-GFP ) . Interestingly , Brm antibody staining detected a high level of endogenous Brm proteins in ISCs/EBs ( GFP− cells in Figure 2A–A′′′ and GFP+ cells in Figure 2B–B′′′ ) , and some ee cells ( co-labeled by prospero , a conserved homodomain transcription factor ) , whereas a relatively low level of Brm protein was detected in ECs ( GFP+ cells in Figure 2A–A′′′ ) . 10 . 7554/eLife . 00999 . 006Figure 2 . Brm is required for EC differentiation . ( A–B′′′ ) Adult guts of wild-type Myo1A-Gal4/UAS-GFP;tubGal80ts ( A–A′′′ ) and esgGal4/UAS-GFP ( B–B′′′ ) were immunostained with Brm antibody ( indicated with arrows ) to show the endogenous Brm protein level in the different cell types . ( C–H′ ) Adult female midguts differentiation measured via the esgts F/O system . Transgenes were induced for 2 days ( C–E′ ) or 5 days ( F–H′ ) . esgts F/O-Brm ( D , D′ and G , G′ ) promoted the formation of ECs , while esgts F/O-Brm RNAi ( E , E′ and H , H′ ) blocked the EC differentiation . ECs are marked by PDM-1 ( red ) and arrows . ( I ) Female posterior midguts were scored for GFP+ and PDM-1+ EC cells in the same region near the Malpighian tubules . The results represent the mean ± SEM , n = 10 for each genotype . ( J ) A schematic diagram of the regulation of Brm activity in intestinal homeostasis . ISCs divide asymmetrically to an EB and an ISC . EBs then differentiate into ECs or ee cells . Cell-type-specific markers are indicated . In normal state ( left side ) , Brm is expressed at a high level in nuclei of ISCs , EBs , and some ee cells , and at a low level in nuclei of ECs . The different Brm protein levels in nuclei are marked by red ( ISCs , EBs , and ee cells ) or pink ( ECs ) . Decrease of Brm protein level in ISCs reduces the ISC proliferative ability and inhibits EC differentiation ( right ) . See also Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 00610 . 7554/eLife . 00999 . 007Figure 2—figure supplement 1 . Brm is required for EC differentiation . ( A–C′ ) The differentiation of adult female midguts was analyzed using the esgts F/O system . Transgenes were induced in the midgut for 13 days . esgts F/O-Brm ( B and B′ ) promoted EC formation , while esgts F/O-Brm RNAi ( C and C′ ) blocked the EC differentiation . PDM-1 marked the EC cells ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 007 On the basis of these findings , we examined the role of Brm in ISC differentiation by overexpression or knockdown of Brm in the ISCs using a lineage induction system , esgts F/O . In this lineage tracing system , progenitor cells and their newborn progenies express Gal4 and UAS-linked Gal4 targets , including the UAS-GFP marker ( Jiang et al . , 2009 ) . PDM-1 , a marker for fully differentiated ECs , is used to identify ECs ( Xu et al . , 2011 ) . Overexpressing Brm for 2 days generated new EC-like GFP+ cells with large nuclei ( Figure 2D , D′ ) , whereas the wild type control group and the Brm RNAi group only contained GFP+ cells with small nuclei ( Figure 2C , C′ , E , E′ , I ) . It is implicated that high levels of Brm lead to precocious differentiation of ISCs . After 5-day or even 13-day induction , large mature ECs were formed in wild-type midguts , while Brm RNAi suppressed ISC proliferation and EC differentiation in experimental midguts ( compare Figure 2H , H′ with Figure 2F , F′ , I , and Figure 2—figure supplement 1A–C′ ) , suggesting that Brm is essential for ISCs and EBs to differentiate into ECs . In summary , the knockdown of Brm by RNAi blocks ISC proliferation and EC differentiation . Interestingly , in addition to its role in ISC proliferation under normal physiological context , Brm is also required for damage-induced midgut regeneration . Previous studies have reported that the feeding of dextran sulphate sodium ( DSS ) causes midgut cell proliferation via the disruption of basement membrane organization and increases in the intestinal stem cell division without affecting the final EB differentiation ( Amcheslavsky et al . , 2009 ) . It is plausible to think that Brm also exerts an effect on DSS-induced midgut cell proliferation , as it is required for midgut cell proliferation . Indeed , when Brm RNAi was expressed , DSS-induced increase of GFP+ cells was blocked ( compare Figure 3C , C′ , D , D′ with Figure 3A , A’ , B , B′ ) , suggesting that Brm is required in ISCs for DSS-induced proliferation . Of note , we did not observe dramatic change in ISC proliferation when overexpressing Brm in these GFP+ cells with or without DSS treatment ( compare Figure 3E , E′ , F , F′ with Figure 3A , A′ , B , B′ ) . 10 . 7554/eLife . 00999 . 008Figure 3 . Brm was required for midgut regeneration . ( A–F′ ) Adult flies expressing esg80ts-Gal4/UAS-GFP ( esg80ts ) ( A–B′ ) , Brm RNAi ( esg80ts-Brm RNAi ) ( C–D′ ) or esg80ts-Gal4/UAS-GFP-Brm ( esg80ts-Brm ) ( E–F′ ) were treated with glucose or DSS . Glucose solution with 3% DSS was fed to the flies ( B–B′ , D–D′ , and F–F′ ) for 3 days before guts dissection . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 008 Previous studies implicated that the Hpo pathway effector Yki functions as a driver of proliferation in both ECs and ISCs and damage-induced ISC proliferation via both cell-autonomous and non-cell-autonomous mechanisms ( Karpowicz et al . , 2010; Ren et al . , 2010; Shaw et al . , 2010 ) . Considering that Brm is involved in DSS-induced ISC proliferation ( Figure 3 ) , we tested whether there is a functional link between Brm and Yki–Sd transcriptional complex . To this end , mass spectrum ( MS ) analysis was first performed . Co-immunoprecipitation ( Co-IP ) experiments were performed in S2 cells to pull down the endogenous Yki or Sd protein using antibodies , and the pull-down products were sent for MS analysis . Several Brm complex components were found in the MS results , including Brm , Osa , Bap60 , Bap55 , and Mor ( Table 1 , see Table 1 for MS details ) . Consistent with the results of Yki MS analysis ( Table 1 ) , we found that Yki and Brm coimmunoprecipitated with each other when Myc-tagged Yki ( Myc–Yki ) and V5-tagged Brm ( Brm–V5 ) were coexpressed in S2 cells ( Figure 4A ) . We also verified the interaction between Brm and Sd using Co-IP in S2 cells . Results showed that overexpressed HA-tagged Sd ( HA–Sd ) interacted with the endogenous Brm ( Figure 4B ) . Sd also coimmunoprecipitated with Mor and Osa but not Bap60 when they were coexpressed in S2 cells ( Figure 4—figure supplement 1A ) . In addition , we checked the cellular localization of these proteins in S2 cells . The majority of overexpressed Brm and overexpressed Sd were located in the nucleus ( Figure 4—figure supplement 1B–D′′′ ) , whereas cytoplasmic–nuclear localization of Yki was not affected by Brm coexpression ( data not shown ) , implicating that Brm complex does not promote the nuclear localization of Yki to influence the transcriptional activity of Yki–Sd complex . 10 . 7554/eLife . 00999 . 009Table 1 . Mass spectrum analysis resultsDOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 009Protein descriptionMolecular functionPep countUnique Pep countYki mass spectrum Brahma ( Brm ) ATP-dependent helicase115 OsaDNA binding54Sd mass spectrum Brahma associated protein 55kD ( Bap55 ) Structural constituent of cytoskeleton85 Brahma associated protein 60kD ( Bap60 ) Protein binding53 Brahma ( Brm ) ATP-dependent helicase44 Brahma associated protein 155 kDa ( Mor ) Protein binding11To determine whether there are physical interactions between Yki/Sd transcriptional complex and Brm complex and gain further understanding of the regulation mechanism of Brm in regulating ISC proliferation , we immunoprecipitated endogenous Sd or Yki protein in S2 cells using generated rabbit anti-Sd or anti-Yki antibodies , respectively , followed by mass spectrometry ( MS ) analysis . The corresponding proteins of Brm complex identified in association with Yki ( Yki mass spectrum ) or Sd ( Sd mass spectrum ) are listed with the number of peptides identified by mass spectrometry . 10 . 7554/eLife . 00999 . 010Figure 4 . Sd and Yki interact with Brm complex components . ( A ) Interaction between overexpressed Myc–Yki and Brm–V5 was detected in S2 cells . Myc–Yki or Brm–V5 was immunoprecipitated with anti-Myc or anti-V5 antibodies . ( B ) Association between HA–Sd and endogenous Brm in vitro . S2 cells were transfected with the HA–Sd . The arrow indicated HA–Sd coimmunoprecipitated with endogenous Brm . ( C–H ) Wild-type male wings ( C ) or hemizygous male wings of null allele brm2/+ ( D ) , or hypomorphic allele sd1/Y ( E ) , or double-mutant combinations of sd1/Y; brm2/+ ( F ) , or hypomorphic allele osa2/+ ( G ) , or combinations of sd1/Y; osa2/+ ( H ) . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 01010 . 7554/eLife . 00999 . 011Figure 4—figure supplement 1 . Brm complex associates with Sd . ( A ) Sd interacted with Brm , Osa in one direction while in both directions with Mor . The interaction between Sd and Bap60 was not detected . The asterisk marked the band of heavy chain of IgG and the arrow marked the band of Osa . ( B–D′′′ ) Overexpressed Sd and Brm localized in the nuclei of S2 cells . Cells were immunostained with indicated antibodies , HA ( red ) , Flag ( green ) and DAPI ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 011 The genetic interaction between brm mutant ( brm2 ) and sd hypomorphic allele ( sd1 ) in adult fly wings was examined . Strong mutations in sd cause lethality , while hypomorphic mutant sd1 flies are viable with a scalloped wing phenotype ( compare Figure 4E with Figure 4C ) . Single-mutant brm2 fly wings are normal ( Figure 4D ) . Interestingly , the double-mutant combination of sd1 and brm2 shows a strong enhancement of the sd1 phenotype ( compare Figure 4F with Figure 4E ) . A similar enhanced phenotype was also found in the flies with sd1 and osa2 , a hypomorphic allele of a Brm complex subunit ( compare Figure 4H with Figure 4E , G ) . These observations indicate that brm and osa genetically interact with sd and contribute to the wing vein alternation phenotype . Together with the biochemical results , these results suggest that Brm complex plays a crucial role in Yki–Sd mediated function . To further test whether Yki-mediated ISC proliferation depends on Brm , we examined the requirement of Brm activity during Yki–Sd induced ISC proliferation . Overexpression of either Yki or SdGA , an active form of Sd ( Zhang et al . , 2008 ) , under the control of esg80ts resulted in an increase in GFP+ and PH3+ cell numbers ( compared Figure 5C , C′′ , E , E′′ with Figure 5A , A′′ , K ) , suggesting an enhancement of ISC proliferation . Interestingly , Yki overexpression resulted in pronounced hyperplasia of intestine with a thicker intestinal epithelium composed of a multi-layer of tightly packed cells ( Staley and Irvine , 2010 ) ( compare Figure 5H with Figure 5G ) , whereas SdGA expression did not induce such a phenomenon ( compare Figure 5J with Figure 5G ) , suggesting that Yki and Sd may have distinct mechanisms in regulating ISC proliferation . When Brm was knocked down , ISC proliferation was greatly suppressed ( Figure 5K , and compare Figure 5B , B′′ , D , D′′ , F , F′′ with Figure 5A , A′′ , C , C′′ , E , E′′ ) with a decreased Dl signal intensity ( Figure 5B′ , D′ ) , and the formation of thicker intestinal epithelium induced by Yki overexpression was inhibited ( Figure 5I ) . Moreover , similar results were obtained by MARCM analysis of brm2 . Overexpression of Yki in control MARCM clones resulted in a significant increase in the cell numbers and in the formation of large clones ( Figure 5L , L′ , N ) , whereas this Yki-induced proliferation was completely blocked in the brm2 clones ( Figure 5M , M′ , N ) . Taken these results together , the depletion of Brm compromised Yki or SdGA overexpression induced ISC proliferation , indicating that Brm functions downstream of Yki–Sd to maintain ISC proliferative ability . 10 . 7554/eLife . 00999 . 012Figure 5 . Knockdown of Brm blocks Yki/SdGA-induced ISC proliferation . ( A–F′′ ) Adult flies expressing esg80ts ( A–A′′ ) , esg80ts-Brm RNAi ( B–B′′ ) , esg80ts-Yki ( C–C′′ ) , esg80ts-Yki+Brm RNAi ( D–D′′ ) , esg80ts-SdGA ( E–E′′ ) , esg80ts-SdGA +Brm RNAi ( F–F′′ ) were cultured at 29°C for 8–9 days . Midguts were dissected and immunostained for Dl ( red ) and DAPI ( blue ) . White arrows indicated the EBs , and yellow arrowheads indicated the ISCs . ( G–J ) Images show an optical cross-section through the center of the intestine , DAPI ( green ) . ( K ) Quantification of PH3 positive mitotic cells of the indicated guts . The results represent the mean ± SEM , n = 10 for each genotype . ( L–M′ ) Adult midguts containing nuclear localized GFP-labeled control non-tagged form of Yki overexpressed clones ( L and L′ ) or Yki plus brm2 clones ( M and M′ ) were immunostained for Yki ( red ) and DAPI ( blue ) . Guts were dissected from the adult flies 72 hr after clone induction . ( N ) Quantification of the cell number of Yki or Yki+brm2 clones . 10 guts were counted for each genotype . ( O–P′′′ ) Adult guts of Myo1A-Gal4 UAS-GFP;tubGal80ts control ( O–O′′′ ) or expressing Myo1A-Gal4 UAS-GFP;tubGal80ts-Yki ( P–P′′′ ) were immunostained for Brm ( red ) , Dl ( purple ) , and DAPI ( blue ) . Arrows indicated ISCs with a high endogenous Brm protein level . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 012 Interestingly , when Yki was expressed in ECs using Myo1A-Gal4 to induce non-autonomous ISC proliferation , the number of ISCs/EBs was increased with high levels of Brm in the nucleus ( compare Figure 5P–P′′′ with Figure 5O–O′′′ ) , suggesting that Yki-induced non-autonomous ISC proliferation also induces high levels of Brm in nuclei of ISCs and EBs . Given that Brm physically interacts with Yki–Sd complex and that the function of Yki–Sd in ISC proliferation requires Brm activity , we sought to determine the underlying mechanism by which the Hpo pathway and Brm regulate ISC proliferation . Interestingly , cotransfection of Brm and Hpo in S2 cells resulted in a lower Brm protein level , suggesting that Brm was destabilized in the presence of Hpo ( Figure 6—figure supplement 1A ) . This result raised the concern that the Hpo pathway might regulate Brm activity by controlling its protein stability . Furthermore , Brm cleavage event in which a small protein band at about 100 kD was detected ( Figure 6A ) in the presence of Hpo upon MG132 treatment . To detect whether this small band represents a cleaved Brm fragment , we generated a Brm construct with a Flag tag at the N-terminus and a V5 tag at the C-terminus ( referred to as Flag-Brm-V5 ) . Using this construct , we were able to identify a Flag-tagged N-terminal cleavage product about 100 kD and a V5-tagged C-terminal cleavage product about 130 kD in the presence of Hpo ( Figure 6A ) . Considering that the molecular weight of full length Brm is about 230 kD , it is possible that Hpo regulates Brm stability by inducing Brm cleavage at only one site . We also found that this cleavage action depended on Hpo protein in a dose-dependent manner , since increasing the dose of Hpo plasmids resulted in an accumulation of the cleaved Brm product and a decrease in the full length Brm protein ( Figure 6B ) . A truncation of Hpo without kinase activity ( Hpo-C ) did not induce such cleavage ( Figure 6—figure supplement 1B , lanes 1 and 2 ) , indicating that Hpo kinase domain but not C-terminal regulatory domain induces Brm cleavage . In addition , Hpo-induced Brm cleavage was blocked in the presence of Yki–Sd ( Figure 6C ) , suggesting that it was regulated by downstream events of the Hpo signaling pathway . 10 . 7554/eLife . 00999 . 013Figure 6 . Brm is cleaved at the D718 site by Hpo-induced caspase . ( A ) Flag-Brm-V5 was transfected with or without Myc-Hpo . Western blots ( anti-Flag or anti-V5 ) of IP samples were performed to detect the N- or C-terminus of Brm . MG132 was treated 6 hr before harvesting the cells . Arrows indicated the full length Brm ( top ) and the N- ( bottom ) , C- ( middle ) terminal cleaved product of Brm . ( B ) 3 μg of Flag-Brm was cotransfected with different dosages of Hpo plasmids in S2 cells , MG132 was treated 6 hr before harvesting the cells . ( C ) Cotransfected Flag-Brm and Myc-Hpo together with Sd/Yki or in the presence of caspase inhibitor Z-VAD-FMK , the cleaved Brm fragments were unable to be detected . Z-VAD-FMK was added to a final concentration of 10 mM for 6 hr . ( D ) S2 cells were transfected with Myc-Hpo and Flag-Brm with HA-Diap1 . ( E ) Flag-Brm and Myc-Hpo were cotransfected in S2 cells to induce the cleavage of Brm . Inhibitors of Caspase 3 , 8 , 9 , 10 were added to block the cleavage in a final concentration of 10 mM for 6 hr . Asterisk indicates IgG bands ( loading control ) . ( F ) BrmD718A mutation blocked Hpo-induced Brm cleavage in S2 cells . ( G ) A schematic representation of Brm deletions and mutations . Brm-D1 to D4 were the deletions that were used to map the cleavages site of Brm . Brm-D718A/D726A/D728A/D731A/D740 , 741A were the mutants generated for mapping the cleavage sites . The novel caspase recognition motif ( DATD ) in Brm is indicated by a single blue underline including D718 residue . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 01310 . 7554/eLife . 00999 . 014Figure 6—figure supplement 1 . Brm is cleaved by Hpo-induced caspase . ( A ) The protein level of Brm was decreased when cotransfected with Myc-Hpo ( lane two ) , but not Myc-Yki ( lane three ) . ( B ) Hpo N-terminal kinase domain 1–342 aa ( Hpo-N ) induced Brm cleavage , but not Hpo C-terminal regulatory domain 343 aa-end ( Hpo-C ) ( lane one and lane two ) . Brm deletions , D1 and D2 , failed to produce the 100 KD band when Hpo existed ( lane three and lane four ) . Brm deletions were immunoprecipitated with anti-Flag antibody . ( C ) Brm deletion D3 still was cleaved when Hpo was cotransfected , but the D4 could not . The full length Brm and the N-terminal cleaved fragment of Brm are marked by arrows . ( D–E′′′ ) Overexpressing Hpo rather than Hpo kinase dead ( Hpo-KD ) induced the activated caspase 3 signal . Wing discs expressing wild-type UAS- Hpo ( D–D′′′ ) or UAS-Hpo-KD ( E–E′′′ ) under control of hhGal4 driver were immunostained with activated casepase 3 antibody ( Cas3 , green ) , Flag antibody ( red ) and Ci antibody ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 014 On the basis of above observations , it is feasible to consider that Hpo activates downstream caspases to cleave Brm . In fact , previous studies have implicated a role of the Hpo signaling pathway in caspase activation and cell apoptosis ( Verghese et al . , 2012 ) . This function of Hpo was verified by detecting active caspase 3 expression in wing discs overexpressing Hpo or Hpo-KD ( a kinase dead form of Hpo ) . As shown in Figure 6—figure supplement 1D–E′′′ , the cleaved caspase 3 , a functional read-out of initiator caspase activity , was detected only in Hpo overexpressed discs but not in the Hpo-KD overexpressed discs . Taken together , we speculated that Hpo-induced caspase activity might regulate the protein level of Brm . To verify this idea , we used the unspecific caspase inhibitor , Z-VAD-FMK , to test whether the activities of caspases are required for Hpo-induced Brm cleavage . We found that the 100 kD cleaved Brm product disappeared on the addition of Z-VAD-FMK ( Figure 6C ) , suggesting that the inhibition of caspase activities blocks Brm cleavage . Furthermore , it is known that the Drosophila inhibitor of apoptosis protein , Diap1 , which is a transcription product of the Hpo pathway target genes ( Zhang et al . , 2008 ) , inhibits caspase activity . Diap1 was cotransfected with Brm and Hpo in S2 cells to inhibit caspase activity . Interestingly , we found that Diap1 cotransfection inhibited Hpo-induced Brm cleavage ( Figure 6D ) , indicating that the Hpo regulates Brm cleavage by inducing caspase activity . To further study the function of caspases during Brm cleavage in details , Hpo and Brm were cotransfected in S2 cells in the presence of inhibitors of mammalian caspase 3 , 8 , 9 , 10 , respectively ( no commercial Drosophila caspase inhibitors were available ) . As shown in Figure 6E , the addition of inhibitor of caspase 3 or caspase 10 completely abolished Hpo-induced Brm cleavage , whereas the addition of other caspase inhibitors only partially affected the cleavage reaction as revealed by the presence of the 100 kD Brm protein fragment . Caspase 10 is an initiator in the extrinsic death-receptor-mediated cell death ( Wachmann et al . , 2010 ) , and caspase 3 is the effector caspase generally believed to carry out the cleavage of nuclear protein substrates . These results suggest that Drosophila homologs of caspase 3 and caspase 10 play important roles in Hpo-induced Brm cleavage . In an attempt to identify the cleavage site of Brm , two Brm deletion forms , D1 ( Δ601–800 aa ) and D2 ( Δ694–768 aa ) , were generated based on previous observations of N- and C-terminal cleavage products ( Figure 6G ) . No cleavage reaction was detected for these two Brm deletion forms ( Figure 6—figure supplement 1B ) . Mapping of the D2 form using two other deletion forms of Brm , D3 ( Δ694–711 aa ) and D4 ( Δ712–729 aa ) , indicated that D3 was cleaved but not D4 ( Figure 6—figure supplement 1C ) , suggesting that the cleavage site locates within the region of amino acid 712–729 . Although no canonical caspase 3 tetra-peptide cleavage site DEVD was found in this region , several aspartic acids that potentially serve as the caspase cleavage sites were identified . To validate these sites , individual aspartic acids were mutated to alanine separately . Interestingly , Brm mutant carrying aspartic acids to alanine mutation at D718 site ( BrmD718A ) does not undergo cleavage ( Figure 6F ) . In conclusion , Brm protein stability was regulated by Hpo-induced caspase-dependent cleavage at the D718 site . Given the finding that BrmD718A was a cleavage resistant Brm mutant ( Figure 6F ) , we wondered whether BrmD718A is an active form of Brm . To test the function of BrmD718A , we expressed BrmD718A under the control of esg80ts in ISCs/EBs . An upregulation of ISC/EB ( GFP+ ) and PH3+ cell numbers was detected in guts expressing BrmD718A mutant ( compare Figure 7B–B′ with Figure 7A–A′ , J ) , whereas expressing wild-type Brm induced a mild increase in the ISC/EB numbers and PH3+ cell numbers ( Figure 7G , G′ , J ) . On coexpression of BrmD718A and Yki , the number of PH3+ cells was further increased , suggesting that ISC proliferation is promoted ( Figure 7C–D′ , J ) . Furthermore , analysis of 5-bromodeosyuridine ( BrdU ) incorporation in midguts showed that BrmD718A overexpression greatly enhanced BrdU ectopic expression , whereas Brm RNAi resulted in a lower proliferative activity ( Figure 7—figure supplement 1A–D′ ) . Altogether , these results indicate that BrmD718A promotes ISC proliferation . 10 . 7554/eLife . 00999 . 015Figure 7 . The cleavage resistant mutant BrmD718A promotes ISC proliferation with antagonistic ability against Hpo activity . ( A–H′ ) Adult guts of esg80ts control ( A and A′ ) , esg80ts-BrmD718A ( B and B′ ) , esg80ts-Yki ( C and C′ ) , esg80ts-Yki+BrmD718A ( D and D′ ) , esg80ts-Hpo ( E and E′ ) , esg80ts- BrmD718A +Hpo ( F and F′ ) , esg80ts-Brm ( G and G′ ) and esg80ts-Brm+Hpo ( H and H′ ) were immunostained for DAPI ( blue ) . ( I–I′′′ ) Adult midguts containing GFP-labeled MARCM clones of hpo null allele ( BF33 ( 16 ) ) . White arrows indicate the ECs in the BF33 ( 16 ) clones , and yellow arrowheads indicate the ECs outside the clones . ( J ) Quantification of PH3 positive mitotic cells of the indicated guts . The results represent the mean ± SEM , n > 10 for each genotype . ( K ) A model of the regulation of Brm protein stability by the Hpo pathway . The Hpo pathway restricts Brm protein level by inducing the activation of caspase to cleave Brm and/or by inhibiting the expression of Yki–Sd target genes , especially diap1 that inhibits the caspase activity . See also Figure 7—figure supplements 1–4 . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 01510 . 7554/eLife . 00999 . 016Figure 7—figure supplement 1 . The cleavage resistance mutant BrmD718A promotes ISCs proliferation . ( A–D′ ) Adult guts of esg80ts-Gal4 control ( A and A′ ) , esg80ts-Brm ( B and B′ ) , esg80ts-Brm RNAi ( C and C′ ) , esg80ts-BrmD718A ( D and D′ ) were immunostained for BrdU ( red ) . Note that BrmD718A increased the BrdU number . ( E and F′ ) Adult midguts containing GFP-labeled MARCM clones of hpo null allele BF33 ( 16 ) + Brm ( E–E′ ) or hpo null allele BF33 ( 16 ) + Brm RNAi ( F–F′ ) . Brm RNAi blocked the cell proliferation in hpo null allele clones . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 01610 . 7554/eLife . 00999 . 017Figure 7—figure supplement 2 . The cleavage products of Brm have low activity in promoting ISC proliferation . ( A–J ) Adult guts of esg80ts control ( A ) , esg80ts-Brm ( B ) , esg80ts-BrmD718A ( C ) , esg80ts- esg80ts-Brm-N ( D ) , esg80ts-Brm-C ( E ) , esg80ts-Yki ( F ) , esg80ts-Yki+Brm ( G ) , esg80ts-Yki+BrmD718A ( H ) , esg80ts-Yki+Brm-N ( I ) and esg80ts-Yki+Brm-C ( J ) were immunostained for PH3 ( red ) and DAPI ( blue ) . ( K ) Quantification of PH3 positive mitotic cells of the indicated guts . The results represent the mean ± SEM , n > 10 for each genotype . ( L ) Quantification of the cell number of the MARCM clones of the indicated genotypes . Guts were divided into two groups after clone induction: 3 days and 10 days . The results represent the mean ± SEM , n > 10 for each group . ( M–R ) Adult midguts containing nuclear localized GFP-labeled wild-type control clones ( M ) , brm null allele brm2 clones ( N ) , Brm+brm2 clones ( O ) , BrmD718A+brm2 clones ( P ) , Brm-N+brm2 clones ( Q ) and Brm-C+brm2 clones ( R ) were immunostained to show the DAPI ( blue ) . Guts were dissected from the adult flies 10 days after clone induction . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 01710 . 7554/eLife . 00999 . 018Figure 7—figure supplement 3 . Genetic interaction assays between Brm and Yki/Sd in Drosophila eyes . ( A–F′ ) Genetic interaction assays between Brm and Yki in Drosophila eyes . BrmD718A further increased Yki-induced eye overgrowth . Adult eyes of GMR ( A and A′ ) , GMR-Brm ( B and B′ ) , GMR- BrmD718A ( C and C′ ) , GMR-Gal4/UAS-Yki ( D and D′ ) , GMR-Gal4/UAS-Yki+wild type Brm ( E and E′ ) , GMR-Gal4/UAS-Yki+BrmD718A ( F and F′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 01810 . 7554/eLife . 00999 . 019Figure 7—figure supplement 4 . Brm regulates Hpo pathway target genes in wing discs . ( A–D′′ ) Wing discs of hh-Gal4 control ( A–A′′ ) or expressing UAS-Flag-Brm ( B–B′′ ) or UAS-Flag-BrmD718A ( C–C′′ ) or UAS-Flag-BrmK804R ( D–D′′ ) were immunostained to show the expression of flag ( red ) , and Diap1 ( green ) . P-compartment of the wing discs was marked by arrows . Of note , BrmK804R always shows a weak expression in wing discs . ( E–F′ ) Bantam sensor upregulated in P-compartment of the wing discs when BrmK804R was expressed , which stands for decreased Bantam level in the P-compartment . Wing discs of hhBanGFP control ( E and E′ ) or expressing UAS-Flag-BrmK804R ( F and F′ ) were immunostained to show the expression of Bantam sensor ( BanGFP , green ) and Ci ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00999 . 019 To further investigate the function of Brm in ISC proliferation , we overexpressed the truncated form of Brm-N ( 1–717 aa ) or Brm-C ( 718–1639 aa ) in ISCs/EBs under the control of esg80ts . Compared with the wild-type Brm or BrmD718A , Brm-C exhibited a weak influence on ISC proliferation while Brm-N did not show any obvious effect ( Figure 7—figure supplement 2A–E , K ) . On coexpression with Yki , both Brm-N and Brm-C further promoted Yki-induced ISC proliferation , but not as dramatic as Brm or BrmD718A ( Figure 7—figure supplement 2F–J , K ) . To better understand the impact of Brm cleavage on ISC proliferation , rescue experiments were carried out using MARCM approach . We overexpressed Brm , BrmD718A , Brm-N and Brm-C in brm2 MARCM clones and found that all of them were able to partially rescue the growth defect of brm2 clones to different degrees . BrmD718A possessed the strongest rescue ability , as it generated bigger clones that contain more cells ( Figure 7—figure supplement 2P , L ) , while Brm-N and Brm-C only showed weak rescue phenotypes ( Figure 7—figure supplement 2Q , R , L ) . These results indicate that Brm cleavage is important for controlling the stability and activity of Brm during ISC proliferation . According to the in vivo observations above , BrmD718A promotes ISC proliferation and exhibits higher activity than wild-type Brm . We speculated that higher activity of BrmD718A might be due to the resistance of BrmD718A against Hpo signaling regulated cleavage . To verify this hypothesis , we coexpressed Hpo and BrmD718A under the control of esg80ts and found that BrmD718A completely rescued the impairment of ISC proliferation induced by Hpo overexpression . In comparison with the decrease of ISC/EB numbers induced by Hpo overexpression , coexpression of Hpo and BrmD718A exhibited a dramatic increase of ISCs/EBs as well as PH3+ cells in midguts ( compare Figure 7F , F′ with Figure 7E , E′ , J ) . On the other hand , coexpression of wild-type Brm and Hpo only slightly rescued Hpo-induced decrease of ISCs/EBs ( compare Figure 7H , H′ with Figure 7E , E′ ) . In addition , PH3+ cell number was increased when Brm and Hpo were coexpressed ( Figure 7J ) , a phenomenon that might be due to an unidentified feedback mechanism of homeostasis in response to Hpo-induced impairment . To determine whether loss of Hpo expression regulates Brm protein level in midguts , we generated hpo null allele BF33 ( 16 ) ( Jin et al . , 2012 ) MARCM clones in midguts . ECs within the clone regions obtained higher Brm protein levels than ECs outside the clones ( Figure 7I–I′′′ ) , suggesting that Hpo also restricts Brm protein level in ECs . We next expressed Brm or Brm RNAi in BF33 ( 16 ) MARCM clones ( Figure 7—figure supplement 1E–F′ ) . The growth of the hpo null allele clones was not affected by Brm overexpression . Yet , it was suppressed by the knockdown of Brm using RNAi ( Figure 7—figure supplement 1E–F′ ) , indicating that Brm is required in the loss-of-Hpo-induced intestinal proliferation . Taken together , these results indicate that Brm protein level is restricted by Hpo activity . BrmD718A , as an uncleavable form of Brm , bypasses the Hpo restriction to retain its ability to promote ISC proliferation . To further understand how BrmD718A functions to promote proliferation in other tissues , we investigated the genetic relationship of Brm and Yki in adult eyes under the control of GMR-Gal4 driver . Overexpression of UAS-Yki posterior to the morphogenetic furrow using the GMR-Gal4 ( referred to as GMR-Yki ) resulted in dramatic eye overgrowth ( compare Figure 7—figure supplement 3D , D′ with Figure 7—figure supplement 3A , A′ ) . Consistent with the findings in midguts , expressing wild-type Brm did not significantly affect Yki overexpression induced eye overgrowth ( Figure 7—figure supplement 3E , E′ ) , yet expressing BrmD718A clearly enhanced GMR-Yki induced overgrowth ( Figure 7—figure supplement 3F , F′ ) . In addition , overexpression of BrmD718A using hhGal4 caused an upregulation of Diap1 protein levels in the posterior region of the wing discs ( compare Figure 7—figure supplement 4C–C′′ with Figure 7—figure supplement 4A–A′′ ) . Conversely , overexpression of the dominant-negative form of Brm , BrmK804R , resulted in a reduction in Diap1 and Bantam levels ( Figure 7—figure supplement 4D–D′′ and Figure 7—figure supplement 4E–F′ ) . These assays indicate that activated Brm promoted the expression of the Hpo pathway target genes , such as diap1 .
SWI/SNF complex subunits regulate the chromatin structure by shutting off or turning on the gene expression during differentiation ( Roberts and Orkin , 2004 ) . Recently , the findings from several research reports based on the stem cell system reveal important roles of chromatin remodeling complex in stem cell state maintenance ( Lessard et al . , 2007; Ho et al . , 2009 ) . Our study suggested that the chromatin remodeling activity of Brm complex was required for the proliferation and differentiation of Drosophila ISCs . Based on our findings , we propose that Brm is critical for maintaining Drosophila intestinal homeostasis ( Figure 2J ) . High levels of Brm in the ISC nucleus represent high proliferative ability and are essential for EC differentiation; low levels of Brm in the EC nucleus may be a response for homeostasis . Changes in Brm protein levels resulted in the disruption of differentiation and deregulation of cell proliferation . In line with previous findings in human , the cell-type-specific expression of Drosophila homologs BRG1 and BRM were also detected in adult tissues ( Reisman et al . , 2005 ) . BRG1 is mainly expressed in cell types that constantly undergo proliferation or self-renewal , whereas BRM is expressed in other cell types ( Marenda et al . , 2004; Reisman et al . , 2005 ) . These observations indicate that Brm may act similarly as BRG1 and BRM in controlling proliferation and differentiation . It is known that the Hpo pathway restricts cell proliferation and promotes cell death at least in two ways: inhibiting the transcriptional co-activator Yki ( Huang et al . , 2005; Zhang et al . , 2008 ) , and inducing activation of pro-apoptotic genes such as caspases directly ( Verghese et al . , 2012 ) . In our study , we identified a novel regulatory mechanism of the Hpo pathway in maintaining intestinal homeostasis . In this scenario , Brm activity is regulated by the Hpo pathway . In normal physiological conditions , under the control of Hpo signaling , the function of Yki–Sd to promote ISC proliferation is restricted and the pro-proliferation of target genes such as diap1 that inhibits Hpo-induced caspase activity cannot be further activated ( Figure 7K ) . Therefore , Hpo signaling normally functions to restrict cell numbers in the midgut by keeping ISC proliferation at low levels . Yki is enriched in ISCs , but predominantly inactivated in cytoplasm by the Hpo pathway ( Karpowicz et al . , 2010; Staley and Irvine , 2010 ) . The knockdown of Yki in ISCs did not cause any phenotype in the midgut ( Karpowicz et al . , 2010 ) , suggesting that Yki is inactivated in ISCs under normal homeostasis . During an injury , Hpo signaling is suppressed or disrupted , Yki translocates into the nuclei to form a complex with Sd ( Karpowicz et al . , 2010; Ren et al . , 2010; Shaw et al . , 2010 ) , which may allow Yki–Sd to interact with Brm complex in the nucleus to activate transcriptional targets . Of note , the loss-of-function of Brm resulted in growth defect of ISCs , suggesting that Brm is required for ISC homeostasis and possessing a different role of Brm from Yki in the regulation of ISCs . It is possible that the function of Brm on ISC homeostasis is regulated via other signaling pathways by recruiting other factors . Therefore , different phenotypes induced by the loss-of-function of Brm and Yki in midgut might be due to different regulatory mechanisms . Despite its unique function cooperating with Yki in midgut , that Brm complex is essential for Yki-mediated transcription might be a general requirement for cell proliferation . While this manuscript was under preparation , Irvine lab reported a genome-wide association of Yki with chromatin and chromatin-remodeling complexes ( Oh et al . , 2013 ) . These results support our model . Our results also suggest that the interaction between Brm and Yki–Sd transcriptional complex is under tight regulation . The loss of Hpo signaling stabilizes Brm protein , whereas the active Hpo pathway restricts Brm levels by activating Drosophila caspases to cleave Brm at the D718 site and inhibiting downstream target gene diap1 transcription simultaneously . In addition , overexpression of Brm complex components induces only a mild enhancement on midgut proliferation ( Figure 1D , D′ and Figure 1—figure supplement 2E , E′ , G , G′ , I , I′ ) . One possibility is that overexpressing only one of the Brm complex components does not provide full activation of the whole complex; the other possibility is that due to the restriction of the Hpo signaling , as overexpressing BrmD718A mutant protein in ISCs/EBs exhibits a stronger phenotype than expressing the wild-type Brm ( Figure 7B , G ) and coexpression of BrmD718A completely rescues the impairment of Hpo-induced ISC proliferation ( Figure 7F ) . D718A mutation blocks the caspase-dependent Brm cleavage and exhibits high activity in promoting ISC proliferation . We have defined a previously unknown , yet essential epigenetic mechanism underlying the role of the Hpo pathway in regulating Brm activity . It is a novel finding that Brm protein level is regulated by the caspase-dependent cleavage . To focus on the function of Brm cleavage in the presence of cell death signals , we tried to examine the activities of the cleaved Brm fragments . Although in vivo experiments did not show strong activity of Brm N- and C-cleavage products in promoting proliferation of ISCs , the C-terminal fragment of Brm that contains the ATPase domain exhibits a relative higher activity than the N-terminal fragment in ISCs ( Figure 7—figure supplement 2D , E , K ) . The cleavage might induce faster degradation of Brm N- and C-terminus , since it was difficult to detect N- or C-fragments of Brm by Western blot analysis without MG132 treatment . It reveals that the degradation events of Brm including both ubiquitination and cleavage at D718 site can be important for Brm functional regulation under different conditions . To this end , the intrinsic signaling ( s ) may balance the activity of Brm complex through degradation of some important components , such as Brm , to maintain tissue homeostasis . Of note , the cleavage of Brm at D718 is occurred at a novel DATD sequence that is not conserved in human Brm . It has been reported that Cathepsin G , not caspase , cut hBrm during apoptosis ( Biggs et al . , 2001 ) , suggesting that the cleavage regulatory mechanism of Brm is relatively conserved between Drosophila and mammals . In this study , we provide evidence that the Brm complex plays an important role in Drosophila ISC proliferation and differentiation and is regulated by multi-levels of Hpo signaling . Our findings indicate that Hpo signaling not only exhibits regulatory roles in organ size control during development but also directly regulates epigenetics through a control of the protein level of epigenetic regulatory component Brm . In mammals , it is known that Hpo signaling and SWI/SNF complex-mediated chromatin remodeling processes play critical roles in tissue development . Malfunction of the Hpo signaling pathway and aberrant expressions of SWI/SNF chromatin-remodeling proteins BRM and BRG1 have been documented in a wide variety of human cancers including colorectal carcinoma ( Reisman et al . , 2009; Pan , 2010; Watanabe et al . , 2011 ) . Thus , our study that implicated a functional link between Hpo signaling pathway and SWI/SNF activity may provide new strategies to develop biomarkers or therapeutic targets .
The following fly stocks were used: UAS-yki ( Zhang et al . , 2008 ) , UAS-HA-Sd ( Zhang et al . , 2008 ) , UAS-HA-SdGA ( Zhang et al . , 2008 ) , BF33 ( 16 ) ( Jin et al . , 2012 ) , UAS-Flag-Hpo ( Jin et al . , 2012 ) , esg-Gal4/UAS-GFP ( Micchelli and Perrimon , 2006 ) , esg-Gal4/UAS-GFP;TubGal80ts ( Micchelli and Perrimon , 2006 ) , Myo1A-Gal4/UAS-GFP;TubGal80ts ( Micchelli and Perrimon , 2006 ) , w;esgGal4 tubGal80ts UAS-GFP; UAS-flp Act>CD2>Gal4 ( esgtsF/O , a gift from Dr Huaqi Jiang ) , esgtsSu ( H ) Z , brm2 ( Bloomington 3619 ) , FRT80 brm2 , sd hypomorphic allele ( sd1 ) ( Zhang et al . , 2008 ) , osa2 ( a gift from Professor Asgar Klebes ) , UAS-Brm RNAi ( VDRC 37720 , VDRC 37721 , and Bloomington 31712 ) , UAS-Bap60 RNAi ( NIG4303R-1 ) , UAS-Mor RNAi ( VDRC 6969 ) , UAS-Osa RNAi ( VDRC 7810 ) , hhBanGFP , UAS-Bap60 , UAS-Mor , UAS-Osa , UAS-Brm-N , and UAS-Brm-C . Bap60 , Mor , Osa were cloned from the Drosophila cDNA . UAS-Flag-Brm , UAS-Flag-BrmK804R , UAS-Flag-BrmD718A , Brm point mutations and deletions were generated by PCR-based site directed mutagenesis . These cDNA fragments were cloned into the pUAST vector . A pUAST vector with attB sequence inserted upstream of the UAS-binding sites was used to make pUAST-attB-Brm and Brm mutants constructs . All plasmids were verified by DNA sequencing . Transgenic flies carrying these constructs were generated . Mutant clones were made using the MARCM system ( Lee and Luo , 2001 ) . Genotypes for making brm mutant clone: hsflp , tub-Gal4 , UAS-GFPnls; tubGal80 FRT80B/FTR80B brm2 . brm mutant clone expressing Yki transgene: hsflp , tub-Gal4 , UAS-GFPnls;UAS-Yki/+;tubGal80 FRT80B/FTR80B brm2 . Hpo mutant clone expressing Brm transgene: yw UAS-GFP hsflp;FRT42D hpoBF33/FRT42D tub-Gal80; tubulin-Gal4/Brm . Hpo clone expressing Brm RNAi transgene: yw UAS-GFP hsflp; FRT42D hpoBF33/FRT42D tub-Gal80; tubulin-Gal4/UAS-Brm RNAi . Flies were cultured at 25°C . F1 adult flies with appropriate genotypes were subjected to heat shock at 37°C for 1 hr to induce clone at 5-day-old flies . Then , flies were raised at 25°C for 3 or 8 days before dissection . Clones of more than 10 midguts were scored in each group . The experiment using esgGal4 UAS-GFP; tubGal80ts was cultured under 18°C to restrict Gal4 activity . 3-day-old F1 adult flies with appropriate genotypes were then shifted to 29°C for a 7-day incubation to allow inactivation of Gal80ts and expression of the UAS transgenes or RNAi . 20 female adults with correct genotypes were dissected and subjected to immunostaining . For intestinal stem cell lineage tracing experiment , we used the inducible lineage tracing esg 80ts F/O system . 2- to 5-day-old F1 adult flies with correct genotypes cultured at 18°C were then shifted to 29°C to induce the expression of transgenes . Female adult flies ( 5/6-day-old ) were used to perform DSS-treated feeding experiments . Flies were cultured in an empty vial containing a piece of 9 cm2 chromatography paper wet with 3% dextran sulfate sodium ( MP Biomedicals , Santa Ana , California , United States ) in 5% glucose solution for 3 days at 25°C or 29°C . S2 cells were cultured in Drosophila Schneider’s Medium ( Invitrogen , Carlsbad , California , United States ) with 10% fetal bovine serum , 100 U/ml of penicillin , and 100 mg/ml of Streptomycin . Plasmid transfection was carried out using LipofectAMINE ( Invitrogen ) according to manufacturer’s instructions . A construct of ubiquitin-Gal4 was cotransfected with pUAST expression vectors for all transfection experiments . Immunoprecipitation and Western blot analyses were performed according to standard protocols as previously described ( Jin et al . , 2012 ) . Antibodies used were as follows: mouse anti-Myc ( 1:5000; Sigma , St . Louis , Missouri , United States ) , mouse anti-Flag ( 1:5000; Sigma ) , mouse anti-V5 ( 1:5000; Invitrogen ) , mouse anti-HA ( Sigma ) , rabbit anti-Sd ( produced by immunizing rabbits with the peptide of Sd amino acids 208–440 ) , and rabbits anti-Brm ( produced by immunizing rabbits with the peptide of Brm amino acids 505–775 ) . The proteasome inhibitor MG132 ( Sigma ) was solubilized in DMSO and added to a final concentration of 50 μM for 6 hr . Z-VAD-FMK and caspase inhibitors ( R&D systems , Minneapolis , Minnesota , United States ) were added to a final concentration of 10 μM . Immunostaining of intestine and S2 cells were carried out as described ( Ren et al . , 2010; Jin et al . , 2012 ) . Primary antibodies used in this study include mouse anti-Delta ( DSHB ) , mouse anti-Prospero ( DSHB , Iowa City , Iowa , United States ) , rat anti-Ci ( 1:500 ) , rabbit anti-PH3 ( Millipore ) , rabbit anti-Yki ( 1:50 , produced by immunizing rabbits with the peptide of Yki amino acids 180–418 ) , rabbits anti-Brm ( this study ) , mouse anti-Flag ( 1:500; Sigma ) , mouse anti-HA ( 1:500; Sigma ) , mouse anti-GFP ( 1:1000; Santa Cruz , Dallas , Texas , United States ) , rabbit anti-PDM-1 ( 1:2000; a gift from Xiaohang Yang , Zhejiang University , Hang Zhou , China ) . Fluorescent microscopy was performed on a Leica LAS SP5 confocal microscope; confocal images were obtained using the Leica AF Lite system . Images were processed in Photoshop CS . The GFP+ EC cells in esg80tsF/O gut were counted in one field of view of the posterior midgut near the Malpighian tubules using a 40× objective . Adult flies of 7–9 days on food containing BrdU ( 200 μg/ml in PBS ) were mixed into the upper layer and dissected 3 days later . The guts were treated with DNase I for 30 min at 37°C . Thirty 10-cm dishes of S2 cells were collected and washed twice with cold PBS . The cells were equally divided into two samples and lysed in 2 . 5 ml lysis buffer ( Tris HCL pH 8 . 0 , 50 mM , NaCl 100 mM , NaF 10 mM , Na3VO4 1 mM , EDTA 1 . 5 mM , NP-40 1% , glycerol 10% supplied with protease inhibitor cocktail [Sigma] ) , and centrifuged at 14 , 000 rpm for 15 min . Supernatant was transferred to and mixed with 150 μl protein A/G beads ( Santa Cruz Biotechnology ) at 4°C for 1 hr on a rolling mixer . Then , the mixture was centrifuged at 14 , 000 rpm for 1 min . Cell supernatant was transferred to a new tube and stored at 4°C . 0 . 5 mg of Brm antibody was mixed with 0 . 25 ml of wet beads ( use the appropriate antibody/protein A/G combination ) in room temperature for 1 hr on a rolling mixer , using serum mixed with beads as control . The beads were washed using 10 vol of Borate Buffer ( Sodium Borate pH 9 . 0 , 0 . 2 mM ) twice . 10 μl aliquot of beads was stored on ice ( sample 1 ) . The rest of the beads were mixed with solid Dimethyl Pimelimidate Dihydrochloride ( final concentration is 20 mM ) on the rolling mixer for 30 min at room temperature . Another 10 μl aliquot of beads was collected as sample 2 . The rest of the beads were washed twice by equal volume of Ethanolamine after discarding the supernatant . Equal volume of Ethanolamine was added and incubated at room temperature for 2 hr on the rolling mixer , then washed by PBS twice and mixed with the cell supernatant at 4°C for 1 hr on a rolling mixer respectively . The mixture was washed by lysis buffer for three times and stored at −20°C for left experiment . The coupling of antibody to beads was checked by analyzing the sample 1 and 2 on a SDS gel . After checking the coupling efficiency , for Sd MS , two total samples ( serum control and Sd IP sample ) were sent for MS directly . For Yki MS , we ran an SDS-PAGE to separate the proteins of Yki IP sample , and then specific bands , which were absent in IgG IP control , were selected for further MS analysis . For MS fingerprinting , the gel slices were cut out of the preparative Coomassie blue-stained gels , destained with100 mmol/l NH4HCO3/30% ACN , and then dried completely by centrifugal lyophilization . The dried gel slices were rehydrated with a total of 25 ng of sequencing grade , modified trypsin ( Promega , Madison , Wisconsin , United States ) in 100 mmol/l ammonium bicarbonate at 4°C for 2 hr . After 20 ml of 50 mmol/l NH4HCO3 , pH 8 . 3 was supplied , the gel slices were incubated at 37°C for 20 hr . The digest buffer was removed and saved . The gel pieces were then extracted with 200 ml of 60% ACN/0 . 1%TFA for 15 min with sonication , and the supernatant was removed . The extraction was repeated twice . The three extracts plus the first saved digest buffer were pooled and dried completely by centrifugal lyophilization . This in-gel digestion method was mainly performed according to the method described previously ( Yu et al . , 2000; Li et al . , 2005 ) with some modifications as described above . Peptide mixtures of each gel slice were redissolved in 0 . 1%TFA , then desalted and concentrated using Stage Tips as reported ( Rappsilber et al . , 2007 ) . Peptide solution was measured using a LTQ Deca XP system ( Thermo Finnigan , San Jose , California , United States ) . HPLC separation was performed with a capillary LC pump . The flow rate of the pump was at 250 μl/min and was about 2 μl/min after split . The mobile phases used for reverse phase were A: 0 . 1% formic acid in water , pH 3 . 0 , B: 0 . 1% formic acid in ACN . Peptides were eluted using a 2–35% , 35–90% stepped linear gradient of solvent B in 60 min , 90 min following 90% solvent B in 10 min , and 2% solvent B in 30 min for balance . An ESIIT mass spectrometer ( LTQ Deca XP; Thermo Finnigan ) was used for peptide detection . The positive ion mode was employed , and the spray voltage was set at 3 . 4 KV . The spray temperature was set at 200°C for peptides . Collision energy is automatically set by the LTQ Deca XP system . After acquisition of a full scan mass spectrum , 10 MS/MS scans were acquired for the next 10 most intense ions using dynamic exclusion . Peptides and proteins were identified using Turbo Sequest software ( Thermo Finnigan ) , which uses the MS and MS/MS spectra of peptide ions to search against the publicly available Uniport fly database ( Version 2011-05-26 ) . The protein identification criteria that we used were based on Delta ( CN ≥ 0 . 1 ) and Xcorr ( one charge ≥ 1 . 9 , two charges ≥ 2 . 2 , three charges ≥ 3 . 75 ) . | Most tissues can generate new cells to repair damage or replace worn-out cells . The new cells are often generated from stem cells—cells that can either reproduce themselves or mature into other types of cells . In the fruit-fly Drosophila , for example , intestinal stem cells in the midgut are capable of producing more stem cells or they can differentiate to produce immature cells called enteroblasts that go on to become either enterocytes ( the cells that line the gut ) or enteroendocrine cells ( which secrete hormones ) . Researchers have identified a number of signalling pathways that are involved in the proliferation and differentiation of intestinal stem cells in the midgut of fruit flies . These include the Hippo pathway , which is important for regulating both cell proliferation and programmed cell death ( apoptosis ) . Activation of the Hippo protein triggers a cascade of signals that culminate in the regulation of many of the genes involved in cell proliferation , division and apoptosis . Another process that is important for controlling the proliferation and differentiation of cells is chromatin remodelling . Chromatin is the ‘packaging’ that keeps DNA tightly wound within the cell nucleus , and remodelling refers to the structural changes that allow proteins called transcription factors to reach the genes and transcribe them into messenger RNA ( which then leaves the nucleus to generate the protein ) . Now , Jin et al . have explored how the Hippo pathway and chromatin remodelling work together to regulate of stem cells . Using a technique called RNA interference to block the expression of various genes in intestinal stem cells and enteroblasts , Jin et al . found that a protein called Brahma—which is an essential part of a chromatin-remodelling complex—must be present for the stem cells to multiply normally . Jin et al . also showed how the Hippo signalling pathway interacts with chromatin remodelling . Activation of the Hippo pathway inhibits gene expression by preventing two other proteins , Yorkie and Scalloped , from forming a complex in the nucleus . The new work shows that Brahma interacts physically with the Yorkie and Scalloped proteins to regulate the proliferation of the intestinal stem cells . It also shows that the Hippo protein regulates the activity of the Brahma protein by inducing a process called caspase-dependent cleavage . Because many of the proteins involved in these pathways are evolutionarily conserved and expressed in a variety of tissues , these findings may have implications for stem cell function and tissue repair in many species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2013 | Brahma is essential for Drosophila intestinal stem cell proliferation and regulated by Hippo signaling |
Mesial temporal lobe epilepsy ( mTLE ) is the most common focal epilepsy in adults and is often refractory to medication . So far , resection of the epileptogenic focus represents the only curative therapy . It is unknown whether pathological processes preceding epilepsy onset are indicators of later disease severity . Using longitudinal multi-modal MRI , we monitored hippocampal injury and tissue reorganization during epileptogenesis in a mouse mTLE model . The prognostic value of MRI biomarkers was assessed by retrospective correlations with pathological hallmarks Here , we show for the first time that the extent of early hippocampal neurodegeneration and progressive microstructural changes in the dentate gyrus translate to the severity of hippocampal sclerosis and seizure burden in chronic epilepsy . Moreover , we demonstrate that structural MRI biomarkers reflect the extent of sclerosis in human hippocampi . Our findings may allow an early prognosis of disease severity in mTLE before its first clinical manifestations , thus expanding the therapeutic window .
Epilepsy is a devastating disease , with a prevalence of about 1% , which makes it one of the most common neurological disorders . Of particular clinical significance is mesial temporal lobe epilepsy ( mTLE ) , in which seizures arise from mesial temporal limbic structures , as it is the most frequent form of epilepsy in adults ( 40% ) and particularly resistant to pharmacological treatment ( Engel , 2001 ) . Hippocampal sclerosis ( HS ) , characterized by the loss of CA1 and CA3 pyramidal cells , gliosis and associated with granule cell dispersion ( GCD ) , represents the most common pathological hallmark of intractable mTLE ( Thom , 2014; Walker , 2015 ) . HS can emerge secondary to an initial precipitating insult , e . g . status epilepticus ( SE ) , head trauma , febrile seizures or limbic encephalitis ( Engel , 2001 ) . To date , resection of the lesioned , epileptogenic tissue is the only therapeutic intervention . However , not all patients remain seizure-free ( Ryvlin and Kahane , 2005 ) . In this context , non-invasive MRI techniques have become an important method to identify sclerotic tissue as a biomarker in both human patients ( Gomes and Shinnar , 2011; Urbach et al . , 2014 ) and rodent TLE models ( Nehlig , 2011; Shultz et al . , 2014; Sierra et al . , 2015a ) . Up to now , this has been , however , restricted to the chronic stage of epilepsy when seizures have already emerged . A promising notion is to start medical treatment during early epileptogenesis before the first seizure arises . This stands in contrast to current anti-epileptic treatments that target clinical symptoms , namely the seizures themselves , but do not cure the underlying disease . However , progress in this direction is impeded by the impossibility to identify patients before they develop epilepsy and experience their first seizure . In mTLE patients , the period between the initial epileptogenic insult and chronic epilepsy spans many years ( Lukasiuk and Becker , 2014 ) , severely hindering the search for and validation of predictive biomarkers in human studies . Conversely , in animal models , epileptogenesis can be induced within weeks to months ( Lévesque and Avoli , 2013 ) , providing an excellent opportunity to identify putative anatomical and physiological biomarkers of epileptogenesis . Considering that multiple cellular and molecular processes accompany the progression of HS , it may be possible to image early pathological changes during epileptogenesis that may prospect the later state . So far , a few MRI biomarkers have been identified that indicate epilepsy onset or seizure susceptibility secondary to pilocarpine-induced SE ( Roch et al . , 2002; Choy et al . , 2010 ) , traumatic brain injury ( Kharatishvili et al . , 2007; Immonen et al . , 2013 ) or prolonged febrile seizures ( Choy et al . , 2014 ) . However , since distinct unilateral HS is lacking in these animal models , prognostic MRI biomarkers for the most common etiology of pharmacoresistant epilepsy have not been identified yet . To address this issue , we applied longitudinal multi-modal MRI in the intrahippocampal kainate mouse model that replicates the major features of human mTLE ( i . e . the development of HS and of spontaneous recurrent seizures during epileptogenesis; Bouilleret et al . , 1999; Riban et al . ( 2002 ) ; Heinrich et al . , 2011 ) . We monitored distinct features of the pathogenesis focusing on the hippocampus ( Figure 1—figure supplement 2 ) , which is mainly affected in pharmacoresistant mTLE ( Malmgren and Thom , 2012; Cendes et al . , 2014; Walker , 2015 ) . We performed a retrospective correlation of MRI measures with structural changes identified by immunohistology , and determined the prognostic value of imaging biomarkers with respect to the severity of HS . Moreover , we applied multivariate data analysis to evaluate and compare the performance of individual biomarkers to predict also the seizure severity . Our study demonstrates for the first time that the extent of epileptogenesis-associated tissue alterations in the hippocampus directly mirrors the ensuing severity of intractable mTLE .
Little is known about the variability of histopathological changes across kainate-injected mice; yet this information is critical to relate early imaging biomarkers to HS severity in mTLE . Therefore , we quantified the degree of histopathological features associated with HS ( i . e . cell death-associated microgliosis and GCD ) in all injected mice individually ( Figure 1 ) . 10 . 7554/eLife . 25742 . 003Figure 1 . The severity of histological changes associated with HS varies in chronically epileptic mice . ( A ) Representative DAPI-stained sections at different levels along the rostro-caudal hippocampal axis showing the extent of GCD ( arrow indicates the transition to non-dispersed regions ) . ( B ) Corresponding quantitative analysis of the GCD area along the rostro-caudal axis , and ( C ) the calculated mean GCD volume from all analyzed sections tested for individual kainate-injected mice ( one-way ANOVA , Bonferroni’s post-test; ***p<0 . 001; n = 8 ) . ( D ) Representative photomicrographs of NeuN ( turquoise; neurons ) and Iba-1 ( magenta; microglia ) double immunostaining ( upper panel ) and Fluoro-Jade B ( FJB ) staining in consecutive sections . Clusters of amoeboid microglia are tightly associated with FJB-stained dying neurons ( arrows and asterisks ) . ( E ) Quantitative analysis of cell death-associated microglial scarring in different regions ( CA1 ipsi and contra; CA3 ipsi ) . ( F ) Regression analysis for the degree of GCD and the extent of microgliosis ( summed for all regions ) in kainate-injected mice ( n = 8; Pearson’s correlation ) . Kainate-injected mice ( NP10 , NP11 , NP14 , NP26 , NP27 , NP31 , NP34 ) are color-coded . Scale bars in A , 1 mm; in D , 200 µm . ( G ) Schematic of the mouse brain adapted from Witter and Amaral , 2004 . Representative EEG traces of non-epileptic mice ( controls or mice displaying only single epileptic spikes , NP10 ) and one example of an epileptic mouse displaying both epileptic spikes and paroxysmal discharges ( NP31 ) . Horizontal scale bars ( left ) 50 s , ( middle ) 5 s , ( right ) 0 . 5 s; vertical scale bar 2 mV . ( H–I ) Quantitative analysis of the total GCL volume ( summed for all analyzed sections ) and extent of microgliosis for epileptic ( dark grey ) and non-epileptic mice ( light grey ) , respectively . Student’s t-test; **p<0 . 01 , ***p<0 . 001; nnon-epi = 6 , nepi = 7 . All values are presented as the mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 00310 . 7554/eLife . 25742 . 004Figure 1—figure supplement 1 . Longitudinal development of epileptiform activity after kainate injection . ( A–F ) Representative longitudinal EEG recording of the ipsilateral hippocampus of a kainate-injected mouse . Recordings in the same mouse were repeated at different time points during epileptogenesis ( 6 hr , 1 day , 3 days , 1 week , 2 weeks and 3 weeks following injection ) . Detailed images of the respective EEG traces are indicated by blue boxes . Scale bars: Horizontal ( left ) 200 s and ( right ) 5 s , vertical 1 mV . Grey arrow heads above the trace and horizontal lines in blue boxes denote examples for high-amplitude recurrent paroxysmal episodes , i . e . epileptic discharges . ( G ) Quantitative analysis for individual mice ( color-coded ) and H ) for groups . One-way ANOVA , Bonferroni’s post-test , **p<0 . 01 , n1d = 8 , n0 . 5w = 6 , n1w = 8 , n2w = 8 , n3w = 5 . Values are presented as the mean ± SEM . Individual recordings were binned for time points: 1 day ( recordings on day 1 ) , 0 . 5 week ( rec . on days 3–5 ) , 1 week ( rec . on days 6–7 ) , 2 weeks ( rec . on days 12–16 ) and 3 weeks ( rec . on days 17–21 ) after SE . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 00410 . 7554/eLife . 25742 . 005Figure 1—figure supplement 2 . Schematic of the experimental design . Shown is an overview of the workflow applying multi-modal MRI measurements during the time course of epileptogenesis , video-EEG recording in the chronic stage of the disease and post-hoc IHC . Date was retrospectively correlated to probe the predictive value of putative MRI biomarkers . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 005 The extent of GCD ( Figure 1A–C ) and cell death-associated microgliosis ( Figure 1D , E ) varied substantially across kainate-injected mice ( nKA = 8 ) , but was positively correlated with each other ( Figure 1F; r² = 0 . 69 , p<0 . 01 ) . In addition , intrahippocampal EEG recordings validated the presence of epileptiform activity in all kainate-injected mice ( Figure 1G ) . However , one kainate-injected mouse ( NP10 ) had no paroxysmal discharges , even though it exhibited single epileptic spikes , and therefore it was considered as non-epileptic . All epileptic mice showed robust GCD indicated by an increase in the total GCL volume compared to non-epileptic mice ( Figure 1H; non-epi . : 415 . 8 ± 9 . 6 * 106 µm³; epi . : 1379 . 0 ± 164 . 5 * 108 µm³ , p<0 . 01 ) and a high spatial extent of cell death-associated microgliosis ( Figure 1I; non-epi . : 563 . 8 ± 493 . 7 µm; epi . : 6965 . 0 ± 811 . 8 µm , p<0 . 001 ) . To monitor the temporal development of epileptic activity another set of mice was implanted with electrodes directly after kainate injection and recorded at different time points during 2 ( n = 3 ) or 3 weeks ( n = 5 ) of disease progression ( Figure 1—figure supplement 1 ) . Consistent with previous studies in the same animal model ( Riban et al . , 2002; Arabadzisz et al . , 2005; Heinrich et al . , 2011 ) , we observed prolonged seizures during SE , followed by a period exhibiting mainly isolated epileptic spikes and low-amplitude bursts . The first high-amplitude recurrent paroxysmal episodes were recorded consistently after approximately two weeks ( Figure 1—figure supplement 1H; number of epileptic discharges , 1d: 0 . 014 ± 0 . 004; 0 . 5w: 0 . 002 ± 0 . 002; 1w: 0 . 036 ± 0 . 019; 2w: 0 . 240 ± 0 . 065 , p<0 . 01 vs 1w; 3w: 0 . 274 ± 0 . 070 , p<0 . 01 vs 1w ) . In only two out of eight mice regular paroxysmal episodes were already evident at one week after SE . In conclusion , this shows that although the speed of disease progression varies among mice , the chronic epileptic stage is reached within the second week after SE in this animal model . Considering the inter-individual histopathological differences observed in kainate-injected mice during the chronic stage of mTLE , we investigated the extent to which high-resolution T2-weighted imaging reflects SE-induced hippocampal injury and whether observed early tissue damage and anatomical changes during epileptogenesis might predict the subsequent HS severity in chronically epileptic mice ( Figure 2 ) . 10 . 7554/eLife . 25742 . 006Figure 2 . Initial hippocampal damage detected by T2 imaging predicts the degree of HS . ( A1-6 ) Overview of whole-brain T2 intensity maps along the rostro-caudal axis before ( pre ) and at distinct time-points following kainate-induced SE ( 1d , 4d , 8d , 16d and 31d ) . ( B1-6 ) Enlarged view of the ipsilateral hippocampus . Dashed lines denote the hippocampal fissure . Open arrows mark the GCD . ( C–D , E–F ) Direct comparison between T2 images and NeuN ( turquoise ) and Iba-1 ( magenta ) double-stained sections . Upper and lower panels show the septal and temporal regions of the hippocampus , respectively . Arrows indicate the transition from the dispersed to the non-dispersed GCL . ( D1-3 , F1-3 ) High-magnification confocal images corresponding to photomicrographs in D and F display the loss of neurons and accompanied microglial scarring in the CA1 region and the hilus . Principal neurons in the GCL remain intact but are highly dispersed . H , hilus; gcl , granule cell layer . Scale bars: A , 2 mm; B , E , F , 200 µm; F3 , 20 µm . ( G , I ) Quantitative analysis of T2 changes in CA1 and the DG during epileptogenesis plotted for individual animals ( left panel: controls , black , n = 5; kainate-injected mice , color-coded ) , and ( right panel ) statistically tested for the epileptic ( dark grey ) and non-epileptic group ( light grey ) . Source data is provided in ‘Figure 2—source data 1' . Two-way ANOVA; Bonferroni’s post-test; *p<0 . 05 , ***p<0 . 001 , nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . ( H , J ) Corresponding linear regression analysis of T2 intensity at distinct time points during epileptogenesis ( color-coded ) with region-specific histopathological changes in epileptic and non-epileptic mice ( n = 13; Pearson’s correlation; corrected for multiple testing; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 00610 . 7554/eLife . 25742 . 007Figure 2—source data 1 . Summary of T2 metrics . Quantitative values of T2 measurements are listed for individual mice ( saline-injected: N12 , NP13 , NP17 , NP28 , NP29; kainate-injected: NP10 , NP11 , NP14 , NP25 , NP26 , NP27 , NP31 , NP34 ) and longitudinal time points ( pre , 1d , 4d , 8d , 16d , 31d following injection ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 00710 . 7554/eLife . 25742 . 008Figure 2—figure supplement 1 . T2 changes in the piriform cortex and the amygdala during epileptogenesis . ( A1–3 ) Representative T2-weighted images of coronal sections prior to kainate injection and at 1 day and 31 days after SE . Enlarged images show the piriform cortex ( PirC ) and the amygdala ( Amyg ) . Scale bar: 50 µm . ( B–C ) Quantitative analysis of T2 signal intensity in the piriform cortex and the amygdala , respectively . Values are plotted for individual mice ( left panel; controls , black , n = 5; kainate-injected animals color-coded ) and for groups ( right panel; epileptic , dark grey; non-epileptic , light grey ) during epileptogenesis . Two-way ANOVA , Bonferroni’s post-test , nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 008 T2-weighted imaging revealed a biphasic rise in hippocampal T2 intensity , characterized by a transient increase at day 1 , followed by a drop to baseline at day 4 and a subsequent increase at 8 days after SE ( Figure 2A , B ) . The early increase of T2 intensity was observed in the dentate gyrus ( DG ) ( Figure 2I; 1d: 10 . 6% ) and the CA1 region ( Figure 2G; 1d: 26 . 9%; both p<0 . 001 ) , whereas in the late phase it was restricted to the DG ( 8d: 7 . 8%; 31d: 7 . 9% , both p<0 . 05 ) . Accordingly , in chronically epileptic mice both regions showed distinct histopathological changes . Extensive death of CA1 pyramidal cells was associated with dense microglial scarring ( Figure 2F , F1 ) . Moreover , hilar interneurons ( Figure 2F , F3 ) were also lost . Dentate granule cells , however , remained intact but dispersed broadly ( Figure 2F2 ) . Importantly , 1 day following SE , T2 intensity in the CA1 region significantly correlated with cell death-associated microgliosis ( Figure 2H; r² = 0 . 88 , p<0 . 001 ) . Similarly , in the DG both early ( at 1 day ) and subsequent T2 hyperintensities ( at 8–31 days ) were related to the later degree of GCD ( Figure 2J ) . No significant change of T2 intensity was found in the piriform cortex or the amygdala ( Figure 2—figure supplement 1 ) . The dynamics of neurodegeneration during epileptogenesis were further characterized by 1H-MR spectroscopy ( Figure 3 ) . N-acetyl aspartate ( NAA ) , serving as a surrogate marker for neurons , and the neurotransmitters glutamate and GABA were measured to differentiate between excitatory principal neurons and inhibitory interneurons , respectively . As early as 1 day following SE , the concentrations of NAA ( −45 . 3% ) , glutamate ( −35 . 1% ) and GABA ( −39 . 4%; all p<0 . 001 ) decreased substantially ( Figure 3C , E , G ) which was correlated with the later extent of cell death-associated microgliosis ( Figure 3D , F , H; for NAA: r² = 0 . 89; for glutamate: r² = 0 . 93; for GABA: r² = 0 . 77; all p<0 . 001 ) . In contrast to glutamate , GABA concentrations remained correlated also at day 4 ( r² = 0 . 74; p<0 . 01 ) , but this disappeared from 8 days onward , which was accompanied by a gradual increase of GABA above control values after 31 days . Conversely , NAA and glutamate concentrations remained low during subsequent epileptogenesis correlating with microgliosis . We also found an early upregulation of lactate and a delayed rise of myoinositol ( Figure 3I , K ) , serving as markers for micro- and astroglial activation , respectively ( Nehlig , 2011 ) . However , changes in both metabolites showed weak correlations with chronic histopathological changes ( Figure 3J , L ) . 10 . 7554/eLife . 25742 . 009Figure 3 . Early decline of glutamate and GABA predict HS . ( A ) Representative horizontal and coronal T2 images illustrating the region-of-interest in the ipsilateral hippocampus ( turquoise boxes ) for 1H-MR-spectroscopy . N-acetyl aspartate ( NAA ) served as a marker for neurons . Glutamate ( Glu ) and gamma-aminobutyric acid ( GABA ) allowed to estimate the loss of excitatory and inhibitory neurons , respectively . Lactate ( Lac ) and myoinositol ( Myoi ) were used as surrogate markers for microglial and astroglial activation . ( B ) Representative photomicrographs of NeuN ( turquoise , neurons ) and Iba-1 ( magenta , microglia ) double-stained sections from one saline- ( control ) and two kainate-injected mice exhibiting different degrees of hippocampal sclerosis ( NP26 , moderate; NP34 , strong ) for qualitative comparison with the degree of metabolic alterations . Arrow , borders of the GCD; Asterisks , cell loss and microglial scarring in CA1 . Scale bar in A , 2 mm; B , 200 µm . ( C , E , G , I , K ) Quantitative analysis of NAA , Glu , GABA , Lac and Myoi concentrations plotted for individual mice ( left panel; controls , black , n = 5; kainate-injected animals color-coded ) and groups ( right panel; epileptic , dark grey; non-epileptic , light grey ) during epileptogenesis . Source data is provided in ‘Figure 3—source data 1' . Two-way ANOVA; Bonferroni’s post-test; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001; nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . ( D , F , H , J , L ) Corresponding linear regression analysis of metabolite concentrations at distinct time-points during epileptogenesis ( color-coded ) with the extent of cell death-associated microgliosis in epileptic and non-epileptic mice ( n = 13; Pearson’s correlation , corrected for multiple comparison; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 00910 . 7554/eLife . 25742 . 010Figure 3—source data 1 . Summary of 1H-MR metrics . Quantitative values of 1H-MR spectroscopy are listed for individual mice ( saline-injected: N12 , NP13 , NP17 , NP28 , NP29; kainate-injected: NP10 , NP11 , NP14 , NP25 , NP26 , NP27 , NP31 , NP34 ) and longitudinal time points ( pre , 1d , 4d , 8d , 16d , 31d following injection ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 010 These results show that the magnitude of initial neurodegeneration following SE , as quantified by MR biomarkers , predicts the severity of HS in the chronic stage of epilepsy . SE-induced neurodegeneration is known to trigger various cellular responses , including inflammatory processes , synaptic plasticity and migration of surviving neurons and glial cells . To probe our hypothesis that these responses might entail microstructural alterations detectable by MRI , specifically diffusion-weighted imaging ( DWI ) , we first identified the morphological changes of granule cells and radial glia cells , two major cell populations which survive in the sclerotic hippocampus ( Figure 4 ) . We used transgenic Thy1-eGFP mice , in which a subset of dentate granule cells is intrinsically labeled by eGFP , to investigate their dendritic and axonal morphology as well as the progression of GCD . Accordingly , quantitative analysis revealed a continuous increase of the GCL width at day 7 , reaching significance at 14 and 21 days after SE when compared to controls ( Figure 4A , B; control: 67 . 8 ± 2 . 5 µm; 14d: 132 . 6 ± 13 . 8 µm; 21d: 168 . 6 ± 9 . 9 µm , both p<0 . 001 ) . This increase was accompanied by a strong disorganization of their distal dendrites in the molecular layer , whereas proximal dendritic segments residing within the GCL became longer and appeared more organized , possibly due to traction forces mediated by GCD . Furthermore , proximal dendrites significantly thickened as early as day 7 ( control: 1 . 54 ± 0 . 04 µm , 7d: 1 . 86 ± 0 . 12 µm , p<0 . 05 ) , while proximal axons increased in volume only during later stages of epileptogenesis at 14 and 21 days after SE ( Figure 4C , D; control: 0 . 70 ± 0 . 02 , 14d: 1 . 15 ± 0 . 08 µm; 21d: 1 . 63 ± 0 . 09 µm , both p<0 . 001 ) . 10 . 7554/eLife . 25742 . 011Figure 4 . Microstructural alterations in the DG during epileptogenesis . ( A1-6 and C1-6 ) Representative confocal images show changes in the cytoarchitecture of the DG ( double-headed arrows denote the dispersion of the GCL ) and morphological features of individual eGFP-labeled granule cells , respectively ( left panel: Overview; upper and lower right: High-magnification of granule cell dendrites or somata , respectively; red arrowheads , axon initial segments; red arrows , stem dendrites; red open arrows , dendritic swellings; red asterisks , degenerating granule cells . ( B ) Quantification of the GCL width ( ncntrl = 7 , n1d = 3 , n4d = 5 , n7d = 6 , n14d = 3 , n21d = 6 ) and ( D ) the diameter of initial axons and stem dendrites ( ncntrl = 5 , n1d = 3 , n4d = 5 , n7d = 4 , n14d = 3 , n21d = 5 ) . ( E1-6 ) Representative confocal z-plane images of ZnT-3 staining in the DG to determine the dynamics of mossy fiber sprouting ( red arrows; ncntrl = 8 , n1d = 4 , n4d = 5 , n7d = 5 , n14d = 5 , n21d = 9 ) . Locations of granule cell somata are spared ( dotted outlines ) . ( F ) Quantitative analysis of ZnT-3 optical density . ( G1-6 ) Representative confocal images of GFAP-stained sections ( upper panel ) and corresponding 3D-reconstruction in the GCL ( lower panel ) . h , hilus; gcl , granule cell layer; ml , molecular layer; slm , stratum lacunosum moleculare . ( H–I ) Quantitative analysis of the mean and total volume of GFAP-stained processes from radial glia cells in the GCL ( ncntrl = 5 , n1d = 4 , n4d = 5 , n7d = 4 , n14d = 4 , n21d = 5 ) . All statistics were performed with one-way ANOVA , Dunnett’s post-test ( compared to saline controls ) ; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . ( J , K ) Representative sections stained for DAPI ( turquoise ) and GFAP ( magenta ) in controls and chronic epileptic mice ( 21d following kainate injection ) , respectively . Dashed lines denote the borders of the GCL . ( L ) Quantitative analysis for the optical density of GFAP in individual sections from three controls and in sections from three kainate-injected mice exhibiting weak and strong GCD ( one-way ANOVA , Dunnett’s post-test; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001; number of sections , ncntrl = 51 , nw-GCD = 6 , ns-GCD = 14 ) . ( M ) Corresponding linear regression analysis for the integrated GFAP optical density and ( I ) the area of the GCL ( Pearson’s correlation ) . Scale bars in A , 50 µm; in C ( left ) , 100 µm; in C ( right ) , 10 µm; in E ( left ) , 100 µm; in E ( right ) , 10 µm; in G , 30 µm; in K , 200 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 011 Next , we determined the time course of axonal rearrangement within the GCL , by analyzing the optical density of ZnT3-labeled mossy fibers . A few ZnT-3-labeled profiles were already evident at days 4 and 7 in the subgranular region , but a significant increase of ZnT-3 was only detected at 14 and 21 days after SE ( Figure 4E , F; control: 4 . 02 ± 0 . 50 , 14d: 11 . 64 ± 2 . 39 , p>0 . 05; 21d: 14 . 57 ± 2 . 11 , p<0 . 001 ) . Although sprouting of mossy fiber collaterals is impressive , it is important to note that these axons are not uniformly aligned , limiting their detectability by DWI . In contrast , processes of radial glia cells are known to radially transverse the GCL in a highly organized fashion ( Heinrich et al . , 2006 ) . Accordingly , we reconstructed GFAP-labeled radial glia processes in three-dimensional space to perform quantitative morphometry . Interestingly , the mean volume of radial glia processes significantly increased as early as 4 days after SE ( Figure 4G , H , I; control: 6 . 47 ± 1 . 15 µm³ , 4d: 19 . 09 ± 0 . 55 µm³ , p<0 . 001 ) . Compared to controls this increase was further augmented at later time points during epileptogenesis ( 7d: 20 . 59 ± 1 . 37; 14d: 29 . 26 ± 4 . 03; 21d: 28 . 75 ± 2 . 59 , all p<0 . 001 ) . Moreover , the optical density of GFAP was correlated with the size of the GCL in chronically epileptic mice ( Figure 4M; r² = 0 . 89 , p<0 . 0001 ) , showing that radial gliosis is tightly associated with GCD . Taken together , our results demonstrate that during epileptogenesis distinct microstructural alterations emerge at different time points within the DG: Radial gliosis starts early after SE , whereas neuronal hypertrophy , dendritic displacement and synaptogenesis accompany GCD at later stages . Next , we monitored the dynamics of microstructural alterations in vivo using DWI , in order to test whether they could predict the later extent of histopathological changes ( Figure 5 ) . 10 . 7554/eLife . 25742 . 012Figure 5 . Microstructural reorganization quantified by DWI during epileptogenesis predicts disease progression . ( A1-6 ) Representative coronal sections from diffusion-weighted tractography at different time points during epileptogenesis ( before injection = pre; 1d , 4d , 7d , 14d and 31d following SE ) . ( B1-6 ) Enlarged images of the ipsilateral hippocampus at different levels along the rostro-caudal axis . The orientation of computed streamlines is color-coded [dorsoventral ( DV ) , turquoise; mediolateral ( ML ) , red; rostrocaudal ( RC ) , blue] . ( C–D ) Representative tractography image and a Nissl-stained section ( modified from Paxinos and Franklin , The Mouse Brain in Stereotaxic Coordinates , 2001 ) of corresponding brain regions for anatomical comparison . Computed fibers relate to major axonal pathways and brain regions exhibiting highly oriented dendrites ( cc , corpus callosum; cp , cerebral peduncle; Cx , cortex; Dlg , dorsal lateral geniculate nucleus; ec , external capsule; eml , external medullary lamina; Hc , hippocampus; HyTh , hypothalamus; ic , internal capsule; ml , medial lemniscus; opt , optic nerve; Th , thalamus ) . ( E , G ) Enlarged tractography images demonstrating the distinct orientation of streamlines in different hippocampal layer ( dashed lines; cc , corpus callosum; so , stratum oriens; py , pyramidal layer; sr , stratum radiatum; slm , stratum lacunosum moleculare; ml , molecular layer; gcl , granule cell layer; asterisks denote the region of pyramidal cell loss ) . ( F , H ) Corresponding DAPI-stained sections . Scale bars in A , 2 mm; B-D , 500 µm; H ( left ) , 100 µm . ( I , K , M , O ) Quantitative analysis of DWI metrics [mean- ( MD ) , radial ( RD ) , axial diffusivity ( AD ) and fractional anisotropy ( FA ) ] in the DG , plotted for individual mice ( left panel; controls , black , n = 5; kainate-injected animals color-coded ) and for groups ( right panel; epileptic , dark grey; non-epileptic , light grey ) during epileptogenesis . Source data is provided in ‘Figure 5—source data 1' . Two-way ANOVA , Bonferroni’s post-test , **p<0 . 01; ***p<0 . 001; nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . ( J , L , N , P ) Corresponding linear regression analysis of DWI metrics at distinct time points during epileptogenesis with the total GCL volume in epileptic and non-epileptic mice ( n = 13; Pearson’s correlation , corrected for multiple comparison; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . Refer to ‘Figure 5—figure supplement 1' for DWI metrics acquired in CA1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 01210 . 7554/eLife . 25742 . 013Figure 5—source data 1 . Summary of DWI metrics . Quantitative values of DWI measurements are listed for individual mice ( saline-injected: N12 , NP13 , NP17 , NP28 , NP29; kainate-injected: NP10 , NP11 , NP14 , NP25 , NP26 , NP27 , NP31 , NP34 ) and longitudinal time points ( pre , 1d , 4d , 8d , 16d , 31d following injection ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 01310 . 7554/eLife . 25742 . 014Figure 5—figure supplement 1 . DWI changes in CA1 during epileptogenesis are poor predictors for hippocampal sclerosis . ( A , C , E , G ) Quantitative analysis of DWI metrics in the CA1 region , plotted for individual mice ( left panel; controls , black , n = 5; kainate-injected animals color-coded ) and groups ( right panel; epileptic , dark grey; non-epileptic , light grey ) during epileptogenesis . Two-way ANOVA , Bonferroni’s post-test , *p<0 . 05; **p<0 . 01; nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . ( B , D , F , H ) Corresponding linear regression analysis of DWI metrics at distinct time points during epileptogenesis with the extent of microglial scarring in the CA1 region of epileptic and non-epileptic mice ( n = 13; Pearson’s correlation , corrected for multiple comparison; *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 01410 . 7554/eLife . 25742 . 015Figure 5—figure supplement 2 . DWI changes in the piriform cortex and the amygdala during epileptogenesis . ( A ) Representative coronal section of diffusion-weighted tractography images before injection . White box denotes the location of the amygdala ( Amyg ) and the piriform cortex ( PirC ) . ( A1-6 ) Enlarged tractography images of the piriform cortex ( PirC ) and the amygdala ( Amyg ) at different time points during epileptogenesis in the same animal ( before injection , pre and 1d , 4d , 7d , 14d and 31d following SE ) . Scale bar: 50 µm . ( B–E ) Quantitative analysis of the mean diffusivity ( MD ) and fractional anisotropy ( FA ) in the piriform cortex and the amygdala , respectively . Values are plotted for individual mice ( left panel; controls , black , n = 5; kainate-injected animals color-coded ) and for groups ( right panel; epileptic , dark grey; non-epileptic , light grey ) during epileptogenesis . Two-way ANOVA , Bonferroni’s post-test , nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 015 In fact , DWI reflected the extensive changes of hippocampal microstructure during epileptogenesis ( Figure 5A , B ) , characterized by a robust increase of dorsoventrally-oriented streamlines in the GCL and a decrease in the stratum radiatum of CA1 ( Figure 5E , G ) . Correspondingly , fractional anisotropy ( as a measure of tissue organization ) was increased in the DG during late epileptogenesis ( Figure 5O; 16d: 19 . 9% , p<0 . 05; 31d: 28 . 1% , p<0 . 001 ) , but exhibited an early and transient decrease in the CA1 region ( Figure 5—figure supplement 1G; 1d: −17 . 8% , p<0 . 05 ) . Direct histological analysis suggested that these changes indeed correspond to GCD and the loss of CA1 pyramidal cell dendrites , respectively ( Figure 5F , H ) . Detailed quantitative analysis revealed that all DWI metrics ( mean diffusivity , MD; axial diffusivity , AD; radial diffusivity , RD; fractional anisotropy , FA ) progressively increased in the DG and exhibited similar dynamics during epileptogenesis ( Figure 5I , K , M , O ) . Particularly , the rise of AD during intermediate disease progression showed the most reliable correlation with chronic histopathological changes: AD significantly increased as early as 8 days after SE in epileptic mice ( Figure 5M; 8d: 10 . 6% , p<0 . 01; 16d: 19 . 9%; 31d: 18 . 0% , both p<0 . 001 ) and was correlated with the volume of the GCL ( Figure 5N; 8d: r² = 0 . 68 , p<0 . 01; 16d: r² = 0 . 84 , p<0 . 001; 31d: r² = 0 . 66 , p<0 . 01 ) , suggesting that AD is sensitive to microstructural alterations associated with GCD . No significant diffusivity changes were found in the piriform cortex or the amygdala ( Figure 5—figure supplement 2 ) . Given that radial glia cells appear to enlarge and proliferate , concomitantly with the progression of GCD ( Figure 5G–M ) , it is conceivable that dorsoventrally-directed radial gliosis might represent an early anatomical substrate affecting the DWI metrics , in particular diffusivity along the dorsoventral axis ( dvD ) in the DG ( Figure 6 ) . Indeed , dvD increased continuously reaching significance at day 8 ( Figure 6F; 8d: 22 . 7%; 16d: 26 . 6%; 31d: 21 . 7% , all p<0 . 001 ) similar to the rise in AD . Importantly , from 4 days onward the magnitude of dvD as well as its calculated volume was highly correlated with the later GCL volume ( Figure 6G ) and the integrated density of GFAP labeling in the GCL ( Figure 6J; 4d: r² = 0 . 61 , p<0 . 01; 8d: r² = 0 . 85; 16d: r² = 0 . 87; 31d: r² = 0 . 81; all p<0 . 001 ) . This suggests that radial gliosis largely affects early diffusivity changes . However , mossy fiber sprouting and reorganization of granule cells ( Figure 4A–F ) likely contribute to changes of DWI metrics observed during further disease progression . 10 . 7554/eLife . 25742 . 016Figure 6 . Radial gliosis contributes to DWI metrics . ( A , C ) Representative DWI tractography maps of control and epileptic mice . Dashed lines , borders of the hippocampal layers; cc , corpus callosum; so , stratum oriens; py , pyramidal layer; sr , stratum radiatum; slm , stratum lacunosum moleculare; ml , molecular layer; gcl , granule cell layer . Scale bar , 100 µm . ( B , D ) High-magnification confocal images of GFAP staining in the region-of-interest corresponding to A and C ( indicated as white boxes ) . Scale bar , 30 µm . ( E , H ) Quantitative analysis of dorsoventral diffusivity ( dvD ) in the DG and the calculated volume of increased dvD , respectively , plotted for individual mice ( color-coded ) . ( F ) Groups analysis of dvD for epileptic ( dark grey ) and non-epileptic ( light grey ) during epileptogenesis . Source data is provided in ‘Figure 6—source data 1' . Two-way ANOVA , Bonferroni’s post-test , **p<0 . 01; ***p<0 . 001; nnon-epi = 6 , nepi = 7 . Values are presented as the mean ± SEM . ( I ) Quantitative analysis of GFAP optical density plotted for controls ( black , n = 5 ) and individual kainate-injected mice ( color-coded ) . One-way ANOVA , Bonferroni’s post-test , **p<0 . 01; ***p<0 . 001 , number of sections , ncntr l= 40 ( 8 sections from five controls each ) , nNP10 = 8 , nNP11 = 8 , nNP14 = 8 , nNP25 = 8 , nNP26 = 8 , nNP27 = 8 , nNP31 = 8 , nNP34 = 8 . ( G , J ) Linear regression analysis of dvD and the calculated dvD volume at distinct time points during epileptogenesis with the total GCL volume and the integrated density of GFAP staining , respectively , in epileptic and non-epileptic mice ( n = 13; Pearson’s correlation , corrected for multiple comparison; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 01610 . 7554/eLife . 25742 . 017Figure 6—source data 1 . Summary of dorsoventral diffusivity metrics . Quantitative values of dorsoventral diffusivity ( dvD ) measurements are listed for individual mice ( saline-injected: N12 , NP13 , NP17 , NP28 , NP29; kainate-injected: NP10 , NP11 , NP14 , NP25 , NP26 , NP27 , NP31 , NP34 ) and longitudinal time points ( pre , 1d , 4d , 8d , 16d , 31d following injection ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 017 In summary , our results show that during epileptogenesis microstructural reorganization in the DG identifies disease progression and predicts the later severity of GCD associated with HS . Next , we investigated the potential translational value of the observed microstructural changes as an imaging biomarker in humans based on ex vivo DWI scans and subsequent histology on resected hippocampi from seven patients with intractable mTLE ( Figure 7 ) . Quantitative DWI evaluation , however , was only feasible in five hippocampi due to limited tissue integrity and slice thickness . 10 . 7554/eLife . 25742 . 018Figure 7 . Validation of DWI biomarkers in human mTLE . ( A1–C1 ) T2-weighted images of sclerotic human hippocampi from mTLE patients scanned ex vivo . ( A2–C2 ) Corresponding FA maps ( color-coded for orientation: dorsoventral , turquoise; mediolateral , red; rostrocaudal: blue ) . Mean diffusivity ( MD ) and fractional anisotropy ( FA ) values for CA1 and DG are denoted in the lower-left , respectively . ( A3–C3 ) Representative NeuN staining of scanned sections . Different Wyler grades according to the severity of neuronal loss ( W I: mild , W III: moderate; W IV: strong ) . Scale bars , 3 mm; inset , 200 µm . ( D–E ) Representative confocal images of hippocampal sections ( Wyler I and IV ) double immunolabeled for NeuN ( turquoise , neurons ) and GFAP ( magenta , astrocytes and radial glia cells ) or ( F–G ) NeuN and Synaptoporin ( magenta , mossy fiber synapses ) , respectively , revealing differences in the microstructure of the human DG between Wyler grades . ( D1–E1 , F1–G1 ) GFAP and Synaptoporin staining alone . ( D2–E2 , F1–G2 ) NeuN staining displayed together with reconstructed radial glia processes and mossy fiber boutons within the GCL , respectively . ( H–I , J–K ) Quantitative analysis for MD and FA in CA1 and in the DG , respectively ( nWI = 2 , nWIII = 2 , nWIV = 1; no statistical test performed ) . Refer to ‘Figure 7—figure supplement 1' for MRI metrics acquired lower scanning resolution . ( L ) Quantitative analysis for the mean volume of GFAP-labeled radial glia processes as well as ( M ) Synaptoporin-labeled mossy fiber synapses , ( N ) and for GFAP optical density as well as ( O ) the density of Synaptoporin-labeled profiles within the GCL ( nWI = 3 , nWIII = 2 , nWIV = 2; no statistical test performed ) . ( P , Q ) Correlation of FA values with the GFAP optical density and the density of Synaptoporin-labeled profiles , respectively ( nWI = 2 , nWIII = 2 , nWIV = 1; Pearson’s correlation ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 01810 . 7554/eLife . 25742 . 019Figure 7—figure supplement 1 . Comparison of high- and low-resolution ex vivo DWI . ( A1–E1 ) T2-weighted images and corresponding FA maps of human hippocampi from mTLE patients scanned ex vivo at high ( 200 × 200×200 µm³ ) , and ( A2–E2 ) at low resolution ( 0 . 5 × 0 . 5×1 . 5 µm³ ) . FA maps are color-coded for orientation: dorsoventral , turquoise; medio-lateral , red; rostro-caudal: blue . ( F–G , H–I ) Quantitative analysis for mean diffusivity ( MD ) and fractional anisotropy ( FA ) in CA1 and in the DG at low scanning resolution . Scale bar: 2 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 01910 . 7554/eLife . 25742 . 020Figure 7—figure supplement 2 . Imaris-based 3D-reconstruction . ( A ) Volume-rendered confocal z-stack of GFAP staining in mouse ( C ) human , and ( E ) Synaptoporin in human DG ( upper panel ) . Corresponding surface reconstruction of immunolabeled profiles within a region-of-interest in the GCL ( lower panel ) . gcl , granule cell layer; iml , inner molecular layer . ( B , D , F ) List of parameters used for Imaris-based surface reconstruction . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 020 Similar to our results obtained in mice , human epileptic hippocampi with strong HS ( i . e . Wyler grade III or IV ) exhibited elevated MD and FA values in the DG , while FA was reduced in the CA1 region ( Figure 7H–K ) . These changes could still be detected at lower spatial resolutions achievable in a clinical setting ( Figure 7—figure supplement 1 ) . Moreover , quantitative morphometry revealed that the density and volume of GFAP-labeled radial glia cell processes also increased with the Wyler grade ( Figure 7L , N ) demonstrating that radial gliosis accompanies HS also in humans . Conversely , only the volume but not the density of Synaptoporin-labeled mossy fiber boutons appear to increase with the Wyler grade ( Figure 7M , O ) . Importantly , FA values in the GCL were positively correlated with the density of radial glia processes ( Figure 7P; r² = 0 . 78 , p<0 . 05 ) , suggesting that DWI biomarkers sensitive to microstructural changes in the DG are applicable in intractable mTLE in humans . In our MRI dataset several biomarkers were inter-correlated ( Figure 8A ) , presumably because they originate from similar physiological processes and/or are mathematically dependent . In order to clarify which biomarkers are interchangeable , and which are more suitable to predict the severity of histopathological changes , we performed a principal component analysis ( PCA ) . Mapping principal components ( PCs ) of the biomarker dataset onto histopathological changes associated with HS , we found a strong linear relationship between PC1 and GCD ( r2 = 0 . 86 , p<0 . 01 ) , microgliosis ( r2 = 0 . 68 , p<0 . 05 ) and a trend for radial gliosis ( Figure 8B–D ) . Higher order PCs were not significantly correlated with histopathology . Control animals formed a well separated group on the PC1-axis , indicating that PC1 is not only suitable to predict the histopathological changes in the epileptic group , but also to distinguish epileptic from healthy animals . The first two PCs accounted for 58% of the variation ( PC1 , 38%; PC2 , 20% ) . To estimate how much information the individual markers share with the first two PCs , we correlated each biomarker with PC1 and PC2 and identified four distinct clusters ( Figure 8E ) . The largest cluster was especially interesting due to its strong correlation with PC1 ( r2 = 0 . 73 , black variables ) . In fact , a subset of biomarkers of this cluster was remarkably close to PC1 = 1 , comprising early changes ( T2 in CA1 and 1H-MR for NAA , glutamate and GABA ) , but also intermediate changes ( DWI metrics ) during epileptogenesis . Early alterations in diffusivity grouped into a separate cluster ( white variables ) and were represented almost exclusively by PC2 . Conversely , changes in myoinositol , early and late lactate or FA , and intermediate GABA ( light grey variables ) were represented by PC1 ( negative correlation ) and PC2 ( positive correlation ) . Similarly , intermediate NAA , glutamate and FA as well as early-to-intermediate changes of T2 intensity in the DG were correlated with PC1 and PC2 ( dark grey variables; positive correlation with both PCs ) . 10 . 7554/eLife . 25742 . 021Figure 8 . PCA-based evaluation of MRI biomarkers . ( A ) Correlation matrix of imaging parameters used for PCA . Only data of kainate-injected mice was analyzed . Positive and negative correlation coefficients are color-coded in red and blue , respectively . ( B ) Pearson’s correlation analysis of PC1 scores and total GCL volume , ( C ) extent of microgliosis and ( D ) radial gliosis ( i . e optical density of GFAP in the GCL ) in epileptic mice ( color-coded , n = 7 ) . Controls ( grey , n = 5 ) are plotted but not included in the analysis . ( E ) Correlation plot showing the similarity of individual MRI variables with PC1 and 2 . Clusters of variables , indicated as numbers , are grey scale-coded . Arrows denote the population vector of the corresponding cluster . Coefficients are indicated as circles ( small = 0 . 5; large = 1 . 0 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 021 Finally , we tested whether our PCA-validated MRI biomarkers also relate to the seizure burden in chronic stages of mTLE . Therefore , we identified seizure-like episodes ( Figure 9A; spike-and-wave trains >10s ) in intra-hippocampal EEG recordings of epileptic mice ( n = 6 ) using an automated algorithm . Across animals we found a considerable variability with respect to their individual seizure frequency ( Figure 9B; NP11 = 0 . 73 ± 0 . 09; NP14 = 0 . 47 ± 0 . 11; NP25 = 0 . 30 ± 0 . 05; NP27 = 0 . 73 ± 0 . 01; NP31 = 0 . 64 ± 0 . 10; NP34 = 0 . 45 ± 0 . 04 seizure-like episodes / min ) . Correlating the seizure frequency with PC1 revealed an inverse relationship between both parameters ( Figure 9; r² = 0 . 94 , p<0 . 01 ) , suggesting that our identified biomarkers are also predictive for the seizure severity in chronically epileptic mice . 10 . 7554/eLife . 25742 . 022Figure 9 . PCA-verified biomarkers predict the seizure severity . ( A ) Representative EEG recordings from the ipsilateral hippocampus showing three examples of identified seizure-like episodes . Scale bars: Horizontal 5 s , vertical 2 mV . ( B ) Quantitative analysis of seizure-like episodes for individual epileptic mice ( color-coded; note that EEG data is lacking for NP26 ) . One-way ANOVA , Bonferroni’s post-test , *p<0 . 05 , **p<0 . 01 , number of recordings: nNP27 = 5 , nNP11 = 7 , nNP31 = 5 , nNP14 = 4 , nNP34 = 5 , nNP25 = 7 . Values are presented as the mean ± SEM . ( C ) Pearson’s correlation analysis of PC1 scores and the frequency of seizure-like episodes ( n = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 02210 . 7554/eLife . 25742 . 023Figure 9—figure supplement 1 . Experimental design to translate MRI biomarkers into the clinic . Schematic illustrating an outline for the clinical translation of experimentally-identified biomarkers . Multimodal MRI is repeatedly applied early after the initial precipitating injury ( IPI ) until the onset of clinical seizures . Cross-correlation and PCA identifies interconnected MRI parameters and validates their predictive value with respect to MRI-identified HS and seizure onset in chronic mTLE . Subsequently , the change of MRI metrics ( d ) early after IPI can be assigned to the probability ( p ) of epilepsy onset and severity . Most robust MRI biomarkers can then be routinely used in clinics to prognose acquired mTLE , which allows to start anti-epileptogenic treatment early before clinical manifestation of epilepsy . DOI: http://dx . doi . org/10 . 7554/eLife . 25742 . 023 In summary , our PCA results corroborate the idea that T2-weighted imaging and 1H-MR spectroscopy are useful tools for early prognosis ( 1–4 days after the precipitating insult ) of later epilepsy severity , whereas DWI is informative at subsequent stages of disease development .
The present study provides a comprehensive , time-resolved analysis of MRI biomarkers for pharmacoresistant mTLE . The identification of biomarkers and their prognostic value was based on the variable severity of histopathological and pathophysiological changes among intrahippocampally kainate-injected mice . Importantly , this well-regarded mTLE model replicates all the major pathological hallmarks of the human disease ( Bouilleret et al . , 1999; Lévesque and Avoli , 2013 ) , comprising SE-induced acute unilateral hippocampal cell loss , progression of HS and network reorganization , as well as the emergence of spontaneous recurrent seizures in the chronic epileptic stage . Although the initial epileptogenic insult triggering mTLE in humans is apparently different ( Lukasiuk and Becker , 2014 ) , HS is the common pathology in both human mTLE and animal models ( Wieser and ILAE Commission on Neurosurgery of Epilepsy , 2004; Thom , 2014 ) , and is thought to be critically involved in seizure generation ( Pallud et al . , 2011; Krook-Magnuson et al . , 2015; Walker , 2015 ) . Similar to the human pathology , we found inter-individual differences in the disease severity of epileptic mice , determined by the degree of pyramidal cell death-associated microgliosis and GCD , as well as the frequency of paroxysmal episodes in chronic epilepsy . In a parallel set of experiments we also show that these seizure-like events typically occur in a highly recurrent fashion around the second week after SE , which is in line with previous observations in the same model ( Riban et al . , 2002; Arabadzisz et al . , 2005; Heinrich et al . , 2011 ) . Thus , we consider the emergence of recurrent paroxysmal episodes at around two weeks after SE as the onset of the chronic phase . Remarkably , during the first two weeks – typically considered as the latent phase - the dynamics of DWI metrics in the hippocampus , its histopathological reorganization and the development of epileptic activity appeared similar . We would like to emphasize , however , that it was unfeasible to directly relate MRI changes to the progression of epileptic activity , since chronically implanted electrodes would have massively interfered with MRI measurements . Probing our hypothesis that initial SE-induced hippocampal damage might predict later histopathological severity , we performed longitudinal high-resolution T2-weighted imaging and 1H-MR spectroscopy at 7T using a mouse brain adapted cryo-coil . We found an acute and transient increase of hippocampal T2 intensity . Accordingly , previous studies have shown that pyramidal cells and hilar neurons are massively lost 1 day following SE ( Heinrich et al . , 2006; Marx et al . , 2013 ) , demonstrating the spatial and temporal specificity of the observed T2 hyperintensity with respect to cell death . In line with our interpretation , increased T2 intensity is thought to represent acute cell loss , subsequent edema , and gliosis ( Jackson et al . , 1993; Wall et al . , 2000 ) . An acute rise of hippocampal T2 intensity had already been demonstrated at 4 . 7T by Bouilleret et al . ( 2000 ) following intrahippocampal kainate application , however , the authors did not assess their prognostic value with respect to disease severity . Our present analysis reveals that this signal increase predicts the degree of later HS . Similarly , early T2 changes were also found in the hippocampus of pilocarpine-injected rats predicting its volume reduction in chronic stages of epilepsy ( Choy et al . , 2010 ) . In contrast , pilocarpine-injected mice exhibited increased hippocampal T2 values after the first week of epileptogenesis , indicating that the extent and time course of SE-induced hippocampal damage also depends on the disease model ( Kharatishvili et al . , 2014 ) . Beyond animal studies , transient T2 hyperintensity was reported in children 3 or 5 days after experiencing prolonged febrile seizures ( Scott et al . , 2002 , 2003; Shinnar et al . , 2012 ) . Importantly , this rise of T2 intensity predicted the emergence of HS in longitudinal studies , but its relation to epilepsy onset is yet unclear ( Provenzale et al . , 2008; Lewis et al . , 2014 ) . Given that T2 hyperintensity is not inevitably linked to cell loss ( Dubé et al . , 2004; Thom et al . , 2005 ) , 1H-MR spectroscopy represents a practical complementary method to monitor tissue damage . Reduced NAA levels can successfully identify the sclerotic hippocampus in both human epilepsy patients ( Ng et al . , 1994; Hetherington et al . , 2002 ) and animal models ( Tokumitsu et al . , 1997; Gomes et al . , 2007; Filibian et al . , 2012 ) , which is in line with our results showing an early and long-lasting decrease of NAA in the epileptogenic hippocampus . Moreover , we show that , as early as 1 day following SE , reduced NAA correlated with the extent of cell loss-associated microgliosis . In addition , we quantified alterations in the concentration of the major neurotransmitters synthesized in excitatory principal cells ( glutamate ) and inhibitory interneurons ( GABA ) . Both neuron populations are known to be substantially diminished in the sclerotic hippocampus ( Marx et al . , 2013; Thom , 2014 ) . Consistent with these findings , we found that glutamate and GABA rapidly decrease following SE . However , one has to bear in mind that during and early after SE dramatic changes occur in the hippocampus , including cerebral edema and compensatory changes that affect 1H-MR metrics at early time points after injury . Similar to NAA , the early decrease of glutamate levels was long-lasting and predicted the degree of HS . Conversely , GABA levels increased again at 8 days onwards . Taking into account that hippocampal neurodegeneration is thought to be progressive in rodent mTLE ( Pitkänen et al . , 2002; Nairismägi et al . , 2004 ) ; and own unpublished observations and in human patients ( Kälviäinen et al . , 1998; Briellmann et al . , 2002a ) , both long-lasting glutamate decline as well as the restoration of GABA levels following SE appear counterintuitive . We hypothesize that during epileptogenesis the sprouting of glutamate and GABA-producing mossy fibers ( own unpublished observations ) likely obscures ongoing neurodegeneration on the molecular level . Accordingly , microstructural reorganization subsequent to hippocampal cell loss was monitored by DWI determining the speed and orientation of water diffusion along biological membranes ( Mori and Zhang , 2006 ) for example of axonal fibers ( Harsan et al . , 2010 , 2013 ) . In the epileptic hippocampus of several rodent epilepsy models DWI successfully identified changes of water diffusion , which is discussed to primarily relate to mossy fiber sprouting and reorganization of myelinated axons ( Kharatishvili et al . , 2007; Kuo et al . , 2008; Laitinen et al . , 2010; Sierra et al . , 2015a , 2015b; Salo et al . , 2017 ) . However , it is important to note that the above mentioned epilepsy models do not fully reflect the histopathological hallmarks of pharmacoresistant mTLE ( i . e . , unilateral HS including GCD ) complicating a translation to the human disease . Using the intrahippocampal kainate mouse model we demonstrate that the degree of HS-associated microstructural alterations can be predicted by the progressive increase of diffusivity in the DG as early as 4 days following SE . Similar results were obtained in epileptic rodents , either after traumatic brain injury or after systemic pilocarpine injection ( Kharatishvili et al . , 2007 , 2014 ) , suggesting that DWI biomarkers identified in our study are robust among different models . In rats undergoing traumatic brain injury the increase of hippocampal mean diffusion spans approximately three months ( Kharatishvili et al . , 2007 ) , indicating a longer duration of disease progression in this model . Conversely using pilocarpine-injected mice , Kharatishvili et al . ( 2014 ) showed that the apparent diffusion coefficient in the hippocampus increased within the first week following SE , predicting the spiking frequency , however , the relation to seizure frequency remained unclear . Moreover , several studies using systemic epilepsy models also revealed T2 or diffusivity changes in the piriform cortex or the amygdala ( Roch et al . , 2002; Choy et al . , 2010 , 2014; Kharatishvili et al . , 2014 ) , suggesting that in these animal models both , the hippocampus as well as extrahippocampal brain regions sensitive to excitotoxic injury are affected . We did not observe these changes , highlighting the key role of the lesioned hippocampus in driving epileptogenesis in the intrahippocampal kainate mouse model . As discussed by Gomes and Shinnar ( 2011 ) , an attractive biomarker should ideally infer the progression of epileptogenesis and not only its presence . Indeed , DWI changes in the DG steadily progressed over the first 2 weeks of epileptogenesis , resembling the aggravation of epileptiform activity during this period ( Heinrich et al . , 2011 ) . Importantly , the fact that early MRI biomarkers are strongly correlated to later biomarkers makes epileptogenesis as a whole predictable . Early changes ( i . e . cell death identified by T2 and 1H-MR ) might scale more or less linearly to the later disease development ( i . e . , microstructural reorganization monitored by DWI ) , however , it remains to be determined whether different MRI dynamics might also reflect different times of seizure onset . Given this interconnectivity of MRI biomarkers , our PCA results validate early changes in T2 , 1H-MR and subsequent changes in DWI metrics as highly prognostic for both , later HS severity and seizure frequency . Interestingly , PC1 is positively correlated with histopathological changes associated with HS , but negatively with seizure frequency . This data indicates that more severe early hippocampal injury and subsequent tissue reorganization lead to the development of stronger HS and lower seizure burden , compared to a milder , but still epileptogenic insult . We hypothesize that hippocampal damage needs to exceed a certain threshold until epilepsy develops , whereas further cell loss might attenuate epilepsy severity because fewer neurons are able to participate in seizure generation . Investigating the anatomical correlates of DWI changes further , our in-depth morphological analysis clearly suggests that the early ( 4 to 8 days ) rise of DWI values is largely driven by radial gliosis , whereas in later stages ( 8 days onward ) neuronal hypertrophy , mossy fiber sprouting and translocation of granule cell dendrites might also contribute . Our results are in line with Budde et al . ( 2011 ) and Salo et al . ( 2017 ) showing that FA increases with glial scarring in the perilesional cortex following cerebral trauma or the CA3 region following SE , respectively . The notion that changes in DWI metrics are influenced by gliosis is further supported by our observation that DWI changes in the DG were accompanied by T2 hyperintensity , known to correlate best with the number of local glia cells ( Briellmann et al . , 2002b ) . Conversely , in systemic epilepsy models reorganization of myelinated fibers in the molecular layer , but not of local astrocytes , appear to effectively drive DWI changes in the DG , particularly with respect to FA and AD ( Salo et al . , 2017 ) . Considering that in these models GCD is not present , we conclude that the region-specific orientation of glial cells in the DG and their reorganization during GCD formation is critical to affect DWI metrics . Accordingly , as a prerequisite for radial glia cells to affect diffusion anisotropy , they traverse the GCL perpendicularly and respond quickly to epileptiform activity with hypertrophy , branching and proliferation ( Heinrich et al . , 2006; Sierra et al . , 2015c ) . In turn , we show that processes of these glia cells are still dorsoventrally organized in the dispersed septal DG , thus contributing to the observed increase in dvD , AD and FA . Indeed , we demonstrated that FA in the DG of human mTLE patients relate to the increase of radial gliosis at higher Wyler grades . Our finding is supported by previous studies revealing that the radial glia processes traverse the GCL and are correlated with the extent of GCD and HS in patients ( Fahrner et al . , 2007 ) . Moreover , our analysis demonstrated that MD and FA are inferior to AD and dvD regarding their ability to predict HS-associated microstructural reorganization in the DG . Previous animal studies attributed increased diffusivity and anisotropy to mossy fiber sprouting ( Kharatishvili et al . , 2007; Parekh et al . , 2010 ) and reorganization of myelinated fibers ( Laitinen et al . , 2010 ) , but in these studies histological validation was performed only in the chronic stage of epilepsy . Our histological analysis expands this view revealing that significant mossy fiber sprouting is restricted to the later stages of epileptogenesis ( around 14 days onward ) , which is consistent with previous observations for the CA2 region ( Häussler et al . , 2016 ) . Although it is well conceivable that mossy fiber sprouting alters DWI metrics during this period , earlier changes ( 4 to 14 days ) particularly of AD and dvD likely correspond to radial gliosis . In vivo DWI studies performed in human mTLE consistently reported an increase of MD , but a decrease of FA in the sclerotic hippocampus ( Thivard et al . , 2005; Salmenpera et al . , 2006; Liacu et al . , 2010 ) which can be explained by the expansion of the extracellular space secondary to cell death and the subsequent loss of tissue organization . However , these studies addressed diffusivity changes in the whole hippocampus , disregarding the differential effects of neuronal reorganization in hippocampal subfields ( Janz et al . , 2017 ) . Consistent with this notion , we found an opposing trend of FA values for the CA1 region and the DG in resected hippocampi from mTLE patients with high Wyler grades . Similarly , in epileptic mice we encountered an initial decrease of FA in the CA1 stratum radiatum , but an increase in the GCL , clearly demonstrating the importance of hippocampal subfield imaging with respect to DWI biomarkers . However , one has to bear in mind the high spatial resolution used to detect these subfield-specific changes . Our DWI protocol employed a planar resolution of 60 × 60 µm² to image the mouse hippocampus spanning a cross-sectional area of about 2 . 5 × 1 . 5 mm² . In contrast to our setup optimized for the mouse brain , a clinical setup might reach 800 µm isotropic resolution in a 1 hr scan using state-of-the-art imaging sequences at 7T ( Heidemann et al . , 2012 ) . Considering the much larger cross-sectional area of the human hippocampus ( about 8 × 8 mm² ) , it should become feasible to retrieve accurate DWI metrics from hippocampal subfields in a clinical setting . Indeed , degeneration of the perforant path in aging humans has already been shown at 3T ( Yassa et al . , 2010 ) . Moreover , imaging of subfield-specific pathology in human TLE was successfully performed at 3T and at 7T at planar resolutions of 1000 × 1000 µm² and 500 × 500 µm² , respectively ( Goubran et al . , 2016 ) . These technical advances provide the foundation for the application of our imaging biomarkers in a clinical setting with the ultimate goal to identify and even prevent epileptogenesis before seizure onset . In this context , longitudinal MRI experiments similar to the FEBSTAT study ( Lewis et al . , 2014 ) will be essential , to determine the dynamics of hippocampal alterations in humans who just suffered from prolonged seizures ( e . g . , due to SE , head trauma , febrile seizures or encephalitis ) . These studies would require repeated MRI measurements over many years , starting with short intervals early after the initial precipitating injury and lasting until the first clinical seizures arise ( Figure 9—figure supplement 1 ) . MR measurements would need to rely on multiple imaging modalities to integrate the most promising MRI biomarkers evaluated in the present study ( i . e . , T2 hyperintensity in CA1 , changes in hippocampal NAA , glutamate and GABA levels and AD , dvD and FA increase in the DG ) . We propose that the quantitative relationship between these biomarkers will be highly informative for the course of epileptogenesis and may allow a prognosis for subsequent disease progression and early pharmacological intervention .
Experiments were performed on adult ( 9–12 weeks ) male C57Bl/6N wildtype ( Charles River , Sulzfeld , GER ) or transgenic Thy1-eGFP mice in which eGFP is expressed under the control of the Thy1 promoter ( RRID:IMSR_JAX:007788 , M-line , C57BL/6 background; Feng et al . , 2000 ) . Each animal represents an individual experiment , performed once . A total of 13 wildtype and 25 transgenic mice were used for longitudinal MRI experiments and complementary histological investigation at individual time points , respectively . Additionally , 8 Thy1-eGFP mice were used for longitudinal EEG recordings . Mice were kept at room temperature ( RT ) in a 12 hr light/dark cycle providing food and water ad libitum . All animal procedures were in accordance with the guidelines of the European Community’s Council Directive of 22 September 2010 ( 2010/63/EU ) and were approved by the regional council ( Regierungspräsidium Freiburg ) . Mice received a single , unilateral kainate injection into the hippocampus as described previously ( Heinrich et al . , 2006; Häussler et al . , 2012 ) . In brief , anesthetized mice ( ketamine hydrochloride 100 mg/kg , xylazine 5 mg/kg , atropine 0 . 1 mg/kg body weight , i . p . ) were stereotaxically injected with 50 nL of a 20 mM kainate solution ( Tocris , Bristol , UK ) in 0 . 9% saline into the right dorsal hippocampus ( coordinates relative to bregma: anterioposterior = −2 . 0 mm , mediolateral = −1 . 5 mm , and relative to the cortical surface: dorsoventral = −1 . 8 mm ) . Controls were injected with 0 . 9% saline . After recovery from anesthesia , behavioral SE was verified by observing mild convulsions , chewing or rotations . Four mice , which died between the fourth and sixth day following kainate injection , were excluded from the study . Seven resected hippocampal specimens from mTLE patients ( mean age: 36 . 8 ± 8 . 8 years ) were used in this study . All patients had experienced pharmacoresistant complex partial seizures and underwent amygdalohippocampectomy to achieve seizure control . Patients were treated according to the Epilepsy Surgery Program of the University Medical Center Freiburg . Informed consent was obtained from all patients . Tissue selection was approved by the Ethics Committee at the University Medical Center Freiburg . Classification of Ammon's horn sclerosis was carried out by a clinical neuropathologist according to Wyler ( Hermann et al . , 1992 ) : Grade 1 , slight ( <10% ) or no loss of pyramidal cells in the cornu ammonis 1–4 ( CA1-4 ) ; Grade 2 , gliosis and moderate loss ( 10–20% ) of CA1 , CA3 and/or CA4 pyramidal cells; Grade 3 , gliosis with >50% pyramidal cell loss in CA1 , CA3 and CA4 , but sparing CA2; Grade 4 , gliosis with >50% pyramidal cell loss involving CA1-4 . All MR measurements were performed on a 7T preclinical MRI system equipped with a mouse head adapted 1H transmit-receive CryoProbe ( MR system: BioSpec 70/20 , software ParaVision 6 . 0 , Bruker , Ettlingen , Germany ) . For animal handling , a dedicated mouse bed ( Bruker ) was used that provided a three point-fixation system ( tooth-bar and ear-plugs ) , isoflurane anesthesia and stabilization of the body temperature of the mice . Respiration was measured by a pressure sensor and isoflurane anesthesia ( 1 . 5% ) was adjusted to keep the respiration rate at 50–60 breaths/min during the scans . At the beginning of each session , a B0 field map ( isotropic resolution 0 . 3 × 0 . 3 × 0 . 3 mm3 ) was acquired and used to improve the B0 field homogeneity inside the mouse brain ( calculation of the local map-based shims ) , followed by adjustments of the local frequency and reference power ( flip angle ) . This process was repeated a second time to optimize the B0 field inside the brain ( linewidth of the whole brain water signal 30–50 Hz ) . For spectroscopy the B0 field was optimized according to the region-of-interest ( line width of the unsuppressed water signal inside the voxel 8–12 Hz ) . For EEG recording after MR measurements ( 31 days after saline or kainate injection ) , mice were anesthetized and were stereotaxically implanted with platinum-iridium wire electrodes ( Ø 125 µm; World Precision Instruments , Berlin , GER ) in both hippocampi . Coordinates relative to bregma: anterioposterior = −2 . 0 mm , mediolateral = +1 . 4 and −1 . 4 mm , and relative to the cortical surface: dorsoventral = −1 . 6 mm . Two stainless-steel screws implanted above the somatosensory cortex were used as reference and ground . Subsequently , mice were connected to a miniature preamplifier ( Multi Channel Systems , Reutlingen , GER ) . Signals were amplified ( 1000-fold , bandpass 1 Hz to 5 kHz ) and digitized ( sampling rate 10 kHz; Power1401 analog-to-digital converter; Spike2 software , RRID:SCR_000903 , Cambridge Electronic Design , Cambridge , UK ) . Mice were recorded at five consecutive days , 2–3 hr each day , to check for epileptiform activity . For longitudinal EEG recordings , eight mice were chronically implanted with platinum-iridium wire electrodes into the ipsilateral hippocampus directly after kainate injection and recorded ( same parameters as described above ) at 6 hr , 1 d , 3–5 d , 6–7 d , 12–16 d and 17–21 d after injection . Epileptic discharges identified by trains of high-amplitude population spikes ( ≥5 s; see Figure 1—figure supplement 1E , F ) were quantified by two blinded observers using Spike2 . The mean of the values obtained by both observers was used for statistical analysis . Each mouse represents the biological and the number of recordings per mouse the technical replicate . Only recordings from electrodes located in the molecular or granule cell layer ( verified in DAPI-stained sections ) were used for further evaluation . Mice were anesthetized 36 days after kainate injection and transcardially perfused with 0 . 9% saline followed by 4% paraformaldehyde in 0 . 1 M phosphate buffer ( PB; pH 7 . 4 ) for 5 min . Postfixation was performed in the same fixative overnight at 4°C . Brains were subsequently cut ( coronal plane , 50 μm ) on a vibratome ( VT1000S , Leica , Bensheim , GER ) . Each mouse represents the biological and the number of brain sections per mouse the technical replicate . Resected human specimens were first trimmed to blocks of about 5 mm thickness and subsequently immersion-fixed in 4% paraformaldehyde in 0 . 1 M PB overnight at 4°C , transferred to 30% sucrose at 4°C overnight , shock-frozen and finally stored at −80°C . Prior to ex vivo DWI scans , tissue blocks were thawed in 0 . 1 M PB at RT . For subsequent histology , tissue blocks were further cut into 100 µm serial sections on a VT1000S vibratome ( Leica ) . Each resected hippocampus represents one biological and the number of sections per hippocampus the technical replicate . For immunofluorescence staining , free-floating serial sections were pre-treated with 0 . 25% TritonX-100 in 1% bovine serum albumin for 1 hr ( mouse ) or 2 hr ( human ) . Subsequently , serial tissue sections were incubated with the following primary antibodies at 4°C for 24 hr ( mouse ) or 48 hr ( human ) : Guinea pig anti-NeuN ( 1:500; RRID:AB_2619988 , Synaptic Systems , Göttingen , GER ) , rabbit anti-ZnT-3 ( 1:2000; RRID:AB_10894885 , Synaptic Systems ) , rabbit anti-Synaptoporin ( 1:1000; RRID:AB_2619748 , Synaptic Systems ) , rabbit anti-Iba-1 ( 1:1000; RRID:AB_839504 , Wako Chemicals , Neuss , GER ) or rabbit anti-GFAP ( 1:500; RRID:AB_10013482 , Dako , Hamburg , GER ) . For detection , Cy2- , Cy3- or Cy5-conjugated secondary donkey anti-guinea pig or goat anti-rabbit antibodies ( 1:200; RRID:AB_2340462 , RRID:AB_2338000 and RRID:AB_2307385 , Jackson ImmunoResearch Laboratories Inc . , West Grove , USA ) were used . All secondary antibodies were applied for 3 hr ( mouse ) or 12 hr ( human ) at room temperature followed by rinsing in 0 . 1 M PB for 6 × 15 min ( mouse ) or 6 × 1 hr ( human ) . Counterstaining was performed with DAPI ( 4‘ , 6-diamidino-2-phenylindole; 1:10 . 000 , Roche Diagnostics GmbH , Mannheim , GER ) . Sections were mounted on glass slides and coverslipped with ProLong Gold ( Molecular Probes , Invitrogen , Carlsbad , USA ) the next day . Fluoro-Jade B ( FJB , Millipore , Schwalbach , GER ) staining was performed to monitor cell death . Sections were mounted on gelatin-coated glass slides and incubated with 0 . 06% potassium permanganate solution for 15 min followed by 0 . 0004% FJB solution for 30 min . Sections were then rinsed in xylene and coverslipped with Hypermount ( Thermo Fisher Scientific , Dreieich , GER ) . To quantify the severity of HS , two parameters ( cell loss in CA1-3 and GCD ) were investigated using an AxioImager2 microscope ( Zeiss , Göttingen , GER ) . Using a 10x objective ( Plan-APOCHROMAT , Zeiss ) , photomicrograph composites were acquired with a digital camera ( MR605 , Zeiss ) from Iba1- and/or DAPI-stained sections , processed with Zen software ( RRID:SCR_013672 , Zeiss ) and analyzed with Fiji ImageJ software ( RRID:SCR_002285 , Schindelin et al . , 2012 ) . The individual extent of pyramidal cell death was estimated by measuring the spatial extent of microgliosis in the cornu ammonis , since activated , amoeboid microglia build a dense scar tightly restricted to the region of pyramidal cell loss ( Figure 1D ) . For each mouse eight to twelve Iba-1 stained sections were categorized into five levels ( L1-5 ) according to the rostocaudal axis: Approximate coordinates relative to bregma , L1 = −0 . 9 to −1 . 6 mm , L2 = −1 . 6 to −2 . 3 mm , L3 = −2 . 3 to −2 . 9 mm , L4 = −2 . 9 to −3 . 6 mm , L5 = −3 . 6 to −4 . 0 mm . The mean length of microglial scarring was calculated for each level and summed up . The degree of GCD , which is characterized by a broadened cell layer and reduced cell density , was assessed by measuring the area of the GCL in DAPI-stained sections ( Figure 1A ) and calculating the overall volume , taking in to account the slice thickness of 50 µm . Since the number of available sections varied across mice ( 51 ± 7 sections ) , possibly due to slightly different brain sizes or non-uniform shrinking of perfused tissue , we normalized the rostocaudal axis to 2500 µm ( 50 sections per animal x 50 µm slice thickness ) . This ensured that correlations with MRI measures actually represent their relationship to HS-associated anatomical changes rather than the variability in the number of sections . In addition , we estimated individual degree of radial gliosis by quantifying the optical density of GFAP in the GCL of the septal hippocampus ( see parameters below ) using Fiji ImageJ , followed by multiplying its mean with the calculated GCL volume . Considering that GCD is not present in more temporal regions of the kainate-injected hippocampus , the mean of GFAP control values was used for calculating its integrated density in the non-dispersed GCL . Microstructural alterations in the DG were analyzed by confocal laser scanning microscopy in MR-scanned wildtype mice and in complementary experiments using transgenic Thy1-eGFP mice . For each animal , four to eight confocal stacks were acquired in two sections with an Olympus FV10i ( Olympus Deutschland , Hamburg , GER ) at high resolution ( 60x oil-immersion objective , 1024 × 1024 pixels; 4x frame-average; confocal aperture 1 airy unit; z-step size: 0 . 5 µm ) . Laser power and detector sensitivity were kept constant for each staining to allow comparison between samples . For quantitative morphometry , image stacks were transferred to Imaris 7 . 7 . 1 software ( RRID:SCR_007370 , Bitplane AG , Zurich , CH ) . Morphological changes of eGFP-labeled dentate granule cells ( 10–20 per z-stack ) were investigated by measuring the diameter at three positions along the dendrite and axon segments close to the soma . Changes in the density and volume of GFAP-labeled processes from radial glia cells and Synaptoporin-labeled mossy fiber boutons were inferred by surface reconstruction of a predefined region-of-interest within the GCL ( Figure 7—figure supplement 2 ) . Additionally , for each sample the optical density of GFAP within the GCL was quantified in six to eight confocal planes taken from the section surface using Fiji ImageJ . Similarly , the amount of sprouted mossy fibers was determined by measuring the optical density of ZnT-3 within the GCL . For optical density measurements the intensity spectrum was optimized for contrast by histogram equalization to obtain quantitative values representing more the actual density of labeled profiles than differences in labeling intensity itself . Due to the apparent anatomical changes in the kainate-injected hippocampus blinded observation was unfeasible . However , all analyses were performed semi-automatically , and both MRI and histological data were collected independently and combined subsequently to avoid a putative bias . Data was tested for statistical significance with Prism five software ( RRID:SCR_002798 , GraphPad Software Inc . , La Jolla , USA ) . Comparison of two groups was performed with an unpaired Student’s t-test ( two-tailed ) . When more than two groups were compared either one-way ANOVA followed by Dunnett's or Bonferroni’s post-test , or two-way ANOVA followed by Bonferroni’s post-test were used to correct for multiple comparison . Significance thresholds were set to: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . For all values , mean and standard error of the mean ( SEM ) are given . Correlations were tested using Pearson’s correlation ( slope significantly non-zero , confidence interval set to 95% , corrected for multiple comparisons ) . PCA was performed with Python ( RRID:SCR_008394 , Python Software Foundation , Beaverton , USA ) . All statistical results are summarized in Supplementary file 1 . The MRI dataset has been made available ( https://osf . io/7gmvn/ ) . The source code and user information for the custom seizure detection algorithm can be obtained from Ulrich Egert upon request . Contact: egert@imtek . uni-freiburg . de | Roughly one percent of people in the world suffer from epilepsy , a disorder in which individuals experience seizures due to abnormal electrical activity in the brain . Seizures can vary from brief episodes of amnesia or déjà-vu to convulsions and loss of consciousness . In adults , the most common form of epilepsy is known as temporal lobe epilepsy . As the name suggests , this type of epilepsy originates in a region of the brain called the temporal lobe , usually within a structure called the hippocampus . Many patients who develop temporal lobe epilepsy will have experienced a head injury or infection earlier in life that damaged their hippocampus . However , damage to the hippocampus does not always lead to epilepsy . Moreover , many years may pass between the damage and the onset of regular seizures . While some patients find that anti-epileptic drugs can control their seizures , others experience no benefit . For these patients , the only effective treatment is to remove the damaged brain tissue . At present , there is no way of knowing which patients with a damaged hippocampus will go on to develop temporal lobe epilepsy . To identify the deciding factors , Janz , Schwaderlapp , Heining et al . treated mice with a toxin that can damage the hippocampus . After roughly two weeks most of the mice were experiencing regular seizures . Imaging the animals’ brains during this two week period revealed that mice whose hippocampi showed more severe cell death shortly after exposure to the toxin surprisingly developed a milder form of epilepsy . The same was also true for animals whose hippocampi showed signs of being extensively reorganized . Further experiments show that samples of hippocampal tissue from the brains of human patients with temporal lobe epilepsy also showed these same cellular features . The next step is to test whether these changes can be used to predict which patients with hippocampal damage will develop epilepsy later in life . Identifying at-risk individuals would allow them to be treated earlier and hopefully prevent them from developing epilepsy in the first place . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Early tissue damage and microstructural reorganization predict disease severity in experimental epilepsy |
How myoblast populations are regulated for the formation of muscles of different sizes is an essentially unanswered question . The large flight muscles of Drosophila develop from adult muscle progenitor ( AMP ) cells set-aside embryonically . The thoracic segments are all allotted the same small AMP number , while those associated with the wing-disc proliferate extensively to give rise to over 2500 myoblasts . An initial amplification occurs through symmetric divisions and is followed by a switch to asymmetric divisions in which the AMPs self-renew and generate post-mitotic myoblasts . Notch signaling controls the initial amplification of AMPs , while the switch to asymmetric division additionally requires Wingless , which regulates Numb expression in the AMP lineage . In both cases , the epidermal tissue of the wing imaginal disc acts as a niche expressing the ligands Serrate and Wingless . The disc-associated AMPs are a novel muscle stem cell population that orchestrates the early phases of adult flight muscle development .
Stem cells populations can expand exponentially through symmetric division or asymmetrically divide to self-renew and produce a daughter cell , which differentiates to contribute to tissue formation or regeneration ( Morrison and Kimble , 2006; Micchelli and Perrimon , 2006; Mandal et al . , 2007; Maurange et al . , 2008; Knoblich , 2001; Egger et al . , 2008; Doe , 2008; Sousa-Nunes et al . , 2010; Brand and Livesey , 2011 ) . Many mature tissues such as colon , prostate , lung , muscle and brain use adult-specific stem cells in tissue maintenance or regeneration ( Reya et al . , 2001; Gonzalez , 2007 ) . The molecular mechanisms of regulating stem-cell proliferation have been studied in neural- , intestinal- , hematopoietic- and epithelial-stem cells of vertebrates and invertebrates ( Tulina and Matunis , 2001; Barker et al . , 2007; Ohlstein and Spradling 2007; Takashima et al . , 2008; Farkas and Hutter , 2008; Reichert , 2011; Homem and Knoblich 2012 ) . The regulation of stem-cell proliferation is usually through signals from a ‘stem-cell niche’ ( Mandal et al . , 2007; Brack et al . , 2008; Chen et al . , 2012; Cordero et al . , 2012 ) . Significant progress in understanding these mechanisms of regulation of stem-cell proliferation and self-renewal has been made in Drosophila . For example , in the developing optic lobe , stem cell-like neuroepithelial cells first increase in number through symmetric divisions and are then transformed into so-called neuroblasts which undergo self-renewing asymmetric divisions to generate differentiated neurons and glial cells in this part of the brain ( Maurange et al . , 2008; Egger et al . , 2011 ) . Recent work has identified signaling pathways and niches required for stem cell proliferation ( Egger et al . , 2010; Ngo et al . , 2010; Reddy et al . , 2010; Orihara-Ono et al . , 2011; Wang and Rudnicki , 2011 ) . Drosophila flight muscles are formed from adult muscle precursors ( AMPs ) ( Currie and Bate , 1991; Fernandes et al . , 1991; Roy and VijayRaghavan , 1999 ) . Myogenesis occurs in two phases; an embryonic one , which makes the muscles required for the larval life ( Bate et al . , 1991 ) while a postembryonic phase leads to formation of muscle required for the adult ( Fernandes et al . , 1991; Roy and VijayRaghavan , 1998; Sudarsan et al . , 2001 ) . The AMPs , lineal derivatives of the mesoderm , are generated embryonically and proliferate postembryonically ( Bate et al . , 1991; Fernandes et al . , 1991; Roy and VijayRaghavan , 1999 ) . Little is known about the cellular and molecular mechanisms by which the AMPs proliferate and to give rise to the large number of cells which are needed to contribute to the massive adult flight muscles . During late embryogenesis the AMPs required for the formation of flight muscles are set aside in the mesothoracic segment ( T2 ) and those required for haltere muscle development in the metathoracic segment ( T3 ) ( Sudarsan et al . , 2001; Roy et al . , 1997 ) . The numbers of AMPs at this early stage in T2 and T3 are same but the AMPs in T2 proliferate profusely while those in T3 far less . Studies on the ‘four-winged-fly’ have clearly shown the key role played by the wing-disc ectoderm in regulating myoblast proliferation ( Fernandes et al . , 1994; Dutta et al . , 2004; Roy and VijayRaghavan 1997 ) . Yet , the mechanisms that regulate the amplification of muscle precursors to generate large ‘pools of myoblasts’ , a feature common to adult muscles in the fly as well as to vertebrate skeletal muscles , ( Sudarsan et al . , 2001 ) have not been studied in the fly or indeed other systems . In this report , we use clonal MARCM ( Yu et al . , 2009 ) techniques to study the proliferative activity of AMPs during postembryonic development . We focus on the AMPs associated with the wing imaginal disc in the second thoracic segment , which give rise to the large indirect flight muscles . We show that an initial amplification of the number of these AMPs occur through symmetric divisions and is followed by a switch to asymmetric divisions , in which the AMPs self-renew and generate postmitotic myoblasts required for the formation of adult myofibers . The sequential nature of these two division modes results in a change in the arrangement of AMP lineages from an initially monostratified layer adjacent to the wing disc epithelium to a markedly multistratified layer comprising both AMPs and their post mitotic myoblast progeny . While the initial amplification of AMPs through symmetric divisions is controlled by Notch signaling , the switch to the subsequent asymmetric division mode of AMP division additionally requires Wingless . In both cases the epidermal tissue of the wing imaginal disc acts as a stem cell niche and provides the ligands , Serrate and Wingless , for the two signaling pathways that operate in the AMPs . We identify the AMPs as a novel muscle stem cell population whose proliferation pattern orchestrates the building of the large flight muscles in Drosophila .
AMPs are lineal descendants of progenitor cells , derived from proneural gene-expressing equivalent groups in the embryonic mesoderm ( Ruiz-Gómez and Bate , 1997; Carmena et al . , 1998 ) . In the thoracic segments of the embryo , each of these progenitor cells divides asymmetrically to generate a muscle founder cell involved in embryonic myogenesis and an AMP which has active Notch ( N ) signaling and continues to express the mesodermal marker Twist ( Twi ) . In contrast to their muscle founder cell siblings , the AMPs do not differentiate in the embryo; rather they are set aside and at the end of embryonic development become associated with the imaginal discs ( Bate et al . , 1991 ) . The Twi-expressing AMPs associated with the wing imaginal discs co-express Vestigial ( Vg ) in late embryonic stages ( Sudarsan et al . , 2001 ) . In the early embryonic stage ( Stage11 ) Wg induced Dll ( Distalless ) expression leads to separation of the primordium for different discs and those for wing imaginal discs start expressing Vg and results in defining of the population for wing disc primordium ( Cohen et al . , 1993; Campos-Ortega , 1997; Kubota et al . , 2000 ) . To determine the number of AMPs present in the thoracic segments at the end of embryogenesis , we identified these cells using Twi-GAL4 > UAS-mCD8::GFP . In each thoracic segment a total of approximately 10 cells were Twi-positive and dorsally located indicating that these correspond to the dorsal AMPs ( Figure 1A ) ( Williams et al . , 1991; Creig and Akam et al . , 1993 ) . To further characterize the extent of post-embryonic proliferation of the flight-muscle AMPs , we determined the number of Twi-positive cells associated with the T2 wing disc at different larval stages . Immediately after larval hatching ( 24 hr AEL ) a total of 10 ( ±2 ) Twi-positive cells were found on the wing disc implying that these had not yet initiated proliferative divisions ( Figure 1B ) . At the late second instar stage ( 72 hr AEL ) the number of Twi-positive cells associated with the wing disc had increased to approximately 250 ( ±15 ) and at the late third instar stage ( 144 hr AEL ) approximately 2500 ( ±90 ) Twi-positive cells were found on the wing disc indicative of intense proliferative activity ( Figure 1C–E ) . We conclude that the 10 AMPs present on the wing disc at the end of embryogenesis and the beginning of larval development initiate a proliferation process during subsequent larval stages which results in some 2500 Twi-positive cells at the end of larval development ( Figure 1E , F , G ) . These findings suggest that each embryonically born AMP , on the average , would give rise to a lineage of approximately 250 cells during larval development . We next investigated the cellular mechanisms that make this remarkable proliferation possible . 10 . 7554/eLife . 03126 . 003Figure 1 . Wing disc associated AMPs proliferate during larval development to reach a population size of 2500 . ( A ) Stage 17 embryo ( ∼22 hr After Egg Laying ( AEL ) showing cluster of 10 AMPs in thoracic segment ( T2 ) marked with Twist Gal4 > UAS mCD8GFP , Vg ( anti-Vestigial , red ) and TO-PRO3 ( A nuclear stain , blue ) , Similar numbers of Twi positive cells are seen in each segment . n = 5 Scale bar , 10 μm . ( B–E ) Wing imaginal discs from early first ( ∼24 hr AEL ) n = 5 . Scale bar , 10 μm , late second instar ( ∼72 hr AEL ) n = 10 and third instar stage ( ∼120 hr AEL , n = 10 and ∼144 hr AEL , n = 10 ) stained for Twi ( anti-Twist , green ) and TO-PRO3 ( A nuclear stain ) showing increase in the number of AMPs during the larval instars . Scale bar , 50 μm . ( F ) Schematic showing AMPs , marked in green color , in T2 region of stage 17 embryo and subsequently in the presumptive notum of the first instar , second instar and late third instar wing imaginal disc . ( G ) A sharp increase is seen in the number of AMPs in first ( I ) and second ( II ) instars ( Till 72 hr AEL ) ( Depicted as red line ) . After 72 hr AEL ( Early third instar ) till the end of third instar ( 144 hr AEL ) , the rate of increase of the AMP population is less sharp . The dotted blue line depicts the extrapolation of the early rate of growth . The graph depicts the average number of cells and the bar represents the standard error . For first instar ( 24 hr ) n = 5 , late second instar ( 72 hr ) n = 10 , mid third instar ( 120 hr ) n = 10 and late third instar ( 144 hr ) n = 10 . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 003 To determine the mechanisms of AMP proliferation we used twin-spot MARCM , a genetic labeling technique which makes it possible to trace cell lineage and determine proliferation patterns by independently labeling the two paired sister cell siblings derived from a given cell division ( Yu et al . , 2009; O'Brien et al . , 2011 ) . In twin-spot MARCM , heat shock induces a timed mitotic recombination event in the precursor cell which results in differential labeling of its two daughter cell clones with heritable expression of green fluorescent protein ( GFP ) in one cell and red fluorescent protein ( RFP ) in the other . Symmetric cell divisions will result in daughter cell clones of RFP and GFP of equal number of cells; asymmetric clonal amplification should result in daughter cell clones of GFP and RFP of unequal size . Twin-spot MARCM clones were induced in proliferating AMP lineages during larval development and recovered in the third instar . Clones were induced using the mesoderm-specific Dmef2-Gal4 driver ( Ranganayakulu et al . , 1998 ) . Wing discs containing labeled clones were co-labeled for Twi-immunoreactivity . Twin-spot MARCM clones induced between the early first instar and the late second instar invariably resulted in two relatively large , differentially labeled daughter cell clones that were similar in size and contained a similar number of cells ( Figure 2A–C ) . This indicates that mitotically active cells in the AMP lineages divide symmetrically during early larval development ( 30 hr–72 hr AEL ) . In contrast , twin-spot MARCM clones induced during larval development from the early third instar ( ∼75 hr AEL ) onward always resulted in two differentially labeled daughter cell clones of different number of cells in each clone ( Figure 2D–F ) . Notably , one of the two daughter cell clones invariably comprised one large labeled cell with elongated morphology , whereas the other daughter cell clone contained up to 10 ( ±3 ) smaller cells . This indicates that mitotically active cells in the AMP lineages have differential clonal potential and divide during the third larval instar stage ( 72 hr–144 hr AEL ) , with each stem cell giving rise to one stem cell through self-renewal as well as a sibling cell which could differentiate further . 10 . 7554/eLife . 03126 . 004Figure 2 . AMP proliferation involves initial symmetric and subsequent asymmetric clonal amplification . ( A ) Third instar wing imaginal disc ( ∼144 hr AEL ) showing a Twin-spot MARCM clone induced by a single15 m heat shock at 37°C in the early first instar ( ∼24 hr AEL ) . The GFP and RFP twin spots are of the same size . Anti-Twist marks all the descendants of AMP lineages . ( B ) Schematic depiction of clonal generation of symmetric clone from the AMP from the early first instar till late second instar stage . N = 10 early clones were examined all with a similar result showing the twin spots of the same size . ( C ) Quantitation of the clone showing exactly same number ( 136 ) of cells in both red and green clone of twin-spot . ( D ) Asymmetric clone of AMP lineage recovered from a single15 m heat shock at 37°C clonal induction in early third instar ( ∼75 hr AEL ) . Clone shows a single large red cell and six small green cells ( zoomed image showed in top most right corner ) . Scale bar 50 μm . ( E ) Schematic depiction of clonal generation of asymmetric clone from the AMP in the early third instar onwards . N = 12 late clones were examined all with a similar result . ( F ) Quantitation of clone showing one red and six green cells from the twin-spot marking experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 004 Taken together , these findings indicate that proliferating cells in the AMP lineage divide in a symmetrical mode during early development such that 10 AMP progenitors in the early first larval instar stage give rise to a total of approximately 250 cells at the end of the second larval instar stage . In the third larval instar stage , these myogenic progenitor cells divide to produces up to 10 progeny to generate a total of approximately 2500 myogenic cells . This division is different from the earlier mode , as the division outcome is asymmetric as the majority of the clone is essentially due to proliferation of the one of the cell . Thus , by transiting through a sequence of symmetric cell divisions followed by asymmetric cell divisions , each embryonically generated AMP can generate a lineage of approximately 250 myogenic progeny during larval development . During the first and second larval instars , AMPs labeled with the Twi are located in a single monostratified layer adjacent to , and in contact with , the wing disc epithelium ( Figure 3A , B , B′ ) . From the early third larval instar onward , labeled cells become organized in multistratified 2-3 cell layers on the wing disc ( Figure 3C , C′; Video 1 ) . To determine if mitotic activity takes place in specific layers , cells were co-labeled for phospo-histone-3 ( PH3 ) immunoreactivity and Dmef2-Gal4 was used to drive UAS mCD8::GFP . Throughout larval development , PH3 positive cells were only observed in the cell layer immediately adjacent to the wing disc epithelium . Notably , even in the multistratified organization manifested in the third larval instar , PH3 positive cells were seen only in cells located adjacent to the epithelium and not in cells located in more distal layers ( Figure 3D–D″ ) . This implies that almost all mitotically active cells are located proximal to the disc epithelium , while the cells in the more distal layers are likely to be postmitotic . Taken together with the findings on cell division modes in the AMP lineages , these results also suggest that both symmetric and asymmetric cell divisions only take place in cells located adjacent to the wing disc epithelium . 10 . 7554/eLife . 03126 . 005Figure 3 . Proliferating AMPs are located in a monolayer adjacent to the wing disc epithelium . ( A–B′ ) Optical section of late second instar wing imaginal disc ( ∼72 hr AEL , schematic in A with AMPs in green ) , showing monostratified ( single layer ) arrangement of AMPs ( marked in green and denoted as M for mesoderm , B ) stained for Twi ( anti-Twist , green ) , Dcad2 ( anti-Dcad2 , red ) ( Disc-epithelium [E] ) and TO-PRO3 ( a nuclear stain , blue ) . n = 10 , Schematic in B′ . ( C–C′ ) Late third instar wing disc showing multistratified ( 3–4 layers ) of AMPs in presumptive notum region . n = 20 Scale bar 50 μm . AMPs are stained for Twi ( anti-Twi , green ) , epithelium ( anti-Dcad2 , red ) and TOPRO ( Nuclear stain , blue ) ( D–D″ ) A late third instar wing imaginal disc showing multistratified arrangement of AMP using Dmef2-Gal4 > UASmCD8GFP ( anti-GFP , green ) and co-stained for mitotic marker PH-3 ( anti-phosphohistone , red ) and TO-PRO3 ( Nuclear stain , blue ) . C′ shows a schematic representation . D′ shows an optical section of image D showing active mitotic divisions ( red dotted circles ) exclusively in the layer most-proximal to the disc epithelium ( marked as E ) . n = 20 . D″ represents this schematically . ( E , F and G ) Late second instar wing discs ( ∼70 hr AEL ) marked for centrosomin-GFP ( CNN-GFP , a pericentrosomal marker , green ) using Dmef2-Gal4 , PH3 ( anti-phosphohistone , red ) and TO-PRO3 ( Nuclear stain , blue ) . In each panel , E is the Epidermis and M the AMPs . ( H–H′ ) Schematic ( H ) of a late second instar disc ( as shown in F ) showing active mitotic division ( Black arrow ) parallel to the disc epithelium . Pie-chart representation ( H′ ) ( Blue: 10°–20° with 0° being parallel to the epidermis . n = 6 preps . ( I , J and K ) Third instar wing imaginal disc ( ∼140 hr AEL ) showing a more orthogonal orientation in AMP marked using centrosomin-GFP ( CNN-GFP ) ( a centrosome marker , green ) using Dmef2-Gal4 , PH3 ( anti-phosphohistone , red ) and TO-PRO3 ( Nuclear stain , blue ) . n = 10 . Scale bar , 10 μm . ( L–L′ ) A schematic representation ( L ) and pie-chart representation ( L′ ) of late third instar disc showing mitotic division axis ( black arrow , in L ) of AMP ( green ) oblique to disc surface ( blue ) . The angle of division at this stage is in the range of 50–80° as represented in L′ . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 00510 . 7554/eLife . 03126 . 006Video 1 . Showing multilayered arrangement of myoblasts ( related to Figure 3 ) . 3D reconstruction of third instar wing disc showing AMP lineage stained with anti-Twi ( green ) and all disc nuclei stained using TOPRO ( blue ) showing stacked arrangement . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 006 Since twin-spot MARCM experiments indicate that a transition from a symmetrical to an asymmetrical mode of cell division occurs between the second and third larval instar stage , we focused on the mitotically active cells in the layer adjacent to the wing disc epithelium and examined the orientation of the axis of division at these two stages . Centrosome position and , hence , the axis of cell division were revealed by a GFP-tagged peri-centrosomal ( CNN-GFP ) label ( Megraw et al . , 2002 ) and Dmef2-Gal4 was used to drive UAS-CNN-GFP expression . Throughout the second instar stage , the axis of cell division in the mitotically active cells is oriented parallel to the surface of the wing disc epithelium such that both daughter cells remain in contact with the epithelium ( Figure 3E–H′ ) . In contrast , from early third instar stage onward , the axis of cell division shows an oblique to orthogonal orientation relative to the disc surface ( Figure 3I–L′ ) . Consequently , only one of the two daughter cells remains in contact with the disc epithelium while the other daughter cell loses epithelial contact and , hence , might contribute to the population of cells located in layers distal to the disc epithelium . Given that proliferative mitotic activity in the AMP lineages only takes place in the cells located next to the epithelium we performed experiments to confirm that the cells in the more distal layers are indeed their postmitotic progeny . For this , we first performed EdU ( 5-Ethynyl-2′-deoxyuridine ) incorporation experiments to visualize cells actively engaged in DNA replication as well as their progeny . EdU incorporation was carried out in early third instar larva ( in a 5 hr pulse ) and visualized in AMP cells on wing discs co-labeled with for Twi immunoreactivity . When cells were examined directly after EdU incorporation ( pulse , no chase ) , labeling was only found in cells adjacent to the disc epithelium ( Figure 4A , B ) . This result confirms the finding that mitotically active cells are located next to the epithelium . If the incorporation of EdU was followed by 48 hr without EdU ( pulse , chase ) and the cells then examined , labeling was seen in all layers . However , in this case , the majority of labeled cells were in the most distal layer , less labeled cells were in an intermediate layer and the lowest number of labeled cells was found in the layer adjacent to the disc ( Figure 4C , D ) . This indicates that the pulse of EdU labeling incorporated into the ( replicating ) cells adjacent to the disc epithelium was now manifest in the ( non-replicating ) cells located in more distal layers . This , in turn , implies that the cells in the more distal layers are the postmitotic progeny of the proliferating cells that remain in contact with the disc epidermis . 10 . 7554/eLife . 03126 . 007Figure 4 . Proliferating AMPs generate clones of post mitotic myoblasts localized in distal layers . ( A ) Optical section of third instar wing disc dissected after 5 hr Edu , a thymidine analogue ( 5-ethynyl-2′-deoxyuridine ) pulse and stained for Edu ( red ) , Twi ( anti-Twist , green ) and Dcad2 ( anti-Dcad2 , blue ) revealing presence of Edu in some AMP lineage . n = 10 . Scale bar , 50 μm . ( B ) Sagittal section of presumptive notum region of disc ( shown in A ) showing Edu labeling of AMP lineage only next to disc epithelium surface . ( C ) Third instar disc dissected after 5 hr Edu pulse and 48 hr chase in pulse free media and stained for Edu ( red ) , Twi ( anti-Twist , green ) and Dcad2 ( anti-Dcad2 , blue ) showing labeling in the most distal layer of AMP lineages . n = 12 . ( D ) Sagittal section of C showing labeling in maximum labeling in distal layers and minimum in the layer next to epithelium . ( E ) A MARCM clone generated in second instar stained for GFP ( anti-GFP , green ) and PH3 ( anti-Phosphohistone , red ) , TO-PRO3 ( blue ) . The clone shows a single PH3 positive cell ( shown in red bracket ) in the cluster of other clonal progenies . n = 10 . ( F–F’ ) Optical section ( F ) and schematic ( F’ ) of figure E , showing the presence of PH3 positive AMP ( red bracket ) next to the disc epithelium ( dotted line ) . ( G–I ) The MARCM clone showing clonal progenies marked with GFP in the proximal most ( G ) , middle ( H ) and distal most ( I ) layers with reference to the disc epithelium . The graph ( N ) shows increase in the cell size from G to J . n = 12 Scale bar 50 μm . ( K–N ) The optical sections of late third instar disc stained for Twi ( anti-Twist , green ) and TO-PRO3 ( blue ) and the quantitation ( M ) showing trend of increase nuclear size in the layers distal most in reference to epithelium . n = 10 . Scale bar 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 007 To further investigate the clonal nature of this proliferative process in the AMP lineages , we next carried out MARCM labeling experiments using the Dmef2-Gal4 driver . MARCM clones induced in the second larval stage and recovered in the third larval stage typically comprised 100–200 labeled cells arranged in a loose cluster and extending from the layer next to the disc epithelium to the most distal layer of cells ( Figure 4E ) . Interestingly , within each clone the labeled cells in the distal layer were larger in size and had a more elongated morphology than those located in the innermost layer next to the disc ( Figure 4G–N; Video 2 ) . To determine if these clones comprised mitotically active cells , we combined MARCM labeling with PH3 immunolabeling . In all cases in which co-labeling was seen , only one cell within the clone was PH3-positive and this cell was always located adjacent to the disc epithelium ( Figure 4F , F′; Video 3 ) . 10 . 7554/eLife . 03126 . 008Video 2 . MARCM clone showing variation of cell size ( related to Figure 4 ) . 3D reconstruction of AMP lineage marked by membrane tethered GFP ( mCD8::GFP ) revealing size differences in the myoblasts at different distances from disc epithelium . All nuclei marked by TOPRO ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 00810 . 7554/eLife . 03126 . 009Video 3 . MARCM clone showing PH3 association with respect to layers ( related to Figure 4 ) . Single cell in a clone ( anti-GFP , green ) showing active mitotic division ( anti-Phospho histone 3 , red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 009 Taken together , these findings indicate that proliferation in the AMP lineages occurs in two phases during larval development . In early larval development ( first and second instar ) , AMPs undergo a marked amplification in cell number through symmetric divisions . These divisions are oriented parallel to the wing disc epithelium and hence the AMPs form a monolayer of proliferating cells that remain in contact with the disc epithelium . In late larval development ( third instar ) AMPs switch to an asymmetric mode of proliferative cell division in which they self-renew and generate postmitotic daughter cells . These asymmetric divisions are oriented obliquely to the wing disc surface such that the self-renewing daughter cells remain in contact with the disc epithelium while the postmitotic daughters become located in the more distal layer of non-replicating myoblasts . Previous work has shown a requirement for Notch ( N ) in adult flight muscle development; in N mutants a depletion of wing disc myoblasts is seen ( Anant et al . , 1998 ) . This suggests that N signaling might be involved in the proliferative activity of AMP lineages . To investigate this further , we first determined if N is expressed in the ( Twi-positive ) cells on the larval wing disc using an anti-NICD ( Notch Intracellular Domain ) antibody . Immunocytochemical analysis shows that all of the Twi-positive cells are co-immunolabeled by the anti-NICD antibody ( Figure 5A–C , F–H ) . This co-immunolabeling was seen in all larval instar stages and in all myogenic cell layers associated with the wing disc . 10 . 7554/eLife . 03126 . 010Figure 5 . Notch signaling is required for proliferative activity of AMP lineages . ( A–C ) Late third instar disc stained for Notch ( anti-Notch intracellular C-terminal domain , [NICD , red] ) , Twi ( anti-Twist , green ) and TO-PRO3 ( blue ) . In this figure , E denotes disc-epithelium and M is for mesoderm ( green ) . ( D–E ) Quantification of number of PH3 positive AMPs in the Notch donwregulation using Dmef2-Gal4 , TubGal80ts > UAS Notch RNAi ( D ) and Notch up regulation using Dmef2-Gal4 , TubGal80ts > UAS NICD background ( E ) . For both experiments Gal80 repression was relieved from early second instar till late third instar by shifting from 18°C to 29°C . All graphs are Mean ± Standard Error ( Student's t test ) . n = 12 . p-values < 0 . 001 . ( F–N ) Optical section of wing discs stained for Notch ( anti-Notch intracellular C-terminal domain , ( NICD , red ) , Twi ( anti-Twist , green ) and TO-PRO3 ( blue ) in control ( F–H ) ( Dmef2-Gal4 , TubGal80ts > Canton-S ) , in Dmef2-Gal4 , TubGal80ts > UAS Notch RNAi ( I–K ) , Dmef2-Gal4 , TubGal80ts > UAS NICD ( L–N ) . The multistratified layered arrangement of AMP lineages is lost in Notch down regulation ( J ) while in Notch up regulation ( M ) it increases with respect to number of layers , in comparison to control ( G ) . Gal80 repression was relieved from early second instar till late third instar by shifting from 18°C to 29°C . n = 20 . Scale bar , 50 μm . E denotes disc-epithelium and M is for mesoderm ( marked by anti-Twist , green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 010 To determine the role of N in these Twi-positive cells , we used the Dmef2-Gal4 driver together with UAS-N RNAi to down-regulate N in the AMPs and then assayed mitotic activity using PH3 immunoreactivity in late third instar stage . ( Gal80ts was used to limit N-RNAi to the second and third larval instar to avoid lethality . ) A significant reduction in the number of mitotically active cells was observed; in the third instar stage only half the number of PH3-positive cells were seen in knockdown vs control experiments ( Figure 5D ) . Similar findings were obtained in experiments in which a dominant negative form of N was expressed using the Dmef2-Gal4 driver ( data not shown ) . Both the number and the layered organization of the ensemble of Twi-positive cells on the wing disc were reduced in these knockdown experiments ( Figure 5I–K ) . Conversely , overexpression of an activated form of N in the AMP using Dmef2-Gal 4 to drive UAS-NICD in second and third larval instar stages revealed a marked increase in mitotically active cells as assayed in late third instar stage . In these experiments the number of PH3-positive cells in the overexpression experiments was approximately twice as large as in controls ( Figure 5E ) . Correspondingly , both the number and the layered organization of the Twi-positive cells on the disc were increased in these overexpression experiments ( Figure 5L–N ) . Interestingly , in contrast to wild type controls , mitotically active cells were occasionally seen in more distal cell layers in N overexpression experiments ( data not shown ) . These findings indicate that proliferative activity in the AMP lineages during larval development is dependent on N signaling . Since mitotic activity in the wild type is only seen in the layer of cells adjacent to the wing disc this further implies that N signaling acts to effect proliferation in the AMPs in this layer . The N ligand Serrate ( Ser ) has been shown to be present in the dorsal ( presumptive notum ) region of the wing disc in which the AMP lineages develop ( Speicher et al . , 1994 ) . This suggests that Ser might be the ligand for the N signaling required in AMP proliferation . To investigate this , we first determined if Ser expression in the disc epithelium occurs in close proximity to AMPs in the adjacent myogenic cell layer . Immunocytochemical labeling of the wing disc epithelium ( with anti-DCad2 ) and of the AMP lineages ( with anti-Twi ) together with a Ser-specific marker ( Ser-LacZ ) reveals a close apposition of Ser-expressing wing disc cells and the AMP cell layer immediately adjacent to the disc epithelium ( Figure 6A–H′; Video 4 ) . This close apposition was observed in all larval stages . 10 . 7554/eLife . 03126 . 011Figure 6 . Serrate located in the wing disc epithelium is necessary for AMP proliferation . ( A ) Serrate-lacZ ( anti-beta Gal ) , a reporter for Serrate expression , visualized in disc epithelium of the late third instar . ( B ) Saggital section shows Serrate being expressed specifically in the disc epithelium . ( C–D ) Dcad2 expression ( anti-Dcad2 ) marking the disc epithelium . ( E–F ) AMP lineages viewed by using Twi staining ( anti-Twist ) . ( G–H ) The merge shows the expression of Serrate ( red ) exclusively in the disc epithelium as Dcad2 ( blue ) and in close proximity of first layer AMP lineages . E-wing disc epithelium , M-AMP lineages . Scale bar 50 μm . H′—Schematic depiction of expression patterns of Serrate , Dcad2 and Twist showing only one layer of AMP out of 3–4 ( As shown in F ) is in direct contact with Serrate producing epithelium surface . ( I ) Quantification of number of PH3 positive AMPs in disc epithelium specific Serrate knock down using Ap Gal4 driving UAS Serrate RNAi . Gal80 repression was relieved from early second instar till late third instar by shifting from 18°C to 29°C . All graphs are Mean ± Standard Error ( Student's t test ) . n = 15 . p-values < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 01110 . 7554/eLife . 03126 . 012Video 4 . Serrate staining in epidermis and close association with myoblasts ( related to Figure 6 ) . Serrate- lacZ ( anti-beta-gal , red ) showing serrate expression in wing disc epithelium ( anti-Dcad 2 , blue ) and only one layer of myoblasts are in contact with epithelium due to multilayered arrangement . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 012 The proximity of Ser in the wing disc epithelium to mitotically active AMPs suggests that epithelial Ser could be the ligand that activates N signaling in the AMPs . To investigate if wing disc-derived Ser is required for N-mediated proliferative activity of AMPs , we conditionally knocked down Ser specifically in the dorsal wing disc region of second instar stages with Ap-Gal4 driving UAS-Ser RNAi using the TARGET system ( McGuire et al . , 2003; Martín and Morata , 2006 ) and assayed mitotic activity based on PH3 immunolabeling in the late third instar stage ( The Ap-Gal4 Gal4 expression in other regions in the larva has no direct effect on these results when perturbation were performed during mentioned time scale ) . A marked reduction in the number of mitotically active cells was observed; in the third instar stage less than half the number of PH3-positive cells were seen in knockdown vs control experiments ( Figure 6I ) . These results indicate that the Ser located in the wing disc epithelium is required for the mitotic activity of cells in the adjacent AMP layer . Given the concomitant requirement of N for mitotic activity of these cells , this implies that epithelial Ser serves as a ligand for the activation of N receptor-mediated signaling in the myogenic AMP lineages . This in turn suggests that the Ser-expressing cells in the wing disc epithelium represents a transient signaling niche for N activation of AMP proliferation . While the activation of N signaling by Ser is required for mitotic activity in AMP lineages , it is unlikely to mediate the switch from symmetric to asymmetric divisions that occurs during late larval development , since expression of N and Ser in these lineages occurs throughout larval development . Numb ( Nb ) is an intracellular protein , which binds to N and causes degradation of this receptor ultimately leading to down-regulation of N signaling ( Frise et al . , 1996; Couturier et al . , 2012 ) . Previous work on mitotic activity of muscle progenitors during embryonic development has shown that Nb acts in asymmetric cell division by down regulating N activity in one of the progenitor’s two daughter cells; in this cell downregulation of N activity results in specification of muscle cell fate , while continued N activity in the other daughter cell maintains progenitor cell fate ( Ruiz-Gómez et al . , 1997; Carmena et al . , 1998 ) . To determine if Nb is involved in asymmetric cell divisions of AMPs we first carried out immunolabeling studies of Nb expression during different larval instar stages . These experiments show that Nb is expressed in AMP lineages during the third instar stage but not in the two preceding larval stages ( data not shown ) . Throughout the third instar stage , Nb is expressed in all of the myogenic cells on the wing disc , however expression is strongest in the cells of the more distal layers and relatively weak in the cells adjacent to the disc epithelium; this contrasts with N expression which is more uniformly distributed in all of these cell layers ( Figure 7A–D ) . The fact that Nb is not expressed in AMP lineages during their symmetric division phase , but is present in these cells during their asymmetric division phase suggests that Nb might be required for the transition to asymmetric cell divisions in AMP lineages . 10 . 7554/eLife . 03126 . 013Figure 7 . Numb expressed in the third instar stage is required for asymmetric cell divisions in AMP lineages . ( A–D ) The late third instar disc stained for expression of Numb ( anti-Numb , red ) , Notch ( anti-NICD , white ) and Dmef2-Gal4 > UAS mCD8GFP ( anti-GFP , green ) . N = 6 . The expression of Numb can be seen in patches ( A ) in contrast to Notch ( B ) which stains most of the AMP lineages marked by using Dmef2-Gal4 > UAS mCD8GFP ( C ) . The merge ( D ) shows the expression of Notch ( white ) in myoblasts lineage ( green ) along with Numb ( red ) . ( E ) Quantification of number of PH3 positive AMPs in Numb knock down ( Dmef2-Gal4 , TubGal80ts > UAS Numb RNAi ) showing significant increase in the total proliferating AMPs . Gal80 repression was relieved from early third instar till late third instar by shifting from 18°C to 29°C . All graphs are Mean ± Standard Error ( Student's t test ) . n = 10 . p-values < 0 . 001 . ( F–K ) Twinspot MARCM in Numb RNAi ( F–H ) and in Notch upregulation ( I–K ) backgrounds , generated in early third instar , showing loss of asymmetry or symmetric clones in contrast to wild type asymmetric clone as shown in Figure 2D–F . n = 5 . Scale bar , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 013 To investigate this , we carried out twin-spot MARCM labeling experiments and used Dmef2-Gal4 driving UAS-Nb RNAi to knockdown Nb in the proliferating AMP lineages . Clones were induced in the early third instar stage and recovered in the late third instar stage . These twin-spot MARCM clones invariably consisted of two relatively large , differentially ( RFP and GFP ) labeled daughter cell clones that were similar in size and cell number indicative of symmetric divisions ( Figure 7F–H ) . These clones were similar to the corresponding twin-spot clones obtained in wild type first and second instar stages but strikingly different from those obtained in the wild type third instar ( see Figure 2 ) . This indicates that in the absence of Nb , cell divisions in AMP lineages are symmetric and not asymmetric in the third instar stage , implying that Nb expression is necessary for the transition from the symmetric to the asymmetric division mode manifest in the third instar . Given that Nb inhibits N signaling , upregulation of N should have similar phenotypic effects as downregulation of Nb . To investigate this , we repeated the twin-spot MARCM labeling experiments but used Dmef2-Gal4 driving UAS-NICD to constitutively activate N in the proliferating AMP lineages . Clones induced in the early third instar stage and recovered in the late third instar stage again consisted of two relatively large , differentially labeled daughter cell clones that were similar in size and cell number indicating a symmetric division mode ( Figure 7I–K ) . Since both downregulation of Nb and constitutive upregulation of N lead to continued symmetric cell divisions in the third instar stage , we postulate that Nb , through downregulation of N , is required for asymmetric proliferative divisions in the AMP lineages . While above mentioned experiments indicate that Nb expression in AMP lineages is important for the transition from symmetric to asymmetric proliferation divisions , they do not identify the signal that controls the onset and maintenance of Nb expression in the third instar stage . Wingless ( Wg ) is a signaling protein that can act to control developmental patterning and growth ( Sharma and Chopra , 1976; Zecca et al . , 1996; van Amerogen and Nusse , 2009 ) . During larval development of the wing disc , Wg is expressed in the prospective wing blade region in all instar stages , however , expression in the prospective notum region , in which the developing AMP lineages reside , is only manifest from the early third instar onwards ( Phillips and Whittle , 1993; Tomoyasu et al . , 2000 ) . Previous work on postembryonic myogenesis has shown that Wg signaling from this region of the larval wing disc epithelium is involved in elaboration and maintenance of myoblast diversity ( Sudarsan et al . , 2001 ) . To investigate if Wg signaling from the notum region of the disc epithelium might be involved in the induction of Nb expression in third larval instar AMP lineages , we first characterized the spatial expression pattern of Wg in the epithelium relative to the Twi-positive myogenic cells in immunolabeling studies of third larval instar stages . Throughout the third instar stage , notal Wg expression is seen in a stripe of epithelial cells that are closely apposed to the adjacent population of Twi positive cells ( Figure 8A–H; Video 5 ) . To determine if Wg signaling from this spatially restricted stripe of epithelial cells can influence all of the Twi-positive cells in the myogenic layers , we used immunolabeling to visualize beta-catenin activity in wild type vs notal Wg loss-of-function using Wg ( ts ) /Wg ( Sp-1 ) alleles . In these experiments , notal Wg loss-of-function resulted in complete loss of beta-catenin activity in all Twi-positive cells implying that inductive Wg signaling from the notum region of the wing disc affects all myogenic cells in the third larval instar ( Figure 8I–N ) . We next carried out similar Wg loss-of-function experiments together with Nb immunolabeling to investigate if inductive Wg signaling is required for onset and maintenance of Nb expression in the third instar myogenic cells . Wg loss-of-function in the wing disc notum resulted in complete loss of Nb expression in all Twi-positive cells ( Figure 8O–V ) . We conclude that Wg signaling in the wing disc notum is required for induction and maintenance of Nb expression in the myogenic AMP lineages in the third larval instar stage . 10 . 7554/eLife . 03126 . 014Figure 8 . Wingless signaling from the wing disc epithelium induces Numb expression in third instar AMP lineages . ( A–H ) Third instar wing disc stained for Dcad2 ( anti-Dcad2 , Blue ) , Wg ( anti-Wg , red ) and Twi ( anti-Twist , green ) demonstrating a prominent longitudinal stripe of Wg expression in disc epithelium . n = 20 . ( H′ ) Schematic of the merge ( H ) depicting disc epithelium , as a source of Wg production and dispersal subsequently leading to Wg signaling ( red arrows ) activated in all AMPs . ( I–N ) AMP lineages ( anti-Twi , green ) stained for Beta-catenin ( down stream molecule of Wg pathway ) ( anti-Beta catenin , red ) in Canton-S ( I–K ) and in Wg ( ts ) /Wg ( Sp-1 ) alleles ( loss of function alleles of Wg gene ) ( L–N ) . Loss of Beta-catenin in Wg ( ts ) /Wg ( Sp-1 ) shows absence of Wg activation in AMPs which also leads to decrease in number of AMP lineages . Presence of Beta-catenin in all AMPs clearly points towards Wg action at a distance far from the disc epithelium , the source of Wg . n = 10 . ( O–R ) The optical section of third instar wing disc showing Numb expression ( anti-Numb , red ) very prominent in the distal layer of AMPs , marked by Notch ( anti-Notch , green ) ( also in Figure 5A–C ) . ( S–V ) Wg loss of function ( Wg ( ts ) /Wg ( Sp-1 ) ) results in total disappearance of Numb in AMPs . n = 10 . Scale bar , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 01410 . 7554/eLife . 03126 . 015Video 5 . Wg staining in Canton-S ( related to Figure 8 ) . Late third instar wing disc showing Wg ( anti-Wg , red ) expression in disc epithelium ( anti-Dcad2 , blue ) . The myoblasts forms a close association with Wg expression . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 015 In addition to its effects on Nb expression , notal Wg loss-of-function also results in the reduction of mitotic activity in the AMP lineages . Thus , both the number of PH3 positive cells and the extent of the multilayered organization are reduced in Wg downregulation compared to wild type ( Figure 9A–J , Figure 9—figure supplement 1 ) . To understand the possible direct link of the induction of Wg signaling and change in the orientation of the cell division , we down regulated Wg using DN-TCF with Dmef2-Gal4 and probed for division axis using CNN-GFP . The axis of the division in this background is parallel to the epithelium , similar to that seen in the early instars ( Figure 9—figure supplement 2 ) . Wg , in the presumptive notum is expressed only from the late second instar ( Alexandre et al . , 2014 ) , therefore this result shows that down-regulation of Wg prevents the switch in cell-division orientation from parallel to the epithelium to perpendicular to the epithelium . This suggests that Wg signaling from the wing disc niche might have dual effects on neighboring myogenic cells , namely the induction of Nb expression and the regulation of proliferation . 10 . 7554/eLife . 03126 . 016Figure 9 . Loss of Wg results in reduction of mitotic activity and layered arrangement in AMP lineage . ( A–H ) Third instar wing disc showing loss of multilayered arrangement in Wg loss of function back ground ( Wg ( ts ) /Wg ( Sp-1 ) ) . In control disc ( Canton-S ) the multistratified arrangement can be distinctly seen which clearly disappears in Wg loss of function background . n = 8 . ( I ) Quantification of number of PH3 positive AMPs in epithelium specific Wg knockdown using ApGal4 > UAS Wg RNAi showing significant decreases in comparison to control ( Canton-s ) . All graphs are Mean ± Standard Error ( Student's t test ) . p-value < 0 . 001 , n = 10 . ( J ) AMP specific perturbations ( using Dmef2-Gal4 ) of Wg pathway downstream molecules ( TCF ) showing changes in mitotic activity in comparison to control . The activation of Wg pathway by overexpressing activated TCF ( UAS TCF 24 ) leads to significant increase in mitotic activity . Gal80 repression was relieved from early second instar till late third instar by shifting from 18°C to 29°C . All graphs are Mean ± Standard Error ( Student's t test ) . n = 15 . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 01610 . 7554/eLife . 03126 . 017Figure 9—figure supplement 1 . Membrane tethered Wg perturbs multistratified arrangement of AMPs . ( A–H ) The third instar wing imaginal discs from wg{KO , Nrt-Wg} ( tethered Wg , non-secretory form ) ( E–H ) shows significant reduction in myoblasts layers ( anti-Twist , green ) and also beta-catenin ( anti-Beta-catenin , red ) activation in comparison to control disc ( A–D ) . n = 8 . Scale bar , 50 μm . ( I ) The graph showing numbers of actively dividing AMPs in wg{KO , Nrt-Wg} genotype and in control discs . The proliferation of AMPs is significantly less than control wing discs ( Mean ± Standard Error ( Student's t test ) . n = 10 . p-values < 0 . 001 ) . ( J ) Intensity of Numb expression in third instar wing discs . The expression is highest in the AMPs farthest from the epithelium . Number of discs used for analysis , n = 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 01710 . 7554/eLife . 03126 . 018Figure 9—figure supplement 2 . Wg down regulation alters division axis of third instar AMPs . AMP specific perturbation ( Using Dmef2-Gal4 ) of Wg pathway downstream molecule ( DN-TCF ) results in an axis of division ( marked by CNN-GFP , green ) parallel to disc epithelium ( wing disc epithelium-E ) in the third instar ( A , B and C are three representative examples ) . n = 10 . Scale bar , 10 μm . Gal80 repression was relieved from early second instar till late third instar by shifting from 18°C to 29°C . ( D ) A pie chart representation of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 01810 . 7554/eLife . 03126 . 019Figure 10 . Model proposed . In the early instars ( I and II ) AMPs exhibit symmetric division along epithelium and Serrate-Notch signaling plays major role at this stage . In the third instar onwards the axis of cell division in AMPs changes to orthogonal orientation and expression of Wg in disc epithelium along with Serrate-Notch signaling regulates AMP proliferation . Wg signaling potentially regulates Numb , which inhibits Notch leading to asymmetric divisions . DOI: http://dx . doi . org/10 . 7554/eLife . 03126 . 019
Early studies in Calliphora ( Crossley , 1972 ) and elegant clonal analysis and transplantation experiments in Drosophila ( Brower et al . , 1981; Lawrencea , 1982 ) established mesodermal cells attached to the wing imaginal disc as the source of muscle progenitors for the large indirect flight muscles of the adult thorax of dipteran insects . These ‘adepithelial’ cells were subsequently shown to express the mesodermal marker Twist ( Thisse et al . , 1988; Currie and Bate , 1991; Fernandes et al . , 1991 ) . Twist labeling of the AMPs on the late-third instar wing imaginal disc reveal a large number of cells ( Currie and Bate , 1991; Fernandes et al . , 1991 ) , which this study identifies as about 2500 . The haltere ( T3 ) disc and the wing ( T2 ) disc , start off with a similar numbers of myoblasts in the late embryo . Studies on muscle development in the ‘four-winged fly’ , where the function of the homeotic gene Ultrabithorax ( Ubx ) is removed from the haltere disc , showed that AMP population is controlled by the disc-epidermis . When the haltere ectoderm is transformed to T2 with the AMPs continue to retain a T3 identity , the AMPs nevertheless transform to a T2 population size ( Fernandes et al . , 1994; Roy et al . , 1997 ) . Despite these early studies examining the role of signaling from the disc-ectoderm ( Sudarsan et al . , 2001 ) in flight muscle specification ( Schnorrer et al . , 2010; Schönbauer et al . , 2011 ) , the mechanism of population expansion have not been hitherto investigated . This is an important general problem as it addresses how diversification of internal tissues in segmented animals takes place . Our results show that , in early development the AMPs on the wing-disc go through a phase of exponential amplification from about 10 cells to 250 . This amplification , where the daughters of each stem are both stem cells , takes place by the division of cells parallel to the disc-epithelium stem-cell niche and requires Ser-N signaling . As the expression of Wg in the presumptive notum known to occur in the late second instar larva ( Tomoyasu et al . , 2000; Alexandre et al . , 2014 ) changes in the AMPs proliferative behaviour are seen . A switch to asymmetric cell- division results in a ‘self-renewed’ stem cell and a post-mitotic sibling poised to differentiate to contribute to multinucleate muscles . The plane of cell division is no longer parallel to the ectoderm , resulting a stratified layer of cells , with the disc-proximal layer consisting of stem cells . Our findings support a model in which the embryonically generated AMPs acting as postembryonic muscle stem cells generate adult-specific myoblasts in a two-step process ( Figure 10 ) . AMPs initially manifest a symmetric cell division mode , which serves to amplify the progenitor pool in the first and second larval stages , and subsequently transit to an asymmetric cell division mode during the third larval stage during which they self-renew and generate the postmitotic myoblasts required for adult muscle formation . In both steps , signaling is required from the notum region of the wing disc epithelium , which acts as a transient niche . Thus , Ser localized in epithelial cells of the disc is necessary for activation of N signaling in AMPs throughout larval development , and diffusible Wg from the disc epithelial cells in the third larval instar is necessary for the expression of Nb in AMP lineages and , hence , for their transition from symmetric to asymmetric division modes . The idea of epithelium tissue as transient niche for the regulation of proliferation is something similar to that known in intestinal stem cell ( Mathur et al . , 2010 ) . In this paper authors provide evidence for the role of progeny of ISC progenitor as a very transient signaling center regulating stem cell proliferation via decapentaplegic . Since these findings reveal remarkable parallels to other tissue stem cell systems in Drosophila , we consider the disc-associated AMPs to be a novel type of muscle stem cell that orchestrates the early phases of adult flight muscle development in Drosophila . This is a newly identified and key role for Wg in the presumptive notum . Its other roles include a requirement for expression of the muscle-attachment gene stripe ( sr ) , in the presumptive notum ( Ghazi and VijayRaghavan , 2003 ) and for maintenance of the indirect-flight muscle marker Vg , in AMPs ( Sudarsan et al . , 2001 ) . Wg causes this switch in cell- division pattern by the transcription of Nb in the stem-cell layer , and its asymmetric distribution by well-established mechanisms ( Skeath and Doe , 1998; Couturier et al . , 2012 ) thereby inhibiting Notch activation in the post mitotic progeny . Many types of tissue stem cells , in a manner similar to that shown here for the AMPs , have the capability of proliferating through symmetric cell divisions , thereby expanding the stem cell pool , and then switching to asymmetric cell divisions . In Drosophila optic lobe development , a set of 30–40 epithelial progenitor cells generated in the embryo begin to proliferate in a symmetric cell division mode during early larval development to generate a neuroepithelial cell pool of approximately 700 cells . These neuroepithelial cells sequentially transform into neuroblasts , which generate differentiated neural cells of the medulla through asymmetric divisions . This transition from symmetric to asymmetric divisions involves cellular events such as the reorientation of the mitotic spindle and multiple molecular signaling processes including the dynamic regulation of Notch signaling ( Egger et al . , 2010 , 2011 ) . Since this transition from symmetrical ‘proliferative’ divisions to asymmetrical ‘differentiative’ divisions is also manifest in neural stem cell pools of the developing retina and neural tube in vertebrates , it may be a general strategy for generating a large final number of differentiated progeny from a small initial number of progenitors ( Kriegstein et al . , 2006; Borrell and Reillo , 2012 ) . The Drosophila adult intestine and that of the mouse are other examples of use of both modes of cell amplification in response to adult environmental inputs such as diet and consequent insulin-pathway signaling ( Lopez-Garcia et al . , 2010; O'Brien et al . , 2011; Yilmaz et al . , 2012 ) . Our results are consistent with the population of cells proximal to the epithelium being stem cells: These disc-proximal cells self-renew and also give rise to siblings , which differentiate . However , as of now , we do not have a marker that specifically labels these proposed stem cells . In this situation our results are also consistent with the view that the myoblasts are a pool of cells , in which only those closest to the signaling center receive signals that impact their division . One of the characteristics of the relationship between stem cells and their niche is the orientation of cell-division . This ensures the balance between self-renewal and differentiation as progeny , which lose contact with the signaling niche , is canalized to a particular fate , whereas those still in contact remain as self-renewing stem cells . In germline stem cell ( GSC ) in the ovarioles and testis , the contact of stem cells with the tightly regulate stem cell number and coordinated maturation of the progeny to form gametes . Loss of regulation in this axis results in mis-regulated cell number amplification ( López-Onieva et al . , 2008; Wang et al . , 2008; Losick et al . , 2011 ) . The intestinal stem cell niche in Drosophila and Mus have a cryptic stem cell niche where a differentiating progeny signal back to the stem cell through EGF signaling , acting as mitogen ( Jiang and Edgar , 2009 ) . Neuroblasts in Drosophila are widely considered as neural stem cells as they are shown to go through a series of self-renewing divisions while siblings differentiating to form neurons . Remarkably , this is also seen in the culture , suggesting that this is an intrinsic property and not niche dependent . However , surrounding glia in vivo are proposed to have a regulatory effect on the stem cell abilities of these cells ( Chai et al . , 2012 ) . There have also been previous studies in Drosophila that address how stem cell division relates to regulation of population size . It is known from the studies in germline stem cell ( Gilboa and Lehmann , 2006 ) and neural stem cells ( Egger et al . , 2011 ) there is change in division rates during larval development . In the early instars there is an amplification that slows down in the third instar . The ovary of adult Drosophila has 16–18 units , ovarioles , which are formed during the larval stages . Each ovariole contains two or three germline stem cells , which are in contact with somatic cells that regulate their establishment , maintenance and differentiation . For example , the developmental origin of the ovaries are the primordial germ cells ( PGCs ) , which till the third instar contribute to most of the cells required for gonad formation and in pupal stage separate into ovarioles ( Gilboa and Lehmann , 2006 ) . The numbers of PGCs double every 24 hr after hatching till the mid-third instar after which it slows down , finally forming 100 cells starting from12 cells ( Gilboa and Lehmann , 2006 ) . Such examples and our results tempt us to speculate that an early amplification and a later slowing down maybe a general feature of regulation of stem cell proliferation in Drosophila . Our results and the underlying mechanisms could also be of general applicability in understanding myogenesis in other contexts . In vertebrates , the Hox identity of the ectoderm could influence proliferation in the mesoderm , in a manner similar to that observed by us , to create the appropriate population of progenitors for the morphogenesis of muscles of very different sizes . Our study suggests the testable hypothesis that vertebrate myoblast pools also develop by symmetric and asymmetric division of muscle progenitor stem cells . In vertebrate muscle , satellite cells form a quiescent muscle stem-cell population important for the regeneration ( Brack and Rando , 2012; Cooper et al . , 1999; Le Grand and Rudnicki , 2007; Vasyutina et al . , 2007 ) . A recent study ( Konstantinides and Averof et al . , 2014 ) of limb regeneration in the crustacean , Parhyale hawaiensis identified satellite like cells ( SLCs ) expressing Pax3/7 genes expressed in vertebrate muscle satellite cells ( Kassar-Duchossoy et al . , 2005 ) . This study along with others show a common mesodermal origin of adult muscle stem cells in crustaceans and chordates though the precise developmental origins of these remains a mystery ( MAURO , 1961; Bryson-Richardson and Currie , 2008; Sambasivan and Tajbakhsh , 2007 ) . Our study raises the interesting possibility that muscle satellite cells are ‘unfused’ stem cells kept aside during early development . We also speculate that the novel stem cells that we observe could be satellite cells of flight muscle , used for repair upon muscle damage in a manner similar to that seen in vertebrates .
Fly stocks were obtained from the Bloomington Drosophila Stock Centre ( Indiana , USA ) and , unless otherwise stated , were grown on standard cornmeal medium at 25°C . The following transgenic lines were used . For wild type Twin-spot MARCM experiments males of genotype +; FRT 40 A , UAS mCD8::GFP , UAS rCD2 RNAi; Dmef2-Gal4 were crossed to females Hsflp/Hsflp; FRT 40A , UAS rCD2::RFP , UAS GFP RNAi . For Notch over expression and Numb downregulation twinspot MARCM experiments males of genotype +; FRT 40 , UAS mCD8::GFP , UAS rCD2 RNAi; Dmef2-Gal4 were crossed to females of genotype Hsflp/Hsflp; FRT 40A , UAS rCD2::RFP , UAS GFP RNAi/Cyo Act-GFP; UAS NICD/TM6 or Hsflp/Hsflp; FRT 40A , UAS rCD2::RFP , UAS GFP RNAi/Cyo Act-GFP; UAS Numb RNAi/TM6 Tb respectively . For MARCM experiments females of genotype Hsflp/Hsflp; FRT 42 B , Tub Gal80 were crossed to males of genotype +; FRT 42 B UAS mCD8::GFP/Cyo Act-GFP; Dmef2-Gal4 . In knockdown and overexpression experiments: +; +; Dmef2-Gal4 , Gal80ts . UAS Notch RNAi ( Bloom , 35213 ) . UAS Numb RNAi ( Bloom , 35045 ) . UAS Wg RNAi ( 13352; VDRC , Austria ) . UAS Serrate RNAi ( 108348; VDRC , Austria ) . UAS NICD . UAS DN Notch . UAS DN TCFΔN . UAS ( TCF ) 24/Cyo GFP . In mitotic spindle orientation experiments females of genotype +; +; Dmef2-Gal4 were crossed to males of genotype P{UASp-GFP-Cnn1}26-1 , w1118 . Serrate lacZ9 . 1 ( Bachmann and Knust , 1998 ) . Wg Sp-1/Cyo GFP . Wgts . Notchts . wg{KO , Nrt-Wg} ( Alexandre et al . , 2014 ) . Wing discs were dissected from first instar ( 24–30 hr AEL ) , second instar ( 48–55 hr AEL ) and third instar larval ( 72 hr onwards ) stages and then fixed in 4% paraformaldehyde diluted in Phosphate buffered saline ( PBS pH-7 . 5 ) . Immunostaining was performed according to Ghazi et al . ( 2000 ) . In brief , samples were then subjected to two washes of 0 . 3% PTX ( PBS + 0 . 3% Triton-X ) and 0 . 3% PBTX ( PBS + 0 . 3% Triton-X + 0 . 1 %BSA ) for 15 m each . Primary antibody staining was performed for overnight at 4°C on shaker and secondary antibodies were added following four washes of 0 . 3% PTX . Excess of unbound secondary antibodies were removed by two washes of 0 . 3% PTX following which samples were mounted in Vectashield mounting media . For immunostaining Anti Wg ( Mouse , 1:100 , DSHB ) , Anti-Twist ( Rabbit , 1:100 , kindly provided by S Roth , University of Cologne ) , Anti-NICD ( Notch intracellular C-terminal domain ) ( Mouse , 1:100 , DSHB ) , Anti-Numb ( Rabbit , 1:100 , kindly provided by Juergen Knoblich , IMBA , Vienna ) , Anti-GFP ( Chick , 1:500 , Abcam , Cambridge , UK ) , Anti-CD2 ( Mouse , 1:100; Serotec , Raleigh , NC , USA ) , TO-PRO-3-Iodide ( 1:1000 , Invitrogen ) , Anti-DCAD2 ( Rat , 1:200 , DSHB ) , Anti-Beta Gal ( Mouse , 1:50 , DSHB ) , Anti-Phospho histone 3 ( Rabbit , 1:100 , Millipore ) . Secondary antibodies ( 1:500 ) from Invitrogen conjugated with Alexa fluor-488 , 568 and 647 were used in all staining procedures . Olympus FV 1000 confocal point scanning microscope was used for image acquisition , which were processed using ImageJ software . Quantification of division axis was essentially performed as explained in the Egger et al . ( 2007 ) . To generate twin-spot MARCM clones , a single heat shock of 15 m at 37°C was given to specifically staged larvae and then larvae were dissected and wing discs were removed . The samples were then processed for antibody staining as mentioned earlier . ImageJ software was used to determine the number of cells in each clone ( Rasband WS , ImageJ U S . National Institutes of Health , Bethesda , Maryland , USA , http://imagej . nih . gov/ij/ , 1997–2012 ) . For Edu labeling , larvae were aged for 24 , 48 and 72 hr after hatching , on standard cornmeal media and then pulse labeled for 5 hr on Edu ( 0 . 2 mM final concentration ) mixed cornmeal media ( Daul et al . , 2010 ) . Half of the larvae from the 72 hr stage were separated , dissected and processed for immunolabeling in the ‘no chase’ cohort . Remaining larvae were transferred to standard cornmeal media without Edu and allowed to develop until wandering third instar stage . Wing discs were then dissected and processed for immunolabeling studies . In both cohorts , Edu detection was performed according to the Click-iT Edu labeling kit ( Invitrogen ) . | Muscle tissues must grow and change to accommodate the needs of an animal at various stages in its life . For example , fruit flies begin life as larvae and their muscles must help them move their soft bodies . Later , when the flies mature into adults , the muscles must provide power for flight and support for the insects' external skeletons . Like other animal tissues , muscles develop from non-specialized stem cells which at first have the potential to become almost any cell type , but later change to become more specialized . Studies of fruit flies , in particular , have yielded insights on how pools of stem cell are created and regulated . Fruit flies are small and easier to study than larger organisms , and as a result , scientists have learned a lot about their genetics and cell biology . Gunage et al . have now identified the stem cell pools that develop into flight muscle tissue , and found that these cells were set aside for the muscles when the fruit fly embryo was still developing . Fruit flies have large forewings that power flight , and small modified hindwings ( called halteres ) that help the insect to balance when flying . Gunage et al . reveal that a small , but similar , number of cells are set aside to make both both the tiny muscles that will move the halteres and the much larger flight muscles that move the forewings . However , the cells that contribute to the flight muscles divide to give far more muscle progenitor cells than their haltere counterparts , and make a couple of thousand cells that eventually fuse to form muscle fibers . Gunage et al . looked at how the flight muscle progenitors multiplied by genetically engineering some of the stem cells in fruit fly larvae so that when each cell divided , its two daughter cells would fluoresce with different colors . One daughter cell would glow green and the other glow red . Gunage et al . found that at first the cells multiply equally , with half the new cells coming from a ‘red’ stem cell and the other half from a ‘green’ cell—meaning that the number of cells increases exponentially . Later , the balance shifted so that either more red cells than green cells were produced , or vice versa . This results in a ‘linear’ increase in number of muscle progenitor cells . Furthermore , Gunage et al . identified the proteins that orchestrate the switch from equal to unequal multiplying of these cells at the different times points in the fruit flies’ development . The next challenge is to see if these stem cells that form the muscles are also available for repair of mature muscle tissue after it is damaged . If this is so , these stems cells might perform a similar function to muscle satellite cells , which are found in the mature muscles of mammals and other vertebrates . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2014 | Identification of a new stem cell population that generates Drosophila flight muscles |
Dentate neural stem cells produce neurons throughout life in mammals . Sonic hedgehog ( Shh ) is critical for maintenance of these cells; however , the perinatal source of Shh is enigmatic . In the present study , we examined the role of Shh expressed by hair follicles ( HFs ) that expand perinatally in temporal concordance with the proliferation of Shh-responding dentate stem cells . Specific inhibition of Shh from HFs or from epithelial sources in general hindered development of Shh-responding dentate stem cells . We also found that the blood–brain barrier ( BBB ) of the perinatal dentate gyrus ( DG ) is leaky with stem cells in the dentate exposed to blood-born factors . In attempting to identify how Shh might be transported in blood , we found that platelets contain epithelial Shh , provide Shh to the perinatal DG and that inhibition of platelet generation reduced hedgehog-responsive dentate stem cells .
Neural stem cells in the dentate gyrus ( DG ) consistently provide new excitatory neurons to modulate hippocampal circuitry and disrupted neurogenesis is linked to multiple neurological and psychiatric diseases ( Zhao et al . , 2008 ) . Stem cell niches in the adult dentate subgranular zone are established primarily by Sonic hedgehog ( Shh ) -responsive radial glial cells , which appear at embryonic day ( E ) 17 in mice ( Ahn and Joyner , 2005; Li et al . , 2013; Li and Pleasure , 2014 ) . In the adult , Shh plays a crucial role in maintaining the dentate stem cell niche and driving neurogenesis ( Machold et al . , 2003; Ahn and Joyner , 2005; Han et al . , 2008; Favaro et al . , 2009 ) . Starting at the end of the first week of life in mice , Shh is provided locally by cells in the dentate hilus ( Li et al . , 2013 ) ; however , Shh-expressing cells are not present in the DG when dentate stem cells first appear ( Ahn and Joyner , 2005; Garcia et al . , 2010; Li et al . , 2013 ) . There are several ways by which Shh-responsive cells might be found in the dentate prior to birth . Our previous work showed that Shh from outside the dorsal forebrain is crucial for the establishment of the dentate stem cells in the dentate , and that at earlier embryonic stages , the Shh ligand is produced by the amygdala and supplied to the adjacent ventral dentate neuroepithelium and these stem cells then migrate to populate both the temporal and septal dentate just before birth in mice ( at E17 . 5–18 . 5 ) ( Li et al . , 2013 ) . Since the postnatal dentate stem cells require Shh signaling to maintain stemness and division ( Ahn and Joyner , 2005; Choe and Pleasure , 2012 ) , the dentate stem cell niche should employ ways of supplying Shh from different sources after establishing the stem cells in the germinal area before birth . What structures supply Shh to the dentate after stem cells leave the germinative zone in the ventral dentate neuroepithelium and before the hilar mossy cells begin to produce Shh at the end of the first week of life ? Skin morphogenesis and homeostasis are regulated by hair follicles ( HFs ) , each of which is a small structure stocked with cells producing factors such as platelet-derived growth factors ( Pdgfs ) and Shh ( St-Jacques et al . , 1998; Chiang et al . , 1999; Karlsson et al . , 1999; Fuchs , 2007; Blanpain and Fuchs , 2009 ) . The perinatal interaction of the epithelial and dermal mesenchymal cells establishes the stem cell niche for the HFs through crosstalk of a variety of signaling molecules ( Rendl et al . , 2005; Nowak et al . , 2008 ) , and the development of HFs exposes new morphogens to nearby stem cell niches including mesenchymal stem cells and hematopoietic stem cells ( HSCs ) . Adult HSC niches form from embryonic HSCs that transiently reside in the liver by migration of the stem cells into the bone marrow . During late embryonic development , the skull is an active site for hematopoiesis , temporally coinciding with neurogenesis in the cortex ( Medvinsky et al . , 2011; Li et al . , 2012 ) . Calvarial mesenchymal cells condense to form the skull vault through intramembraneous ossification , and bone marrow mesenchymal cells play a critical role of recruiting HSCs from the circulation by secreting Cxcl12 , a critical homing signal for HSCs ( Méndez-Ferrer and Frenette , 2007; Lo Celso et al . , 2009; Méndez-Ferrer et al . , 2010; Greenbaum et al . , 2013 ) . Thus , the osteoblastic cells in the skull niche control hematopoiesis including megakaryopoiesis in the context of skeletal homeostasis ( Pallotta et al . , 2009 ) . Megakaryocytes , one of the HSC lineages , produce bone matrix components , cytokines and growth factors and mutant mice , which fail to release platelets from megakaryocytes such as Gata1 and Nfe2 knockout mice have abnormal bone mass ( Kacena et al . , 2004 , 2005 , 2006 ) . Activation of platelets leads to release of contents including Tgfβ1 , implying a messenger role for megakaryocytes to convey signals from the bone marrow and mesenchymal stem cell niches into the rest of the organism , particularly in places and locations with leaky blood vessels during development ( Levine et al . , 1993 ) . Interestingly , morphogens like Shh are also carried by blood-derived cells . T lymphocytes shed microvesicles containing Shh and Shh anchored in the microvesicles is functionally active in new blood vessel formation ( Agouni et al . , 2007; Soleti and Martinez , 2009; Benameur et al . , 2010 ) . Thus , the HSC generated cells may be critical for delivery of morphogens via the developing vascular system . HFs in the head skin are established perinatally , coinciding with expansion of calvarial and dermal mesenchymal cells covering the developing brain . The blood–brain barrier ( BBB ) matures as early as embryonic day ( E ) 15 . 5 in most forebrain areas ( Daneman et al . , 2010 ) except for a few areas , including the DG where the BBB matures postnatally . This raises a possibility that the HF stem cell niche signals interact with dermal/calvarial HSCs and the developing neurovascular units of the DG . In the present study , we provide evidence that HF stem niche signals such as Shh control the dentate stem cells by utilizing platelets as a delivery system in the early postnatal period .
Shh signaling is critical for ventral forebrain development in early embryogenesis and the signaling pathway becomes restricted within the neural and glial stem cell niches at the end of embryogenesis . Embryonically produced dentate granule neurons and dentate stem cells originate from the ventricular zone of the DG , whereas the adult dentate has hedgehog-responsive stem cells that reside in the dentate subgranular zone ( Altman and Bayer , 1990; Ahn and Joyner , 2005; Li et al . , 2013 ) . Since Shh is not detected in the dorsal forebrain when the adult dentate stem cells appear before birth , we examined putative sources of Shh that might contribute to Shh delivery via the dentate vasculature . To gain insight into the anatomy of Shh signaling in the head , we examined Gli1-GFP transgenic mice expressing GFP in hedgehog-responding cells . The GFP + hedgehog-responding cells of a Gli1-GFP GENSAT transgenic mouse line were obvious in the forming HFs ( Figure 1A , arrow heads ) of the dermis at E15 . 5 when the dermal mesenchymal cells condense before the appearance of calvarial bones , which showed GFP expression at later ages ( Figure 1A , red arrows ) . From E17 . 5 onward , the DG showed GFP + dentate progenitors and their descendants ( Figure 1A , yellow arrows ) . Despite the expansion of dentate Gli1-GFP + cells , the expression of Shh , however , was not detected in the dorsal cortex . Perinatally , Shh expression was rather restricted in the ventral forebrain such as around the third ventricle and in the entorhinal cortex ( Figure 1B ) . Interestingly , the HFs , expanding dramatically after E17 . 5 , were the geographically closest Shh-expressing cells to the DG when examined using the perinatal mouse head ( Figure 1C ) . 10 . 7554/eLife . 07834 . 003Figure 1 . Hedgehog signaling is restricted in the dermal mesenchyme and dentate stem cells . ( A ) Expression of Gli1-GFP shows the hedgehog-responding cells in the dermal mesenchyme ( red arrows ) , and hair follicles ( HFs ) ( arrow heads ) and the dentate ( yellow arrows ) at the late embryogenesis . ( B ) Expression of Sonic hedgehog ( Shh ) is restricted in the HFs ( boxes ) and the periventricular area of the third ventricle ( arrows ) and the entorhinal cortex ( arrow head ) . ( C ) High-power images of Shh expression in the HFs of boxed areas in ( B ) . Scale bars: A , B = 400 μm , C = 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 003 HFs act to produce hairs by being a niche for stem cells and by expressing secreted morphogenic molecules like Bmps , Wnts , Pdgfs , and Shh ( Karlsson et al . , 1999; Huelsken et al . , 2001; Suzuki et al . , 2009 ) . The vascular plexus and close location of HFs to the dermal and calvarial HSC niche led us to hypothesize that the follicular Shh could affect forebrain development and be a source of Shh ligand to the brain . To examine this possibility , we used Krt14-Cre , a dermis-specific Cre line , to conditionally inhibit the expression of Shh from the HF . The Krt14-Cre;Shhflx/flx mutants showed loss of Ptch1 , a downstream target gene of hedgehog signaling , and Shh expression in the skin; however , Shh expression in the ventral forebrain was unaffected ( Figure 2A and Figure 2—figure supplement 1 ) . We stained the DG with Ki67 ( a marker of dividing dentate progenitors ) , Pdgfrα ( a meningeal [white dashed lines] , and oligodendrocyte precursor marker—in the dentate , a few oligogenic progenitors exist at E17 . 5 and more at P1 in the DG ) and Reelin ( a Cajal-Retzius cell marker ) from E17 . 5 to P1 , when the mutant died , and observed significant decreases of Ki67 + dentate progenitors in the Krt14-Cre;Shhflx/flx mutant compared to the heterozygous Krt14-Cre;Shhflx/+ or control littermates ( Figure 2B , B′ , B″ ) . Prox1 staining of the dentate granule neurons showed more restricted localization of neurons in the upper blade of the mutant DG ( Figure 2B ) . The decline of Ki67 + dentate progenitors in the Krt14-Cre;Shhflx/flx mutant coincides with the requirement of hedgehog signaling for the expansion of dentate stem cells just before birth ( Ahn and Joyner , 2005 ) . However , these straightforward data do not exclude indirect involvement of hair follicular Shh in the expansion of dentate progenitors . 10 . 7554/eLife . 07834 . 004Figure 2 . Conditional inhibition of dermal Shh expression led to reduced dentate progenitors . ( A ) Krt14-Cre was used to conditionally delete Shh expression ( Krt14-cre;Shhflx/flx , in short Mut in this figure ) . Expression of Ptch1 , a downstream target gene of Shh signaling , and Shh showed absence of Shh expression in the mutant HFs . The right panel shows Ptch1 expression in the forebrain at E17 . 5 . ( B ) Expression of a dentate progenitor marker Ki67 at E17 . 5 and P1 . Pdgfrα and Reelin show meningeal cells and Cajal Retzius cells outlining the dentate gyrus ( DG ) , respectively . Prox1 shows dentate granule neurons at E18 . 5 . ( B′ , B″ ) Plots show Ki67 + cells in the dentate at E17 . 5 ( B′ ) and P1 ( B″ ) . Student t-test was used to determine the significant difference between groups . Scale bars: A = 200 μm , B = 100 μm . p = 0 . 03 ( B′ ) , <0 . 05 ( B″ ) . Dashed lines were used to outline the dentate . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 00410 . 7554/eLife . 07834 . 005Figure 2—figure supplement 1 . A representative image for in situ hybridization of Ptch1 using P1 Krt14-Cre;Shhflx mice . Krt14-Cre;Shhflx/flx mutant shows reduction of Ptch1 in the skin but not in the ventral midline ( scale bar = 200 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 005 Krt14-Cre-mediated Shh deletion compromised the survival of pups after birth because of the general loss of Shh widely in the epidermis . By utilizing an Msx2-Cre expressed in the skin area covering the forebrain ( Choe et al . , 2012 ) , we generated conditional mutants that survived to adulthood showing diminished epidermal Shh expression . As observed in Krt14-Cre;Shhflx/flx mutants , the Msx2-Cre;Shhflx/flx mutants also showed decreased numbers of dentate progenitors labeled by Ki67 that coincided with the appearance of hedgehog-responsive dentate progenitors ( Figure 3A , A′ , B , B′ ) . One hour BrdU labeling at P3 confirmed the diminished cell proliferation to one third of normal in the mutant dentate ( Figure 3C , C′ ) . Reduction of Shh expression was maintained in the skin; however , Shh + cells in the dorsal forebrain started to appear at P3 including in the hippocampal hilus ( Figure 3D ) ( Li et al . , 2013 ) . Expression of Msx2-Cre ( Figure 3D′ ) and Krt14-Cre was restricted in the skin perinatally ( Figure 3—figure supplement 1 ) with reduction of SHH expression in the DG of Msx2-cre;Shhflx/flx mutants ( Figure 3—figure supplement 2 ) . To examine whether the effect of Shh is specific to the dentate progenitors , we counted 1 hr labeled BrdU + nuclei in the cortex by dividing the cortex into four bins at E17 . 5 and three bins at P1 ( See the scheme in Figure 3E′ , E′ ) . The BrdU + dividing cells in the cortex did not show a significant difference at both time points ( Figure 3E , E′ , E″ ) . This result supports the idea that Shh from the HFs selectively affects dentate progenitors . 10 . 7554/eLife . 07834 . 006Figure 3 . Msx2-Cre-mediated inhibition of dermal Shh expression reduced postnatal dentate progenitors . ( A–C ) Dentate progenitors were stained with Ki67 ( A , B ) or BrdU ( 1 hr , ( C ) ) from Msx2-Cre-mediated conditional inhibition of Shh expression at E16 . 5 ( A ) and P3 ( B , C ) . ( A′ , B′ , C′ ) The decrease of the dentate progenitors is presented at P3 from staining Ki67 or BrdU ( 1 hr ) using embryos from six different litters ( n = 6 ) . ( D ) In situ hybridization of Shh at P3 shows the decrease of hair follicular Shh expression in the mutant ( Msx2-Cre; Shhflx/flx ) . ( E ) Progenitors in the cortical subventricular zone ( SVZ ) were stained with 1 hr BrdU labeling . ( E′ , E″ ) Cortical BrdU expressing cells were measured by dividing the cortex into four ( A-B-C-D , E17 . 5 ) or three ( A-B-C , P1 ) bins from the pial layer to the ventricle as depicted in drawings on the right panel . Scale bars: A , B , C = 100 μm , D = 400 μm , E = 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 00610 . 7554/eLife . 07834 . 007Figure 3—figure supplement 1 . Two Cre reporter mice were used to reveal the expression of Krt14-Cre and Msx2-Cre in the skin . For E18 Krt14-Cre and P1 Msx2-Cre , ROSA-LacZ reporter mice were used for X-gal staining ( scale bar = 200 μm ) . A GFP-stained image for the sagittal section of P2 Msx2-Cre with a ROSA-Yfp reporter also show restricted Cre activities in the HFs and the skin ( scale bar = 1 mm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 00710 . 7554/eLife . 07834 . 008Figure 3—figure supplement 2 . Immunostaining for Shh ( Epitomics ) and PECAM ( BD Pharmigen , blood vessels ) shows perivascular and dentate localization of Shh at P3 . Shh expression was reduced in both vascular and dentate areas of Msx2-Cre;Shhflx/flx mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 008 To examine whether the decrease in proliferating cells reflects a loss from the hedgehog-responding population , we bred the Ptch1-LacZ hedgehog-signaling reporter line to Msx2-Cre;Shhflx mutants . At P3 when Ptch1-LacZ + cells were visible in the dentate , the Msx2-Cre;Shhflx/flx;Ptch1-LacZ showed a third of the X-gal stained cells in the dentate compared to their heterozygous littermates ( Figure 4A , A′ ) . Cells expressing Sox2 marking dentate progenitor cells also decreased in the P1 mutants ( Figure 4—figure supplement 1 ) . That , hedgehog signaling was reduced in these mice was obvious when examining the skin on the scalp ( Figure 4B ) . Since the Msx2-Cre;Shhflx/flx mutant mice survived , unlike the Krt14-Cre;Shhflx/flx mutants , we examined the dentate progenitors at P10 , when the dentate subgranular zone has been established and in young adults at P40 . We examined Lef1+ or Blbp + radial glial stem cells , which comprise the pool of the dentate stem cells , and Ki67 + transit amplifying cells ( TACs ) ( Choe and Pleasure , 2012 ) . At P10 and P40 , both radial glial and TACs were reduced in the mutant . This was surprising since the hilar mossy cells have clearly started to express Shh by the first week of birth ( Figure 4C , C′ ) ( Li et al . , 2012 ) . At P50 , doublecortin ( DCX ) -positive immature dentate neurons were also significantly reduced in the mutants ( p < 0 . 005 , Figure 4—figure supplement 2 ) . This implies that the loss of hedgehog signaling from the skin likely affects the later development of the dentate niche in a long-lasting way even after local Shh expression appears . We suspect this is due to a critical window for epithelial supplied Shh , but the potential mechanisms for the long-lasting effect need to be examined in future works . 10 . 7554/eLife . 07834 . 009Figure 4 . Reduced hedgehog-responding dentate progenitors in the postnatal mutant . ( A ) To examine hedgehog signaling in the dentate , Ptch1-LacZ transgenic reporter mice were bred to Msx2-Cre;Shhflx/ mice . X-gal staining of the P3 DG shows reduced Ptch1+ cells in the mutant . ( A′ ) X-gal + cells were quantified from sections obtained from three litters ( n = 3 ) . ( B ) Reduced hedgehog signaling in the skin was revealed by X-gal staining of skin tissues obtained from Msx2-Cre;Shhflx/+;Ptch1-LacZ and Msx2-Cre;Shhflx/flx;Ptch1-LacZ mice at P3 . ( C ) Dentate progenitors were stained at P10 and P40 with BrdU ( 1 hr ) , Blbp , Lef1 ( glial progenitors ) , Ki67 ( intermediate progenitors ) . ( C′ ) Numbers of marker positive cells were plotted . Four different litters were used to count cells from the DG ( n = 4 ) . Student t-test was used to address the statistical significance . ** , p < 0 . 05; *** , p < 0 . 001 . Scale bars: A , C = 100 μm , B = 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 00910 . 7554/eLife . 07834 . 010Figure 4—figure supplement 1 . Expression of a neural stem cell marker , Sox2 ( green ) , in the P1 DG . Blbp ( red ) staining was used to mark the boundary of the DG , which shows perivascular glial cells and a few migrating glial cells in the infrapyramidal blade . Sox2-positive neural stem cells were scattered in the developing DG but a few Sox2-positive cells were detected in the upper blade of the Msx2-Cre;Shhflx/flx mutant DG ( A , scale bar = 100 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 01010 . 7554/eLife . 07834 . 011Figure 4—figure supplement 2 . P50 Msx2-Cre;Shh mice were used to stain DCX to visualize the immature newly born neurons in the DG . A rabbit anti-DCX antibody ( Abcam ) was used for the detection of newly born immature neurons of Msx2-Cre;Shhflx mice at P50 . The upper blade of the DG was used to count DCX positive cells ( scale = 100 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 011 The remaining and quite important question is how hair follicular Shh is able to affect the development of the brain and whether there is some transport mechanism that leads to transfer to the dentate . To gain insight into this question , we examined what types of cells respond to hedgehog near HFs using the Ptch1-LacZ mice . At P3 , Ptch1-LacZ + cells were enriched in the cranial suture , a front of bone growth ( Figure 5A , arrow ) . Ptch1-LacZ + cells were also observed in the calvarial bone stained with Alizarin red as well as in dermal mesenchymal cells ( Figure 5B ) . A closer examination of the hedgehog-responding cells in sections of E17 Gli1-CreERt2;Rosa-Yfp embryos 48 hr after Tamoxifen induction revealed that GFP + cells were Pdgfrβ+ and overlapped with Vimentin+ and Desmin + perivascular cells ( Figure 5C , C′ ) . Dermal mesenchymal cells expand rapidly perinatally and the mesenchymal cell numbers were reduced after loss of hair follicular Shh ( Figures 2A and 3D ) . Since dermal mesenchymal cells produce morphogenic proteins that could indirectly affect dentate progenitors , we conditionally ablated hedgehog signaling in the dermal mesenchymal cells using Pdgfrb-Cre , a mesenchymal Cre driver ( aka , Pdgfrb-Cre;Smoflx/flx ) . Inhibition of dermal mesenchymal hedgehog signaling dramatically reduced the size of the skull at P3 , which was not observed in either Krt14-Cre;Shhflx/flx and Msx2-Cre;Shhflx/flx mutants ( Figure 5D ) . This implies the involvement of Indian hedgehog ( Ihh ) signaling as it was previously reported that Ihh has a positive role for intramembraneous ossification of the skull ( St-Jacques et al . , 1999; Lenton et al . , 2011 ) . In these mutants , there was a slight but smaller magnitude decrease in Ki67 + dentate progenitors at E17 . 5 and P1 ( Figure 5E , E′ ) . These results suggest that hedgehog signaling in dermal mesenchymal cells might mediate some role in dentate expansion; however , the change in dentate progenitors in the Shh mutants , Krt14-Cre;Shhflx/flx and Msx2-Cre;Shhflx/flx was much greater implying that an indirect role via mesenchymal cells and skull dysplasia isn't the primary explanation . 10 . 7554/eLife . 07834 . 012Figure 5 . Development of perinatal dentate progenitors by inhibition of hedgehog signaling in the dermal mesenchyme . ( A ) Ptch1-LacZ expression in the calvarium is presented . Arrow indicates X-gal staining of Ptch1-LacZ in the calvarial suture . ( B ) Ptch1-LacZ expression was detected in the fronts of developing calvarial bones ( arrows ) . Ptch1-LacZ + dentate progenitors are obvious at P3 ( arrow head ) . Fetal mouse heads were stained with X-gal and Alizarin red to counter-stain the calvarial bone . ( C ) Hedgehog-responding cells and their descendants in the calvarial and dermal mesenchymes are presented using E17 . 5 Gli1-CreERt2;Rosa-Yfp embryos that was injected with TM at E15 . 5 . Sections were stained for GFP to label hedgehog-responding cells with mesenchymal markers such as Pdgfrα , Pdgfrβ ( dermal mesenchyme , meninges ) , and Sp7 ( calvarial mesenchyme ) . Inset shows co-localization of GFP and Pdgfrβ . ( C′ ) GFP + cells were stained with pericyte markers such as Desmin and Vimentin . GFP + vascular cells are noted ( arrows ) . ( D ) Hypoplasic skull bone development in the Pdgfrb-Cre;Smoflx/flx mutant at P3 . Lines indicate the length of skull bones . ( E ) Dentate progenitors were stained for Ki67 using E17 . 5 and P1 Pdgfrb-Cre;Smoflx/+ and Pdgfrb-Cre;Smoflx/flx embryos . Dentate blood vessels were counter-stained with PECAM to outline the dentate ( dashed lines ) . ( E′ ) Four different litters were used to measure the decrease of dentate progenitors in the mutant ( n = 4 ) . Student t-test was used to test the significant difference of the number of Ki67 + cells . p values are presented in the graph . Scale bars: B , E = 200 μm , C , C′ = 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 012 Adult neural stem cells reside in a structure termed by some the ‘neurovascular niche’ ( Shen et al . , 2008; Tavazoie et al . , 2008 ) . It is known that the BBB forms as early as E15 . 5 ( Daneman et al . , 2010 ) and is reinforced later by astrocytes . We wondered whether the dentate BBB matures according to the same time frame as the rest of the cortex or whether the dentate BBB might still be leaky at perinatal stages when dentate stem cells populate the dentate . We examined the integrity of the perinatal dentate BBB first using 70 KDa biotin-conjugated dextran dyes perfused into E17 . 5 embryos and P1 pups followed by staining with streptavidin to visualize the integrity of the dentate blood vessels . At E17 . 5 and P1 , the dentate blood vessels were leaky , with 70 KDa dye shedding into the adjacent cells ( Figure 6A ) . To examine the formation of glial-vascular units by newly born dentate cells , Gli1-CreERt2 mice were injected with TM at E17 . 5 and pups were analyzed at P2 . Staining the Cre reporter Rosa-Yfp revealed the blood vessels wrapped by endfeet of hedgehog-responding dentate cells ( Figure 6B ) suggesting the involvement of Shh-responsive dentate stem cells in the formation of the dentate BBB . Gli1-GFP pups were used to perfuse biotin-conjugated cadaverine dyes to visualize the transcytosis of the dye in the dentate blood vessels at P5 ( Figure 6C ) . The neuronal cells adjacent to GFP + dentate cells showed uptake of perfused cadaverine dyes released from the leaky blood vessels of the dentate . These results imply that at these perinatal stages , the glia-like dentate stem cells are involved in the organization of the dentate BBB and that these stem cells could be exposed to blood-born factors . Perfusion of 70 KDa dyes into P5 pups also showed weak integrity of dentate blood vessels compared to blood vessels in CA1 pyramidal zone ( Figure 6D , D′ ) . We thus hypothesized that the hair follicular Shh could reach the dentate stem cells through the leaky BBB . 10 . 7554/eLife . 07834 . 013Figure 6 . The weak integrity of the BBB in the fetal dentate . ( A ) E17 . 5 embryos and P1 pups were perfused with biotin-conjugated 70 KDa dextran and sections of the DG were stained for biotin . Arrows indicate the dextran outside blood vessels . ( B ) Gli1-CreERt2 pups ( TM at E17 . 5 ) were used to stain hedgehog-responding cells ( GFP+ ) and blood vessels ( PECAM ) . Hedgehog-responding Gli1+ dentate progenitors have endfeet wrapping blood vessels forming neurovascular units . Magnified images on the right panels show representative GFP + cells . Yellow dashed lines mark the meningeal blood vessels of the DG . ( C ) P5 Gli1-GFP pups were perfused with biotin-conjugated cadaverine to reveal the area of the DG with leaky blood vessels . GFP + hedgehog-responding cells are surrounded by the dentate granule neurons uptaking the dyes ( arrows ) . Dashed lines mark the meningeal blood vessels of the DG . ( D ) P5 CD1 pups were perfused with biotin-conjugated 70 KDa dextran and sections were stained for biotin . ( D′ ) Biotin signals from the section of ( D ) were used to measure the permeability of blood vessels in the hippocampus ( CA1 ) , and the DG ( DG ) . The area ( 100 μm2 ) surrounding blood vessels was selected to count the transcytosed dextran dyes ( n = 6 ) . ( E ) Staining for fibrinogen ( Fib ) , a marker for leaky blood vessels , was conducted at E17 . 5 and P1 in the DG . Arrows indicate the leaky blood vessels . ( F ) CD41 , a marker for platelets , was used to stain the dermal platelets ( including megakaryocytes ) and circulating platelets in sagittal sections at P1 . The boxed area is presented as a high-power image on the right . Student t-test was used to address the statistical significance . *** , p < 0 . 001 . Scale bars: A , D , F = 200 μm , B , C , E = 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 013 However , there seemed very little prospect that Shh might be freely diffusible in blood , so we assumed that there must be a carrier of some sort . Platelets are a good candidate as blood-born messengers to mediate transport of hair follicular signaling molecules to the brain considering the presence of the head HSC niche underlying the dermal HFs . Binding of fibrinogen with its receptors on platelets leads to the activation of platelets and secretion of granule proteins from platelets' fragments in so-called microparticles ( Offermanns et al . , 1997 ) . Staining for fibrinogen revealed strong expression in the dentate at E17 . 5 to P1 suggesting possible involvement of platelets and their fragments in the dentate and maturation of the dentate BBB ( Figure 6E ) . A marker of platelets , CD41 , showed circulating megakaryocytes and platelets within the dermis ( Figure 6F , arrow ) and in the dentate blood vessels and dentate at P1 ( Figure 6F ) . These results provide evidence of a plausible interaction between dentate glial cells wrapping dentate blood vessels and circulating platelets . To further examine hedgehog signaling and platelet formation , Gli1-CreERt2;Rosa-Yfp mice were injected with TM at E15 . 5 and analyzed two days later . Abundant GFP+;DAPI- cells were indeed found near the calvarial suture and the dentate blood vessels and the cells were CD42d + indicating that they are ( pro ) platelets ( Figure 7A ) . We examined a variety of other blood cell markers to determine the nature of GFP+;DAPI- cells . Hematopoietic progenitor markers such as CD34 and CD45 did not stain GFP + cells , while these cells were stained with platelet markers such as CD41 , CD42d , and CD61 suggesting that platelets are generated from cells responsive to hedgehog signaling ( Figure 7B ) . To examine whether the platelets contain Shh , fifty μl of embryonic blood was streaked on glass slides to stain platelets with Shh antibodies . Platelets from E15 . 5 did not show Shh expression , but platelets from E17 . 5 embryonic blood showed strong Shh expression ( Figure 7C , C′ ) . Western blot analysis of immunoprecipitates obtained with anti-CD41 antibodies using E17 . 5 embryonic blood samples to partially purify platelets showed that Shh precursors are present in platelets ( Figure 7C″ ) . The detection of Shh precursors in platelets implies that Shh could be packaged into platelets similar to other factors that are packaged into platelets without being expressed by megakaryocytes ( Harrison and Cramer , 1993 ) considering that Shh gene expression is lacking in megakaryocytes ( data not shown ) . To test this possibility , blood streaks from Krt14-Cre;Shh mutant embryos at E17 . 5 were used to detect Shh and we found loss of Shh in the platelets from the mutant embryonic blood supporting the idea that platelets take up Shh from dermal sources ( Figure 7D , D′ ) . To confirm the result , we utilized five commercially available anti-Shh antibodies to repeat the immunoprecipitation experiment ( Figure 7—figure supplement 1 ) and localize Shh in the microvesicles of platelets using a super-resolution microscope ( Figure 7—figure supplement 2 ) . Both results support that platelets contain SHH proteins , but Shh gene expression was not detected in the platelets as determined by using Shh-Cre with Ai14 reporter mice ( Figure 7—figure supplement 3 ) . Since we specifically detected unprocessed Shh proteins from our immunoprecipitation experiments , we cultivated epidermal cells from the P1 head skin to collect microvesicles from the conditioned media . The microvesicles from the conditioned media contained unprocessed Shh proteins as compared from the Shh produced in the cells ( Figure 7—figure supplement 4 ) and this result implies that hair follicular Shh can be released as unprocessed form in microvesicles to be taken up by carrier cells such as platelets . It needs to be further studied how proteins from other sources are packaged into embryonic platelets at specific ages; however , these results support that platelets could carry Shh from the dermis to the DG . 10 . 7554/eLife . 07834 . 014Figure 7 . Platelets and their neighboring cells respond to hedgehog signaling in the dermal and the dentate blood circulation . ( A ) Pregnant Gli1-CreERt2;Rosa-Yfp mice were injected with TM at E15 . 5 and embryos were collected at E17 . 5 to co-stain hedgehog-responding cells and a platelet marker , Cd42d . Top panels show the GFP+;CD42d + platelets ( arrow ) in the front of calvarial bone growth . Bottom panels show GFP+;CD42d + platelets ( arrow ) in the blood vessels covering the DG . Boxed areas are presented as a high-power image on the right . ( B ) E17 . 5 embryos ( Gli1-CreERt2;Rosa-Yfp , TM at E15 . 5 ) were stained with Nestin , CD34 ( hematopoietic , mesenchymal progenitor cells ) , CD45 ( hematopoietic cells not in platelets ) , and platelet markers such as CD42d , CD41 , and CD61 . Arrows indicate GFP+;CD42d + platelets in the dorsal dermal mesenchyme . Bottom panels present high-power images of cells ( arrows ) . ( C ) Blood streaks from E15 . 5 and E17 . 5 embryos were used to stain CD41 + platelets and Shh . Anti-rabbit IgG ( Rb-IgG ) was used for the negative control of Shh staining . ( C′ ) Shh signal intensities from CD41 + platelets were measured to show the increase of Shh in the platelets at E17 . 5 ( n = 6 ) , which correlates with the expansion of hair follicular Shh . ( C″ ) Western blot analysis was conducted using E17 . 5 blood samples immunoprecipitated with CD41 antibodies . Three different blood samples were loaded . ( D ) Blood streaks were obtained from E17 . 5 Krt14-Cre and Krt14-Cre;Shhflx/flx embryos to stain Shh and CD41 . ( D′ ) Ratio of CD41 + platelets without Shh from total CD41 + platelets was measured ( n = 6 ) . Student t-test was used to address the statistical significance . *** , p < 0 . 0001 . Scale bars: A = 200 μm , B , C , D = 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 01410 . 7554/eLife . 07834 . 015Figure 7—figure supplement 1 . Plasma samples from perinatal mice ( E18 – P3 ) were immunoprecipitated using anti-CD41 antibodies followed by Dynabead-conjugated Protein A ( Life Technologies ) . Immunoprecipitated protein extracts were detected using the Shh antibodies used in N-SIM imaging ( Figure 7—figure supplement 2 ) . Rabbit anti-Shh ( #2207 ) was obtained from Cell Signaling Technology and did not detect Shh in cell staining experiment ( Figure 7—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 01510 . 7554/eLife . 07834 . 016Figure 7—figure supplement 2 . Super-resolution images of Shh + platelets were taken from N-SIM ( Nikon ) equipped with a 100X ( N . A . 1 . 49 ) oil objective and 405- , 488- , and 561-nm lasers . Anti-Shh antibodies showed localization of Shh ( green ) in the microvesicles of platelets co-stained for CD41 or CD42d ( red ) , Table lists applied antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 01610 . 7554/eLife . 07834 . 017Figure 7—figure supplement 3 . To examine Shh gene expression in platelets during development , Shh-Cre;Ai14 embryos were stained for RFP and CD41 ( platelets ) . Megakaryocytes or platelets did not express Shh-Cre as revealed by a Cre reporter , Ai14 ( Jax ) in the representative tissues such as the lung , liver , whisker , and skin ( scale bar = 100 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 01710 . 7554/eLife . 07834 . 018Figure 7—figure supplement 4 . Dermal cells were isolated from P1 head skin harboring HFs and further expanded in 10% FBS/DMEM/F12 ( 50:50 ) media . To remove microvesicles in the media , FBS was precleared by ultracentrifugation at 120 , 000 for 90 min . Cells were cultured for 2 days and condition media were collected by ultracentrifugation at 120 , 000 for 90 min followed by lysis in the 1× Laemmli sample buffer ( Bio-Rad ) . Whole-cell lysates ( WCLs ) were directly lysed in the 1× Laemmli sample buffer for the Western blot analysis . Detection of Shh was done by using a rat anti-Shh antibody ( R&D Systems ) . Unprocessed Shh was detected in the microvesicle collected from the conditioned media . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 018 Nfe2−/− mutant mice fail to produce platelets from megakaryocytes ( Shivdasani et al . , 1995 ) , so we used these mice to test the hypothesis that reduced platelets could affect the Shh-responding population in the perinatal DG . Nfe2−/− mutants showed reduced numbers of platelets in the dentate and the dermal cells adjacent to the HFs at P1 ( Figure 8A ) . Blood streaks from P1 Nfe2 mutant also showed that Shh+;CD41 + platelets were reduced to about 30% of normal ( Figure 8A , A′ ) . These results indicate that Nfe2−/− mutants have significant reductions in circulating platelets containing Shh . Embryonic Nfe2−/− mutants were examined to determine if there was an effect on the dentate . At E16 . 5 , the number of Ki67 + dentate progenitors was not changed ( Figure 8B , B′ ) ; however , we found that the Ki67 + or Lef1+ dentate progenitors at P1 and Ki67 + dentate progenitors at P3 were significantly reduced in the Nfe2−/− mutants ( Figure 8B′ , C , D , D′ ) . This indicates a potential perinatal specific role for platelets in the dentate and raises the question of whether the reduced dentate progenitors were contributed from Shh-responding dentate stem cells . We bred Nfe2 mice to Ptch1-LacZ , to visualize hedgehog-responding dentate stem cells . At P3 , the Ptch1-LacZ + cells were reduced to 50% ( Figure 8E , E′ ) . Consistent with this result , the expression of Gli1-GFP + DG progenitors and their descendants in Nfe2−/− mutant pups at P1 were also reduced to about 50% ( Figure 8—figure supplement 1 ) . The effect on dentate progenitors in the mutant was still seen at P5 , which was the latest age for mutant survival ( Figure 8F , F′ ) . These data are consistent with our hypothesis that platelets are important for generation of perinatal Shh-responding dentate cells . 10 . 7554/eLife . 07834 . 019Figure 8 . Diminished hedgehog-responding fetal dentate progenitors in a platelet mutant , Nfe2−/− . ( A ) Heterozygote and mutant Nfe2 pups at P1 were used to stain CD41 ( platelets ) and Shh . Dashed line was used to outline the DG . Embryonic dermis ( skin ) and blood streaks were used for Shh and CD41 staining . HF = a hair follicle in the skin . ( A′ ) The ratio of CD41+;Shh + platelets to total Shh + cells was measured ( n = 6 ) . ( B , C ) Nfe2 embryos at E16 . 5 or pups at P1 were used to stain dentate progenitors ( Ki67 ) and Pdgfrα+ meninges or Reelin + Cajal Retzius cells were co-stained to outline the embryonic DG . ( B′ ) Numbers of dentate progenitors ( Ki67 + or Lef1+ ) were plotted ( n = 6 ) . ( D , D′ ) The Ki67 + dentate progenitors were stained at P3 . The plot shows the decreased Ki67 + dentate progenitors in Nfe2 mutants ( n = 6 , D′ ) . ( E , E′ ) The Ptch1-LacZ + dentate progenitors were stained at P3 using Nfe2;Ptch1-LacZ pups . The plot shows the decreased Ptch1-LacZ + dentate progenitors in Nfe2 mutants ( n = 3 , E′ ) . ( F , F′ ) The dentate progenitors ( Ki67 , Lef1 , Blbp ) were stained at P5 when a few Nfe2 mutant survived . The plot shows the decreased dentate progenitors in the Nfe2 mutant ( n = 3 , F′ ) . Dashed lines denote the outline of the DG . Student t-test was used to address the statistical significance . ** , p < 0 . 05 , *** , p < 0 . 001 , **** , p < 0 . 0001 . Scale bars: all = 200 μm except A ( skin and blood streak ) = 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 01910 . 7554/eLife . 07834 . 020Figure 8—figure supplement 1 . P1 pups from Nfe2 mutant mice with a Gli1-GFP reporter allele were stained for Blbp ( marks glial cells surrounding the DG at P1 ) and GFP ( marks Shh-responding cells ) ( A , scale bar = 100 μm ) . Nfe2 mutant pups showed significantly reduced ( p = 0 . 0027 , n = 4 ) Gli1-GFP positive cells ( A′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07834 . 020
Early ventral patterning is governed by Shh secreted from the ventral neurons and floor plate cells , and development of the dorsal neural tube requires the absence of Shh signaling ( Echelard et al . , 1993; Ruiz i Altaba et al . , 1995 ) . However , after neural tube closure , the zona limitans intrathalamica cells , cerebellar Purkinje neurons , and the tectal plate start to express Shh and regulate expansion of the dorsal brain structures ( Ruiz i Altaba et al . , 2002 ) . From E14 . 5 to E17 . 5 , weak expression of Shh was detected in the layer V cortical projection neurons away from the Gli gene expression domains , which are exclusively detected in the germinal area ( Dahmane et al . , 2001 ) . During late stage of dorsal brain development , Shh is known to function as a mitogen for Shh-responsive precursors in the cortex , the hippocampus , and the cerebellum ( Dahmane , 1999; Wechsler-Reya and Scott , 1999 ) . However , deletion of dorsal Shh expression using Emx1-Cre or NeuroD6-Cre resulted in only a modest effect on the cortex and more significant but postnatal effect in the DG ( Komada et al . , 2008; Li et al . , 2013 ) . In the ventral forebrain , cells in the striatum and the amygdala strongly express Shh and contribute to the generation of dorsal neuronal cells ( Machold et al . , 2003; Li et al . , 2013 ) . Shh expression becomes broadly detected after birth in the dorsal forebrain including dentate hilar neurons ( Machold et al . , 2003; Li et al . , 2013 ) . A prominent source of Shh in the adult forebrain is restricted expression in the layer 5 cortical projection neurons , and the neuronal Shh primarily regulates neuronal microcircuit formation and the development of astrocytes ( Garcia et al . , 2010; Harwell et al . , 2012 ) . Development of the DG begins about E13 and extends for two weeks , continuing after birth . The loss of Tbr2 , a nuclear regulator of intermediate neuronal progenitors , fails to generate postnatal DG ( Hodge et al . , 2012 ) and Tbr2 mutants have postnatal abnormalities of the DG similar to the DG of Smo mutants ( Machold et al . , 2003 ) . This result implies that Shh signaling might be involved in neurogenesis through production of intermediate neuronal progenitors . One of the Shh effects could be transforming embryonic dentate stem cells to an intermediate neuronal progenitor-generating derivative of dentate stem cells . This idea somewhat explains the perinatal appearance of Shh-responsive dentate stem cells ( Ahn and Joyner , 2005 ) , the proliferative effect of Shh to the late-born dentate stem cells , and the expansion of Shh-responsive radial glial stem cells during the early postnatal period ( Lai et al . , 2003; Choe and Pleasure , 2012; Li et al . , 2013 ) . The neuroepithelial cells are located near the ventricle , a rich source of trophic factors , but Shh-responding perinatal dentate stem cells need transient cell–cell interactions to maintain their stemness while producing dentate cells and migrating to the SGZ . To produce a large volume of dentate granule neurons and finally form a ‘V’ shaped SGZ to generate the dentate blades with new granule neurons , migrating dentate stem cells should be continuously exposed to mitogenic factors . The migrating dentate stem cells could take advantage of other sources of stem cell factors like the meninges ( Choe et al . , 2013 ) but stem cells ultimately travel far from the meninges but still need access to mitogens . It is well known that platelets are thrombotic anuclear cell fragments that regulate hemostasis , wound healing , and tissue repair ( Gawaz and Vogel , 2013 ) . After vascular damage , activated platelets plug the wound and maintain hemostasis . Growth factors from platelets have been known to promote growth of fibroblasts , hepatocytes , glial cells , hippocampal neurons , and endothelial cells , and platelets are the best source of purifying growth factors such as PDGF , HGF , VEGF , PAF , TGFβ1 , and BDNF ( Kohler and Lipton , 1974; Westermark and Wasteson , 1976; Nakamura et al . , 1986; Miyazono et al . , 1988; Clark et al . , 1992; Mohle et al . , 1997 ) . Although embryonic platelets appear before forebrain neurogenesis ( Tober et al . , 2007 ) , the embryonic role for these cells has not been fully studied . Platelets are a source of various cytokines and growth factors and it is not surprising that platelets functions not only in the pathological conditions like thrombosis and neuronal degenerative diseases but also in physiological situations like embryonic development . Nonhemostatic and developmental functions of platelets are best understood for their roles in embryonic vascular patterning . Meis1 mutant embryos lack platelets and show defective separation of blood and lymphatic vessels , while CLEC-2 receptors on platelets regulate lymphatic endothelial SLP-76 signaling and specify the lymphatic vessels from the blood vessels ( Bertozzi et al . , 2010; Carramolino et al . , 2010 ) . In a pathological condition like stroke , platelet lysates injected into the ventricle showed a neuroprotective effect with increases in neurogenesis and angiogenesis ( Hayon et al . , 2013 ) . These results show that a plethora of biologically active contents in platelets can play a modulatory role under various physiological and pathological conditions and therapeutic benefits of platelets are broadly possible from this feature of platelets ( Stellos and Gawaz , 2007; Mazzucco et al . , 2010 ) . Proliferation of adult dentate stem cells proceeds in perivascular microenvironements ( Palmer et al . , 2000 ) and the platelets' contents including Shh , Vegf , Tgfβ could control stem cell behavior depending on the vascular condition . In a pathological condition like vascular dementia , the wound in the blood vessels could open up the hole for platelets to release contents to the endothelial cells and the dentate cells . Since Shh could affect differentiated dentate granule neurons ( Petralia et al . , 2013 ) , the shedding of platelets through the wounded blood vessels could not be beneficial to the adult dentate system like perinatal DG . Duration of Shh exposure time contributes to the proliferative capacity of progenitor cells and the specification of differentiated neurons such as dopaminergic neurons ( Hayes et al . , 2013 ) and olfactory interneurons ( Ihrie et al . , 2011 ) . Shh encapsulated in circulating microparticles secreted from various cell types in circulation were known to regulate vascular homeostasis and inflammation ( Soleti and Martinez , 2009; Puddu et al . , 2010; Angelillo-Scherrer , 2012 ) , and in this study , we further extended the role of Shh in the serum and provide a way for dentate stem cells to get constitutively exposed to Shh in the perivascular niche during stem cell migration and dentate development . This also helps to explain the expansion of migrating dorsal dentate stem cells from the ventral DG in the absence of definitive Shh-expressing cells just after birth ( Li et al . , 2013 ) . It is possible that Shh from platelets could affect not only the DG but also other regions like the cortex but the effect of Shh could be masked by other contents of platelets . Elucidation of the interplay among platelets' contents during formation of the dentate neural stem cell niche will bring new aspects of communication between HSCs and neural stem cells and it will shed light on the therapeutic use of platelets for neurodegenerative vascular diseases . Recent studies showing that youthful systemic circulation rejuvenates aged stem cells in heterochronic-aged parabionts provides a fascinating additional implication of our study ( Conboy et al . , 2005; Villeda et al . , 2011; Conboy and Rando , 2012 ) . The role of vehicles that convey molecular cues into the aged brain stem cell niche may provide means to treat diseased neural stem cells . One such vehicle could be platelets , as shown during perinatal periods in this study . To understand whether aging of hematopoietic niches disturbs contents of platelet's granules , which could exacerbate aged stem cell niches in systemic manner , will give potential insights into molecular cues present in young circulation . Moreover , the most interesting phenomenon of neural stem cells is the maintenance of stemness , so exposing stem cells to consistent supplies of nutrients from various sources such as ventricles , dermal HF niches , and HSC niches could help balance consistent neurogenesis . In pathological conditions in which the cerebrovascular integrity is perturbed such as neuroinflammatory diseases and Alzheimer's diseases ( Bell et al . , 2012; Sengillo et al . , 2012 ) , taking advantages of platelets as a vehicle to transfer druggable materials into the leaky area may become a translational approach . | Although most of the neurons in the brain have been made by the time we are born , new neurons develop throughout life in part of the brain called the hippocampus . These neurons are thought to help with learning and forming memories . Conditions such as depression and Alzheimer's disease have been linked to not being able to produce enough new neurons . The neurons develop from a pool of stem cells in part of the hippocampus . A protein called Sonic Hedgehog ( Shh ) helps to ensure there are enough stem cells and control when they develop into new neurons . The brain cells that produce Shh in adult mice do not appear until a week after birth , by which point the stem cells are already present and generating neurons . This has led scientists to question where these cells get Shh from around the time of birth . One idea is that cells outside of the brain contribute the Shh such as hair follicles—the structures that hairs grow out of—in the scalp . Hair follicles produce Shh , develop at around the same time as the brain stem cells , and are known to regulate the development of other nearby stem cells . So , Choe et al . conducted a series of experiments in genetically engineered newborn mice and found that the brain stem cells multiply at around the same time that the hair follicles start to produce Shh . Furthermore , reducing the amount of Shh produced by the hair follicles hampered the growth of these stem cells and caused fewer neurons to develop from the stem cell pool . These results raised the question of how Shh gets from the hair follicles to the stem cell pool in the developing brain . In adult animals , a barrier exists between the brain and the blood supply to protect the brain from infection . However , parts of this barrier are still leaky before birth , which might allow blood cells to carry Shh to the brain . Cloe et al . found that platelets—the blood cells responsible for clotting—are able to carry Shh to the brain stem cell pool . Further experiments showed that preventing platelets from forming caused fewer stem cells to develop . The suggestion that Shh from the epithelium—the tissue layer that hair follicles are found in—is able to signal to the brain during a specific window of time raises several questions that require further study . Does epithelial Shh also signal to other organs during embryonic or postnatal development ? Does injury to the nervous system that increases the permeability of the blood–brain barrier lead to the delivery of Shh to the brain via the circulation in adult animals ? | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"developmental",
"biology",
"neuroscience"
] | 2015 | Epithelial cells supply Sonic Hedgehog to the perinatal dentate gyrus via transport by platelets |
Heterochromatin , a highly compact chromatin state characterized by histone H3K9 methylation and HP1 protein binding , silences the underlying DNA and influences the expression of neighboring genes . However , the mechanisms that regulate heterochromatin spreading are not well understood . In this study , we show that the conserved Mst2 histone acetyltransferase complex in fission yeast regulates histone turnover at heterochromatin regions to control heterochromatin spreading and prevents ectopic heterochromatin assembly . The combined loss of Mst2 and the JmjC domain protein Epe1 results in uncontrolled heterochromatin spreading and massive ectopic heterochromatin , leading to severe growth defects due to the inactivation of essential genes . Interestingly , these cells quickly recover by accumulating heterochromatin at genes essential for heterochromatin assembly , leading to their reduced expression to restrain heterochromatin spreading . Our studies discover redundant pathways that control heterochromatin spreading and prevent ectopic heterochromatin assembly and reveal a fast epigenetic adaptation response to changes in heterochromatin landscape .
Eukaryotic genomic DNA is folded with histones and non-histone proteins in the form of chromatin , which regulates every aspect of DNA metabolism , including transcription , replication , recombination , and DNA damage repair . Chromatin is classified into euchromatin , which is gene rich and actively transcribed , and heterochromatin , which is gene poor and highly compacted ( Grewal and Jia , 2007 ) . Heterochromatin preferentially forms at repetitive DNA elements in order to limit transcription and recombination at these regions to maintain genome integrity . It also forms at developmentally regulated genes to regulate their expression in response to developmental cues and external stimuli . Heterochromatin tends to spread into neighboring regions , leading to the inactivation of genes in a sequence-independent manner ( Talbert and Henikoff , 2006; Wang et al . , 2014 ) . Therefore , the sites of heterochromatin formation and extent of heterochromatin spreading need to be tightly controlled to prevent improper gene silencing , and misregulation of heterochromatin assembly has been linked to many human diseases , especially various types of cancers ( Geutjes et al . , 2012 ) . Heterochromatic regions generally have distinct chromatin signatures such as histones that are hypoacetylated and methylated at histone H3 lysine 9 ( H3K9me ) , and the enrichment of HP1 family proteins ( Rea et al . , 2000; Bannister et al . , 2001; Lachner et al . , 2001; Nakayama et al . , 2001 ) . Formation of heterochromatin requires the concerted actions of a diverse group of histone-modifying proteins , such as H3K9 methyltransferases and histone deacetylases ( HDACs ) , and is divided into three distinct steps: establishment , spreading , and maintenance ( Grewal and Moazed , 2003; Rusche et al . , 2003 ) . Heterochromatin is established at nucleation centers through the targeting of histone-modifying activities by transcription factors or non-coding RNAs ( Cohen and Jia , 2014 ) . Subsequently , heterochromatin spreads into neighboring regions , mostly via a network of interactions among chromatin proteins , resulting in the formation of large heterochromatin domains independent of the underlying DNA sequences ( Talbert and Henikoff , 2006; Cohen and Jia , 2014 ) . Once these domains are formed , they can maintain themselves also through interactions among chromatin proteins even in the absence of the initiation signal ( Moazed , 2011; Ragunathan et al . , 2014 ) . The formation of heterochromatin has been extensively studied in fission yeast , which uses highly conserved histone-modifying enzymes and chromatin proteins for heterochromatin assembly , such as the SUV39 family histone H3K9 methyltransferase Clr4 , the HP1 homologue Swi6 , and HDACs Sir2 and Clr3 ( Grewal and Jia , 2007 ) . There are four types of heterochromatin identified in fission yeast: constitutive heterochromatin at repeat regions such as centromeres , telomeres , and the silent mating-type region ( Grewal and Jia , 2007 ) ; facultative heterochromatin islands at a subset of meiotic genes ( Hiriart et al . , 2012; Zofall et al . , 2012; Tashiro et al . , 2013; Egan et al . , 2014 ) ; HOODs ( heterochromatin domains ) at sexual differentiation genes and retrotransposons in response to the misregulation of the exosome ( Yamanaka et al . , 2013 ) ; and transient heterochromatin at convergent genes ( Gullerova and Proudfoot , 2008 ) . These locations use distinct pathways to recruit histone-modifying activities to form heterochromatin . The establishment of constitutive heterochromatin at repetitive DNA elements requires the RNA interference ( RNAi ) pathway , a phenomenon also vastly conserved in eukaryotes ( Moazed , 2009; Lejeune and Allshire , 2011; Castel and Martienssen , 2013 ) . The DNA repeats are transcribed and the transcripts are processed by the RNAi machinery into small interfering RNAs ( siRNAs ) , which target the Clr4 complex ( CLRC , consisting of Clr4 , Cul4 , Rik1 , Raf1 , and Raf2 ) to repeat regions to initiate H3K9me . In addition , DNA binding factors , such as telomeric shelterin and stress-activated ATF/CREB family proteins Atf1/Pcr1 , also directly recruit histone-modifying activities to establish constitutive heterochromatin at telomeres and the silent mating-type region , respectively ( Jia et al . , 2004; Kim et al . , 2004; Kanoh et al . , 2005; Tadeo et al . , 2013 ) . The formation of facultative heterochromatin islands at meiotic genes requires RNA binding protein Mmi1 , Zinc finger protein Red1 , and the exosome ( Hiriart et al . , 2012; Zofall et al . , 2012; Tashiro et al . , 2013; Egan et al . , 2014 ) . Mmi1 binds to RNA transcripts containing DSR ( determinant of selective removal ) sequences and recruits the RNA-induced transcriptional silencing ( RITS ) complex and the Red1-Mtl1 complex , which directly interacts with Clr4 complex , to initiate H3K9me at meiotic genes ( Harigaya et al . , 2006; Zofall et al . , 2012; Lee et al . , 2013; Egan et al . , 2014 ) . HOODs are formed at sexual differentiation genes and retrotransposons in response to exosome malfunction or changes in environmental conditions and requires RNAi , polyA polymerase Pla1 , and PolyA binding protein Pab2 ( Lee et al . , 2013; Yamanaka et al . , 2013 ) . Convergent genes generate overlapping transcripts during the G1 phase of the cell cycle , which induce the formation of transient heterochromatin through the RNAi pathway ( Gullerova and Proudfoot , 2008 ) . The spreading of heterochromatin requires Swi6 and the chromodomain of Clr4 , both of which bind to H3K9me and position Clr4 to methylate neighboring nucleosomes ( Hall et al . , 2002; Zhang et al . , 2008; Al-Sady et al . , 2013 ) . The reiteration of H3K9 methylation and recruitment of Clr4 by H3K9me , either directly through the chromodomain or indirectly through Swi6 , results in the ‘inch-worm’-like spreading of heterochromatin from nucleation centers into large chromosome domains ( Talbert and Henikoff , 2006; Wang et al . , 2014 ) . Some heterochromatin regions are flanked by DNA sequences termed boundary elements , which block heterochromatin spreading ( Wang et al . , 2014 ) . In other cases , heterochromatin borders are determined by the local balance of heterochromatin and euchromatin proteins , which tends to differ between cells . Therefore , precise regulation of heterochromatin spreading is essential to maintain stable gene expression profiles . One of the best-known negative regulators of heterochromatin spreading is Epe1 as epe1∆ results in heterochromatin spreading beyond its normal boundaries as well as ectopic heterochromatin formation ( Zofall and Grewal , 2006; Trewick et al . , 2007; Zofall et al . , 2012; Ragunathan et al . , 2014 ) . Loss of Epe1 also bypasses RNAi for pericentric heterochromatin assembly by strengthening heterochromatin spreading ( Trewick et al . , 2007 ) . Epe1 contains a JmjC domain , which is frequently associated with histone demethylase activity . Although no demethylase activity has been detected for Epe1 ( Tsukada et al . , 2006 ) , genetic evidence is consistent with Epe1 being a H3K9 demethylase and conserved catalytic residues are essential for Epe1 function ( Trewick et al . , 2007; Ragunathan et al . , 2014 ) . The Mst2 complex is similar in composition to budding yeast NuA3 and mammalian HBO1/MOZ/MORF complexes ( Wang et al . , 2012 ) . It is a highly specific histone H3K14 acetyltransferase that cooperates with Gcn5 to regulate global H3K14 acetylation levels ( Nugent et al . , 2010; Wang et al . , 2012 ) . The formation of heterochromatin is negatively correlated with H3K14 acetylation ( Sugiyama et al . , 2007; Motamedi et al . , 2008 ) , and mst2∆ bypasses the requirement of the RNAi pathway for pericentric heterochromatin assembly through modulating H3K14ac levels at heterochromatin ( Reddy et al . , 2011 ) . Moreover , mst2∆ strengthens silencing at telomeres ( Gomez et al . , 2005 ) . These results suggest that Mst2 complex functions to antagonize heterochromatic silencing , although the mechanism by which it affects heterochromatin assembly is unknown . The ability to bypass RNAi requires ablating the enzymatic activity of the Mst2 complex ( Reddy et al . , 2011 ) . It was proposed that Mst2-mediated H3K14 acetylation regulates histone turnover at heterochromatin regions and the loss of such activity preserves parental histone modifications to promote heterochromatin maintenance ( Reddy et al . , 2011 ) , although the ability of Mst2 to regulate histone turnover has not been directly tested . In this study , we show that Mst2 regulates histone turnover at heterochromatin regions and that loss of Mst2 results in heterochromatin spreading at telomeres and heterochromatin islands where boundaries are absent . We also found that mst2∆ epe1∆ cells are initially very sick due to heterochromatin spreading-mediated inactivation of essential genes , suggesting that Mst2 and Epe1 function redundantly in regulating heterochromatin spreading . Interestingly , these cells quickly recover by forming ectopic heterochromatin at the clr4+ locus to mitigate the negative effects of heterochromatin . Disrupting heterochromatin assembly at the clr4+ locus results in ectopic heterochromatin formation at the rik1+ locus , which encodes another subunit of the Clr4 complex required for H3K9me . These results demonstrate that promiscuous heterochromatin assembly generates epigenetic mutations that provide fast adaptions to heterochromatin stress .
To directly examine the role of the Mst2 complex in regulating histone turnover , we generated a Flag-tagged version of histone H3 driven by the urg1 promoter at the endogenous urg1+ locus , which can be quickly induced by the addition of uracil into the growth medium at levels far below the endogenous histone H3 ( Watt et al . , 2008 ) ( Figure 1A ) . To prevent replication-dependent histone incorporation , we blocked the cell cycle with hydroxyl urea ( HU ) before induction of H3-Flag expression ( Figure 1B ) . We found that pericentric dh repeat was associated with lower amounts of H3-Flag in wild-type cells compared with RNAi mutant dcr1∆ ( Figure 1C ) , suggesting that histone turnover rates increase when heterochromatin is compromised . In addition , the incorporation of H3-Flag was reduced in epe1∆ dcr1∆ cells , as observed previously ( Figure 1C ) ( Aygun et al . , 2013 ) . In mst2∆ dcr1∆ cells , H3-Flag incorporation was reduced to wild-type levels ( Figure 1C ) , suggesting that the Mst2 complex indeed regulates histone turnover at heterochromatin . 10 . 7554/eLife . 06179 . 003Figure 1 . Mst2 counteracts heterochromatin assembly . ( A ) Western blot analysis of H3-Flag levels . Samples were taken at indicated times after the addition of uracil , and Western blot analyses were performed with Flag and H3 antibodies . ( B ) Schematic diagram of the histone turnover assay . ( C ) Enrichment of H3-Flag at pericentric dh sequence as an indicator of histone turnover rates . The values are normalized to a region within the silent mating-type locus with background histone turnover ( Aygun et al . , 2013 ) . Error bars represent standard deviation of three experiments . ( D ) ChIP–chip analyses of H3K9me2 levels across the genome . ( E ) ChIP–chip data of H3K9me2 levels around the telomere IL , centromere I , the silent mating-type region , and the mei4+ locus . ( F ) Mst2 is not required for boundary function at IRC1R . Serial dilution analysis were performed to measure the expression of IRC1R::ura4+ reporter . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 00310 . 7554/eLife . 06179 . 004Figure 1—figure supplement 1 . ChIP–chip data of H3K9me2 levels around centromere II , centromere III , telomere 1R , telomere 2L , and telomere 2R . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 004 To further examine the role of the Mst2 complex in regulating heterochromatin assembly , we performed Chromatin Immunoprecipitation coupled with DNA microarray ( ChIP–chip ) analyses of H3K9me2 levels across the fission yeast genome . In wild-type cells , H3K9me2 was mainly present at centromeres , telomeres , and the silent mating-type region ( Figure 1D ) . There were also a few heterochromatic islands with low levels of H3K9me2 ( Figure 1D ) . Although less heterochromatic islands were identified compared to a recent study ( Zofall et al . , 2012 ) , our results are consistent with that of an earlier one ( Cam et al . , 2005 ) . The discrepancies might be caused by the use of batches of antibody with different sensitivity or different data processing methods . In mst2∆ cells , constitutive heterochromatin domains at centromeres and the silent mating-type region were in good agreement with wild-type cells , but telomeric heterochromatin showed significant spreading into chromosome arms ( Figure 1D , E and Figure 1—figure supplement 1 ) , consistent with previous findings that mst2∆ strengthens silencing at telomeres ( Gomez et al . , 2005 ) . Interestingly , there are a number of additional small H3K9me2 peaks scattered across the genome , most of which are also present in epe1∆ cells ( Figure 1D , E , and Supplementary file 1 ) ( Zofall et al . , 2012 ) . Therefore , Mst2 also prevents ectopic heterochromatin assembly , similar to Epe1 . We observed only minor heterochromatin spreading in telomeric regions in epe1∆ cells compared with a previous study ( Zofall et al . , 2012 ) , which might be due to the presence of two epigenetically stable subpopulations of cells with different effects on heterochromatin assembly ( Trewick et al . , 2007 ) . The difference between pericentric regions , telomeres , and heterochromatin islands is the presence of well-defined boundary elements at pericentric regions that block heterochromatin spreading ( Wang et al . , 2014 ) . We found that mst2∆ has no effect on boundary activity of an inverted repeat at the pericentric region , IRC1R , which requires Epe1 and the double bromodomain protein Bdf2 for function ( Figure 1F ) ( Wang et al . , 2013 ) , suggesting that Mst2 regulates heterochromatin spreading only in the absence of boundaries . Since mst2∆ and epe1∆ have similar phenotypes in heterochromatin assembly and each bypasses the RNAi pathway for pericentric heterochromatin functions ( Trewick et al . , 2007; Reddy et al . , 2011 ) , we generated mst2∆ epe1∆ cells to examine their epistatic relationship . All freshly generated mst2∆ epe1∆ cells formed very small colonies , suggesting a strong negative genetic interaction between these two mutants ( Figure 2A and Figure 2—figure supplement 1 ) , consistent with high throughput epistasis mapping ( Roguev et al . , 2008; Ryan et al . , 2012 ) . Moreover , abolishing the enzymatic activity of Mst2 ( mst2-E274Q or nto1∆ ) or Epe1 ( epe1-H374A and epe1-Y307A ) resulted in similar sickness ( Figure 2—figure supplement 1 ) , suggesting that the enzymatic activities of Mst2 and Epe1 have redundant functions . Double mutant of mst2∆ bdf2∆ had no defects in growth ( Figure 2—figure supplement 1 ) , suggesting that the boundary activity of Epe1 is not involved in genetic interaction with Mst2 . 10 . 7554/eLife . 06179 . 005Figure 2 . A suppressor mutation confers normal growth of mst2∆ epe1∆ cells . ( A , D , E , H ) Tetrad dissection analysis of the indicated genetic crosses . Pictures are examples of colonies derived from the same tetrad containing all individual genotypes , after one replication for a total of 6 days growth . ( B ) Serial dilution analysis of indicated strains . Cells were grown in rich medium overnight before dilution analyses were performed . ( C ) The growth curve of indicated strains . ( F ) Workflow to introduce mst2∆ and epe1∆ into the deletion library . ( G ) Left , a representative image of colony growth was shown . Middle , colonies were assigned scores between 0 and 3 , as indicated . Right , list of identified heterochromatin mutants that confer fast growth . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 00510 . 7554/eLife . 06179 . 006Figure 2—figure supplement 1 . Tetrad dissection analysis of the indicated genetic crosses . ( A ) Genetic cross between mst2∆ and epe1∆ to show the frequency of mst2∆ epe1∆ colonies ( red boxes ) . Each column is a tetrad . Genotypes are indicated on the right . WT , wild type; m , mst2∆; e , epe1∆; me , mst∆ epe1∆; ? , genotype can not be assigned . Red letters indicate no colony growth detected . ( B ) Representative examples of colonies derived from the same tetrad containing all individual genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 00610 . 7554/eLife . 06179 . 007Figure 2—figure supplement 2 . Tetrad dissection analysis of indicated genetic crosses . Pictures are representative examples of colonies derived from the same tetrad containing all individual genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 007 Interestingly , cells from the small mst2∆ epe1∆ colonies grew comparably to wild-type cells ( Figure 2B , C ) , suggesting that the accumulation of either a genetic or more intriguingly , an epigenetic suppressor leads to quick and persistent production of normally growing mst2∆ epe1∆ cells . We crossed independent clones of recovered mst2∆ epe1∆ cells ( mst2∆ epe1∆* , which denotes the presence of a suppressor ) and found that all resulting progenies grew normally from the beginning ( Figure 2D ) . Moreover , when we crossed mst2∆ epe1∆* cells with wild-type cells , half of the resulting mst2∆ epe1∆ colonies were small and the other half were normal ( Figure 2E ) , suggesting that changes associated with a single genomic locus was responsible for the recovery of these cells and also ruling out the possibility that the initial growth defects were the result of these cells going through meiosis . To identify the suppressor , we performed two successive rounds of crosses to introduce mst2∆ and epe1∆ into the fission yeast deletion library and measured the initial growth of triple mutants before accumulation of a suppressor ( Figure 2F ) . Interestingly , a number of heterochromatin assembly mutants allowed robust growth of mst2∆ epe1∆ cells . Such mutants included deletions of components of the CLRC histone H3K9 methyltransferase complex ( clr4∆ , raf1∆ , and raf2∆ ) , an HP1 protein ( swi6∆ ) , and histone deacetylases ( sir2∆ and clr3∆ ) ( Figure 2G and Supplementary file 2 ) . In contrast , none of the RNAi factors were identified in our screen . Individual crosses also confirmed that heterochromatin mutants conferred normal growth to mst2∆ epe1∆ cells ( Figure 2H and Figure 2—figure supplement 2 ) . Such data suggest that the effect of mst2∆ epe1∆ on cell growth is possibly the result of misregulation of heterochromatin . To determine whether there are any global changes in heterochromatin organization , we decided to perform ChIP–chip analysis of H3K9me2 levels across the genome in mst2∆ epe1∆ cells . However , the quick generation of epigenetic suppressors in mst2∆ epe1∆* cells prevented us from directly testing the reason for the initial growth defects . Because H3K9me functions upstream of Swi6 localization and the silencing function of H3K9me requires Swi6 , we reasoned that examining H3K9me2 levels in mst2∆ epe1∆ swi6∆ cells might show the misregulation of heterochromatin assembly that resembles early stages mst2∆ epe1∆ cells . Indeed , in mst2∆ epe1∆ swi6∆ cells , the H3K9me2 domains at constitutive heterochromatin regions such as centromeres showed significant expansion , even when boundary elements are present ( Figure 3A , B ) . In addition , many additional peaks of H3K9me2 were detected across the genome , at levels comparable to constitutive heterochromatin domains ( Figure 3A , B , Figure 3—figure supplement 1 , and Supplementary file 1 ) . Most , but not all , of these additional sites correspond to previously described heterochromatin islands . Compared to wild-type , mst2∆ , or epe1∆ cells , these heterochromatic islands were also greatly expanded ( Figure 3B and Figure 3—figure supplement 1 ) . Given that Swi6 also contributes to heterochromatin spreading ( Hall et al . , 2002; Al-Sady et al . , 2013 ) , heterochromatin probably spreads over even longer distances when Swi6 is present . Interestingly , essential genes reside within or near some of the expanded H3K9me2 domains ( Figure 3B and Figure 3—figure supplement 1 ) , suggesting that misregulation of heterochromatin spreading inactivates essential genes and causes the initial sickness of mst2∆ epe1∆ cells . 10 . 7554/eLife . 06179 . 008Figure 3 . Increased heterochromatin spreading in mst2∆ epe1∆ cells leads to growth defects . ( A ) ChIP–chip analyses of H3K9me2 levels across the genome . ( B ) ChIP–chip data of H3K9me2 levels at centromere I , mei4+ , mcp7+ , and clr4+ locus . ( C ) Tetrad dissection analysis of the indicated genetic cross . ( D ) ChIP-qPCR analysis of H3K9me2 levels at indicated locations , normalized against act1+ . Error bars represent standard deviation of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 00810 . 7554/eLife . 06179 . 009Figure 3—figure supplement 1 . ChIP–chip data of H3K9me2 levels at heterochromatin islands . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 009 If heterochromatin spreading is the cause of the initial growth defects of mst2∆ epe1∆ cells , we expect that mutations blocking heterochromatin spreading will abolish such an effect . The spreading of heterochromatin requires Swi6 as well as the chromodomain of Clr4 , which binds to pre-existing H3K9me to allow the modification of adjacent nucleosomes ( Hall et al . , 2002; Zhang et al . , 2008; Al-Sady et al . , 2013 ) . Our data that mst2∆ epe1∆ swi6∆ allows normal growth ( Figure 2H ) is consistent with such a hypothesis . However , Swi6 is required for heterochromatin spreading as well as heterochromatin-mediated silencing , making it difficult to definitively assess the contribution of heterochromatin spreading in this process . We therefore tested a mutation within the Clr4 chromodomain , W31G , which affects the binding of Clr4 to H3K9me to block heterochromatin spreading ( Zhang et al . , 2008 ) . Indeed , mst2∆ epe1∆ clr4-W31G cells showed no initial growth defects ( Figure 3C ) and heterochromatin expansion is prevented as indicated by ChIP analysis of H3K9me2 levels outside of centromere I boundary and at the mei4+ locus ( Figure 3D ) . ChIP–chip analysis also showed that the patterns and levels of H3K9me2 in mst2∆ epe1∆* cells were more similar to those in wild-type cells with high levels of H3K9me2 at constitutive heterochromatic regions and low levels of H3K9me2 at heterochromatic islands , and much less heterochromatin spreading compared with mst2∆ epe1∆ swi6∆ cells ( Figure 3A , B ) . Most significantly , H3K9me2 was enriched at a 5 kilobase region covering clr4+ and an adjacent gene meu6+ , at levels comparable to constitutive heterochromatin regions in independent clones tested ( Figures 3A , B , and 4A , and data not shown ) . Consistent with the fact that H3K9me is associated with gene silencing , both clr4+ mRNA and Clr4 protein levels were reduced in mst2∆ epe1∆* cells ( Figure 4B , C ) . 10 . 7554/eLife . 06179 . 010Figure 4 . Inheritance of ectopic heterochromatin at the meu6-clr4 locus . ( A , E ) ChIP-qPCR analysis of H3K9me2 levels at the clr4+ coding region , normalized against act1+ . Error bars represent standard deviation of three experiments . ( B , F ) qRT-PCR analysis of clr4+ mRNA levels , normalized against act1+ . Error bars represent standard deviation of three experiments . ( C ) Western blot analyses of Flag-Clr4 and Tubulin protein levels . ( D ) Tetrad dissection analysis of indicated genetic crosses . ( G ) Serial dilution analysis to measure the expression of otr::ura4+ reporter . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 010 To test if the epigenetically silenced clr4+ locus is the suppressor of mst2∆ epe1∆* cells , we crossed mst2∆ epe1∆* ( containing a methylated clr4+ locus ) with Flag-clr4+ cells ( containing an unmethylated Flag-clr4+ locus ) . We found that the resulting mst2∆ epe1∆ clr4+ daughters ( inheriting the methylated clr4+ locus ) grew normally , whereas the mst2∆ epe1∆ Flag-clr4+ daughters ( inheriting the unmethylated Flag-clr4+ locus ) were sick ( Figure 4D ) . Moreover , ChIP analysis showed that the mst2∆ epe1∆ clr4+ progeny also inherited H3K9me2 associated with this locus ( Figure 4E ) , and clr4+ mRNA levels were low ( Figure 4F ) . Therefore , the inheritance of a silenced clr4+ locus allows cells to avoid the negative effects on cell growth imposed by mst2∆ epe1∆ . In contrast , wild-type cells inheriting the methylated clr4+ locus lost H3K9me2 , clr4+ mRNA levels were significantly restored , and these cells exhibited no defects in the silencing of a pericentric otr::ura4+ reporter gene ( Figure 4E , F , G ) . These results suggest that the silencing of clr4+ is epigenetic and not due to changes in DNA sequence . They also suggest that the continued absence of Mst2 and Epe1 is required to maintain H3K9me2 at the clr4+ locus . We then examined whether any of the known heterochromatin assembly pathways are required for heterochromatin assembly at the clr4+ locus in mst2∆ epe1∆ cells . We found that mst2∆ epe1∆ dcr1∆ cells also quickly recovered and H3K9me2 levels were similar at the clr4+ locus in mst2∆ epe1∆* and mst2∆ epe1∆ dcr1∆* cells , suggesting that RNAi is not required for heterochromatin assembly at clr4+ , even though clr4+ is in a convergent orientation with meu6+ ( Figure 5—figure supplement 1 ) . In addition , H3K9me2 levels persisted in mst2∆ epe1∆ mmi1∆* and mst2∆ epe1∆ pab2∆* cells ( Figure 5—figure supplement 1 ) , suggesting that heterochromatin assembly is not through Mmi1-mediated facultative heterochromatin assembly pathway or Pab2-mediated assembly of HOOD , even though meu6+ is a meiotic gene . Therefore , heterochromatin assembly at clr4+ differs from known heterochromatin assembly pathways . Due to the severe growth defects associated with red1∆ or rrp6∆ , we were unable to generate triple mutant strains with mst2∆ epe1∆ and whether these factors are involved in H3K9me2 at the clr4+ locus in mst2∆ epe1∆ cells is unknown . The domain of H3K9me2 in mst2∆ epe1∆* cells includes clr4+ and meu6+ , with its center within meu6+ coding region . Interestingly , RNA sequencing analysis showed that meu6+ is not expressed in vegetative growing cells and clr4+ transcript runs through the entire meu6+ gene ( Figure 5A ) ( Wang et al . , 2015 ) . We therefore replaced the entire meu6+ open reading frame with a kanMX6 cassette , in the same transcription orientation as meu6+ ( Figure 5B ) . Given that meu6+ is right next to clr4+ , we first examined whether this manipulation affects Clr4 function . We found that meu6∆::kanMX6 has no silencing defects at otr::ura4+ , and clr4+ mRNA and Clr4 protein levels were similar to those of wild-type cells ( Figure 5—figure supplement 2 ) . ChIP analysis showed that H3K9me2 was abolished at the clr4+ locus in mst2∆ epe1∆ meu6∆* cells ( Figure 5C ) , and clr4+ mRNA and Clr4 protein levels were similar to wild-type cells ( Figure 5D , E ) . We also generated a meu6-Flag::KanMX6 strain , which preserves the coding sequence of meu6+ , and found that H3K9me2 levels at the clr4+ locus were also abolished in mst2∆ epe1∆ meu6-Flag::KanMX6 cells ( Figure 5—figure supplement 3 ) , suggesting that the meu6+ coding sequence is not responsible for heterochromatin assembly at this locus , but rather the insertion of the KanMX6 cassette disrupts a heterochromatin initiation signal . Given that the clr4+ transcript has a long 3′-UTR and that both meu6∆::KanMX6 and meu6-Flag::KanMX6 alter the 3′-UTR , heterochromatin assembly at the clr4+ locus likely requires the intact 3′-UTR of clr4+ . 10 . 7554/eLife . 06179 . 011Figure 5 . Blocking heterochromatin formation at the clr4+ locus in mst2∆ epe1∆ cells results in ectopic heterochromatin assembly at the rik1+ locus . ( A ) RNA sequencing data of the meu6-clr4 region . ( B ) Schematic diagram of the meu6∆::kanMX6 construct . ( C , I ) ChIP-qPCR analysis of H3K9me2 levels at the clr4+ or rik1+ coding region , normalized against act1+ . Error bars represent standard deviation of three experiments . ( D , J ) qRT-PCR analysis of clr4+ or rik1+ mRNA levels , normalized against act1+ . Error bars represent standard deviation of three experiments . ( E ) Western blot analyses of Flag-Clr4 protein levels . ( F ) The growth curve of indicated strains . ( G ) ChIP–chip analyses of H3K9me2 levels across the genome in recovered mst2∆ epe1∆ meu6∆* cells . ( H ) ChIP–chip data of H3K9me2 levels at clr4+ and rik1+ loci . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 01110 . 7554/eLife . 06179 . 012Figure 5—figure supplement 1 . Dcr1 , Mmi1 and Pab2 are not required for heterochromatin assembly at the clr4+ locus in mst2Δ epe1Δ* cells . ( A ) Serial dilution analysis of indicated strains . Cells were grown in rich medium overnight before dilution analyses were performed . ( B ) The growth curve of indicated strains . ( C ) ChIP-qPCR analysis of H3K9me2 levels at the clr4+ coding region , normalized against act1+ . Error bars represent standard deviation of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 01210 . 7554/eLife . 06179 . 013Figure 5—figure supplement 2 . meu6∆ has no effect on clr4+ expression under normal conditions . ( A ) Serial dilution analysis was performed to measure the expression of otr::ura4+ . ( B ) qRT-PCR analysis of clr4+ mRNA levels , normalized against act1+ mRNA . Error bars represent standard deviation of three experiments . ( C ) Western blot analysis to measure Flag-Clr4 protein levels . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 01310 . 7554/eLife . 06179 . 014Figure 5—figure supplement 3 . Meu6-Flag abolished H3K9me in mst2∆ epe1∆ cells . ( A ) Schematic diagram of meu6-Flag . ( B ) ChIP-qPCR analysis of H3K9me2 levels at the clr4+ and rik1+ coding region , normalized against act1+ . Error bars represent standard deviation of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 014 If the silenced clr4+ is the suppressor in mst2∆ epe1∆* cells , we expect that abolishing H3K9me2 at the clr4+ locus will result in the failure of mst2∆ epe1∆ cells to recover . Interestingly , mst2∆ epe1∆ meu6∆* cells were able to recover to some extent although they grew at a slower rate compared to wild-type or mst2∆ epe1∆* cells ( Figure 5F ) . To understand how mst2∆ epe1∆ meu6∆* cells recover without silencing of clr4+ , we performed ChIP–chip analysis of H3K9me2 levels in mst2∆ epe1∆ meu6∆* cells ( Figure 5G ) . The distribution of H3K9me2 is on the whole similar to that of mst2∆ epe1∆* cells , with two major exceptions . First , H3K9me2 was indeed abolished from the entire meu6+-clr4+ region ( Figure 5G , H ) . Second , an additional peak of H3K9me2 appeared at a 7-kb region that includes two genes in convergent orientation: rik1+ , which encodes a component of the Clr4 complex and is required for Clr4 function ( Nakayama et al . , 2001; Hong et al . , 2005; Horn et al . , 2005; Jia et al . , 2005 ) , and an uncharacterized gene SPCC11E10 . 09c+ ( Figure 5G , H , I ) . As expected , rik1+ mRNA levels were significantly reduced in mst2∆ epe1∆ meu6∆* cells compared to wild-type or mst2∆ epe1∆* cells ( Figure 5J ) . It is likely that the reduction of rik1+ expression allows mst2∆ epe1∆ meu6∆* cells to grow to some extent by decreasing global H3K9me and heterochromatin assembly . We suspect that mst2∆ epe1∆ meu6∆* cells recovered less well than mst2∆ epe1∆* cells because heterochromatin forms less efficiently at the rik1+ locus , which might explain why independent mst2∆ epe1∆* clones all preferentially silenced clr4+ ( data not shown ) . Moreover , H3K9me2 was also enriched at the rik1+ locus in mst2∆ epe1∆ meu6-Flag* cells ( Figure 5—figure supplement 3 ) , suggesting that the formation of heterochromatin at this locus is not random . It is interesting to note that the enzymatic activity of Clr4 is much higher than that of its mammalian counterparts ( Rea et al . , 2000 ) . Therefore , Mst2 and Epe1 are likely evolved to counteract such a hyperactive H3K9 methyltransferase . Clr4 has an arginine at residue 406 , which corresponds to a histidine in its mammalian and Drosophila homologues ( Figure 6A ) . Conversion of the histidine to arginine ( H320R ) makes mammalian SUV39H1 a hyperactive histone methyltransferase ( Rea et al . , 2000 ) . We generated a R406H mutation in Clr4 and found that this mutation resulted in a drastic reduction of Clr4 enzymatic activity in vitro ( Figure 6B ) . When introduced into the endogenous clr4+ locus , R406H moderately affected silencing of a pericentric otr::ura4+ reporter , indicating that high activity of Clr4 is required for heterochromatin assembly in wild-type cells ( Figure 6—figure supplement 1 ) . Interestingly , mst2∆ epe1∆ clr4-R406H had no initial growth defects ( Figure 6C ) , no H3K9me2 at the clr4+ locus ( Figure 6D ) , and clr4+ transcript levels were not affected ( Figure 6E ) . Therefore , a less active Clr4 can also mitigate the effects of simultaneous loss of Mst2 and Epe1 . 10 . 7554/eLife . 06179 . 015Figure 6 . The high activity of Clr4 leads to growth defects of mst2∆ epe1∆ cells . ( A ) Sequence alignment of part of Clr4 homologues . * indicates R406 of Clr4 . ( B ) In vitro histone methyltransferase assays were performed with recombinant GST-Clr4 or GST-Clr4-R406H together with a histone H3 ( 1–21 ) peptide . For the control reaction , no Clr4 was added . ( C ) Tetrad dissection analysis of the indicated genetic cross . ( D ) ChIP-qPCR analysis of H3K9me2 levels at the clr4+ coding region , normalized against act1+ . Error bars represent standard deviation of three experiments . ( E ) qRT-PCR analysis of clr4+ mRNA levels , normalized against act1+ . Error bars represent standard deviation of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 01510 . 7554/eLife . 06179 . 016Figure 6—figure supplement 1 . The effect of clr4-R406H on silencing of otr::ura4+ . Right , qRT-PCR analysis of ura4+ RNA levels , normalized against act1+ . Error bars represent standard deviation of three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 016
The formation of heterochromatin and its subsequent spreading result in silencing of large chromosomal domains in a sequence-independent manner . Therefore , the sites of heterochromatin assembly and the extent of heterochromatin spreading are generally precisely controlled to maintain stable gene expression patterns . In addition to the diverse pathways that accurately initiate heterochromatin assembly , anti-silencing activities also play essential roles in limiting heterochromatin spreading to shape the chromatin landscape . Our results reveal a novel function of the Mst2 complex in regulating histone turnover at heterochromatin regions to counteract heterochromatin spreading . Loss of Mst2 bypasses the requirement of RNAi for pericentric heterochromatin assembly ( Reddy et al . , 2011 ) , increases heterochromatin spreading and silencing at telomeres ( Figure 1 ) ( Gomez et al . , 2005 ) , and increases the efficiency of ectopic heterochromatin assembly ( Ragunathan et al . , 2014 ) , which is phenotypically very similar to epe1∆ . Biochemically , the Mst2 complex is a highly specific histone H3K14 acetyltransferase and mutations resulting in the loss of its enzymatic activity , such as mst2-E274Q or nto1∆ ( Wang et al . , 2012 ) , also resulted in slow growth when combined with epe1∆ , suggesting that its enzymatic activity is required for counteracting silencing . H3K14 mutants have a direct effect on heterochromatin assembly independent of its acetylation state , making it difficult to directly address whether H3K14 is the only target of Mst2 in regulating heterochromatin spreading ( Mellone et al . , 2003; Reddy et al . , 2011; Alper et al . , 2013 ) . Therefore , it remains possible that Mst2 modifies heterochromatin assembly factors to regulate histone turnover and counteract silencing . Our results also revealed the functional redundancy of Mst2 and Epe1 in regulating heterochromatin spreading , which explains why heterochromatin spreading only occurs in a small population of cells and requires the overexpression of Swi6 to be efficiently detected ( Noma et al . , 2006; Wang et al . , 2013 ) . In the absence of both Mst2 and Epe1 , heterochromatin spreading increases significantly , leading to the inactivation of essential genes and severe growth defects . Such strong survival pressure results in the selection of cells that can establish heterochromatin at the clr4+ locus , leading to reduced transcription of clr4+ and decreased Clr4 protein levels , thus allowing cells to reach a new equilibrium where heterochromatin assembly at regular locations is intact while the negative effects of heterochromatin spreading are mitigated ( Figure 7 ) . Although we cannot test the epigenetic profiles of individual cells , it is possible that mst2Δ epe1Δ cells initially generate varied epigenetic profiles , and only cells containing H3K9me at the meu6-clr4 locus are clonally selected due to its beneficial effects on cell growth . The quick generation of this epigenetic suppressor also benefited from the stabilization of ectopic heterochromatin domains in mst2∆ epe1∆ cells . Once established , such ectopic heterochromatin can be inherited , but can also be quickly erased to allow cells to adapt to new conditions . Interestingly , when H3K9me at the clr4+ locus is blocked , the survival pressure instead selects cells that can establish heterochromatin at the rik1+ locus to similarly reduce heterochromatin-forming abilities . The flexibility in heterochromatin assembly allows cells and their subsequent generations to efficiently cope with changes in heterochromatin levels . 10 . 7554/eLife . 06179 . 017Figure 7 . A model of the negative feedback of heterochromatin assembly . DOI: http://dx . doi . org/10 . 7554/eLife . 06179 . 017 The formation of heterochromatin at the clr4+ locus does not employ any of the known heterochromatin assembly pathways . The intact 3’-UTR region of clr4+ seems to initiate heterochromatin through a novel mechanism , although such a mechanism must be very inefficient in wild-type cells given that no H3K9me is observed . It is interesting to note that the 3′-UTR of clr4+ overlaps the entire meu6+ gene , which is arranged in a convergent orientation . The expression of meu6+ is extremely low during vegetative growth , but is upregulated during meiosis , which might interfere with Clr4 expression , resulting in the loss of heterochromatin islands as seen in cells under conditions that induce meiosis ( Zofall et al . , 2012 ) . It has long been known that heterochromatin assembly in fission yeast is tightly regulated to prevent promiscuous heterochromatin assembly . For example , the amount of heterochromatin proteins such as Swi6 is limiting , and ectopic heterochromatin assembly can only succeed when Swi6 is overexpressed or when endogenous heterochromatin structures are compromised to release silencing proteins ( Iida et al . , 2008; Kagansky et al . , 2009; Tadeo et al . , 2013 ) . Moreover , sub-telomeric heterochromatin regions , which contain no boundaries and no essential genes , can serve as ‘sinks’ to absorb extra heterochromatin proteins ( Tadeo et al . , 2013 ) . These extensive pathways that limit heterochromatin assembly might be a response to the high enzymatic activity of Clr4 , which is likely required for heterochromatin assembly in an otherwise highly compact and active genome . Our results provide an additional layer of control for cells to monitor heterochromatin levels through a negative feedback mechanism that uses the potentially promiscuous nature of heterochromatin assembly to induce heterochromatin formation at genes encoding heterochromatin assembly factors , thus ensuring the epigenetic stability of the genome . Our results fit into a growing body of evidence demonstrating that epigenetic regulation of gene expression enables cells to adopt a wide variety of phenotypes to adapt to external or internal stresses ( Heard and Martienssen , 2014 ) . Compared to genetic mutations , epigenetic mutations provide much faster responses . Most importantly , the effects are reversible , allowing easy reversion to normal epigenetic profiles when external stimuli disappear . In cancer cells , such epigenetic variations might result in the inactivation of tumor suppressor genes during tumorigenesis and might also enable tumor cells to survive certain therapies ( Sharma et al . , 2010; Kreso et al . , 2013 ) . Therefore , our work sheds light on the mechanisms underlying how a relatively stable heterochromatic profile is maintained both under normal conditions and upon heterochromatin stress and will guide future efforts to combat epigenetic adaptations that interfere with cancer treatment .
Detailed genotypes of strains used are listed in Supplementary file 3 . Strains containing meu6∆ or meu6-Flag and urg1-hht2-Flag were constructed by a PCR-based module method . Genetic crosses were used to construct all other strains . For serial dilution plating assays , ten-fold dilutions of mid-log-phase cultures were plated on the indicated medium and grown for 3 days at 30°C . A two-step cross scheme was employed to avoid the accumulation of suppressors before colony growth measurement . Query strains ( mst2∆::natMX6 and epe1∆::hphMX6 ) were first separately mated with the fission yeast deletion library arrayed in 384 strains/plate format with the aid of the Singer RoToR HDA pinning robot as previously described ( Tadeo et al . , 2013 ) . After mating and selection , the resulting haploid double mutant libraries containing individual gene deletions with either mst2∆ or epe1∆ were mated again , and haploid triple mutants were selected on YES medium supplemented with antibiotics to measure cell growth . ChIP analyses with H3K9me2 antibody ( Abcam , Cambridge , MA ) were performed as described previously ( Tadeo et al . , 2013 ) . Quantitative real-time PCR ( qPCR ) was performed with Maxima SYBR Green qPCR Master Mix ( Fermentas , Grand Island , NY ) in a StepOne Plus Real-Time PCR System ( Applied Biosystems , Grand Island , NY ) . DNA serial dilutions were used as templates to generate a standard curve of amplification for each pair of primers , and the relative concentration of the target sequence was calculated accordingly . An act1 fragment was used as reference to calculate the enrichment of ChIP over WCE for each target sequence . Oligos used are listed in Supplementary file 4 . For histone turnover assay , cells were cultured in EMM–uracil medium , and then arrested for 4 hr by 20 mm HU , followed by the addition of 0 . 25 mg/ml uracil to induce the expression of H3-Flag , before ChIP analysis was performed . ChIP–chip analyses were performed according to the ‘Agilent Yeast ChIP-on-chip Analysis’ protocol . The microarray used was an Agilent S . pombe Whole Genome ChIP-on-chip Microarray with additional probes that encompass centromeres , which were originally absent from the array due to the repetitive nature of these DNA sequences . At least two repeats were performed for each microarray experiment . To control for the experimental variation , the average of top 20 probes was set to 1 before averaging the results . For heterochromatin islands , the cutoff of H3K9me2 levels is 0 . 2 . Microarray data have been deposited in the GEO database under accession number GSE60521 . Total cellular RNA was isolated from log-phase cells using MasterPure yeast RNA purification kit ( Epicentre , Madison , WI ) according to the manufacturer's protocol . Quantification with qRT-PCR was performed with Power SYBR Green RNA-to-CT one-step Kit ( Applied Biosystems ) . RNA serial dilutions were used as templates to generate the standard curve of amplification for each pair of primers , and the relative concentration of target sequence was calculated accordingly . An act1 fragment served as reference to normalize the concentration of samples . The concentration of each target gene in wild type was arbitrarily set to 1 and served as references for other samples . Oligos used are listed in Supplementary file 4 . | The DNA in the nucleus of a cell is wrapped around histone proteins to form a compact structure known as chromatin . Chromatin's structure can control how the genes in DNA are expressed . Loosely packed chromatin contains active genes , whereas densely packed chromatin ( also called ‘heterochromatin’ ) contains silenced genes that are not expressed . The assembly of DNA into heterochromatin needs to be carefully controlled . Otherwise , the DNA next to heterochromatin regions can become densely packed as well ( via a process called ‘heterochromatin spreading’ ) , and the genes within this DNA are incorrectly silenced . Incorrect gene silencing is often associated with diseases such as cancer . Cells add chemical groups onto the histone proteins to influence how chromatin is compacted . Densely packed chromatin contains histones with many methyl groups but few acetyl groups . A protein called Epe1 , which potentially removes methyl groups , helps to prevent heterochromatin spreading in yeast cells . Wang et al . found that an enzyme called Mst2 , which adds acetyl groups onto histones , also limits heterochromatin spreading and prevents extra heterochromatin from assembling at undesirable locations . Wang et al . then generated yeast cells that lacked both Epe1 and Mst2 . At first , these cells were sickly and unable to grow , because several essential genes were incorrectly silenced due to rampant heterochromatin spreading . However , the cells quickly overcame this growth defect by gaining an additional mutation . Normally mutations occur through changes in DNA sequences . However , Wang et al . found that the cells acquired this mutation by packing a gene required for heterochromatin assembly into heterochromatin . This in turn stopped more chromatin from becoming packed too densely . Changes to chromatin can also be passed on to the yeast's offspring , and such a change could help the offspring to better cope with changes in heterochromatin levels . Future work could test how often inheritable changes to chromatin modification help organisms adapt to environmental stresses , or if similar changes allow cancer cells to become tolerant to anticancer drugs . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2015 | Rapid epigenetic adaptation to uncontrolled heterochromatin spreading |
During Drosophila embryonic nervous system development , neuroblasts express a programmed cascade of five temporal transcription factors that govern the identity of cells generated at different time-points . However , these five temporal genes fall short of accounting for the many distinct cell types generated in large lineages . Here , we find that the late temporal gene castor sub-divides its large window in neuroblast 5–6 by simultaneously activating two cell fate determination cascades and a sub-temporal regulatory program . The sub-temporal program acts both upon itself and upon the determination cascades to diversify the castor window . Surprisingly , the early temporal gene Kruppel acts as one of the sub-temporal genes within the late castor window . Intriguingly , while the temporal gene castor activates the two determination cascades and the sub-temporal program , spatial cues controlling cell fate in the latter part of the 5–6 lineage exclusively act upon the determination cascades .
During nervous system development , neural progenitor cells often undergo stereotyped changes in their competence , evident by the sequential and programmed generation of distinct neuronal and glial sub-types ( Okano and Temple , 2009; Pearson and Doe , 2004 ) . In the Drosophila embryonic central nervous system ( CNS ) , neuroblasts ( NBs ) sequentially expresses the transcription factors , Hunchback ( Hb ) > Kruppel ( Kr ) > POU-homeodomain factors Nubbin and Pdm2 ( Pdm ) > Castor ( Cas ) > Grainy head ( Grh ) ( Baumgardt et al . , 2009; Brody and Odenwald , 2000; Isshiki et al . , 2001; Novotny et al . , 2002 ) . These factors temporally alter NB competence to determine the types of neurons and glia born at each step of lineage progression ( Kohwi and Doe , 2013; Li et al . , 2013 ) . However , because Drosophila NB lineages can generate an array of different cell types , the instructive capacity of five temporal genes falls short of explaining the diversity observed ( Baumgardt et al . , 2009; Tsuji et al . , 2008 ) . Studies suggest that this regulatory challenge is solved by the activity of the so-called sub-temporal genes , which act in cascades downstream of the temporal genes , do not feedback on the temporal genes , and play a role in sub-dividing larger temporal competence windows ( Baumgardt et al . , 2009; Benito-Sipos et al . , 2011 ) . Downstream of temporal cues , the specification of cell fate is subsequently controlled by determination genes , referred to as terminal selector genes , that activate repertoire ( s ) of terminal cell fate genes e . g . , neurotransmitters and ion channels ( Hobert , 2008; Wenick and Hobert , 2004 ) . The terminal selectors have been found to often act in combinatorial codes to dictate final and unique cell fate ( Allan and Thor , 2015; Baumgardt et al . , 2007; Enriquez et al . , 2015; Sharma et al . , 1998; Thor et al . , 1999 ) . In addition , terminal selectors may act in cascades denoted coherent feedforward loops ( FFLs ) ( Mangan and Alon , 2003; Mangan et al . , 2003 ) . FFLs are common in E . coli and yeast gene regulatory networks ( Alon , 2007 ) , but have also been identified in animals , including in both Drosophila and C . elegans ( Baumgardt et al . , 2009; Baumgardt et al . , 2007; Etchberger et al . , 2009; Johnston et al . , 2006 ) . However , how temporal and sub-temporal genes intersect with terminal selector FFLs to dictate cell fate is poorly understood . The Apterous ( Ap ) neurons of the Drosophila ventral nerve cord ( VNC ) constitute a group of interneurons expressing the LIM-HD factor Apterous ( Ap ) ( Lundgren et al . , 1995 ) . Because of a multitude of antibody markers and genetic tools available for Ap neurons , these cells have been subject to a number of studies of cell fate specification . Ap neurons can be subdivided into; ( 1 ) dorsal Ap neurons ( dAp ) that are a dorsal bi-lateral row of Ap neurons generated in abdominal and thoracic segments by NB4-3 , and ( 2 ) the Ap cluster that are a bi-lateral group of four Ap neurons , denoted Tv1-Tv4 , that are generated consecutively by NB5-6T in thoracic segments ( Figure 1 ) ( Baumgardt et al . , 2007; Gabilondo et al . , 2016; Park et al . , 2004 ) . Two out of four Ap cluster cells have a neuropeptidergic cell fate; the Tv1/Nplp1 and Tv4/FMRFa cells ( Baumgardt et al . , 2007; Benveniste et al . , 1998; Park et al . , 2004 ) , while Tv2 and Tv3 are Ap interneurons . All four cells express Ap and the transcriptional co-factor Eyes absent ( Eya ) ( Miguel-Aliaga et al . , 2004 ) . Two related terminal selector FFLs operate in Ap cluster cells to dictate Nplp1 or FMRFa cell fate , col>ap/eya>dimm>Nplp1 and ap/eya/dac>dimm/BMP>FMRFa ( Allan et al . , 2005 , 2003; Baumgardt et al . , 2007; Miguel-Aliaga et al . , 2004 ) . Each cell type-specific FFL cascade is triggered by specific temporal and spatial inputs established during lineage progression . The spatial input , conferred by body position , consists of the combinatorial action of the Hox homeotic gene Antennapedia ( Antp ) , the Hox co-factors extradenticle ( exd ) and homothorax ( hth ) , as well as the homeobox gene ladybird early ( lbe ) ( Gabilondo et al . , 2016; Karlsson et al . , 2010 ) . The temporal input is mediated by Cas and Grh; the last two factors in the Hb>Kr>Pdm>Cas>Grh temporal cascade ( Baumgardt et al . , 2009 ) . Initially , these spatial and temporal inputs activate the two terminal FFL cascades in all four Ap cluster neurons . However , Cas triggers several additional genes in the neuroblast NB5-6T that act to sub-divide the broader Cas window . These genes were denoted sub-temporal genes , and consist of squeeze ( sqz ) , nab and seven up ( svp ) , which act to repress collier ( col; Flybase knot ) in the generic Tv2/3 neurons and the Tv4 cell ( Baumgardt et al . , 2009; Benito-Sipos et al . , 2011 ) . The repression of col in the Tv2/3 and Tv4 neurons prevents those cells from being specified into Tv1/Nplp1 neurons . However , in spite of the identification of the three sub-temporal genes sqz , nab and svp , the precision of Ap cluster cell specification clearly indicated the existence of additional players . 10 . 7554/eLife . 19311 . 003Figure 1 . Kr affects Nplp1 expression in Tv1 cells . ( A–B ) Whole VNCs of control and Kr mutants , at AFT , reveal loss of Nplp1 expression in the dAp cells , but also in the Tv1 cells . ( C–D ) Ap cell clusters at AFT , showing an expression of Eya , Dimm , FMRFa and Nplp1 , in control ( C ) and Kr mutants ( D ) . In mutants , while Eya is normally expressed in four cells , Dimm and Nplp1 expression is lost in the Tv1 cell . Dimm and FMRFa expression in the Tv4 cell is not affected in Kr mutants compared to control . ( E–F ) Nab and Col expression , in the NB5-6T at St14 , is similar in control and Kr mutants . ( G–H ) In Tv1 cells ( dashed circles ) at St14 ( distinguishable from Tv2 cells by positive Eya but negative Nab expression ) Col expression is similar in control and Kr mutants . ( I–J ) At late St16 , while Eya is un-affected , Col and Dimm expression is lost in the Tv1 cell in Kr mutants . ( K–L ) While Dac is usually expressed in three out of four cells in the Ap clusters at AFT , we find that in Kr mutants Dac expression is observed in all four Ap cluster cells . ( M ) Dac is significantly upregulated in Kr mutants when compared to control ( ***p<0 . 001 , nctrl = 7 clusters , nKr = 10 clusters , Chi-square test ) . ( N–Q ) Timeline of Kr expression in control shows late onset of Kr expression specifically in the Tv1 cell . ( N ) Kr is not expressed in the NB5-6T or ( O ) the Tv1 cell at St14 . ( P ) At St16 , Kr expression commences in Tv1 ( dashed circle ) , which is characterized by maintenance of strong Col expression . ( Q ) Variably , Kr expression in Tv1 is maintained into AFT , identifiable by co-staining for Eya and Nplp1 . ( R ) Model of NB5-6T , showing the early Kr expression commencing at St9 and persisting until St11 , together with the Eya positive postmitotic Ap cluster cells , out of which Tv1 starts to express Kr at St16 , to ultimately specify into the Nplp1 positive Tv1 cells . Genotypes: ( A , C , E , H , J , L , O , P , Q ) OregonR . ( B , D , F , I , K , M ) Kr1 , KrCD . ( N ) lbe ( K ) -EGFP . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 003 In a recent study , we identified Kr as being important to trigger a final cell fate specification cascade in the neuropeptidergic dAp neurons . We found that Kr acts in a predicted and typical early temporal role in the NB4-3 lineage , to initiate the final dAp specification ( Gabilondo et al . , 2016 ) . However , upon further analysis of Kr mutants we found that the neuropeptidergic cell fate of the Nplp1 cell in the Ap cluster generated by NB5-6T is also lost . Here , we find that Kr mutants show a specific loss of Col , Dimm and Nplp1 expression in the Tv1 cell , while specification of the other three Ap cluster neurons is unaffected . We find that Kr is expressed in a second phase in the NB5-6T lineage; postmitotically in the Tv1/Nplp1 cell . The late temporal gene cas activates Kr in Tv1 , while the sub-temporal genes sqz and nab , which are also activated by cas , repress Kr . In contrast , Kr does not repress sqz or nab , but rather acts to suppress svp in the Tv1 cell , thereby preventing svp from repressing Col and Dimm . Hence , an initial generic Ap cluster cell fate , triggered by cas and spatial cues , is sub-divided by four sub-temporal genes , also activated by cas , that act on each other and on the two terminal selector FFLs . This study reveals that the sub-division of a broader temporal competence window is controlled by an elaborate sub-temporal regulatory program triggered by one and the same temporal factor , which also triggers the terminal selector FFLs . It furthermore demonstrates that a member of the canonical Drosophila temporal cascade can act at two stages of a lineage . Finally , it shows that sub-temporal genes can act both on the terminal selector FFLs and on each other .
We recently identified Kruppel ( Kr ) as being important for the development of the NB4-3 lineage , and for specification of the dAp/Nplp1 neurons generated in this lineage ( Gabilondo et al . , 2016 ) . However , upon analyzing Kr mutants we noted that Kr not only affects Nplp1 expression in dAp cells , but also in the Tv1 cells of the Ap cluster ( Figure 1A–D ) . To begin addressing the role of Kr in Tv1 specification , we analyzed the expression of several Tv1 specification factors: Collier ( Col; an EBF/COE factor ) , Ap ( LIM-HD ) , Eyes absent ( Eya; nuclear phosphatase ) and Dimmed ( Dimm; bHLH ) ( Allan et al . , 2005; Baumgardt et al . , 2007; Hewes et al . , 2003; Miguel-Aliaga et al . , 2004; Park et al . , 2004 ) . While Ap and Eya were unaffected , Dimm expression was lost from Tv1 ( Figure 1C–D ) . In contrast , Dimm and FMRFa expression in the Ap4 cell was unaffected ( Figure 1C–D ) . With regards to Col , expression was unaffected at the onset of expression in the NB and in Tv1/Ap1 ( Figure 1E–H ) . However , by stage 16 ( St16 ) we noted that Col was lost from the Tv1 cell ( Figure 1I–J ) . We also analyzed expression of the transcriptional co-factor Dachshund ( Dac ) , important postmitotically for Tv4/FMRFa fate ( Miguel-Aliaga et al . , 2004 ) . Dac is normally expressed in Tv2 , Tv3 and Tv4 , but in Kr mutants Dac expression expands into the Tv1 cell ( Figure 1K–L ) . The sub-temporal factor Nab ( Baumgardt et al . , 2009 ) was not affected in Kr mutants ( Figure 1E–H ) . Kr acts at an early stage in the canonical temporal gene cascade ( Isshiki et al . , 2001; Kohwi and Doe , 2013 ) , and was found to be expressed during St9-St11 in NB5-6T ( Baumgardt et al . , 2009 ) ( Figure 1R ) . We were therefore intrigued by the late and restricted specification phenotype in the Tv1 cells observed in Kr mutants . These effects prompted us to analyze the expression of Kr in the NB5-6T neuroblast at St14 and in the Ap cluster at St14 onward to late embryonic stage ( air-filled trachea stage; AFT ) . In agreement with previous Kr expression analysis ( Baumgardt et al . , 2009 ) , we did not detect Kr expression in the NB5-6T or in the Tv1 cell at St14 ( Figure 1N–O ) . However , by St16 Kr expression was evident in the Tv1 cell ( Figure 1P ) . With some variability , Kr expression is maintained into stage AFT , and overlaps with the onset of Nplp1 expression ( Figure 1Q ) . In summary , Kr shows two phases of expression in the NB5-6T lineage; an early phase , in line with its known role as an early temporal factor , and a postmitotic phase in the Tv1 neuron ( Figure 1R ) . We observe highly selective effects on the Ap cluster specification in Kr mutants , specifically affecting the Tv1 neuron; normal Nab and Col expression in the NB , normal Eya and Ap expression in the Ap cluster cells , ectopic Dac expression in Tv1 , subsequent loss of Col expression in Tv1 , and failure to turn on Dimm and Nplp1 expression in Tv1 , while Dimm and FMRFa expression in Tv4 is unaffected . The selective expression of Kr in the postmitotic Tv1 neuron , and the importance of Kr for Tv1 specification , prompted us to ectopically express Kr , by crossing UAS-Kr to the late Ap neuron specific driver ap-Gal4 ( Allan et al . , 2005 ) . Misexpression of Kr in all four postmitotic Ap neurons resulted in the ectopic expression of Col , Dimm and Nplp1 in other Tv neurons ( Figure 2A–D ) . This is in line with previous findings ( Baumgardt et al . , 2007 ) , showing that maintained expression of Col in all four Ap cluster neurons results in ectopic Dimm and Nplp1 expression . The conversion of Ap cluster neurons to a Tv1 cell fate was accompanied by the loss of FMRFa expression ( Figure 2E–H ) , but not by the loss of Dac or Nab expression ( Figure 2I–L ) . 10 . 7554/eLife . 19311 . 004Figure 2 . Kr misexpression triggers Tv1 fate . ( A–D ) Postmitotic misexpression of UAS-Kr driven by ap-Gal4 results in ectopic expression of Dimm , Col and Nplp1 in the Ap cluster cells . ( C–D ) Quantification of Dimm and Col cells in the Ap cluster shows a significant increase when Kr is misexpressed ( ***p<0 . 001; n ≥ 12 clusters; Chi-square test +/- SEM ) . ( E–H ) Kr misexpression triggers ectopic expression of Nplp1 in the Ap cluster , but loss of FMRFa expression . ( G–H ) Quantification of FMRFa and Nplp1 expressing cells in the Ap cluster ( ***p<0 . 001; n ≥ 9 clusters; Chi-square test +/- SEM ) . ( I–L ) Misexpression of Kr , does not affect Dac or Nab expression . ( M–O ) Kr rescue by UAS-Kr driven from elav-Gal4 restores Dimm and Nplp1 specifically to Tv1 . ( P ) Cross-rescue of Kr with UAS-col from elav-Gal4 rescues Kr mutants , evident by Dimm and Nplp1 expression in the Ap clusters . In both rescue and cross-rescue , driving each UAS transgene from elav-Gal4 triggers supernumerary Tv1 cells . ( Q ) Quantification of the number of cells expressing Nplp1 in the Ap cluster ( ***p<0 . 001; n ≥ 9 clusters; Chi-square test +/-SEM ) . Genotypes: ( A , E , K , M ) OregonR . ( B , F , J , L ) ap-Gal4/UAS-Kr; +/UAS-Kr . ( I ) UAS-GFP/ap-Gal4 . ( N ) Kr1 , KrCD . ( O ) elav-Gal4/+; Kr1 , KrCD; UAS-Kr/+ . ( P ) Kr1 , KrCD; UAS-col/elav-Gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 004 Because Col is essential for activating Dimm and Nplp1 , the loss of their expression in Kr mutants may be a direct consequence of the loss of Col expression . To test this idea , we attempted to 'cross-rescue' Kr with UAS-col . First , as a control experiment , we attempted to rescue Kr with UAS-Kr , using the elav-Gal4 driver line , which is expressed in the NB5-6T at St12 and onward ( Karlsson et al . , 2010 ) . Indeed , we noted robust rescue of Dimm and Nplp1 in the Ap clusters , and as anticipated from the Kr misexpression experiments ( above ) there were supernumerary Tv1 neurons ( Figure 2M–O , 2Q ) . Next , we cross-rescued Kr with UAS-col , and observed rescue of Tv1 neurons , evident by the expression of Dimm and Nplp1 ( Figure 2P , Q ) . As previously , described ( Baumgardt et al . , 2007 ) , ectopic expression of col from elav-Gal4 resulted in supernumerary Eya , Dimm and Nplp1 cells ( Figure 2P , Q ) . These results suggest that the role of Kr in the Tv1 neuron is not to act directly upon Nplp1 , but rather to ensure the maintenance of Col , and thereby ensure the propagation of the col>ap/eya>dimm>Nplp1 feedforward cascade leading to Tv1/Nplp1 terminal differentiation . The selective expression of Kr in the Tv1 cell is key for the discriminated specification of the Tv1 cell fate from that of the other Ap cluster cells . But what activates Kr exclusively in the Tv1 neuron ? To address this , we first analyzed Kr expression in ap , eya and col mutants , critical Tv1 cell fate determinants ( Baumgardt et al . , 2007 ) , but did not observe any change in Kr expression in these mutant backgrounds ( Figure 3—figure supplement 1 ) . Next , we analyzed ladybird early ( lbe ) and Antp mutants , both of which are critical for NB5-6 identity ( lbe ) and for Ap cluster development ( both ) ( Gabilondo et al . , 2016; Karlsson et al . , 2010 ) . Since Antp and lbe show a loss of the Ap cluster markers Eya , Ap , Col , Dimm and Nplp1 , we identified the NB5-6T lineage based on the specific expression of reporter genes under the control of an enhancer fragment from the ladybird early gene ( lbe ( K ) ) ( Figure 3—figure supplement 1 ) ( Baumgardt et al . , 2007; De Graeve et al . , 2004 ) . In order to accurately monitor the Ap cluster cells in control , Antp and lbe mutant backgrounds , we used the non-affected Ap cluster markers Nab and Cas to thereby identify the Tv1 cells ( Figure 3—figure supplement 1 ) . At St16 , when Kr is robustly expressed in Tv1 , Nab is expressed in three out of four Ap cells and overlaps with Cas expression in Tv2/3 ( Baumgardt et al . , 2009 ) . The Tv1 cell resides in immediate proximity to the two Cas/Nab double positive Ap cluster cells , and is identifiable by Kr and lbe-GFP expression , combined with the lack of Nab and Cas . Analysis of Kr expression in the Tv1 cell in Antp and lbe mutants , using this marker combination , revealed that Kr is not affected in either mutant ( Figure 3—figure supplement 1 ) . Next , we addressed Kr expression in mutants for the castor ( cas ) temporal gene . cas plays a key role during NB5-6T lineage development , and regulates col , and thereby a number of Tv1 determinants , including Ap , Dimm and Nplp1 ( Baumgardt et al . , 2009 ) . However , in cas mutants Eya is still expressed in the Ap cluster cells ( Baumgardt et al . , 2009 ) . This allowed us to assess Kr expression in the Ap cluster cells within the NB5-6T lineage . We found that cas mutants often displayed loss of Kr in the Tv1 cells ( Figure 3A–C ) . This was accompanied by a reduction in the number of Kr-expressing cells in the entire NB5-6T lineage ( Figure 3D ) . The reduction of Kr expression in the entire NB5-6T lineage in cas mutants was mirrored by extensive overlap between Cas and Kr in the NB5-6T lineage in the wild type background ( Figure 3F ) . Moreover , scanning the entire thoracic VNC we observed Cas/Kr overlap in a number of other cells ( Figure 3E ) . Analyzing cas mutants in the entire thoracic segments , we observed striking reduction in Kr expression throughout these segments ( Figure 3G–H ) . 10 . 7554/eLife . 19311 . 005Figure 3 . castor regulates Kr . ( A–B ) Kr expression in the Ap cluster cells is frequently lost in cas mutants at St16 . Ap cluster cells were identified using Eya as a marker for the Ap cluster and lbe ( K ) -eGFP as a marker for the NB5-6T lineage . ( C–D ) Quantification of Eya/Kr cells and Kr/GFP cells in the Ap cluster and NB5-6T lineage , respectively ( ***p<0 . 001 , n ≥ 18 clusters , Chi-square test ± SEM , for ( C ) , and Student's t-test ± SEM , for ( D ) ) . ( E–F ) Kr expression overlaps with Cas in the NB5-6T ( F ) , as well as in many cells in the entire thorax ( arrowhead marks overlapping cells , asterisk marks non-overlapping cells ) . ( G–H ) Kr expression is globally reduced in the thoracic region T1-T3 in cas mutants . Genotypes: ( A , K ) lbe ( K ) -EGFP . ( B , L ) lbe ( K ) -EGFP /+; casΔ3/casΔ1 . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 00510 . 7554/eLife . 19311 . 006Figure 3—figure supplement 1 . Ap cluster cell fate determinants do not regulate Kr in Tv1 . ( A–B ) Staining for Eya , Col and Kr in ap mutants , at St16 , has no effect on Eya , Col and Kr expression . ( C–D ) Staining for Eya , GFP , Dimm and Kr in eya mutants , at St17 , shows loss of Eya and Dimm expression , but Kr expression is unaffected in the Ap cluster . ( E–G ) col mutants do not show any effects on Kr expression in the Ap cluster at St 16 ( n = 18 lineages; Student's t-test ± SEM ) . ( H–J ) Staining for GFP , Kr , Nab and Cas in control , Antp and lbe mutants , at St16 , shows that Kr expression is not affected in the Tv1 cell . ( K ) Cartoon depicting the Ap cluster composition at St16 , with the markers analyzed for control , Antp and lbe mutants . ( L–M ) grh mutants show normal Dimm and Kr expression in Tv1 at St16 . Genotypes: ( A ) OregonR . ( B ) apP44/apP44 . ( C ) UAS-GFP/ap-Gal4 . ( D ) eyaCli/eyaDf; ap-Gal4/UAS-nmEGFP . ( E , H , L ) lbe ( K ) -EGFP . ( I ) lbe ( K ) -EGFP/+; Antp12/Antp25 . ( J ) lbe ( K ) -EGFP/+; lbDf/lbe12C005 . ( M ) grhDf3064/grh370 , lbe ( K ) -Gal4; UAS-nmEGFP . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 006 Finally , we analyzed Kr expression in mutants for the last gene in the temporal gene cascade; grainy head ( grh ) , which is known to be expressed during the latter part of NB5-6T lineage development , and important for Tv4/FMRFa cell fate ( Baumgardt et al . , 2009 ) . However , we did not observe any effect on Kr expression in grh mutants ( Figure 3—figure supplement 1 ) . We conclude that Kr is activated in Tv1 by cas , and strikingly , that this regulatory connection extends into many late postmitotic cells in the VNC . In contrast , neither Antp nor lbe regulate Kr in the Tv1 neuron . We find that Kr is specifically expressed in the Tv1 cell , and that Kr is necessary and sufficient , within the Ap cluster to specify Tv1 cell fate . Kr acts by ensuring maintenance of Col expression , thus allowing for the col>ap/eya>dimm feedforward cascade to progress and finally activate Nplp1 expression . We find that cas activates Kr in the Tv1 cell . However , Cas expression spans the latter half of the NB5-6T lineage , covering all four Ap neurons ( Baumgardt et al . , 2009 ) , and hence does not readily explain Kr expression specifically in Tv1 . Previous studies revealed that down-regulation of Col in Tv2 , Tv3 and Tv4 is under control of a cas>sqz>nab FFL cascade , where down-regulation of Col is controlled by sqz/nab ( Baumgardt et al . , 2009 ) . Because of the timed delay in the cas>sqz>nab loop after an initial postmitotic phase ( Baumgardt et al . , 2009 ) , this ensures that Col expression is only maintained in Tv1 , and down-regulated in the other three Ap neurons . How does Kr relate to the cas>sqz>nab feedforward loop ? As described above , Nab is not affected in Kr mutants or by Kr misexpression ( Figures 1E–H , 2K–L ) . Thus , the cas>sqz>nab feedforward loop is not affected by Kr . In contrast , in both sqz and nab mutants , we observed one extra Kr/Dimm expressing cell in the Ap cluster ( Figure 4A–E ) . While sqz is expressed by all four Ap cluster neurons , the timing-delay in the cas>sqz>nab loop results in Nab expression only in three latter born cells; Tv2 , Tv3 and Tv4 . Misexpression of nab suppresses the Col maintenance and results in loss of Dimm and Nplp1 from the Tv1 neuron ( Baumgardt et al . , 2009 ) . In line with these findings , we addressed the effects of misexpressing nab in the Tv1 neuron , from elav-Gal4 . We observed that Kr expression was absent from the Ap cluster in nab misexpression ( Figure 4K–M ) . 10 . 7554/eLife . 19311 . 007Figure 4 . squeeze and nab regulate Kr . ( A–C ) At St16 , both sqz and nab mutants show additional Dimm and Kr expressing cells in the Ap cluster . ( D–E ) Quantification of Dimm and Kr expressing Ap cluster cells in sqz and nab mutants ( ***p<0 . 001; n ≥ 17 clusters; Student's t-test +/-SEM for ( D ) and between Ctrl and sqz in ( E ) ; Chi-square test for Ctrl and nab in ( E ) ) . ( E–H ) Similar to St16 , at AFT , both sqz and nab mutants show an increase in the numbers of Dimm expressing Ap cluster cells . ( I–J ) Quantification of Dimm/Ap and Svp/Ap cells ( ***p<0 . 001; n ≥ 14 clusters; Chi-square test; +/-SEM ) . ( K–L ) Misexpression of nab from elav-Gal4 results in a significant decrease in Kr expressing Ap cluster cells , at St16 . ( M ) Quantification of Kr/Eya Ap cluster cells ( ***p<0 . 001; n ≥ 16 clusters; Chi-square test ) . ( N–O ) nab misexpression from elav-Gal4 at AFT shows a reduction of Dimm expressing and an increase in Svp expressing Ap cluster cells . ( P–Q ) Quantification of Dimm/Ap and Svp/Ap cluster cells ( ***p<0 . 001; n ≥ 10 clusters; Chi-square test; ± SEM ) . ( R ) Cartoon showing the Ap cluster in control , sqz and nab mutants , as well as nab misexpression , at AFT . Control Ap clusters show stereotyped expression of Nplp1 , Col and Kr in the Tv1 cell . Dimm is present in both neuropetidergic Tv1/Nplp1 and Tv4/FMRFa cells . While Tv2-Tv4 cells all express nab , svp expression is restricted to Tv2 and Tv3 . In nab and sqz mutants one of the 'generic' Tv2/Tv3 Ap neurons is converted into a Tv1/Nplp1 cells , by expression of Kr , which in turn suppresses svp expression , which allows for col and dimm expression . The misexpression of nab shows the reciprocal phenotype of the mutants , with one extra Svp expressing cell , loss of Kr , Col , Dimm and Nplp1 . Genotypes: ( A , F , K , O ) OregonR . ( B , G ) sqzie/sqzie . ( C , H ) nabR52/nabR52 . ( L , P ) UAS-nab/+; elav-Gal4/+ . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 007 We conclude that the sub-temporal genes sqz and nab act to restrict the expression of Kr to the Tv1 neuron , while , Kr does not reciprocally regulate the cas>sqz>nab feedforward loop . Kr does not repress the cas>sqz>nab feedforward loop , playing out in Tv2 , Tv3 and Tv4 . In addition , Kr can be cross-rescued by col , and hence is not critical for the col>ap/eya>dimm>Nplp1 feedforward cascade . What , then , is the mechanism by which Kr ensures maintenance of Col in Tv1 cells ? Previous studies found that svp mutants displayed ectopic Col , Dimm and Nplp1 expression in the Ap cluster , and that postmitotic misexpression of svp from ap-Gal4 could suppress expression of Col , Dimm and Nplp1 ( Benito-Sipos et al . , 2011 ) . Svp is initially expressed by all four Ap cluster cells ( St15-16 ) , but is rapidly confined to Tv2 and Tv3 , where it is maintained into AFT ( Benito-Sipos et al . , 2011 ) . The reasons for the maintenance of Svp expression in Tv2 and Tv3 were previously not addressed . We find that in sqz and nab mutants , at stage AFT , there is loss of svp expression in one cell and also one extra Dimm expressing cell ( Figure 4F–J ) . Conversely , nab misexpression results in Svp expression in one extra cell coupled with loss of Dimm expression in one cell ( Figure 4N–Q ) . Hence , sqz and nab control Ap cluster diversification also by ensuring Svp expression in Tv2 and Tv3 . We next addressed the connection between Kr and svp . We found that in Kr mutants , Svp is expressed in one additional Ap cluster cell . The coincidental loss of Dimm and Nplp1 expression leads us to conclude Svp is activated in the Tv1 cell ( Figure 5A–E ) . Conversely , Kr misexpression postmitotically in all four Ap neurons resulted in complete loss of Svp expression ( Figure 5F–H ) . 10 . 7554/eLife . 19311 . 008Figure 5 . Kr represses the sub-temporal gene svp . ( A–B ) Kr mutants at AFT , reveals ectopic Svp expression in the Tv1 cell , accompanied by loss of Dimm expression . ( C–E ) Quantification of Svp/Ap , Dimm/Ap and Nplp1/Ap expressing cells ( ***p<0 . 001; n ≥ 16 clusters; Chi-square test ± SEM ) . ( F–G ) Misexpression of Kr from the late and Ap cluster specific driver ap-Gal4 , results in loss of Svp expression from Ap2/3 , at AFT . ( H ) Quantification of Svp/GFP or Svp/Kr expressing cells ( ***p<0 . 001 , n ≥ 12 clusters , Chi-square test ) . ( I–J ) svp mutants show increased number of Kr expressing cells in the Ap cluster . ( K ) Quantification of Kr/Eya expressing cells ( ***p<0 . 001 , n ≥ 12 clusters , Student's t-test ± SEM ) . ( L–M ) Misexpression of svp from elav-Gal4 suppresses Dimm expression , but has no effect on Kr . ( N–O ) Quantification of Kr/Eya and Dimm/Eya expressing cells ( ***p<0 . 001; n ≥ 12 clusters; Chi-square test ± SEM ) . ( P–R ) Both svp and svp , Kr double mutants show an increase of Dimm and Nplp1 expressing cells in the Ap cluster . ( S–T ) Quantification of Dimm/Eya and Nplp1/Eya cells ( ***p<0 . 001 , n ≥ 12 clusters , Student's t-test ± SEM in ( S ) and Chi-square test +/-SEM in ( T ) ) . Genotypes: ( A , I , L , P ) OregonR . ( B ) Kr1 , KrCD/Kr1 , KrCD . ( F ) ap-Gal4/+;UAS-nmEGFP/+ . ( G ) ap-Gal4/UAS-Kr; UAS-Kr/+ . ( J , Q ) svp1/svp1 . ( M ) elav-Gal4/UAS-svp . ( R ) Kr1 , KrCD/Kr1 , KrCD; svp1/svp1 . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 00810 . 7554/eLife . 19311 . 009Figure 5—figure supplement 1 . Kr regulates Svp expression in other NB lineages than NB5-6T . ( A ) Staining for GFP and Gsb-n shows no overlap between the GFP signal from ham-Gal4 and Gsb-n , showing that ham-Gal4 is not expressed in NB5-6T . ( B–D , F–G ) Staining for GFP , Svp and Kr in ( B , F ) control , ( C ) Kr misexpression , and ( D , G ) Kr mutants . ( C ) Kr misexpression triggers loss of Svp expression in the lateral-most cells of the ham-Gal4 lineage . ( D ) In Kr mutants the number of Svp positive cells are slightly increased , however not significant . ( E ) Quantification of Svp positive cells in control , Kr misexpression and Kr mutants , at stage 16 , shows a significant reduction of Svp positive cells when Kr is misexpressed from ham-gal4 ( ***p<0 . 0001 , nCtrl = 40 lineages , nUAS-Kr = 18 lineages , students t-test , ± SEM ) . Quantification of Svp positive cells in the ham-Gal4 lineage in Kr mutants reveals no significant differences ( p=0 . 14 , nCtrl = 40 lineages , nKr = 18 lineages , students t-test , ± SEM ) . ( F–G ) Comparison of the number of Svp positive cells in the intermediate region between the ham-Gal4 lineages ( white dashed boxes ) , reveals additional Svp positive cells in Kr mutants . ( H ) Quantification of the number of Svp cells in the ham-Gal4 intermediate regions reveals significant increase in Svp expressing cells ( ***p<0 . 0001 , nCtrl = 11 segments , nKr = 12 segments , students t-test , +/- SEM ) . Genotypes: ( A , B , F ) ham-Gal4 , UAS-nm-GFP . ( C ) ham-Gal4 , UAS-GFP/UAS-Kr . ( D , G ) Kr1 , KrCD/Kr1 , KrCD;ham-Gal4 , UAS-GFP . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 009 These findings prompted us to address if the role of Kr as a repressor of svp is a more general feature during VNC development . To this end , we utilized a ham-Gal4 driver ( GMR80G10-GAL4 ) . This driver expresses in several NBs of rows 3 and 4 , and hence does not overlap with NBs of row 5 and 6 , including NB5-6T , evident by its non-overlap with the row 5 and 6 marker Gsb-n ( Figure 5—figure supplement 1 ) . The ham-Gal4 lineages contain Kr and Svp expressing cells , but similarly to NB5-6T , these are non-overlapping ( Figure 5—figure supplement 1 ) . This provided a scenario to test for the repressive action of Kr onto Svp outside the NB5-6T lineage . Misexpression of Kr under the control of ham-Gal4 resulted in striking and significant loss of Svp expressing cells in ham-lineages ( Figure 5—figure supplement 1 ) . To address if Kr is also necessary for Svp repression in these lineages , we placed the ham-Gal4 reporter line into a Kr mutant background . While we observed a trend of increased numbers of Svp expressing cells in Kr mutants , this was not significant ( Figure 5—figure supplement 1 ) . However , we noted an apparent increase in Svp expressing cells in the region in between the ham-Gal4 lineages ( Figure 5—figure supplement 1 ) , and quantification of this region did reveal a significant increase in Svp expressing cells in Kr mutants ( Figure 5—figure supplement 1 ) . We conclude that the role of Kr as a repressor of svp extends into many , albeit not all , NB lineages in the developing VNC . Analyzing svp mutants , we observed ectopic Kr expression within the Ap cluster ( Figure 5I–K ) . However , when we misexpressed svp from elav-Gal4 , while we could reproduce the complete loss of Dimm expression from the Ap cluster , we found no loss of wildtype Kr expression in Tv1 ( Figure 5L–O ) . Prompted by these findings , we analyzed svp , Kr double mutants . We found that double mutants showed the typical extra expression of Dimm and Nplp1 observed in svp mutants ( Figure 5P–T ) . These results demonstrate that sqz-nab ensures maintained expression of Svp in Tv2/3 , by repressing Kr . svp in turn acts as a potent repressor of col and dimm expression , and is necessary but not sufficient to repress Kr . The key role Kr plays in the Tv1 cell is to suppress Svp expression , to safeguard the expression of col and dimm , and thereby ensure the propagation of the col>ap/eya>dimm>Nplp1 feedforward cascade leading to Tv1/Nplp1 terminal differentiation . We find that the main role of Kr with regards to specifying Tv1 fate is to repress svp . Is this interaction restricted to NB5-6T and the Ap cluster , or is it a global phenomenon ? To address this we misexpressed Kr pan-neuronally , using elav-Gal4 . Intriguingly , we find that Kr strongly suppresses Svp expression broadly in the VNC ( Figure 6A–B , E ) . 10 . 7554/eLife . 19311 . 010Figure 6 . Kr refines the combinatorial effects of lbe-col . ( A–B ) Single misexpression of Kr results in reduced numbers of Svp positive cells in thoracic segments T1-T3 compared to control . ( C ) Combinatorial misexpression of col with lbe results in ectopic Ap and Svp expressing cells . ( D ) Triple misexpression of col , lbe and Kr still results in ectopic Ap cells , but the number of Svp expressing cells is significantly reduced when compared to control . ( E ) Quantification of Svp expressing cells ( **p<0 . 01; ***p<0 . 001; Student's t-test; n = 3 thoracic regions per group; ±SEM ) . ( F–M ) Single , double and triple co-misexpression of col , lbe and Kr at AFT from elav-Gal4 . ( F–H ) Single misexpression of col or lbe results in ectopic Eya , Dimm and Nplp1 cells , when compared to control . ( I–J ) Single misexpression of Kr or double misexpression with lbe results in ectopic Eya and Nplp1 expression , but not of Dimm , when compared to control . ( K ) Co-misexpression of col and Kr results in ectopic Eya , Dimm and Nplp1 cells . ( L ) Co-misexpression of col and lbe results in extensive ectopic Eya , Dimm and Nplp1 expression . ( M ) Triple co-misexpression of col , lbe and Kr results in ectopic Eya , Dimm and Nplp1 cells , with Dimm cell numbers increased in comparison to double misexpression . ( N ) Quantification of Dimm expressing cells in thoracic segments T1-T3 ( *p<0 . 05; **p<0 . 01; ***p<0 . 001; n = 3 thoracic regions; Student's t-test +/- SEM ) . Genotypes: ( A , F ) OregonR . ( B , I ) elav-Gal4; UAS-Kr/+; UAS-Kr/+ . ( C , L ) elav-Gal4/UAS-col , UAS-lbe . ( D , M ) UAS-Kr; UAS-col , UAS-lbe/elav-Gal4 . ( G ) elav-Gal4/UAS-col . ( H ) elav-Gal4/UAS-lbe , ( J ) UAS-lbe/+ , UAS-Kr/elav-Gal4 . ( K ) UAS-Kr/+ , UAS-col/elav-Gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 19311 . 010 We recently found that lbe and col can act in a potent combinatorial code to activate Ap cluster cell fate throughout the VNC ( Gabilondo et al . , 2016 ) . Indeed , co-misexpressing lbe and col together result in striking ectopic Ap and Eya expression ( Figure 6C , L ) . However , only a subset of these ectopic Ap cluster cells express Dimm and Nplp1 , and are bona fide Tv1 neurons ( Figure 6L ) . This observation , together with previous studies ( Baumgardt et al . , 2009; Baumgardt et al . , 2007; Gabilondo et al . , 2016 ) , indicate that while lbe and col are sufficient to activate an early 'generic' Ap cluster cell fate , with ectopic Ap and Eya , specification of the different Ap cluster cell fates i . e . , Tv1/Nplp1 , Tv2/3 , Tv4/FMRFa requires several other inputs . On this note , col and lbe co-misexpression also triggered ectopic activation of Svp expression , corroborating the idea that col and lbe act together to trigger a mixture of Ap cluster cell fates ( Figure 6C , E ) . Given the potency of Kr in repressing Svp , we speculated that co-misexpression of lbe and col together with Kr may act to suppress the ectopic Svp expression . Indeed , while triple co-misexpression of Kr , lbe and col still activates extensive ectopic Ap expression , as anticipated , the number of Svp expressing cells is dramatically reduced ( Figure 6D , E ) . These results suggested that the generation of 'generic' Ap cluster cell fates triggered by lbe-col co-misexpression , could be refined by co-misexpression of Kr , and result in a more specific cell fate transformation towards exclusively Tv1/Nplp1 fate . To test this idea , we misexpressed the three different factors alone or in all combinatorial mixes from elav-Gal4 , and scored for Eya , Dimm and Nplp1 expression . To quantify the specification of Tv1 cells in contrast to 'generic' Ap cells , we counted the number of Dimm expressing cells . While single misexpression of lbe or col resulted in increased number of Dimm positive cells compared to control , Kr misexpression alone did not result in ectopic Dimm cells ( Figure 6F–I , N ) . Co-misexpression of lbe and Kr did not result in ectopic Dimm cells ( Figure 6J , N ) . In contrast , combinatorial misexpression of col and Kr increased the number of Dimm cells ( Figure 6K , N ) . As previously reported ( Gabilondo et al . , 2016 ) , we find that col and lbe together are tremendously potent at activating ectopic Eya , Dimm and Nplp1 expression in the thoracic region of the VNC , and Dimm cell quantification shows a four-fold increase compared to control ( Figure 6L , N ) . Finally , adding Kr to this cocktail refines the lbe-col effects , and 'generic' Ap neurons are now exclusively steered towards the Tv1/Nplp1 cell fate , evident by a robust increase in the numbers of Dimm/Nplp1 cells ( Figure 6M , N ) . We conclude that lbe and col co-misexpression , as anticipated , triggers extensive generation of ectopic Ap cluster cells . However , due to the simultaneous activation of Svp in many ectopic Ap cluster cells , a mixture of Ap cluster cell fates is generated . Adding Kr to this misexpression cocktail ensures robust suppression of Svp , and thereby refines the effect of lbe-col into exclusively Tv1 cell fate .
The temporal gene cas plays a pivotal role in the specification process of the different Ap cluster cells due to its activator role on a number of downstream regulators; col , a terminal selector in Tv1 specification , the sub-temporal genes sqz , nab , svp and Kr , as well as the temporal gene grh . Strikingly , cas thus activates both of the two terminal selector FFLs , and the genes required to refine both FFLs . cas activates Kr and svp , but how is Kr expression then restricted to only Tv1 and svp expression to Tv2/3 ? For Kr , restricted expression of sqz , nab and svp in Tv2-Tv4 , all of which suppress Kr , can explain the confined expression pattern of Kr to Tv1 . The gradually restricted expression of svp in Tv2-3 is in turn explained by Kr repressing svp in Tv1 , and by grh repressing svp in Tv4 . However , because grh misexpression is not sufficient to repress svp , it is tempting to speculate that there exists a similar factor to Kr , being exclusively expressed in the Tv4 cell , acting to suppress svp expression in a highly confined manner to ensure FMRFa/Tv4 specification . Besides its activation by cas , col activation requires additional spatial information , provided by lbe , Antp , hth and exd ( Gabilondo et al . , 2016; Karlsson et al . , 2010 ) , which subsequently initializes the generic Ap cluster program , by activating ap and eya . In contrast , cas alone activates grh and the sub-temporal factors , which are then important for the cell diversification , whether by activating or repressing each other’s actions , or the FFLs , or partake in the FFL ( grh ) , in order to allocate the correct cell fate to the four Ap cluster neurons . Remarkably , the four spatial inputs ( lbe , Antp , hth and exd ) act only on col , while the temporal input ( cas ) acts both on col , as well as the temporal and sub-temporal factors ( sqz , nab , svp , Kr and grh ) . It is tempting to speculate that this may point to a general role for spatial versus temporal cues , and may be logically explained by the fact that spatial cues generally do not display the highly selective temporal expression profile necessary for sub-temporal cell diversification ( Figure 7B ) . An unexpected finding in this study pertains to the dual role of Kr , first acting early in the canonical temporal cascade and subsequently late in the sub-temporal cascade , to ensure the specification of the Tv1 cell . The main role of Kr in Tv1 cells is to suppress svp , hence allowing for the maintenance of Col , which itself is critical for the propagation of the terminal FFL , fundamental for Tv1 cell fate . Interestingly , dual expression of Kr , first in the neuroblast and subsequently in neurons , was previously observed in NB3-3 , but the functional role of the second Kr expression pulse was not addressed ( Tsuji et al . , 2008 ) . svp itself also displays a dual expression and function , being expressed early in many NB lineages to suppress hb ( Kanai et al . , 2005 ) , then being re-expressed in several linages , and in NB5-6T it acts to suppress col and dimm ( Benito-Sipos et al . , 2011 ) . With regards to postmitotic activity , another example of a temporal gene acting postmitotically applies to the role of the last temporal gene , grh , which is necessary and sufficient for FMRFa expression in Tv4 cells , and can trigger ectopic FMRFa in Ap neurons when misexpressed postmitotically ( Baumgardt et al . , 2009 ) . Yet in contrast to Kr , grh does not experience a dual expression profile . Hence , with several examples of dual ( Kr and svp ) and progenitor versus postmitotic roles of temporal genes ( Kr and grh ) , it is tempting to speculate that this type of temporal multi-tasking may indeed be a common feature for many temporal genes , both in Drosophila and in higher organisms .
lbe ( K ) -Gal4 ( Baumgardt et al . , 2009 ) . lbe ( K ) -EGFP ( Ulvklo et al . , 2012 ) . elav-Gal4 ( DiAntonio et al . , 2001 ) ( provided by A . DiAntonio ) . casΔ1 and casΔ3 ( Mellerick et al . , 1992 ) ( provided by W . Odenwald ) . UAS-nls-myc-EGFP ( referred to as UAS-nmEGFP ) ( Allan et al . , 2003 ) . UAS-col ( Vervoort et al . , 1999 ) ( provided by A . Vincent ) . grh370 ( Bray and Kafatos , 1991 ) ( provided by S . Bray ) . Kr1 , KrCD ( Isshiki et al . , 2001 ) ( provided by C . Q . Doe ) . nabR52 and UAS-nab ( Terriente Félix et al . , 2007 ) ( provided by F . Diaz Benjumea ) . Df ( lbl-lbe ) B44 and UAS-lbe ( provided by K . Jagla ) . Antp12 ( Abbott and Kaufman , 1986 ) ( provided by F . Hirth ) . From Bloomington Drosophila Stock Center: Antp25 ( BL#3020 ) . lbe12C005 ( BL#59385 ) . elavC155= elav-Gal4 ( BL#458 ) . Df ( 2L ) BSC354 ( eyaDf ) ( BL#24378 ) . elav-Gal4 ( BL#8765 ) . sqzie ( BL#36497 ) . svp1 ( BL#6190 ) . apmd544 ( referred to as apGal4 ) ( BL#3041 ) . grhDf3654=Df ( 2R ) Pcl7B ( BL#3064 ) . sqzie ( BL#36497 ) . ham-Gal4 = GMR80G10-GAL4 ( BL#40090 ) . UAS-col-HA and UAS-myc-lbe were generated by the codon optimization for expression in Drosophila ( http://www . kazusa . or . jp/codon/ ) . EcoRI site and consensus start codon ( Cavener and Ray , 1991 ) was added to the 5’ end . Three different stop codons ( amb , och , opa ) followed by an XbaI site were added to the 3’ end ( see Supplementary file 1 for both sequences ) . DNAs were generated by gene-synthesis ( Genscript , New Jersey ) , and cloned into pUASattB ( Bischof et al . , 2007 ) , as EcoRI/XbaI fragments . DNAs were injected into landing site strains BL#9736 ( 53B ) and BL#9744 ( 89E ) , for UAS-myc-lbe , and BL#9750 ( 65B ) for UAS-col-HA ( BestGene , CA ) . Mutants were maintained over GFP- or YFP-marked balancer chromosomes . As wild-type control OregonR was used . Staging of embryos was performed according to Campos-Ortega and Hartenstein ( Campos-Ortega and Hartenstein , 1985 ) . Guinea pig a-Cas antibodies were generated by inserting a DNA fragment coding for Cas amino acids 171–793 in the Cas-PA protein into plasmid pET16b , followed by an expression in bacteria , gel purification and injection into 4 guinea pigs ( Davids Biotechnologie , Regensburg , Germany ) . The strongest sera was used at 1:500 , and tested for loss of specific nuclear staining in the VNC in cas mutants . Other primary antibodies were: Guinea pig a-Deadpan ( 1:1000 ) and rat a-Dpn ( 1:200 ) ( Ulvklo et al . , 2012 ) . Guinea pig a-Col ( 1:1000 ) , guinea pig a-Dimm ( 1:1000 ) , chicken a-proNplp1 ( 1:1000 ) and rabbit a-proFMRFa ( 1:1000 ) ( Baumgardt et al . , 2007 ) . Rat a-Grh ( 1:1000 ) and rat a-Nab ( 1:500 ) ( Baumgardt et al . , 2009 ) . Rabbit a-Cas ( 1:250 ) ( provided by W . Odenwald ) . Mouse mAb a-Dac dac2–3 ( 1:25 ) , mAb a-Eya 10H6 ( 1:250 ) ( Developmental Studies Hybridoma Bank , Iowa City , IA , US ) . Rabbit a-Kr ( 1:500 ) ( provided by Ralf Pflanz ) , mAb a-Svp ( 1:50 ) ( provided by Y . Hiromi ) . Rabbit a-Ap ( 1:1000 ) ( Bieli et al . , 2015 ) ( provided by D . Bieli and M . Affolter ) . Chicken a-GFP 1:1000 ( Abcam , ab13970 ) . Rat monoclonal α-Gsb-n ( 1:10 ) ( provided by R . Holmgren ) . Zeiss LSM 700 Confocal microscopes were used for fluorescent images; confocal stacks were merged using LSM software or Adobe Photoshop . Statistic calculations were performed in Graphpad prism software ( v4 . 03 ) . Cell counts in Figure 6 were done manually on thoracic segments T1-T3 in 3 VNCs for each group , with ImageJ FIJI and numbers transferred to Graphpad prism . To address statistical significance Student's t-test or in the case of invariant cell numbers contingency tables together with Chi-Square test were used . Images and graphs were compiled in Adobe Illustrator . | As a nervous system develops , stem cells generate different types of nerve cells at different times . This series of events follows a fixed schedule in developing embryos , and even a single stem cell that is removed and then grown outside the body will follow the same schedule . This strongly suggests that stem cells have a built-in clock that controls their development . Studies of the developing nervous system of fruit flies reveal that this clock works by switching genes on in specific sequences , which defines which nerve cells are produced at different stages of development . However , a clock built from the genes that are currently known to be involved in the process is simply not fine-tuned enough to explain how so many different types of nerve cell develop at such precise times . This implies that scientists do not yet know all of the genes that are involved . Using genetic experiments in stem cells from fruit flies , Stratmann , Gabilondo et al . now identify additional clock genes that act to divide broad windows of time during development into smaller , more precise ones . Genes that define broad windows of time switch on the “small window” genes at specific times – a bit like large cogs turning small cogs in a clock . One small window gene , called Kruppel , works at different stages of development and it is possible that other small window genes multi-task in similar ways in other developmental clocks , such as those found in more complex organisms like humans . It is clear that many genes work in sequence in the developing nervous system to ensure that developmental stages happen at precise times . Stratmann , Gabilondo et al . will next investigate the molecular details of this timing , specifically how genes in sequential time windows connect together like cogs in the developmental clock . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Neuronal cell fate diversification controlled by sub-temporal action of Kruppel |
Many primary sensory cilia exhibit unique architectures that are critical for transduction of specific sensory stimuli . Although basic ciliogenic mechanisms are well described , how complex ciliary structures are generated remains unclear . Seminal work performed several decades ago provided an initial but incomplete description of diverse sensory cilia morphologies in C . elegans . To begin to explore the mechanisms that generate these remarkably complex structures , we have taken advantage of advances in electron microscopy and tomography , and reconstructed three-dimensional structures of fifty of sixty sensory cilia in the C . elegans adult hermaphrodite at high resolution . We characterize novel axonemal microtubule organization patterns , clarify structural features at the ciliary base , describe new aspects of cilia–glia interactions , and identify structures suggesting novel mechanisms of ciliary protein trafficking . This complete ultrastructural description of diverse cilia in C . elegans provides the foundation for investigations into underlying ciliogenic pathways , as well as contributions of defined ciliary structures to specific neuronal functions .
Animals must sense and respond to multiple environmental cues over a wide range of signal intensities . The complexity of external cues is reflected in part in the remarkable diversity of sensory neuron morphologies and functions . Many major sensory neuron types contain microtubule ( MT ) -based primary cilia that house signal transduction molecules and are essential for the neurons’ sensory properties ( Perkins et al . , 1986; Inglis et al . , 2007; McEwen et al . , 2008; Ramamurthy and Cayouette , 2009; Pifferi et al . , 2010 ) . Although basic assembly mechanisms , core structures , and overall functions of cilia are highly conserved , sensory neuron cilia are structurally diverse and specialized for their unique roles ( Berbari et al . , 2009; Silverman and Leroux , 2009 ) . For example , vertebrate olfactory neurons contain a tuft of 6–17 cilia that emanate from the dendritic knobs and house olfactory signaling proteins ( Menco , 1980; McEwen et al . , 2008; Pifferi et al . , 2010 ) . Similarly , rod and cone photoreceptors contain highly elaborate ciliary outer segments that differ in morphology , and that localize molecules required for phototransduction ( Berbari et al . , 2009; Yildiz and Khanna , 2012 ) . Sensory cilia morphology and function are further regulated via interactions with non-neuronal cells such as glia ( Young and Bok , 1969; Strauss , 2005; Bacaj et al . , 2008; Procko et al . , 2011 ) . Defects in cilia structure and function lead to altered cellular signaling and contribute to systemic disorders collectively called ciliopathies ( Green and Mykytyn , 2010; Louvi and Grove , 2011; Waters and Beales , 2011 ) . Thus , analysis of sensory cilia architecture , as well as their organization and interaction with supporting cells , is essential for a complete description and understanding of the mechanisms that underlie functional specialization in the sensory nervous system . The nematode Caenorhabditis elegans is an excellent model system in which to explore the molecular and neuronal mechanisms by which animals detect and respond to environmental cues ( e . g . , Bargmann , 2006; Goodman , 2006; Zhang , 2008; Garrity et al . , 2010 ) . The C . elegans hermaphrodite has a compact sensory nervous system comprised of at least 78 neurons of 34 types that differ in their morphologies , connectivities , and sensory properties ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986; White et al . , 1986 ) ( www . wormatlas . org ) . Many of the major chemo- , mechano- and thermosensory neurons are present in sensory organs located in the head and tail of the worm; 60 of these sensory neurons are ciliated ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986; White et al . , 1986; Bargmann , 2006; Inglis et al . , 2007 ) ( www . wormatlas . org ) . As in other organisms , development , and maintenance of correct cilia structure is essential for the unique sensory properties of each neuron type in C . elegans ( Dusenbery et al . , 1975; Lewis and Hodgkin , 1977; Culotti and Russell , 1978; Perkins et al . , 1986 ) . In the head , ciliated sensory neurons are organized in several sensilla , each of which contains different sensory neuron types and associated glial support cells ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986 ) . Specific functions have been assigned to many sensory neurons ( Bargmann , 2006 ) ( www . wormatlas . org ) , providing an experimentally amenable system in which to not only correlate structural and functional specialization , but to also describe in detail the extent of diversity in sensory cilia morphologies and ultrastructures present in a metazoan . Pioneering work in the 1970s and 1980s provided an initial morphological and ultrastructural description of the anterior sensory anatomy of C . elegans ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986 ) . In these studies , serial section transmission electron microscopy ( ssTEM ) of chemically fixed animals was used to describe the structure and organization of head sense organs . In particular , these studies described the morphologically diverse cilia present at the dendritic endings of sensory neuron types in these sensilla ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986 ) . For example , eight amphid organ sensory neurons were shown to contain structurally simple cilia that are directly exposed to the environment through a ‘channel’ formed by glial cells , whereas the cilia of other amphid sensory neurons were shown to be more complex and embedded within glial cell processes ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986 ) . Although these studies provided the basis for extensive further investigations into the sensory biology of C . elegans , two issues prompted us to revisit the description of the wild-type anterior sensory anatomy . First , advances in imaging and tissue preservation techniques have led to the development of methods and instrumentation for ultrastructural imaging that provide far more detailed and higher resolution views than was previously possible ( Rostaing et al . , 2004; Weimer , 2006; Muller-Reichert et al . , 2008; Hall et al . , 2012 ) . As an example , serial section electron tomography ( ssET ) now allows for three-dimensional ( 3D ) reconstruction of structures , and reveals novel features , due to almost isotropic resolution within the reconstruction rather than being limited in the z-direction by the thickness of plastic sections ( McEwen and Marko , 2001; Muller-Reichert et al . , 2010 ) . Second , the ultrastructural features of several sensory cilia have remained poorly described , in part due to limitations in the visualization methods , but also due to issues arising from conventional chemical fixation of biological structures . For instance , the morphology and ultrastructure of cilia in the BAG gas-sensing and FLP nociceptive neurons are poorly described , and the axonemal MT distribution pattern in highly branched cilia , such as those in the AWA chemosensory neurons , have not been fully characterized . Here , we provide a high-resolution morphological and ultrastructural analysis of the anterior endings of sensory neurons and glia in the adult C . elegans hermaphrodite using ssTEM and ssET of high-pressure frozen and freeze-substituted ( HPF-FS ) adult animals . This analysis has now allowed us to robustly model the morphologies of all anterior sensory cilia , comprising 50 of 60 ciliated neurons in the C . elegans adult hermaphrodite . We describe the ultrastructures of diverse cilia types in unprecedented detail including those of previously uncharacterized cilia , and describe new features of cilia–glial interactions . We also characterize their intercellular relationships , and identify important ultrastructural features including MTs , fibers , vesicles , and junctions . Moreover , this high-resolution map reveals three distinct and novel MT distribution processes in branched cilia , suggesting diverse modes of ciliary protein transport and delivery . Together with the extensive analyses of cell specification pathways possible in C . elegans , this complete ultrastructural description of the anterior sensory anatomy will allow for a detailed understanding of the molecular and cellular mechanisms by which morphological diversity is established , as well as the contribution of this diversity to cellular functional specialization .
HPF-FS has been shown to significantly reduce fixation artifacts and morphological deterioration of C . elegans samples for EM as compared to conventional chemical fixation techniques ( Muller-Reichert et al . , 2003 , 2008; Weimer , 2006 ) . Consistent with these observations , we obtained excellent preservation of tissues from adult C . elegans hermaphrodites using HPF-FS ( Figure 1A , Figure 1—figure supplement 1 ) . In particular , we noted that: ( a ) subcellular structures in muscle were maintained , allowing resolution of the paracrystalline organization of actin-myosin fibers in sarcomeres ( Figure 1—figure supplement 1A ) ; ( b ) cell–cell connections as in the electron-dense C . elegans apical junctions ( CeAJs ) ( Labouesse , 2006; Lynch and Hardin , 2009 ) or hemi-desmosome-like junctions were preserved with clearly visible membranes and cytoskeletal filaments ( Figure 1—figure supplement 1B , C ) ( Cox and Hardin , 2004 ) ; ( c ) ciliary membranes exhibited well-defined , smooth structures with clearly resolved bilayers as compared to scalloped membranes observed in chemically fixed samples ( Figure 1—figure supplement 1D ) ; ( d ) doublet and singlet MTs could be observed , allowing identification of different protofilament ( pf ) numbers and inner MT structures ( Figure 1—figure supplement 1D ) ; and ( e ) MT-associated fibers ( Y-links ) could be identified in the ciliary transition zone ( TZ ) , a morphologically distinct structure at the base of cilia that acts as a ciliary diffusion barrier ( Figure 1—figure supplement 1E ) ( Reiter et al . , 2012; Szymanska and Johnson , 2012 ) . We did not observe obvious cell shrinkage or ice damage , which would be indicated by empty spaces surrounding cells , or the destruction or aggregation of filamentous structures such as those of muscle and cell junctions , respectively . 10 . 7554/eLife . 01948 . 003Figure 1 . TEM cross-section with identified ciliary and glial endings . ( A ) Example TEM cross-section of a high-pressure frozen/freeze-substituted ( HPF-FS ) C . elegans hermaphrodite animal ( 70-nm cross-section , 5 . 9 µm from the anterior nose tip ) . Edges surrounding the cuticle were feather-cropped . See Figure 1—figure supplement 1 for examples of preservation of subcellular structures in cilia , muscle , and junctions . ( B ) Endings of identified cells in the indicated bilateral sensilla within the cross-section TEM; only the right side of the animal is labeled . Cell endings are marked with a false-color overlay . Dorsal up , ventral down . Scale bar: 1 µm . See Figure 1—figure supplement 2 for additional cross-section views . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 00310 . 7554/eLife . 01948 . 004Figure 1—figure supplement 1 . Ultrastructural features maintained by HPF-FS . ( A ) TEM cross-section images of striated ventral body wall muscle ( m ) with myosin and actin filaments . Mitochondria ( mt ) with cristae are also shown . ( B ) Cross-section TEM image showing an apical junction ( arrows ) between glial cell processes associated with the CEP neuron ( CEPso–socket; CEPsh–sheath ) . ( C ) Hemi-desmosome-like structures ( arrow ) found within pharyngeal epithelium . ( D ) Cross-section TEM image of amphid channel cilia showing bilayer plasma membranes , doublet MTs ( arrow ) and inner singlet MTs ( arrowhead ) . ( E ) TZ showing Y-link ( arrow ) . Scalebars: 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 00410 . 7554/eLife . 01948 . 005Figure 1—figure supplement 2 . Example TEM cross-section images of the C . elegans nose . ( A ) 3D reconstruction of the anterior endings . The location of sections B–F and the section shown in Figure 1 are indicated . ( B–F ) Anterior cross-section TEM images from two animals ( B , D , F–animal 2; C , E–animal 1 ) showing invariant location of neuronal endings . Only the right side is labeled as in Figure 1 . Scale bars: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 005 We reconstructed the 3D structure of the distal head region ( ∼14 µm from the nose tip ) , containing the sensory endings of many sensory neuron types and glia , for three serially sectioned wild-type adult worms that were grown and prepared under identical conditions . We also imaged selected sections of the head region for five additional wild-type animals in cross-section ( Figure 1—figure supplement 2 ) , and one animal in longitudinal section . For each ssTEM 3D reconstruction , ∼170 cross-sections of 70-nm thickness were collected on grids , while preserving the section sequence from proximal to distal . To fit an entire cross-section into a single EM image would require a relatively low EM magnification , which provides a large field of view but sacrifices image resolution . To instead take advantage of the excellent sample preservation by HPF-FS and thus the inherent increased resolution , we digitally recorded a montage of >100 EM images for each section at high magnification , providing a pixel size well below the resolution limit set by the sample preparation . The individual image tiles of the montage of each section were aligned and stitched together to generate a single 2D cross-section image ( with a size of up to 20K × 20K pixels ) ; images were then sequentially aligned to generate an ssTEM 3D reconstruction ( see ‘Materials and methods’ for additional details ) . We then followed and outlined the cell membranes of the anterior processes of more than 90 cells , including those of sensory neurons and associated glial cells , on these serially aligned cross-sections to generate a complete 3D graphical model of the cellular organization of the C . elegans anterior anatomy ( Figure 2 , Figure 3A , A′ , Video 1 ) . 10 . 7554/eLife . 01948 . 006Figure 2 . Graphical model of a high-resolution ssTEM 3D reconstruction of the anterior sensory endings . 3D reconstruction of the cilia and dendritic endings of anterior sensory neurons modeled from 166 thin serial sections ( 14 µm , starting at the nose ) . Front ( A ) and angled ( B ) profile views are shown . Cell projections are color-coded as indicated in upper right panel . D , dorsal; V , ventral; A , anterior; P , posterior . Scale bar: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 00610 . 7554/eLife . 01948 . 007Figure 3 . Two glial cell types are associated with each ciliated sensillum . ( A and A′ ) 3D reconstruction models of sheath ( sh ) and socket ( so ) glia for left side of the worm for amphid ( AM ) , inner labial ( IL ) , outer labial ( OLL/OLQ ) , and cephalic ( CEP ) glia . Glial types are color coded as indicated . See Figure 3—figure supplement 1 for 3D models of glia associated with individual anterior sensilla . ( B–D ) Distal ( B ) , middle ( C ) and region proximal to the amphid cilia TZ ( D ) . Channel cilia of individual neurons are labeled with the last letter in their names . Arrow indicates the autocellular junction; arrowhead indicates vesicle present in the AMso process . ( C ) The AMsh glial cell process surrounds the proximal regions of amphid neuron cilia; the lumen contains electron-dense matrix ( yellow arrow ) . White arrowhead indicates vesicle in the AMsh process . ( D ) The AMsh process contains extensive Golgi lamellae ( yellow arrowhead ) and secretory vesicles ( white arrowhead ) . Cilia are labeled as above . ( E–G ) Distal ( E ) , middle ( F ) , and region proximal to the TZs ( G ) of the IL neurons . Autocellular junction in the ILso process is indicated by an arrow in ( E ) . The lumen of the ILsh glial cell process surrounding the IL1/IL2 proximal cilia region also contains electron-dense matrix ( yellow arrow in F ) , and a network of ER/Golgi ( yellow arrowheads in F and G ) . Vesicles are indicated by white arrowheads . Scale bars: ( A–D ) 500 nm; ( E–G ) 200 nm . Figure 3—figure supplement 2 shows additional examples of CeAJs among glial and neuronal cells in individual sensilla . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 00710 . 7554/eLife . 01948 . 008Figure 3—figure supplement 1 . 3D reconstruction models of glia associated with different sensilla . ( A ) amphid , ( B ) IL , ( C ) OLL/OLQ , and ( D ) CEP sheath and socket glial endings associate closely with the cilia of their respective neurons . Glial subtypes are color coded . Scale bars: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 00810 . 7554/eLife . 01948 . 009Figure 3—figure supplement 2 . CeAJ connections between glial processes and surrounding cells . ssTEM cross-section images showing CeAJ connections ( arrowheads ) between ( A ) AMso , OLLso and hypodermal ( hyp ) syncytial cells , ( B ) AMso and AMsh , and ( C ) AMsh and amphid neuron cilia at their PCMCs . Scale bars: 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 00910 . 7554/eLife . 01948 . 010Video 1 . ssTEM 3D reconstruction model of all ciliated and non-ciliated anterior sensory endings in the C . elegans adult hermaphrodite . Color codes are as indicated in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 010 We identified each neuron within the bilateral amphid sensilla and reconstructed 3D models of their ciliated endings ( Figure 6 , Video 2 ) . These neurons can be separated into three major categories based on their ciliary morphologies . Eight neurons ( ASE , ADF , ASG , ASH , ASI , ASJ , ASK , ADL ) contain simple rod-like ‘channel’ cilia that terminate within a channel formed by the amphid socket ( AMso ) glial cell and respond largely to aqueous compounds ( Ward et al . , 1975; Perkins et al . , 1986; Bargmann and Horvitz , 1991 ) ( www . wormatlas . org ) . The channel ends as a pore in the cuticle of the lateral lip , exposing the channel lumen to the environment . Three neurons ( AWA , AWB , AWC ) contain elaborate ‘wing’ cilia , which are required for responses to volatile odorants and temperature , and one thermo- and gas-sensing neuron ( AFD ) which ends in a finger-like cilium and multiple microvilli that do not contain MTs ( Ward et al . , 1975; Perkins et al . , 1986; Bargmann et al . , 1993; Mori and Ohshima , 1995 ) ( www . wormatlas . org ) . Although the distal ciliary segments of AWA , AWB , and AWC are embedded within the amphid sheath ( AMsh ) process and do not enter the AMso channel , their PCMCs , TZs , and proximal ciliary middle segments are exposed to the matrix-filled lumen encompassed by the AMsh cell process , which is continuous with the channel lumen , and thus in contact with the environment ( Figure 6B–D , Video 2 ) ( Ware et al . , 1975 ) . In contrast , the endings of the AFD neurons appear to be fully embedded within the AMsh process without contact with the matrix-filled lumen ( Figure 6B–D; Video 2 ) ( Ware et al . , 1975 ) . Reconstruction of the 3D models clearly confirms that amphid cilia weave around and are entwined with each other in a remarkably stereotypical manner ( Figure 6A , Video 2 ) ( Ware et al . , 1975; Perkins et al . , 1986 ) . For example , the proximal dendrites of the AWB olfactory neuron type are positioned in the middle of the amphid dendrite fascicle but their cilia branch out laterally , weaving around the cilia of the AWA , AWC , and ADL amphid neurons ( Figure 6 , Video 2 ) . Similarly , although the ADL and ADF dendrites are located at distinct positions in the amphid bundle , their cilia are closely appositioned in the distal amphid pore ( Figure 6 , Figure 7 ) . 10 . 7554/eLife . 01948 . 015Figure 6 . The amphid sensillum . ( A and A′ ) 3D graphical model of the reconstructed amphid sensilla on the right side indicating the endings of twelve sensory neurons . Scale bar: 1 µm . ( B ) Distal , ( C ) middle , and ( D ) proximal TEM cross-sections of an amphid sensillum . Endings of individual amphid neurons are color coded as indicated . Scale bars: 500 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 01510 . 7554/eLife . 01948 . 016Video 2 . ssTEM 3D reconstruction model of all amphid neuron cilia and associated socket and sheath cell processes . Color codes as indicated in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 01610 . 7554/eLife . 01948 . 017Figure 7 . Morphology and ultrastructure of amphid channel cilia . Reconstructed 3D models of amphid channel cilia containing single ( A and A′ ) or double ( B and B′ ) rods on the left ( A and B ) and right ( A′ and B′ ) sides . Labels indicate approximate positions of cross-sections shown in C–E′ . TZs ( purple ) are indicated . Scale bar: 500 nm . ( C–E′ ) Cross-section TEM images of amphid channel at the level of distal segment ( C and C′ ) , middle segment ( D and D′ ) , and TZ ( E and E′ ) . Individual cilia are identified on the right only by the last letter in their names . Scale bars: 200 nm . See Figure 3—figure supplement 2C for additional views of the TZs and rootlet-like structures at the base of amphid neuron cilia . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 017 The endings of the labial and cephalic neurons extend the furthest anteriorly of all anterior sensory neurons and terminate in cuticle pores at the lip apices . These sensory neurons have been implicated largely in mechanosensation ( Kaplan and Horvitz , 1993; Hart et al . , 1999; Goodman , 2006; Chang et al . , 2011; Chatzigeorgiou and Schafer , 2011; Lee et al . , 2012 ) . The sensory endings of BAG and FLP neurons exhibit a highly complex 3D architecture , which has not previously been characterized in detail . As reported previously ( Ward et al . , 1975; Perkins et al . , 1986 ) , the sensory endings of BAG and FLP are not wrapped by their own set of support glia cells . Instead , there appears to be a role reversal such that the neuronal dendrites wrap or engulf part of the process of the ILSo cell associated with the lateral IL sensilla ( Figure 15 , Video 7 ) ( Ward et al . , 1975; Ware et al . , 1975 ) . 10 . 7554/eLife . 01948 . 034Figure 15 . Ultrastructure of BAG and FLP cilia . ( A ) 3D reconstruction model of BAG ( purple ) and FLP ( green ) sensory endings . BAG and FLP sensory endings are positioned laterally . FLP has an extensive dendritic branching network . Scale bar: 1 µm . ( B and B′ ) 3D reconstruction models of BAG and FLP endings indicating TZs ( gray ) . BAG and FLP ( transparent ) closely ensheath the ending of the ILso glial cell ( yellow-tan ) . Approximate locations of sections shown in C–J are indicated . Arrowheads indicate bulbous structures in FLP dendritic branch . ( C ) Cross-section TEM image showing the far distal end of BAG cilium with several dMTs ( inset ) at the edges of the BAG ciliary branches ( black arrows ) and associating with the ILso process . ( D ) The distal segment of the BAG cilium wraps around a projection of the ILso and contains dMTs ( inset ) segregating preferentially to one side . ( E and F ) Ultrastructures of the BAG TZ ( E ) and rootlet ( arrowhead in F ) . ( G ) The distal segment of FLP cilia has flap-like projections with several dMTs ( arrow ) . ( H ) The middle segment of FLP cilia contains several disorganized dMTs . ( I ) The FLP TZ contains nine dMTs . ( J ) Arrowhead indicates rootlet-like structures in the FLP PCMC . ( K ) Longitudinal ET view of a FLP dendritic branch shows an iteratively bulbous dendrite ( arrows ) that includes 2–3 MTs , which are tightly packed in the ∼60 nm constrictions between bulbs ( inset ) . Scale bars: 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 03410 . 7554/eLife . 01948 . 035Video 7 . 3D reconstruction model of BAG and FLP cilia associated with the ILso process . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 035 In addition to the ciliated anterior sensory neurons described above , the dendritic endings of four non-ciliated neuron types have been described to project with the anterior sensory fascicle bundles ( Figure 16A ) . These include projections from four URA and two URB neurons of unknown function found dorsally and ventrally , two dorsal URX gas-sensing neurons ( Zimmer et al . , 2009; Bretscher et al . , 2011; Busch et al . , 2012 ) and four ( two dorsal and two ventral ) URY neurons that are involved in mate-searching behavior in males and may also be implicated in pathogen sensing ( Pradel et al . , 2007; Barrios et al . , 2012 ) ( Figure 16A ) . Sensory endings of the URA and URB neurons were only present in a few sections in the region examined here and were not modeled further . 10 . 7554/eLife . 01948 . 036Figure 16 . Non-ciliated endings of URX and URY . ( A and A′ ) 3D reconstruction of two dorsal URX ( purple ) and four dorsal and ventral URY ( pink ) dendritic endings . ( Bi–iv ) Selected cross-sections from distal ( Bi ) to proximal ( Biv ) . ( Bi ) The tip of URX ( purple ) terminates around the ILD sensillum with branches around the process of ILsoD . ( Bii ) URY is found apposed to processes of OLQshD and ILshD . ( Biii ) URX and URY endings are found near the OLQ dendrite and also surrounded by processes of OLQsoD , ILsoD , and ILshD . Note singlet MTs in URY . ( Biv ) URX and URY are found sub-dorsally with IL1D , OLQshD , ILshD , OLQsoD , and ILsoD processes as well as hypodermal cells as neighbors . Scale bars: 500 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01948 . 036 The morphological and ultrastructural analyses presented here build on previous work ( Ward et al . , 1975; Ware et al . , 1975; Perkins et al . , 1986 ) to provide a complete 3D description of the anterior sensory anatomy of C . elegans at high resolution . Together , these analyses reveal several novel features of C . elegans cilia in detail that will inform investigations into neuronal functions . For instance , the IL1/2 , OLL/Q , CEP and FLP mechanosensory neurons contain diverse ciliary specializations including the presence of electron-dense disks , deposits of electron-dense material that can be MT-associated , tubular bodies , and flap-like elaborations ( as in FLP ) at the distal ends , spurs and strings that connect the cilium to the cuticle and surrounding cells , unique axoneme MT organization , distribution and protein association , and the presence of striated or non-striated rootlets and rootlet-like structures . In contrast , the polymodal ASH amphid neurons that also respond to mechanical stimuli ( Kaplan and Horvitz , 1993; Hilliard et al . , 2005 ) contain a prototypical axoneme . This diversity in sensory structures may reflect the distinct types and intensities of mechanical forces transduced by different mechanosensory neurons and/or different transduction mechanisms . As an example , we speculate that the lack of specialized cilia structures in ASH may imply a mode of force transduction that does not involve extensive ciliary deflection perhaps due to the tight arrangement of cilia in the amphid channel pore . Our analyses also reveal the ultrastructural basis of the intricate branching patterns present at the sensory endings of many head neurons . The specific contributions of these complex structures to neuronal functions remain to be determined . It can be argued that a larger ciliary surface area enhances sensitivity via increased localization of signal transduction molecules ( analogous to the situation in photoreceptor outer segments ) . However , it is currently unclear whether the complex branching pattern in AWA cilia , the larger wing-like structure in AWC , the many microvilli found in AFD , and the two cilia found per dendrite in ADL , ADF , and AWB , all serve to simply increase surface area , or whether the precise topological organization of signaling molecules within these structures is in addition important for signal transduction . In future , further advances in cilia imaging techniques ( e . g . , Sahl and Moerner , 2013; Su et al . , 2013; Ye et al . , 2013 ) and examination of ciliary protein trafficking pathways in these unique cilia types may allow us to describe not only how these structures are formed , but how their morphologies shape , and are in turn shaped , by neuronal functions .
C . elegans animals were maintained at 20°C on standard nematode growth media plates seeded with E . coli OP50 bacteria . Wild-type C . elegans ( Bristol strain N2 ) was obtained from the Caenorhabditis Genetics Center . Worm samples were prepared by transferring 1-day-old adult hermaphrodite worms to 20% bovine serum albumin ( BSA ) in M9 buffer in the cavity of an aluminum planchette ( type ‘A’ hat; 100 µm deep , Wohlwend , Switzerland ) for high-pressure freezing ( McDonald et al . , 2007 ) . In some cases , animals were treated with 10 mM levamisole and embedded in 2 . 4% low-melting point agarose pads ( Kolotuev et al . , 2009 ) . We noted that levamisole treatment and agarose embedding resulted in shortened channel cilia lacking distal segments; thus , this method was not used to model the distal ends of amphid channel cilia . The flat surface of another planchette ( type ‘B’ hat ) was placed on top to enclose the worms in the planchette’s cavity . The specimen–planchette sandwich was rapidly frozen using a Leica EM HPM100 high-pressure freezing system ( Leica Microsystems , Vienna , Austria ) . Freeze-substitution was performed at low temperature ( −90°C ) over 3 days in a solution containing 1% osmium tetroxide ( 19 , 100 , EMS ) , 0 . 5% glutaraldehyde ( 16 , 530 , EMS ) and 2% water in anhydrous acetone ( AC32680-1000 , Fisher ) using a Leica EM AFS2 freeze-substitution system . The temperature was progressively increased up to 0°C ( 5°C/hr ) , and finally brought to 4°C and maintained at this temperature for 1 hr . Samples were washed four times with anhydrous acetone ( 30 min each ) , then infiltrated and embedded in Araldite 502/Embed-812 Resin ( Araldite—10 , 900 , EMS , Embed-812—14 , 900 , EMS , DDSA—13 , 710 , EMS ) at room temperature and polymerized in an oven at 60°C for 3–4 days . Resin blocks with specimens were in most cases trimmed so that the block face was perpendicular to the longitudinal axis of the worm nose for serial cross-sections , while keeping a small amount of resin around the specimen . For longitudinal images of the nose sensilla , the block face was aligned parallel to the longitudinal axis of the worm nose . Serial ultrathin sections ( 70-nm thickness ) were collected on slot grids covered with Formvar support film , and post-stained with 2% uranyl acetate ( 0379 , Polysciences , Inc . , Warrington , PA ) for 30 min , and Reynold’s lead citrate ( Lead nitrate—17 , 900 , EMS , and Sodium citrate—S-279 , Fisher ) for 15 min . Serial sections were imaged on a Tecnai F20 ( 200 keV ) or F30 ( 300 keV ) transmission electron microscope ( FEI , Hillsboro , OR ) and recorded using a 2K × 2K charged-coupled device ( CCD ) camera at 14 , 500X magnification on the F20 ( 1 . 25 nm pixel size ) , or at 23 , 000X magnification on the F30 ( 1 nm pixel size ) . For large overviews of the cross and longitudinal sections , we acquired montages of overlapping high-magnification images in an automated fashion using the microscope control software SerialEM ( Mastronarde , 2005 ) . After analyzing TEM images of the serial sections , grids , and sections with structures of interest were prepared for ssET by coating both sides of the serial sections on Formvar with 10-nm colloidal gold fiducials ( Sigma-Aldrich , St . Louis , MO ) that were previously incubated for 30 min in 5% BSA ( SC-2323 , Santa Cruz Biotechnology , Inc . ) ( Iancu et al . , 2006 ) . Dual-axis tilt series , that is , two orthogonal tilt series of regions of interest , were acquired by tilting the sample from −60 to +60 with 1° increments using the microscope control software SerialEM ( Mastronarde , 2005 ) on a Tecnai F20 ( 200 keV ) . All images were digitally recorded on a 2K × 2K CCD camera , at 14 , 500X or 19 , 000X magnification , resulting in a pixel size of 1 . 25 nm or 1 . 04 nm , respectively . | To survive , animals must constantly gather information about their surroundings and then decide how to respond . Animals rely on cells called sensory neurons to help them perceive and process this information , and these neurons in most animals have smaller structures called cilia that help them to gather this information . The structures of these cilia can range from simple hair-like rods to complex branched arbors . Defective cilia can lead to cell degeneration and death . Scientists have identified and determined the functions of many of the 60 sensory neurons with cilia in C . elegans , a tiny roundworm with a simple nervous system . These experiments have revealed that the shapes of these cilia are quite diverse , and that the shape determines the type of information the neurons process . Learning more about how cilia are shaped , and how these shapes allow them to perform specific sensory functions , would give scientists a better understanding of how the brain processes sensory information . Doroquez et al . have now taken advantage of advances in imaging technology to generate highly detailed three-dimensional reconstructions of the cilia on 50 neurons in the nose of C . elegans . The experiments involved rapidly freezing the worms , slowly replacing the frozen water molecules with a preservative solution , and then embedding in resin . This allowed Doroquez et al . to slice the samples into very thin sections—some 1400 times thinner than a sheet of paper—and then image them with transmission electron microscopy and electron tomography . Finally , all these images were combined in a computer to produce 3D models of the cilia . The models reveal a wide range of cilia structures , including some that have never been examined in detail before . Doroquez et al . were also able to see detailed structures within the cilia , including compartments that determine which proteins should enter into , or be excluded from , an individual cilium . The models , along with the results of previous studies , suggest that cilia are shaped by genetic factors and also by interactions with the environment . This detailed description of diverse cilia structures should now allow researchers to identify the genes that determine their unique shapes , and explore how specific shapes contribute to specific sensory functions . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | A high-resolution morphological and ultrastructural map of anterior sensory cilia and glia in Caenorhabditis elegans |
Patterns of spatial positioning of individuals within microbial communities are often critical to community function . However , understanding patterning in natural communities is hampered by the multitude of cell–cell and cell–environment interactions as well as environmental variability . Here , through simulations and experiments on communities in defined environments , we examined how ecological interactions between two distinct partners impacted community patterning . We found that in strong cooperation with spatially localized large fitness benefits to both partners , a unique pattern is generated: partners spatially intermixed by appearing successively on top of each other , insensitive to initial conditions and interaction dynamics . Intermixing was experimentally observed in two obligatory cooperative systems: an engineered yeast community cooperating through metabolite-exchanges and a methane-producing community cooperating through redox-coupling . Even in simulated communities consisting of several species , most of the strongly-cooperating pairs appeared intermixed . Thus , when ecological interactions are the major patterning force , strong cooperation leads to partner intermixing .
Biological interactions drive pattern formation at different levels of organization ( Murray , 2003 ) , ranging from developmental patterning within multicellular organisms and biofilms ( Shapiro , 1998; Lewis , 2008; Vlamakis et al . , 2008; Chuong and Richardson , 2009 ) , to ecological patterning within multi-species communities ( Levin , 1992; Rietkerk and van de Koppel , 2008; Momeni et al . , 2011 ) . Patterning , reflecting the relative spatial positioning of individuals with respect to each other , can be critical for the proper functioning of a community . Consider microbial communities: in a synthetic community , three bacterial species , each contributing an essential benefit while simultaneously competing for these benefits , can only grow when they are separated by an intermediate distance ( Kim et al . , 2008 ) ; different types of patterning are correlated with different levels of biofilm growth ( Christensen et al . , 2002; Brenner and Arnold , 2011 ) ; branching colony morphology allows more effective spreading across a nutrient-poor surface ( Levine and Ben-Jacob , 2004 ) ; and in waste treatment granules , the layered pattern of bacteria and archaea is thought to facilitate the sequential degradation of substrates ( Satoh et al . , 2007 ) . Despite the wide-ranging importance of microbial communities in , for example , human health and the biogeochemical cycling of elements , it is still unclear how cell–cell and cell–environment interactions govern the patterning of communities ( Elias and Banin , 2012 ) . Understanding the mechanistic basis of pattern formation from observations of natural communities is stymied by the multitude of cell–cell and cell–environment interactions , as well as environmental variations within and across communities . Thus , it is not uncommon to observe qualitatively different patterns in samples of essentially the same type of community ( Christensen et al . , 2002; Wilmes et al . , 2008; Dekas et al . , 2009 ) . To circumvent the lack of control in natural communities , we employed mathematical and experimental systems to systematically investigate how different types of ecological interactions might lead to distinct community patterning . Interactions can be classified into different ecological types based on their fitness effects on the interacting partners . We focused on the fitness effects rather than the molecular mechanisms of interactions , because diverse molecular mechanisms , ranging from physical associations in cell coaggregates and biofilms ( Kolenbrander et al . , 2010 ) to chemical interactions such as quorum sensing ( Parsek and Greenberg , 2005 ) , toxin warfare ( Vetsigian et al . , 2011 ) , and metabolite supply ( Christensen et al . , 2002 ) , all have fitness consequences which can be positive , neutral , or negative . Among different ecological interactions , we have placed a special emphasis on strong cooperation , interactions with large positive fitness effects on both partners including obligatory cooperation . This is because 1 ) it is important in a wide variety of microbial communities ranging from syntrophic systems critical for nutrient cycling ( Schink , 2002; Falkowski et al . , 2008; McInerney et al . , 2009 ) to pathogenic biofilms ( Kelly , 1980; Kolenbrander et al . , 2010; Elias and Banin , 2012 ) ; and 2 ) the codependence between cooperative partners poses special challenges for isolating and culturing cells ( Schink , 2002; McInerney et al . , 2009 ) . We investigated patterning in three-dimensional communities grown from two fluorescently-marked populations of cells initially randomly distributed on top of a surface ( Figure 1A ) . Starting with a generalized model based on fitness effects of ecological interactions between two populations ( A and B ) occurring at a local scale ( ‘fitness model’ ) , we predicted: 1 ) interactions benefiting at least one partner could potentially allow initially disparate partner ratios to converge over time , and 2 ) unlike other types of ecological interactions that caused partner segregation or layering of one population over the other ( A over B or B over A ) , strongly cooperating partners intermixed by forming patches that successively accumulated on top of each other ( A over B over A over B , etc ) . We tested these predictions experimentally in obligatory cooperative systems including engineered yeast communities and syntrophic methanogenic biofilms . Finally , we used the fitness model to show ‘strongly cooperating partners intermix’ could be generalized to communities consisting of multiple species . 10 . 7554/eLife . 00230 . 003Figure 1 . The fitness model generates two ecological patterning predictions . ( A ) In all simulated and experimental communities ( see ‘Materials and methods’ ) , two populations of cells , marked in red and green , were initially randomly distributed on a surface unless otherwise stated . The two populations engaged in one of the six ecological interactions . Population ratios for the entire community and patterns of vertical cross sections were examined . ( B ) The fitness model predicts that strong interactions beneficial to at least one partner can potentially lead to the convergence of initially disparate population ratios ( Figure 1—figure supplement 2 ) . ( C ) – ( H ) Representative vertical cross-sections of simulated communities , each engaging in one of the six types of ecological interactions , are presented . The fitness effects of ↑and ↓are large compared to the non-zero basal fitness of the recipient ( Figure 1—source data 1 ) , and therefore [↑ ↑] is strong facultative cooperation . ( I ) Vertical cross-sections of single-cell thickness from cooperative communities show significantly more intermixing than those from other communities ( n = 28 sections; p<0 . 01 , Mann–Whitney U test ) . An intermixing index of 6 ( red dotted line ) or above separates cooperative from non-cooperative communities in our simulations . To reduce the correlation of sections sampled from the same community , nearest sections were separated by at least seven sections . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 00310 . 7554/eLife . 00230 . 004Figure 1—source data 1 . Parameter values used in the fitness model . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 00410 . 7554/eLife . 00230 . 005Figure 1—figure supplement 1 . Cell rearrangement in simulations follows experimental observations on cells that are not actively motile . ( A ) To incorporate realistic assumptions about cell rearrangement in three-dimensional communities , we monitored the growth of a single fluorescent yeast cell into a microcolony on top of solid agarose ( Video 1 ) . Starting from one budding cell ( a , hour 0 . 5 ) , cells initially divided and pushed other cells to the side such that all cells remained in the same plane ( b , hour 9 ) . When the group of cells reached approximately 5-cell radius ( c , hour 14 ) , cells in the center could no longer push other cells to the side and instead budded upward , as indicated by higher intensities in the fluorescence images . Continued growth of microcolony forced more cells in the middle to send their progeny to upper layers , while cells close to the edge could still push other cells to the side and remained on the agarose surface ( d , hour 17 . 5 ) . Scale bars are 20 μm . ( B ) In simulations of the diffusion and fitness models , a focal budding cell ( star ) was assumed to bud toward the nearest empty space within its coplanar ( x , y ) confinement neighborhood of n-cell radius ( left panel , magenta circle ) . If the nearest empty space was not unique , a random choice was made . Cells along the path ( yellow rim ) to the nearest empty space shifted positions ( left panel , cyan arrows ) to accommodate the new cell . During rearrangement , if a cell moved to a new position with no cells immediately below it , the cell was lowered until it landed on another cell or the agarose surface . If the confinement neighborhood was filled , the cell budded upward and pushed up all cells above it ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 00510 . 7554/eLife . 00230 . 006Figure 1—figure supplement 2 . The fitness model predicts that convergence of population ratios is possible when an interaction benefits at least one partner . ( A ) The fitness model demonstrated that interactions conferring large benefit to at least one partner could lead to ratio convergence . Simulation parameters are listed in Figure 1—source data 1 . ( B ) We explored conditions for community ratio convergence ( ‘Requirements for steady-state ratios in the six types of communities’ in ‘Materials and methods’ ) by looking for the existence of stable steady states in which the two partners in an interaction neighborhood would reach identical fitness r^G−r^R=0 ( solid circles ) . Open and solid circles denote unstable and stable steady states , respectively . Here , ϕG* is the steady state fraction occupancy of G so that ϕG*:ϕR* is the steady state ratio . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 00610 . 7554/eLife . 00230 . 007Figure 1—figure supplement 3 . Strong mutual antagonism can lead to rapid divergence of population ratios . The fitness model shows that if two partner populations inhibit each other sufficiently strongly compared to their basal fitness , population ratios may diverge rapidly . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 007
To search for ecological patterning rules in microbial communities , we built a three-dimensional fitness model that ignored molecular details and instead focused on the fitness effects of interactions ( ‘The fitness model’ in ‘Materials and methods’ ) . Specifically , two populations of cells , marked as red and green , were initially randomly distributed on a surface ( Figure 1A ) . Cells grew horizontally until sufficiently confined , at which point they grew upward to accommodate the birth of new cells ( Figure 1—figure supplement 1 , Video 1 ) . Thus , no active cell motility was present during community growth . The growth rate of a focal cell was determined by its basal fitness as a single cell and by its interactions with other cells in a defined interaction neighborhood . To reflect the negative fitness effects of intra- and inter-population competition for shared resources ( [∼ ∼] ) , the fitness of the focal cell was decreased as the total population size in the interaction neighborhood increased . In addition , the focal and partner cells affected each other's fitness positively ( ↑ ) , negatively ( ↓ ) , or neutrally ( ∼ ) . The magnitude of fitness effect is quantitative . Thus , to obtain qualitative ecological patterning rules , we focused on strong interactions in which ↑ and ↓ exert fitness effects large enough to be comparable to the recipient's basal fitness . In most simulations using the fitness model , the basal fitnesses of both partners were non-zero , and therefore ↑ represented strong facultative interactions: a participant could survive on its own at its basal fitness but fared much better in the presence of its partner . 10 . 7554/eLife . 00230 . 008Video 1 . Yeast cells bud to the sides when there is available space and bud upward when sufficiently confined ( corresponding to Figure 1—figure supplement 1A ) . To infer the process of cell rearrangement in three-dimensional communities , we monitored how single YFP-fluorescent yeast cells grew into microcolonies on top of solid agarose . Initially , dividing cells pushed other cells to the side such that all cells remained in the same plane . When a cell was sufficiently confined from the sides by other cells ( approximately within a 5-cell radius ) , it could no longer bud to the side and instead budded upward , as indicated by higher intensities in the fluorescence images . Continued growth of microcolony forced more cells in the middle to send their progeny to upper layers , while cells close to the edge could still push other cells to the side and remain on the agarose surface . All images are taken with the same exposure time . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 008 We first analyzed the population composition of communities formed in the fitness model . Simulations ( Figure 1—figure supplement 2 and Figure 1—source data 1 ) and analytical calculations ( ‘Requirements for steady-state ratios in the six types of communities’ in ‘Materials and methods’ ) show that for interactions benefiting at least one partner , including cooperation ( [↑ ↑] ) , commensalism ( [∼ ↑] ) , and exploitation ( [↓ ↑] ) , different initial partner ratios can potentially converge over time ( Figure 1B ) . Ratio convergence requires ‘balanced’ fitness . For example , for A[∼ ↑]B to achieve ratio convergence , the basal fitness of A must be higher than that of B and after gaining the fitness benefit from commensalism , B must be able to grow at least as fast as A . This way , a situation incompatible to ratio convergence , that is , one partner always fitter than the other , does not occur . Using engineered competitive , obligatory commensal , and obligatory cooperative yeast communities ( see below ) , we confirmed that population ratios converged in the commensal and cooperative but not competitive communities ( Figure 2—figure supplement 1 ) . This convergence of population ratios reflects a balance between supply and consumption when growth is limited by supply ( Shou et al . , 2007 ) : if , for instance , the supplier population suddenly increased in relative abundance , then each individual in the consumer population would receive more benefit and grow faster , which would return the ratio to its original value . We next examined patterns in vertical cross-sections of communities simulated by the fitness model ( Figure 1A , parameters in Figure 1—source data 1 ) , because patterning along the x and y directions can depend on the initial spatial distribution of cells whereas patterning along the vertical z direction results from growth under the fitness influences of ecological interactions . In [∼ ∼] , [∼ ↓] , and [↓ ↓] , red and green populations primarily formed columns that are spatially segregated from each other ( Figure 1C–E ) . In [↑ ↓] and [∼ ↑] , frequently one of the populations ( green ) either formed a column or became covered by the partner population ( red ) ( Figure 1F–G ) . Only in cooperation ( [↑ ↑] ) conferring large fitness benefits to both partners , the two partner populations appeared to be extensively ‘intermixed’ , manifested as the two different cell types successively piling on top of each other ( Figure 1H ) . To compare levels of intermixing in different communities , we defined an ‘intermixing index’ as the average number of cell type transitions spanning community height ( ‘Spatial analysis’ in ‘Materials and methods’ ) . Since the intermixing index can be a function of community height , we compared intermixing indexes of simulated communities at equivalent heights . Statistically significant differences were observed between strong cooperation versus other types of interactions ( Figure 1I ) . Thus , we predicted that partner intermixing would distinguish strong cooperation from other ecological interactions . To test the prediction that strong cooperation is the only ecological interaction capable of driving partner intermixing , we engineered yeast communities engaged in competitive ( [∼ ∼] ) , obligatory commensal ( [∼ ↑] ) , and obligatory cooperative ( [↑ ↑] ) metabolic interactions . Competitive communities represented the baseline intra- and inter-population competition common to ‘all’ communities , while commensalism served as the most stringent control to be discriminated against . All engineered yeast communities consisted of two non-mating S . cerevisiae strains , a G strain expressing GFP or YFP and an R strain expressing DsRed ( see ‘Materials and methods’ ) . In competitive communities , prototrophic R and G competed for shared nutrients in agarose and for space ( Figure 2—figure supplement 1A ) . Depending on their genetic backgrounds , the two strains engaged in either equal-fitness or unequal-fitness competition . In obligatory commensal communities , R→A←L took in lysine from the media and overproduced adenine to feed the adenine-requiring G←A ( Figure 2—figure supplement 1B ) . In obligatory cooperative communities ( previously described as Cooperation that is Synthetic and Mutually Obligatory , or “CoSMO” ) , R→A←L overproduced adenine and required lysine while G→L←A overproduced lysine and required adenine ( Shou et al . , 2007 ) ( Figure 2—figure supplement 1C ) . When mixed , the two cooperative strains could form a viable community , growing from low to high densities in synthetic minimal medium ( SD ) lacking adenine and lysine ( Shou et al . , 2007 ) . To predict and extrapolate experimental results of yeast communities , we developed a three-dimensional model based on the consumption , release , and diffusion of metabolites in the yeast communities ( ‘the diffusion model’ in ‘Materials and methods’ ) . Specifically , the diffusion model assumed that metabolites diffused in the community and agarose ( Figure 2—figure supplement 2 ) and that cell growth depended on the local concentration of its limiting metabolite according to Monod's equation ( Monod , 1949 ) . Most parameters in the diffusion model were measured experimentally ( Figure 2—source data 1 ) . We reasoned that if predictions from the diffusion model were consistent with experimental observations , we could use this model to simulate experiments that would be technically difficult to perform . To examine vertical patterning in yeast communities , we first used top-view time-lapse images to infer patterns which were subsequently verified by cryosectioning . This is because confocal and two-photon microscopy cannot penetrate deep into yeast communities ( Váchová et al . , 2009 ) . In competitive communities , whether in the diffusion model or experiments , time-lapse top-views suggested population segregation . For equal-fitness competitive communities ( Figure 2A , left; Video 2 ) , individual cells ( i ) grew into microcolonies ( ii ) which continued to grow and expand until they reached neighboring microcolonies ( iii ) . After this stage , cells were unable to push other cells to the side , and further cell divisions mainly occurred in the vertical z direction ( Figure 1—figure supplement 1 ) . Consequently , columns of primarily a single cell type formed , and top-views of the community remained static ( compare iv and v ) . Competitive communities composed of populations with different fitnesses developed similarly ( Figure 2A , right; Video 3 ) , except that the fitter population expanded in the top view during growth ( compare the green population in iv and v ) . 10 . 7554/eLife . 00230 . 009Figure 2 . Obligatory cooperation , but not competition or obligatory commensalism , results in substantial partner intermixing in engineered yeast communities and in communities simulated using the diffusion model . Competitive communities of strains with equal fitness ( equal-fitness competition , abbreviated as ‘Eq-fitness Comp . ’ ) showed population segregation as suggested by static late-stage top-views ( A , left ) and columnar cross-section patterns ( D , left ) . When competing strains had different fitnesses ( unequal-fitness competition , abbreviated as ‘Uneq-fitness Comp . ’ ) , the fitter population G expanded during growth , as evident in top views ( A , right ) and vertical cross-sections ( D , right ) . In obligatory commensal communities , since one supplier could support the birth of multiple consumers , consumers eventually overgrew and covered suppliers ( top-views in B and vertical cross-sections in E ) . Obligatory cooperative communities showed substantial population intermixing as suggested by alternating cell types in top-views ( C , 6× magnification insets in experiments ) and patchy cross-section patterns ( F ) . Top views of communities from the diffusion model integrate intensity and color over height such that brighter colors represent higher cell numbers and yellowness indicates the simultaneous presence of green and red . Scale bar: 100 μm . All communities started from total 500 cells/mm2 and R:G = 1:1 . ( G ) Quantification of intermixing in experimental ( symbols ) and diffusion-model ( lines ) communities showed that while the intermixing index remained low for commensal ( brown ) and competitive ( grey and black ) communities , it increased linearly with community height in obligatory cooperative ( magenta ) communities . Results from the diffusion model underestimated intermixing indexes because a confined cell was modeled to divide strictly vertically upward ( Figure 1—figure supplement 1 ) without allowing cell movements in horizontal directions ( Figure 3—figure supplement 1F ) . ( H ) A conceptual model illustrates the development of intermixing over time in a strongly cooperative community with 1:1 steady-state population ratio . Local deviations from the steady-state ratio result from asymmetric partner properties and/or stochastic fluctuations in cell numbers ( i ) . The under-abundant population ( red ) grows faster than its neighboring over-abundant partner ( green ) . Due to the spatial localization of large cooperative benefits , cells near population borders grow faster than those farther away . Consequently , cells from the initially under-abundant red population at the border divide the fastest . Progeny that pile on the green partner have more access to cooperative benefits than those who do not ( ii ) , which favors intermixing . When the previously over-abundant partner becomes under-abundant ( iii ) , piling-up in the opposite direction occurs ( iv , green on red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 00910 . 7554/eLife . 00230 . 010Figure 2—source data 1 . Definitions and values of parameters used in the diffusion model . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01010 . 7554/eLife . 00230 . 011Figure 2—figure supplement 1 . In engineered yeast communities , obligatory cooperation and obligatory commensalism allow initially different partner ratios to converge over time . Engineered yeast communities consisted of two non-mating fluorescent strains of yeast , G and R , engaged in a metabolic interaction . ( A ) In competitive communities , two prototrophic strains engaged in equal-fitness or unequal-fitness competition for shared resources . ( B ) In obligatory commensal communities , R→A←L took in lysine from the media and overproduced adenine to feed G←A . ( C ) In obligatory cooperative communities , R→A←L overproduced adenine and required lysine while G→L←A overproduced lysine and required adenine . ( D ) – ( H ) Symbols: experiments; lines: results from the diffusion model . ( D ) In competitive communities , the ratio between two equally-fit populations remained steady ( grey ) , whereas the ratio between two populations with unequal fitness ( G 20% fitter than R ) changed monotonically ( black ) in favor of the fitter population . Population ratios of obligatory commensal ( E ) and obligatory cooperative ( F ) communities converged regardless of their initial values ( different colors representing different initial conditions ) . ( G–H ) Ratio fluctuations observed in experimental cooperative communities were better recapitulated in diffusion models that incorporated additional specifics of the yeast system . ( G ) Incorporating a slower diffusion constant in the community ( ∼20 μm2/s ) than that in agarose ( 360 μm2/s ) allowed ratio convergence to more closely resemble experimental data . ( H ) Incorporating experimental observations , such as better starvation tolerance of adenine-requiring cells compared to lysine-requiring cells and the consequent delay in lysine release compared to adenine release ( Shou et al . , 2007 ) , allowed ratio convergence to more closely resemble experimental data . All communities were initiated at 3000 cells/mm2 on top of a 24-mm-high column of agarose . Population ratios in experiments were measured using flow cytometry ( see ‘Materials and methods’ ) . As predicted by the fitness model , population ratios converged as time progressed in obligatory cooperative and obligatory commensal but not competitive communities both in the experiments and the diffusion model . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01110 . 7554/eLife . 00230 . 012Figure 2—figure supplement 2 . Basic assumptions in the diffusion model . In the diffusion model , community grew on the top surface of an agarose column ( light blue ) . Cell grids ( black borders ) represented cells at different states: live R ( light red ) , live G ( light green ) , dead R ( dark red ) , dead G ( dark green ) , and no cells ( white ) . ( A ) When the diffusion constants of nutrients in the yeast community were assumed to equal those in agarose ( 360 μm2/s ) , nutrient grids ( orange ) were used to simulate the distribution of nutrients . ( B ) When the diffusion constants of nutrients in the yeast community were modeled to be that of the fluorescent dye Sulforhodamine 101 ( ∼20 μm2/s; Figure 2—figure supplement 1G ) , we implemented community diffusion and agarose diffusion grids of 3- and 12-cell grid width , respectively , to accommodate two diffusion constants . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01210 . 7554/eLife . 00230 . 013Figure 2—figure supplement 3 . Cooperative communities exhibit a characteristic patch size associated with the spatial localization of benefits . ( A ) Starting from densities and ratios spanning orders of magnitude , experimental obligatory cooperative yeast communities developed a consistent characteristic vertical patch size λz* of ∼10 to 20 μm . ( B ) The diffusion model shows that when both released nutrients were instantly and uniformly distributed throughout a community ( ‘Coop . Instant Distr . ’ ) , intermixing was diminished compared to when nutrients diffused in the community ( 360 µm2/s , ‘Coop . ’ ) . Excessive release of nutrients ( 200× compared to typical values in Figure 2—source data 1 ) also significantly reduced population intermixing ( ‘Coop . Excessive Release’ ) , since the benefits exceeded the uptake capacity of partner cells and therefore no longer remained localized . All simulations started with 2000 total cells/mm2 and 1:1 population ratio . Box plots show the 25th to 75th percentile range , with the median marked with a line and whiskers extending to the 95% confidence interval . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01310 . 7554/eLife . 00230 . 014Figure 2—figure supplement 4 . Obligatory cooperative yeast partners intermix . Vertical cross-sections in ( A ) and ( B ) were from different growth stages of obligatory cooperative yeast communities starting at total 500 cells/mm2 and 1:1 population ratio . Vertical cross-section in ( C ) and ( D ) were respectively from the center and the edge of an obligatory cooperative yeast community started from a high-density inoculum and allowed to expand to new territories . Areas enclosed in the yellow boxes were 3× magnified in the right panels to show details of intermixing . The top-view schematic insets in ( C ) and ( D ) show the relative position of the corresponding cross-section with regard to the initial inoculum ( dashed line ) . All scale bars are 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01410 . 7554/eLife . 00230 . 015Video 2 . Top views of an equal-fitness competitive community suggest population segregation ( corresponding to Figure 2A , left ) . Competitive communities of strains with equal fitness showed population segregation as suggested by static late-stage top-views . The community started from a uniform distribution of total 500 cells/mm2 and R:G = 1:1 . Intensities of both fluorescent channels in images at different times were normalized to the same maximum value for better representation of patterns throughout growth . Blanks in the video were due to removal of the dish to sample other replicate communities for flow-cytometry or sectioning . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01510 . 7554/eLife . 00230 . 016Video 3 . In unequal-fitness competition , the fitter population gradually covers the less fit population ( corresponding to Figure 2A , right ) . Here , G is fitter than R . The community started from a uniform distribution of total 500 cells/mm2 and R:G = 1:1 . Intensities of both fluorescent channels in images at different times were normalized to the same maximum value for better representation of patterns throughout growth . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 016 In obligatory commensal communities , time-lapse top-views suggested a ‘layered’ pattern with one population covering the other ( Figure 2B; Video 4 ) . We found that in both the diffusion model and experiments , R→A←L initially grew rapidly by consuming lysine in the agarose ( compare i , ii , and iii ) . In contrast , G←A initially grew slowly , presumably limited by the low level of adenine released by R→A←L before R→A←L became abundant ( compare i , ii , and iii ) . During the growth of R→A←L , some G←A cells had been pushed to the top layer of the community ( iii ) . As the R→A←L population continued to expand and release adenine , G←A started to grow rapidly ( compare iii and iv ) . Eventually , R→A←L stopped growing after lysine in the agarose had been depleted . Since the amount of adenine released during the lifetime of every R→A←L cell could support the birth of multiple G←A cells ( Shou et al . , 2007 ) , G←A population outnumbered and covered R→A←L ( v ) . 10 . 7554/eLife . 00230 . 017Video 4 . Top views of an obligatory commensal community suggest population layering ( corresponding to Figure 2B ) . For a detailed explanation of the growth kinetics of the community ( R→A←L[∼↑]G←A ) , please refer to the main text . The community started from a uniform distribution of total 500 cells/mm2 and R→A←L:G←A=1:1 . Intensities of both fluorescent channels in images at different times were normalized to the same maximum value for better representation of patterns throughout growth . Blanks in the video were due to removal of the dish to sample other replicate communities for flow-cytometry or sectioning . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 017 In contrast , time-lapse top-views of obligatory cooperative communities suggested population intermixing ( Figure 2C; Video 5 ) . After plating on agarose lacking adenine and lysine ( i ) , cells in both populations divided once or twice by utilizing metabolites stored in their vacuoles ( Shou et al . , 2007 ) . R→A←L released adenine and entered the death phase while G→L←A continued to grow by utilizing the released adenine ( ii ) ( Shou et al . , 2007 ) . G→L←A entered the death phase and released lysine after a significant delay due to their better starvation tolerance compared to R→A←L ( Shou et al . , 2007 ) . Released lysine supported the growth of surviving R→A←L cells into microcolonies ( iii ) . R→A←L in turn released adenine which promoted growth of nearby G→L←A cells and led to partial covering up of R→A←L microcolonies by rapidly growing G→L←A ( iv and v , insets ) . Abundant G→L←A cells provided enough lysine for rapid growth of local R→A←L cells , which gave rise to patches of R→A←L cells on top of the community ( vi , inset ) . These community growth kinetics were consistent with previous measurements in liquid cultures ( Shou et al . , 2007 ) . 10 . 7554/eLife . 00230 . 018Video 5 . Top views of an obligatory cooperative community suggest populations intermixing ( corresponding to Figure 2C ) . For a detailed explanation of the growth kinetics of the community ( R→A←L[↑ ↑]G→L←A ) , please refer to the main text . The community started from a uniform distribution of total 500 cells/mm2 and R→A←L:G→L←A=1:1 . Intensities of both fluorescent channels in images at different times were normalized to the same maximum value for better representation of patterns throughout growth . Blanks in the video were due to removal of the dish to sample other replicate communities for flow-cytometry or sectioning . The video started at ∼300 hr , after the formation of R→A←L microcolonies . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 018 To confirm that partners intermixed in cooperative but not non-cooperative communities , we obtained vertical cross-sections of competitive , commensal , and cooperative communities at their maximal community heights ( ‘Cryosectioning’ for experimental communities in ‘Materials and methods’ ) . We found that in both the diffusion model and experiments , equal-fitness competitive communities formed segregated columns ( Figure 2D , left ) . Unequal-fitness ( Figure 2D , right ) and obligatory commensal communities ( Figure 2E ) displayed a layered pattern in which the top portion of a community was dominated by one partner . In contrast to competitive and obligatory commensal communities , obligatory cooperative communities appeared to show population intermixing with patches of red and green cells emerging on top of each other ( Figure 2F and Figure 2—figure supplement 4A and B ) . Cooperative intermixing appeared to be less in the diffusion model than in yeast communities . This is presumably because the diffusion model assumed that a confined cell would bud strictly upward whereas in yeast communities , cell divisions could stray to the side ( Figure 3—figure supplement 1F ) . To compare levels of intermixing in different communities , we quantified the intermixing index of each community ( Figure 2G , symbols for experiments and lines for results from the diffusion model ) . As expected , equal-fitness competition resulted in small intermixing indexes at all community heights ( Figure 2G , grey ) . In unequal-fitness competition ( Figure 2G , black ) and obligatory commensalism ( Figure 2G , brown ) , the formation of a layered pattern caused a small increase in the intermixing index that subsequently leveled off . In contrast , the intermixing index in obligatory cooperative communities increased proportionally to community height in both experiments and the diffusion model ( Figure 2G , magenta ) . This proportionality suggested the existence of a characteristic patch size ( ‘Spatial analysis’ in ‘Materials and methods’ ) , denoted λz* , of 10–20 µm . The characteristic patch size was independent of initial conditions ( Figure 2—figure supplement 3A ) . Indeed , a calculation of the patch size based on experimentally-determined release , diffusion , and consumption of exchanged metabolites yielded comparable results ( ‘Calculating the characteristic patch size in cooperative yeast communities’ in ‘Materials and methods’ ) . What caused partner intermixing during cooperation ? Using the diffusion model , we found that if cooperative benefits were not spatially localized because of instant distribution of benefits throughout the community or because of excessive supply levels , intermixing was diminished ( Figure 2—figure supplement 3B ) . Based on this result and based on patterns observed in top-views and vertical cross-sections of cooperative communities , we propose that local deviations from the steady-state ratio coupled with localized large cooperative benefits cause partners to ‘take turns’ to grow , which leads to population intermixing ( Figure 2H ) . In summary , in communities engaging in strong cooperation , but not in communities governed by other types of ecological interactions , the intermixing index increases proportionally as a function of community height . This proportionality is due to a fixed patch size determined by localized nutrient supply and consumption . Under what conditions can we observe partner intermixing in cooperation ? First , we experimentally tested partner intermixing in obligatory yeast cooperative communities initiated under different population ratios and densities . Next , we used the diffusion model to examine communities in which obligatory cooperative partners interacted with different dynamics . Finally , using the fitness model , we tested intermixing in facultative cooperation by varying the relative magnitude of cooperative benefits compared to the basal fitness of the two interacting populations . We utilized yeast communities when experimentally possible and otherwise took advantage of the diffusion and the fitness models . We found that intermixing was insensitive to initial conditions in the yeast obligatory cooperative communities . The initial partner ratio did not significantly affect the level of intermixing ( Figure 3—figure supplement 1D ) . At very high initial cell densities , we observed significant intermixing in communities directly above the inoculation area even in the absence of cooperation ( Figure 3—figure supplement 1E , yellow shading ) . This is because high cell densities put different cell types in close proximity to one another , and cell divisions that were not perfectly straight upward ( Figure 3—figure supplement 1F ) caused intermixing . However , we reasoned that communities beyond the inoculation area might reveal patterns indicative of the underlying interactions , because these regions are formed by cell growth under the fitness influences of interactions . To test this possibility , we spotted cell mixtures at high densities on agarose and allowed the community to expand to new territories beyond the inoculation area . Even though communities directly above the inoculum always appeared highly intermixed ( Figure 3A , ‘Center’ and Figure 2—figure supplement 4C ) , in edge sections , significant intermixing was only observed in cooperative communities ( Figure 3A , ‘Edge’ and Figure 2—figure supplement 4D ) . 10 . 7554/eLife . 00230 . 019Figure 3 . Strongly cooperating populations intermix under a wide range of conditions . ( A ) In engineered yeast communities , even though both obligatory cooperative and non-cooperative communities directly above the high-density inoculation spot showed high population intermixing ( ‘Center’ ) , edge sections ( ‘Edge’ ) of obligatory cooperative communities showed significantly more intermixing than those from non-cooperative communities ( Mann–Whitney U test , p<5 × 10−5 ) . A total of 106 R and G cells at a 1:1 ratio were deposited in an inoculation spot of ∼2 mm2 , corresponding to 10 cell layers . Communities were allowed to grow and expand beyond the inoculation spot on a 4-mm-tall agarose pad of 500 mm2 area . Vertical sections from the edges were taken at a height approximately half of the maximum community height at the center . Box plots show the 25th to 75th percentile range , with the median marked with a line and whiskers extending to the 95% confidence interval . ( B ) In simulations using the fitness model , facultative cooperation conferring smaller fitness benefits required greater community heights to exhibit a significant level of intermixing . The strength of facultative cooperation is shown as the ratio of fitness benefit received from each cooperative partner cell in the interaction neighborhood to the basal fitness of the focal cell . Simulation parameters can be found in Figure 3—source data 1 . Error bars indicate 95% confidence interval . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 01910 . 7554/eLife . 00230 . 020Figure 3—source data 1 . Parameter values used in the fitness model in Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 02010 . 7554/eLife . 00230 . 021Figure 3—figure supplement 1 . Intermixing is observed in obligatory cooperative communities over a wide range of conditions . ( A ) Partner intermixing is insensitive to diffusion constants . In the diffusion model , obligatory cooperative communities with diffusion constants ranging from 20 to 360 µm2/s ( corresponding to the diffusion constants of Sulforhodamine 101 in yeast community and agarose , respectively ) both showed significant partner intermixing during growth . ( B ) Partner intermixing is insensitive to the kinetics of interactions . With or without incorporating details such as significantly better starvation tolerance in adenine-requiring cells compared to lysine-requiring cells and the consequent delay in lysine release compared to adenine release ( Shou et al . , 2007 ) , the diffusion model ( lines ) yielded substantial population intermixing . Diffusion constant within the community was assumed to be 20 μm2/s . In both ( A ) and ( B ) , communities started from 3000 total cells/mm2 at 1:1 population ratio . ( C ) Intermixing does not depend on asymmetry between cooperating partners . We examined symmetric obligatory cooperation by assuming that for the two partners , all live cells released nutrients at a constant and identical rate without delay , and that other properties such as rates of growth , death , and nutrient consumption were also identical . Partner intermixing in this symmetric obligatory cooperative system ( Sym . coop ) was similar to that in asymmetric obligatory yeast cooperative system ( Coop ) . Simulated communities started from total 500 cells/mm2 at 1:1 population ratio . ( D ) Cooperative intermixing is insensitive to initial ratios . Engineered yeast communities were initiated at a total cell density of 3000 cells/cm2 and grown to near the carrying capacity of the agarose column . Box plots show 25th to 75th percentile values of the intermixing index , with the median marked with a line and whiskers extending to 95% confidence interval . To facilitate comparisons , intermixing indexes in communities with different heights were linearly normalized to correspond to a height of ∼40 cell layers ( 200 µm ) . ( E ) Experimental obligatory cooperative communities ( magenta ) showed more intermixing than obligatory commensal ( brown ) , equal-fitness competitive ( grey ) , and unequal-fitness competitive ( black ) communities , except when the initial cell densities exceeded 10% confluence ( yellow shading ) . Large intermixing indexes at high initial cell densities were due to the fact that the budding direction of a cell under spatial confinement was not exactly vertical ( F ) . To facilitate comparisons , intermixing indexes in communities with different heights were linearly normalized to correspond to a height of ∼30 cell layers ( 150 µm ) . Communities were initiated at R:G = 1:1 and grown until near the carrying capacity of the agarose . Box plots show the 25th to 75th percentile range , with the median marked with a line and whiskers representing 95% confidence intervals . ( F ) An equal-fitness competitive community was initiated at confluent density ( 50000 cells/mm2 ) and R:G = 100:1 to ensure that single green cells were laterally confined by red cells . Cross-sections showed that the early progeny of a single cell was distributed in a division zone of ∼3-cell rather than 1-cell width . Lateral budding by late progeny further widens the zone to ∼5-cell diameter in a mature community . Scale bar is 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 021 Intermixing is insensitive to interaction dynamics , so long as large cooperative benefits remain sufficiently localized for both partners . As described above , in the diffusion model , excessive supply amounts or instant distribution of benefits throughout the community diminished intermixing ( Figure 2—figure supplement 3B ) . The former can occur if the local availability of cooperative benefits is not growth-limiting . However , cooperative benefits are unlikely to be available in large excess because of the potential fitness cost of generating benefits and because of competition for these benefits in the partner population . As far as diffusion is considered , the diffusion model showed significant population intermixing in obligatory cooperative communities with diffusion constants varying more than 10-fold ( Figure 3—figure supplement 1A ) . Additional calculations showed that intermixing was largely insensitive to diffusion kinetics , because the characteristic patch size was related to the diffusion constant by 1/5 power ( ‘Calculating the characteristic patch size in cooperative yeast communities’ in ‘Materials and methods’ ) . The diffusion model also showed that intermixing was present with or without a delay in G→L←A supplying lysine benefits to the partner ( Figure 3—figure supplement 1B ) . Furthermore , even though asymmetry between properties of partners ( Figure 2C ) would seem to facilitate intermixing ( Figure 2H ) , asymmetry in partner properties was not required to generate intermixed patterns: partners that grew , died , and released and consumed metabolites at identical rates intermixed ( Figure 3—figure supplement 1C ) . Using the fitness model , we found that in facultative cooperation , small fitness benefits generated less intermixing than large fitness benefits ( Figure 3B ) . This result is intuitive: facultative cooperation should resemble obligatory cooperation if fitness benefits are large for both partners . When only one partner receives a large fitness benefit , facultative cooperation should resemble obligatory commensalism . Finally , small fitness benefits for both partners will make facultative cooperation resemble competition . In facultative cooperation with smaller fitness benefits , intermixing would be apparent if communities could grow to greater heights . Further experiments are required to test these predictions . Together , these results suggest that intermixing relies on spatial localization of cooperative benefits that are sufficiently large to both partners , and is otherwise insensitive to initial conditions or the detailed kinetics of interactions . To test whether cooperative patterning applies to other biological systems , we examined redox-coupling in a two-species methane-producing biofilm consisting of the bacterium Desulfovibrio vulgaris and the archaeon Methanococcus maripaludis . In the absence of sulfate and hydrogen , the two species engage in obligatory mutualism: D . vulgaris ferments lactate and promotes the growth of M . maripaludis by supplying the electron donor H2 , while M . maripaludis promotes the growth of D . vulgaris by consuming the H2 gas which can be inhibitory to D . vulgaris under these conditions ( Figure 4A ) . Similar types of syntrophic interactions leading to methane production typically occur in microbial consortia that digest organic compounds in freshwater sediments , sewage treatment plants , and the guts of ruminants ( Conrad et al . , 1985; Schink , 1997 ) . 10 . 7554/eLife . 00230 . 022Figure 4 . Obligatory cooperation through redox-coupling leads to partner intermixing . ( A ) In the absence of sulfate and hydrogen , the bacterium Desulfovibrio vulgaris ( Dv ) and the archaeon Methanococcus maripaludis ( Mm ) cooperate through redox coupling . Dv ferments lactate and produces mainly acetate , CO2 , and H2 . However , this reaction is not thermodynamically favorable unless H2 is kept at very low concentrations . H2 is used by Mm to reduce CO2 to methane . ( B ) In cooperative biofilms of Dv ( green ) and Mm ( red ) , the intermixing index increased with height . Cell identification relied on FISH ( see ‘Materials and methods’ ) . Scale bar: 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 022 Vertical cross-sections of independent D . vulgaris–M . maripaludis biofilms indeed exhibited increasing intermixing as a function of community height ( Figure 4B ) . Thus , naturally mutualistic microbes cooperating through a coupling mechanism different from metabolite exchange also exhibited a significant level of intermixing . Other known cooperative communities , including those degrading herbicide pollutants ( Breugelmans et al . , 2008 ) and those colonizing teeth ( Palmer et al . , 2001 ) also seemed to display intermixed patterns , although we do not know whether the intermixing indexes of these communities increased linearly with height . In communities with more than two species , indirect interactions can obscure direct interactions ( Wootton , 2002 ) . For example , if A promotes B which inhibits C , it will appear that A inhibits C . This is akin to indirect interactions between gene products in a cell . To test whether intermixing between cooperators was affected by other members of a community , we used the fitness model to simulate communities composed of five interacting species ( Figure 5 and Figure 5—figure supplement 1 ) . We randomly assigned one of the six possible ecological interactions between each pair of populations , and consequently , each network had 10 pair-wise interactions . The fitness effects of ↑ and ↓ were sufficiently large to be comparable to the recipient's basal fitness . A total of 240 pair-wise interactions in 24 independent communities were examined in the fitness model . In most cases ( 26 out of 31 ) , cooperative pairs intermixed ( Figure 5A , D ) . Occasionally , commensal pairs ( Figure 5B , ② [∼ ↑] ③ ) showed substantial intermixing ( 3 out of 58 commensal pairs , Figure 5D ) , or cooperative pairs ( e . g . , Figure 5C , ① [↑ ↑] ③ ) showed little intermixing ( 5 out of 31 , Figure 5D ) , presumably due to interactions through other community members ( Figure 5B , ③ indirectly promoted ② through ④; Figure 5C , ③ promoted ⑤ which inhibited ① ) . These deviations are consistent with the notion that the intermixing index reflects the ‘overall’ interaction between two partners , integrating any additional fitness effects of indirect interactions through other community members . Thus , strongly cooperating partners intermix while deviations from this expectation reflect the presence of strong indirect interactions . 10 . 7554/eLife . 00230 . 023Figure 5 . Most of the strongly-cooperative pairs intermixed in simulated five-species communities . ( A-C ) Examples of networks in which cooperative pairs intermixed ( A ) , non-cooperative pairs intermixed ( B ) , or cooperative pairs did not intermix ( C ) are shown . In the schematic network diagrams , line termini of → , ⊣ , and — represent ↑ , ↓ , and ∼ , respectively; cooperative pairs are highlighted in magenta . Simulations were performed using the fitness model . The basal fitness for each population was chosen randomly from a range spanning 0 . 03–0 . 05/hr . The fitness effect from each partner cell in the interaction neighborhood was either 0 for ∼ , or otherwise randomly chosen to be approximately 2–3% of basal fitness to achieve strong interaction . Initial population ratios were randomly assigned such that no population was initially lower than 5% of the total population . Simulations were run for 10 generations , and vertical cross-sections of the final communities were examined for intermixing . We considered intermixing index exceeding a threshold of 6 ( Figure 1I ) as significant ( red dotted lines ) . The remaining nine panels in Figure 5B and C are provided in Figure 5—figure supplement 1 . ( D ) Quantifying intermixing in a total of 240 interactions from 24 independent communities showed that most of the cooperative pairs intermixed ( magenta ) . Indirect interactions through other community members could make cooperative pairs not intermixed ( grey ) or non-cooperative ( commensal ) pairs intermixed ( brown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 02310 . 7554/eLife . 00230 . 024Figure 5—figure supplement 1 . The complete results of Figure 5B and C ( panel A and B , respectively ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00230 . 024
Patterning is driven by cell–cell and cell–environment interactions . Here , we examined how the net fitness effects of cell–cell interactions could influence patterning . Using defined mathematical and biological systems under controlled spatial environments , we have established two expectations for ecological patterning between two interacting partners . The first expectation is that interactions conferring large fitness benefits to at least one partner ( [↑ ↑] , [∼ ↑] , and [↓ ↑] ) can potentially result in ratio convergence ( Figure 1—figure supplement 2 and Figure 2—figure supplement 1 ) . Ratio convergence confers stability to community composition . The composition of a community is defined to be locally ( or globally ) stable if after small ( or any-size ) deviations from it , either due to initial conditions different from it or perturbations , the community eventually returns to this composition . An analytical calculation based on the fitness model showed that in [↑ ↑] and [∼ ↑] , global stability can be achieved whereas in [↓ ↑] , local stability can be achieved ( Figure 1—figure supplement 2; ‘Requirements for steady-state ratios in the six types of communities’ in ‘Materials and methods’ ) . Experimentally , obligatory cooperation and obligatory commensalism led disparate initial ratios to converge ( Figure 2—figure supplement 1 ) . In contrast , in competition between two populations of equal fitness , population ratio is determined by the initial value until a perturbation resets the value which will remain unchanged until the next perturbation . In competition between two populations of unequal fitness , the fitter population should increase in frequency and therefore , population ratio is unstable . In amensalism and mutual antagonism , population ratios at best have an unstable fixed point ( Figure 1—figure supplement 2B ) . This means that even if the population ratio may seem ‘fixed’ , slight deviations will send it farther and farther away from the original fixed value . The reason for this instability , using mutual antagonism as an example , is because if A inhibits B and B inhibits A ( A[↓ ↓]B ) , then an increase in population A will facilitate inhibition of B and therefore make B less able to inhibit A , further increasing the relative abundance of A . The second expectation is that strong cooperation results in partner intermixing in the direction vertical to the surface of initial colonization . Cooperative intermixing requires spatially localized benefits that are sufficiently large to both partners ( Figure 2—figure supplement 3 and Figure 3B ) , and otherwise appears to be robust against variations in initial conditions ( Figure 2—figure supplement 3A , Figure 3A , and Figure 3—figure supplement 1D ) or reaction dynamics ( Figure 3—figure supplement 1A–C ) . We have observed cooperative intermixing in the fitness model simulating strong facultative cooperation ( Figure 1 ) and in obligatory experimental systems including engineered yeast communities ( Figure 2 ) and syntrophic biofilms ( Figure 4 ) . The intermixing indexes in these communities increased linearly as a function of community height , because these communities exhibited fixed-size patches associated with the spatial localization of benefits ( Figure 2—figure supplement 3; ‘Calculating the characteristic patch size in cooperative yeast communities’ in ‘Materials and methods’ ) . This is in stark contrast with other ecological interactions which lead to segregated or layered patterns ( Figures 1 and 2 ) in which the intermixing index remained low or increased transiently before leveling off . In simulated multi-species communities , strongly cooperative pairs intermixed in most cases; cases where cooperative pairs did not intermix or non-cooperative pairs intermixed were likely caused by strong indirect interactions through other partners . In a theoretical study using a one-dimensional stepping-stone model , strongly cooperating partners also appeared much more mixed than competing partners ( Korolev and Nelson , 2011 ) . Our work defines the expected pattern created by pair-wise interactions if the fitness effects of interactions are the main driving force of patterning . The generality of these ecological patterning conclusions awaits further experimental validations . Many simple communities grown in laboratory environments conform to these findings . For instance , competing bacterial species form a columnar or layered pattern ( Palmer et al . , 2001; Kreth et al . , 2005; An et al . , 2006; Hallatschek et al . , 2007; Bernstein et al . , 2012 ) . Burkholderia sp . LB400 and Pseudomonas sp . B13 ( FR1 ) show spatial segregation as competitors when grown on citrate , but when grown on chlorobiphenyl they engage in metabolic commensalism and exhibit a layered pattern ( Nielsen et al . , 2000 ) . Similarly , when grown on a non-selective carbon source , Comamonas testosteroni WDL7 outcompeted and covered Variovorax sp . WDL1 , but when fed with an aromatic compound as the sole carbon source , competition switched to cooperation and the two populations intermixed ( Breugelmans et al . , 2008 ) . Cooperative oral bacteria species intermixed when grown on saliva ( Palmer et al . , 2001; Periasamy and Kolenbrander , 2009 ) . If cooperative intermixing is suspected , then examining whether the intermixing index increases linearly as a function of community height will be informative to exclude transient increases . We have described the expected patterning between two cell populations if the fitness effects of interactions are the major driving force . These expectations are abstract in the sense that in reality , no interactions can exist in the absence of molecular mechanisms or an abiotic environment . Assigning expected patterns to different types of ecological interactions will allow us to identify deviations from expectations . Determining the causes of such deviations will help us better understand the biology of a community . Fitness effects of ecological interactions rely on molecular mechanisms to manifest themselves . Different types of molecular mechanisms can conceivably alter patterning . For instance , in the fitness and the diffusion models and in the S . cerevisiae communities , cells divide upward when sufficiently confined horizontally . This type of cell rearrangement has also been observed in bacterial colonies ( Kreft et al . , 1998 ) and was adopted to model biofilm growth ( Xavier and Foster , 2007 ) . However , it is conceivable that if at least one population actively moves ( Houry et al . , 2012 ) or grows hyphae to penetrate the entire community , two populations might appear intermixed even if they do not cooperate . In biofilms of Pseudomonas aeruginosa , two populations differing only in fluorescent colors ( Klausen et al . , 2003 ) showed modest intermixing even though the expectation for equal-fitness competition is a columnar pattern with an intermixing index close to zero ( Figure 2D , left ) . This modest intermixing was caused by P . aeruginosa differentiating into non-motile ‘stalk’ cells that anchored to the surface and motile cells that climbed up to form the mushroom-like caps ( Klausen et al . , 2003 ) . Environmental influences can also alter ecological patterning . For instance , if two cooperating populations have very different preferences for oxygen , then the two populations are likely not to intermix and instead form layers in which the aerobic population is exposed to oxygen while the anaerobic population is protected from oxygen . Variability in patterning has been observed within and between communities ( Christensen et al . , 2002; Wilmes et al . , 2008; Dekas et al . , 2009 ) , even if they were grown in laboratory-controlled environments ( Christensen et al . , 2002 ) . Stochastic events such as environmental variability , mutations , or fluctuations in initial conditions can all lead to variable patterns . For instance , if the straight-columnar pattern expected for equal-fitness competition is observed for the majority of community but a layered pattern is observed in occasional locations , then fitter mutants may have arisen in these locations , initiating unequal-fitness competition ( Hallatschek et al . , 2007; Korolev et al . , 2012 ) . For mutually antagonistic interactions , the fitness model showed that population ratios can quickly diverge . Thus , which population eventually dominates depends on the initial population ratio ( Figure 1—figure supplement 3 ) . In this case , stochastic variations in initial conditions can result in dramatically different patterns , giving rise to a phenomenon equivalent to ‘survival of the first’ . Indeed , different patterns have been observed for communities formed by two antagonistic bacteria species ( Kuramitsu et al . , 2007 ) . In summary , our work is conceptually analogous to that of the competitive exclusion principle ( Gause , 1934 ) . The competitive exclusion principle , also known as Gause's law , states that two species competing for the exact same resources cannot stably coexist . Analogous to how the competitive exclusion principle has created a framework to examine forces that cause species coexistence , our work on ecological patterning will hopefully lay the ground for examining mechanisms that shape patterning in microbial communities . We encourage comments , especially those pertaining to the generality of our conclusions , to be posted to the eLife website .
In competitive communities , equal-fitness G and R strains were respectively WS931 ( MATa ste3::kanMX4 trp1-289::pRS404 ( TRP ) -ADHp-venus-YFP ) and WS937 ( MATa ste3::kanMX4 trp1-289::pRS404 ( TRP ) -ADHp-DsRed . T4 ) , both from the S288C background . For unequal-fitness competitive communities , G was replaced by WS1246 ( MATa ho::loxP AMN1-BY Supercontig17 ( 27163-27164 ) ::TDH3p-yEGFP-loxP-kanMX-loxP ) from the RM11 background with a 20% fitness advantage over the S288C background . In commensal communities , R→A←L and G←A strains were respectively WS950 ( MATa ste3::kanMX4 lys2Δ0 ade4::ADE4 ( PUR6 ) trp1-289::pRS404 ( TRP ) -ADHp-DsRed . T4 ) and WS932 ( MATa ste3::kanMX4 ade8Δ0 trp1-289::pRS404 ( TRP ) -ADHp-venus-YFP ) , both from the S288C background . In cooperative communities , R→A←L and G→L←Astrains were respectively WS950 and WS954 ( MATa ste3::kanMX4 ade8Δ0 lys21::LYS21 ( fbr ) trp1-289::pRS404 ( TRP ) -ADHp-venus-YFP ) from the S288C background . For yeast communities , agarose columns were prepared by pouring SD minimal medium ( Sherman , 2002 ) with 2% low melting temperature agarose in 1 . 3-ml deep-well plates ( U96 DeepWell from Nunc , Penfield , NY ) , and after agarose had solidified , adding drops of melted SD agarose to make the surface flat . For commensal yeast communities , SD was supplemented with lysine ( 80 μM final concentration ) . Batch cultures of yeast strains were grown to exponential phase at 30°C in SD with supplements when necessary . Cells were washed free of supplements if any , mixed , and filtered on top of MF membrane ( HAWP04700 from Millipore , Billerica , MA ) . Disks were cut from the membrane using a 6-mm-diameter Harris Uni-Core puncher and transferred to the top of agarose columns , unless otherwise stated . For the coculture of D . vulgaris and M . maripaludis , a CDC reactor ( BioSurface Technologies Corp . , Bozeman , MT ) was used for anaerobic biofilm growth . Biofilm coupon holders were modified to hold glass microscope slides ( Fisher Scientific #12-544-1 , Fisher Scientific , Pittsburgh , PA ) cut to 7 . 6 × 1 . 8 cm . Cocultures were grown in CCM ( Walker et al . , 2009 ) , a modified basal salt medium without choline chloride . Headspace was sparged with anoxic 80% N2:20% CO2 , and the reactor was maintained at 30°C with stirring ( 150 rpm ) . The reactors were inoculated with planktonic coculture , and after cell attachment to the glass slides ( 48 hr ) , biofilms were allowed to develop and grow over time in the presence of planktonic cells , or initially drained of planktonic cells . Patterns in the two experimental regimes were similar . Flow cytometry was performed in a modified 4-laser FACS Calibur machine ( DxP; CyTek Development , Fremont , CA ) . Cells in communities were diluted to ∼106 cells/ml in H2O . Each sample ( 90 µl ) was supplemented with 10 µl of fluorescent bead stock ( Thermo Scientific Fluoro-Max Cat# R0300 at ∼8 × 106/ml , counted using a Z2 Coulter counter and a hemacytometer ) as a reference to determine total cell density and 3 µl of 1 µM ToPro3 ( Invitrogen , Grand Island , NY ) to mark dead cells . The laser and filter configurations for different fluorophores were 50 mW 488 nm laser with 530/30 filter for YFP , 75 mW 561 nm laser with 575/26 filter for DsRed , and 25 mW 639 nm laser with 660/20 filter for ToPro3 . Using an automatic micro-sampling system ( DxP; CyTek Development , Fremont , CA ) samples in 96-well plates were processed for 60 s at a flow rate of 0 . 5–1 µl/s , recording 104–105 events . FlowJo software ( Tree Star , Ashland , OR ) was used to measure the ratio of different populations of fluorescent cell against the bead standard in order to calculate cell densities . A Nikon TE2000 inverted microscope equipped with a Prior stage controller , a Sutter Lambda XL fluorescent lamp , and a Photometric SnapHQ CCD camera was controlled by custom LabVIEW software to auto-focus and acquire images . All images were taken at 10× magnification using a Nikon long working distance CFI Plan Fl objective ( NA 0 . 30 , WD 16 ) . For imaging YFP- and DsRed-tagged strains , ET500/20×-ET535/30m-T515LP and ET545/30×-ET620/60m-T570LP filter cubes were used , respectively . Timelapse imaging took place in a 30°C chamber ( In Vivo Scientific microscope incubator ) . To obtain vertical cross-sections of communities through cryosectioning , we froze communities in liquid nitrogen for 15 s and fixed them in methanol at −20°C . After 20 min , the communities were transferred to a pre-cooled empty plate at −20°C to allow methanol to evaporate , which typically took 4 hr . The communities were embedded in optimal cutting temperature ( OCT ) compound for 10 min at room temperature and subsequently frozen over dry ice and kept at −20°C for sectioning . We also froze down communities without fixing by directly embedding them in OCT ( Piccirillo et al . , 2010 ) and immediately freezing them on dry ice . Results obtained from the two procedures yielded similar conclusions . For sectioning , communities embedded in OCT blocks were mounted on a cryotome . The blade was adjusted to ensure cross-sections of the community were vertical . For each community , typically thirty to fifty 14-μm ( 3-cell-thick ) sections were cut and transferred to glass slides . Cross-sections were imaged using the fluorescence microscope as described above . Only images of cross-sections that were minimally perturbed by the fixing and sectioning processes were included in the analysis . More details of the cryosectioning method can be found in ( Momeni and Shou , 2012 ) . Biofilms of D . vulgaris and M . maripaludis were fixed in 4% paraformaldehyde for 4 hr , and then embedded in polyacrylamide ( Daims et al . , 2006 ) . The embedded biofilms were dehydrated and hybridized in 1 ml buffer solution ( 0 . 9 M NaCl , 20 mM Tris–HCl ( pH 8 ) , 0 . 01% SDS , and 35% deionized formamide ) with 3 ng each of probes EUB338 ( GCT GCC TCC CGT AGG AGT ) 5′ and 3′-labeled with Cy3 and ARCH915 ( GTG CTC CCC CGC CAA TTC CT ) 5′ and 3′-labeled with Cy5 ( Stoecker et al . , 2010 ) for 5 or 8 hr at 46°C in a humid chamber . Next , samples were washed in 50 ml prewarmed washing buffer ( 70 mM NaCl , 20 mM Tris–HCl pH 8 , and 5 mM EDTA ) at 47°C for 20 min , then dipped in ice cold ddH2O and quickly dried with compressed air ( Amann et al . , 1995 ) . Finally , each sample was mounted with Citifluor AF1 antifadent ( Citifluor Ltd . , Leicester , United Kingdom ) and viewed using a Leica TCS SP5 II inverted confocal laser scanning microscope with 488 , 561 and 633 nm lasers and appropriate filter sets for Cy3 and Cy5 . Confocal voxel size was typically 0 . 24 × 0 . 24 × 0 . 49 μm . For both simulated and experimental sections , our unit of analysis was the size of one CCD camera field of view under a 10× objective which has a width of 0 . 7 mm . In any unit of analysis , community height is the value such that 90% of height values are below it . This choice is made to exclude artifacts such as height spikes in simulations . Images of experimental community sections were rotated in ImageJ such that the x–z axes of the frame matched those of the community . We further digitized these images into f ( x , z ) with assigned values of +1 , −1 , and 0 for pixels identified as population 1 , population 2 , and no-signal , respectively . No-signal pixels were defined as having a fluorescence intensity per unit exposure time less than 10–20% above the background in fluorescence channels . For the remaining pixels , the intensity values of 90th percentiles were found in each fluorescence channel . These values were used to normalize the corresponding green- or red-fluorescence intensity for each pixel whose identity was then assigned to be the color with the higher normalized value . In simulated communities with more than two populations , analysis was performed on two focal colors at a time , and in each analysis , pixels identified as other colors were treated as no-signal . To compare levels of intermixing in different communities , we estimated the number of cell type transitions spanning community height . We define the intermixing index as the average number of color changes along community height: let h ( xi ) and c ( xi ) respectively be the local height and the number of color changes along z at the lateral position xi . The intermixing index IM can be calculated as ( 1 ) IM= ( ∑xic ( xi ) h ( xi ) ) / ( ∑xih ( xi ) ) . Intermixing is small for segregated patterns with few color changes along height , and increases when patches of different cell types successively appear on top of each other . Note that this choice of intermixing index yields small values when one population is very rare compared to the other . We weighted c ( xi ) by the local community height h ( xi ) , thus giving more emphasis to taller regions . This is because taller regions have gone through more growth regulated by the fitness effects of interactions . To estimate the vertical patch size λz* , the ratio of height h ( xi ) to number of cell-type layers , 1 + c ( xi ) , was averaged across the community cross-section ( 2 ) λz*= ( ∑xi[h ( xi ) ]2/[1+c ( xi ) ] ) / ( ∑xih ( xi ) ) . For engineered yeast communities , typically 10–20 frames from different locations of each community were included in the analysis to ensure an unbiased representation of community patterns . For biofilms of D . vulgaris and M . maripaludis , vertical cross-sections ∼2 . 4 μm apart were sampled from confocal z-stack images of biofilms . The characteristic patch size can also be calculated as the effective length of interaction between two partners . We calculate how far inside the community released lysine can diffuse before being consumed; similar discussion applies to adenine . Assume one cell has released βL fmole of lysine that diffuses at most a distance l before being consumed . We define the sphere of radius l as the diffusion domain of the release event . The number of consuming cells within the diffusion domain can be estimated as Nu = ( 2l/c ) 3 , where c is the diameter of a cell . The average time nutrient can diffuse in the diffusion domain before being consumed is tc = l2/2D1 , where D1 = 360 μm2/s is the diffusion constant within the community . Assuming that each consuming cell takes up lysine with a rate v ≤ αL/T , with αL being the amount of lysine required for a cell division and T being the minimum doubling time , we haveβL=Nuvtc≤ ( 2lc ) 3αLTl22D1;thus , l≥[βLαLD1Tc34]1/5≈50 μm . Within radius l , the average number of release events within time tc is ( 2l/c ) 3tcdG≈0 . 2 . Thus , the probability of a second release event occurring in the same diffusion domain before nutrient from the first release event has been consumed is low . Therefore , l defines the interaction length scale . Note that diffusion constant contributes by fifth root to l , and therefore , different diffusion constants in the community should not considerably alter patterns , as observed in simulations ( Figure 3—figure supplement 1A ) . The calculated l of 50 μm is larger than the experimental patch size λz* of ∼10 to 20 μm . What could account for this discrepancy ? Experimentally measured diffusion constant of Sulforhodamine 101 in community ( 20 µm2/s ) reduces l to ∼30 µm . In addition , cells may take up more nutrients than what is required for producing one daughter and may store extra nutrients in vacuoles ( Shou et al . , 2007 ) , further reducing the estimated interaction length scale . The individual-based fitness model followed cell growth in a three-dimensional simulation grid of 100 × 100 × 300 cells with periodic boundary conditions along the x and y directions . Consider population i ( i = R or G ) interacting with population j ( j = G or R ) . Without loss of generality , consider a focal cell from population i . The growth of the focal cell is influenced by cells in its cubic three-axial interaction neighborhood ( Deutsch and Dormann , 2004 ) defined by l-cell-width to the left , right , front , back , above , and below . Let ϕi and ϕj be the fraction occupancy of i and j in the interaction neighborhood , respectively . The growth rate of the focal cell is ri=[ri0+rijϕj ( 1−ϕi ) ][1−χ ( ϕi+ϕj ) ] . ri0 is the basal fitness ( growth rate of i without any interactions ) ; rijϕj ( 1−ϕi ) represents the fitness effect on i by j , which increases with partner abundance and decreases with recipient abundance due to intra-population competition for partner; [1−χ ( ϕi+ϕj ) ] reflects intra- and inter-population competition for shared resources with fitness decreasing as the neighborhood becomes more occupied . Cells were inoculated in the bottom surface of the simulation grid . In each simulation time step Δt , the probability of cell division is ri Δt . A cell would divide either to the side if there was space within its ( x , y ) planar confinement neighborhood of n-cell radius or upward otherwise ( Figure 1—figure supplement 1 ) . Parameters used are listed in Figure 1—source data 1 . χ = 0 . 8 , l = 3 , and n = 2 in all cases . See Source code 1 for an example ( and Source code 4 for the MATLAB function ) . The individual-based diffusion model followed actions of cells ( nutrient uptake , cell division , cell death , and possibly release of nutrients ) and the distribution of nutrients in a three-dimensional simulation grid . Since cell division and death occur at a time-scale much longer than diffusion and nutrient uptake , we used a multi-grid scheme in both space and time . In this model , a three-dimensional simulation domain consisted of cell grids representing individual cells and nutrient grids representing nutrient concentrations ( Figure 2—figure supplement 2 ) . Cell actions and nutrient distributions were updated at discrete time steps over the simulation domain . Simulations were typically performed over an agarose domain of 0 . 75-mm length × 0 . 75-mm width × 24-mm depth and a community domain of 0 . 75-mm length × 0 . 75-mm width × 0 . 3-mm height with parameters listed in Figure 2—source data 1 . Nutrient concentrations as a function of space and time are based on the diffusion equation ( 3 ) ∂S∂t=∇· ( D∇S ) −U+Q , with ( 4 ) U=vmSS+KMMnu ( 5 ) Q=ρ nq . Equation ( 3 ) states that S , the amount of limiting nutrient in a diffusion grid , depends on three processes: i ) diffusion of nutrient with diffusion constant D , ii ) uptake of nutrients ( Walther et al . , 2005 ) by cells ( U ) , and iii ) in cooperative communities , release of nutrients by the partner population ( Q ) . In equations ( 4 ) and ( 5 ) , nu and nq are the number of consuming and releasing cells within the diffusion grid , respectively , KMM is the Michaelis-Menten constant for uptake , vm is the maximum uptake rate per cell , and ρ is the release rate per cell . To solve this diffusion equation , we used a finite difference time–domain method ( Crank , 1980 ) with no-flow ( ∂S/∂z=0 ) boundary conditions applied to the top and bottom surfaces of the simulation domain and periodic boundary conditions applied to the four vertical sides of the domain . Cell growth rate r in the model is dictated by Monod's equation ( 6 ) r ( S ) =rmSS+KM , in which rm is the maximum growth rate when nutrients are abundant , S is the concentration of the limiting nutrient , and KM is the S at which half maximal growth rate is achieved . We assume that individual cells take up nutrients with KMM = KM , and once they have accumulated the required amount of the limiting nutrient , cell division occurs . To incorporate realistic assumptions about cell rearrangement upon division in the community , we monitored single yeast cells growing into microcolonies on solid media ( Figure 1—figure supplement 1A ) . Initially , each cell budded in the same plane and pushed others in its immediate neighborhood to the side . Once a cell was completely surrounded on each side by roughly five cells , it budded upward . The same process was implemented in the diffusion model ( Figure 2—figure supplement 2 ) . It should be noted that by forcing the confined cells to bud strictly upward , the model underestimates intermixing . As a result , communities show more vertical features in simulations than experiments ( Figure 2F ) . Temporally , the update time-step for nutrient diffusion and uptake ( ∼1 s ) was smaller than that for cell division and death ( ∼360 s ) . Spatially , cell actions took place on a cell grid with single-cell resolution ( 5 μm ) , while nutrient distributions were followed on a diffusion grid at lower spatial resolution ( ∼15 to 60 μm ) . These values were selected considering the trade-off between simulation time and accuracy , while ensuring the stability of simulations . For instance , at each time-step within a diffusion grid , the total amount of nutrients consumed should be considerably smaller than available nutrients . In other words , assuming that the length of a diffusion grid is nc where c is the length of a cell grid which is equivalent to the size of a cell , there are at most n3 cells in each diffusion grid and ( 7 ) S ( nc ) 3>n3vmSS+KMMdtu , where S is the concentration of the limiting nutrient in the spatial grid , KMM is the Michaelis-Menten coefficient for nutrient uptake , vm is the maximum uptake rate , and dtu is the uptake time-step . After simplifying equation ( 7 ) , we obtain ( 8 ) dtu<S+KMMvmc3 . In the worst case scenario of S being much smaller than KMM , using parameters of our engineered yeast strains , we find dtu ∼ 0 . 5 s . Thus , we chose dtu = 0 . 5–1 s as the time-step for updating the nutrient uptake and diffusion equations . To ensure stability of the finite-difference equations for diffusion , the diffusion grid-size ( nc ) and the time-step ( dtu ) for the diffusion equation have to satisfy ( Iserles , 2009 ) ( 9 ) dtu<12 ( nc ) 2D , where D is the diffusion constant in the region of interest . From this relation , we chose diffusion grid size nc = 50 μm , and consequently , each diffusion grid contains 10 × 10 × 10 cells ( Figure 2—figure supplement 2A ) . Since diffusion is fast within each grid ( ∼1 s ) , this choice of grid size is unlikely to introduce a notable error in our calculations . See Source codes 2 and 3 for examples ( and Source code 4 for the MATLAB function ) . Let ϕR and ϕG be the fraction occupancy of R and G in an interaction neighborhood , respectively . ϕG*is the fraction occupancy of G that leads to equal fitness of the two populations and thereby result in a steady state ratio within the interaction neighborhood . Following the assumptions of the fitness model , the growth rate of each population is ri=[ri0+rijϕj ( 1−ϕi ) ][1−χ ( ϕi+ϕj ) ] where i = R or G and j = G or R . For simplicity , we assume that ϕR+ϕG=1 , which leads to r^i=ri/ ( 1−χ ) =ri0+rijϕj . In two-population cooperative communities , using the simplifying assumption of rRG=rGR=rint>0 , we have r^G−r^R= ( rG0−rR0+rint ) −2rintϕG ( Figure 1—figure supplement 2B ) . Setting r^G−r^R to 0 , the community can achieve a steady-state value of ϕG*= ( rG0−rR0+rint ) /2rint . To satisfy 0<ϕG*<1 , rint>|rG0−rR0| which means that the interaction term has to be strong enough to overcome the difference in the basal fitness of the two populations . Here , ϕG* is stable , since r^G−r^R is positive ( favoring G ) when ϕG<ϕG* and negative ( favoring R ) when ϕG>ϕG* . At ϕG* , two populations grow at the same rate and population ratio is fixed at R:G = ( 1 − ϕG* ) :ϕG* . Similar analysis shows the existence of a stable partner ratio under commensalism G[∼↑]R . A steady-state ratio ϕG* can exist under exploitation . However , initial ratios below the critical value ϕG , c ( Figure 1—figure supplement 2B ) will not converge to ϕG* . Other interactions ( [∼ ∼] , [∼ ↓] , and [↓ ↓] ) do not converge to a stable ratio . Similar conclusions on ratio convergence hold if we assume r^i=ri0+rijf ( ϕj ) for any continuous function f that monotonically increases with ϕj ( proof not shown ) . Results are summarized in Table 1 . | Microorganisms such as bacteria , archaea and tiny eukaryotes are found throughout the biosphere . Some of these microorganisms are pathogens that cause diseases in animals , while others provide nutrients , including essential amino acids and vitamins; there are also microorganisms that have critical roles in recycling elements such as carbon , nitrogen and oxygen in the biosphere . In the natural world , microorganisms interact with their environment and with each other , often competing for space , light and nutrients , but sometimes they act cooperatively , which benefits all parties involved . Microbial communities exhibit spatial patterns that reflect the relative positioning of different microbes in a community . These patterns can be critical for the proper functioning of a microbial community . For example , in the microbial granules that digest organic compounds in waste water , the stratified pattern of different microbial species can be thought of as a sequence of catalysts needed to perform a series of biochemical processing steps . Thus , it is important to understand the mechanisms that drive pattern formation in multispecies communities . Now , through a combination of simulations and experiments , Momeni et al . have identified two features of spatial patterns in two-population microbial communities when pattern formation is driven by fitness effects related to the ecological interactions between cells . First , interactions that confer significant advantages to at least one of the populations can potentially result in the generation of a stable community; the community is stable in the sense that if it is disturbed , it will return to its stable population composition following the disturbance . Indeed , in engineered Saccharomyces cerevisiae communities , very different initial population ratios converged to the same value over time when one strain depended on the other strain , or when the two strains depended on each other , but not when the two strains competed . The second feature applies to microbial communities composed of two cooperating populations: whereas two populations that compete with each other tend to segregate , cooperation results in the members of the two populations mixing together . Momeni et al . observe the formation of such an “intermixed” community in simulations , and also in two experimental systems that involve cooperation—a community containing two different strains of yeast cooperating through metabolite exchange , and a biofilm in which Methanococcus maripaludis , an archaeon that produces methane , cooperates with the bacterium Desulfovibrio vulgaris . These two features of spatial patterning are conceptually similar to the competitive exclusion principle , which states that two species competing for the same resources cannot stably coexist if competition is the sole force at work . This principle has , therefore , encouraged scientists to search for the other forces that must be responsible for the coexistence of different species . Similarly , by predicting the sorts of patterns that will form when the fitness effects of ecological interactions between cells are the only forces at work , Momeni et al . lay the groundwork for investigations into other mechanisms , such as cell–environment interactions and active cell motility , that can govern pattern formation in microbial communities . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2013 | Strong inter-population cooperation leads to partner intermixing in microbial communities |
Cell-cell communication via gap junctions regulates a wide range of physiological processes by enabling the direct intercellular electrical and chemical coupling . However , the in vivo distribution and function of gap junctions remain poorly understood , partly due to the lack of non-invasive tools with both cell-type specificity and high spatiotemporal resolution . Here , we developed PARIS ( pairing actuators and receivers to optically isolate gap junctions ) , a new fully genetically encoded tool for measuring the cell-specific gap junctional coupling ( GJC ) . PARIS successfully enabled monitoring of GJC in several cultured cell lines under physiologically relevant conditions and in distinct genetically defined neurons in Drosophila brain , with ~10 s temporal resolution and sub-cellular spatial resolution . These results demonstrate that PARIS is a robust , highly sensitive tool for mapping functional gap junctions and study their regulation in both health and disease .
Gap junctions are intercellular channels that are expressed in virtually all tissue types in both vertebrates and invertebrates . They allow direct transfer of small molecules such as ions , metabolites and second messengers between connected cells , thereby playing essential roles in various physiological processes , including embryo development , metabolic coordination of avascular organs , tissue homeostasis and synchronizing electrical activity in excitable tissues ( Bennett and Zukin , 2004; Kumar and Gilula , 1996 ) . Defects of gap junction function are linked with a wide array of diseases , including myocardial infarction , hearing loss and hypomyelinating disorder ( Jongsma and Wilders , 2000; Laird , 2010; Söhl et al . , 2005 ) . Studying gap junction coupling ( GJC ) under physiological or disease conditions in complex systems such as the nervous system requires a non-invasive method with both cell-type specificity and high spatiotemporal resolution . Current methods used to monitor GJC include paired electrophysiological recordings ( Bennett et al . , 1963; Furshpan and Potter , 1959 ) , dye microinjection ( Stewart , 1978 ) , fluorescence recovery after photobleaching ( FRAP ) ( Wade et al . , 1986 ) , and local activation of molecular fluorescent probes ( LAMP ) ( Dakin et al . , 2005 ) , all of which are either invasive and/or lack cell-type specificity , thus limiting their use in heterogeneous tissues . Recently , hybrid approaches have been developed in which genetically encoded hydrolytic enzymes or promiscuous transporters are used to introduce small molecule substrates or peptides tagged with a fluorescent label in specific cells ( Qiao and Sanes , 2015; Tian et al . , 2012 ) . Although these methods might provide cell-type-specific investigation of GJC , the requirement of exogenous substrate and the background generated by endogenous enzymes or transporters still make them difficult to apply in vivo . In addition , dye diffusion is an irreversible process and thus these methods are difficult to be applied to examine the same gap junctions repeatedly in order to measure their dynamics , regulations and plasticity ( Dong et al . , 2018 ) . To overcome these limitations , we developed an all-optical approach named PARIS ( pairing actuators and receivers to optically isolate gap junctions ) in which we express an optically controlled actuator in one cell , to generate an electrochemical gradient of specific molecules between two connected cells , and a fluorescent receiver in the adjacent cell , to detect the movement of the molecules across the gap junctions . GJC between the actuator cell ( i . e . expressing actuators ) and the receiver cell ( i . e . expressing receivers ) is detected by a fluorescence increase in the receiver following the optical activation of the actuator ( Figure 1A ) .
At the beginning , we tested several pairs of optical actuators/receivers based on generating/detecting small molecules that can readily diffuse across gap junctions , such as cGMP , Ca2+ and proton ( H+ ) . Our first step was to test whether the actuator/receiver pair can generate a cell-autonomous signal . We found that , when co-expressed in HEK293T cells ( i . e . in the cis configuration ) , neither a light-activated cGMP cyclase BeCylOp ( Gao et al . , 2015 ) paired with a cGMP sensor FlincG3 ( Bhargava et al . , 2013 ) nor the red shifted channelrhodopsin CsChrimson ( Klapoetke et al . , 2014 ) paired with a sensitive Ca2+ indicator GCaMP6s ( Chen et al . , 2013 ) could generate detectable light-induced signal ( Figure 1—figure supplement 1 ) . Interestingly , when we co-expressed a light-gated outward proton pump ArchT ( Han et al . , 2011 ) and a pH-sensitive green fluorescent protein pHluorin ( Miesenböck et al . , 1998; Sankaranarayanan et al . , 2000 ) in HEK293T cells , a 4 s laser illumination at 561 nm elicited a robust increase in pHluorin fluorescence , with the membrane-targeted pHluorin ( pHluorinCAAX ) producing a larger change in fluorescence than the cytosolic pHluorin ( Figure 1—figure supplement 2A , B ) . No light-induced change in fluorescence was observed in cells that co-expressing pHluorinCAAX and the deficient proton-pump ArchTD95N ( Kralj et al . , 2011 ) , or in cells that only express pHluorinCAAX ( Figure 1—figure supplement 2A , B ) . Furthermore , the evoked response is dependent on both the duration and the power of the activating light ( Figure 1—figure supplement 2C–F ) . These results demonstrate that ArchT and pHluorin can function as a pair of proton actuator and proton sensor . We next examined whether PARIS based on ArchT/pHluorin can be used to measure GJC between cultured HEK293T cells , which endogenously express both connexin ( Cx ) 43 and Cx45 , therefore spontaneously form gap junctions between adjacent cells ( Butterweck et al . , 1994; Langlois et al . , 2008 ) . When ArchT and pHluorin were separately expressed in neighboring cells ( i . e . in the trans configuration , see Materials and methods; Figure 1B1 ) , a brief photoactivation of ArchT in the actuator cells ( 4 s , ~0 . 5 mW , indicated by the yellow circle in Figure 1B2 ) faithfully induced a ~ 4 . 3% ∆F/F0 increase in pHluorinCAAX fluorescence in the neighboring receiver cells whereas non-adjacent pHluorinCAAX-expressing cells had no measurable change in fluorescence ( Figures 1B2–B3 ) . Application of carbenoxolone ( CBX , 100 μM ) which blocks gap junctions ( Connors , 2012 ) significantly decreased the light-induced PARIS signal ( Figure 1C ) , confirming that the signal measured in receiver cells is mediated by GJC . Similar to autonomous signals , increasing the duration of the illumination pulse from 1 s to 20 s incrementally increased the PARIS response from ~2% to~20% ( Figure 1D–E ) . A 4 s laser pulse was sufficient to induce a robust PARIS signal ( SNR = 23 ± 8 , Figure 1F ) with a half-rise time of ~10 s ( Figure 1G ) . On the other hand , a 20 s laser pulse induced an ~4 . 3-fold increase in the signal-to-noise ratio compared to 4 s with a half-rise time of ~21 s ( Figure 1F , G ) ; however , the half-decay time did not differ between a 4 s pulse and a 20 s pulse ( t1/2 decay = 61 ± 5s and 67 ± 3 s respectively , Figure 1G ) . We also observed the spatially graded PARIS signals in three receiver cells that are sequentially connected to the actuator cell ( Figure 1—figure supplement 3 ) . Specifically , the directly connected cell had the strongest response , and the thirdly connected cell had the weakest response ( Figure 1—figure supplement 3D ) . We then quantified the ArchT-induced pH change in the actuator cells using the ratiometric pH indicator mTagBFP-pHluorinCAAX generated by fusing the pH-insensitive blue fluorescent protein mTagBFP ( Subach et al . , 2008 ) to the N-terminus of pHluorinCAAX and then calibrating the correlation between pH and the ratio of GFP/BFP fluorescence ( Figure 1—figure supplement 4 ) . Based on a fit to the titration curve , we estimated that a 4 s and 20 s laser pulse induces a transient increase of intracellular pH from 7 . 35 to 7 . 45 and 7 . 80 respectively in actuator cells ( Figure 1—figure supplement 4D–F ) , which allowed us to repeatedly elicit a PARIS signal in specific cells as shown above . Together , these data provide proof-of-principle that PARIS is a robust tool for measuring GJC between connected cells . We have showed that PARIS could detect GJC in a photostimulation-dependent way and sensitive to CBX ( Figure 2A , D1 and Figure 1 ) . Next , we further validated PARIS by patch-clamping the receiver cell in order to record the gap junction-mediated current induced by activating the actuator cell using a laser pulse ( Figure 2B1 ) . Applying increasingly stronger light pulses to the actuator cell yielded time-locked currents in the receiver cell that were blocked by CBX ( Figure 2B2 , D2 ) . In the same cells , voltage steps on the actuator cell also elicited non-rectifying and CBX sensitive currents in the receiver cell ( Figure 2C , D3 ) . Quantification of the group data showed that the CBX inhibition of GJC was independent from the approaches used to activate the actuator cell ( by light or voltage ) and from the signals measured in the receiver cell ( pHluorinCAAX fluorescence or currents ) ( Figure 2E ) . In addition , we performed a head-to-head comparison between PARIS and FRAP—a dye diffusion based methods which detects the gap junction mediated fluorescence recovery after photobleaching ( Figure 2F ) . The PARIS signal was stable for five sequential pulses at 2 min intervals , whereas the FRAP signal decayed considerably over the same time period in terms of both basal fluorescence and SNR ( Figure 2G–I ) . Moreover , the PARIS signal had considerably faster kinetics than FRAP , with a half-rise time of ~21 s compared to ~197 s , respectively ( Figure 2J ) . Phosphorylation has been implicated in the modulation of GJC by affecting the trafficking , assembly/disassembly , degradation and gating of gap junctions ( Laird , 2005; Nihei et al . , 2010 ) . To test whether PARIS could report GJC under different regulations such as protein phosphorylation , we treated PARIS-expressing HEK293T cells with the cAMP analog 8-Br-cAMP , the adenylyl cyclase agonist forskolin , or the protein kinase C ( PKC ) agonist tetradecanoylphorbol acetate ( TPA ) . Compared with the control group , treating cells with TPA significantly inhibited the PARIS signal compared to the control group; in contrast , neither 8-Br-cAMP nor forskolin had obvious effect ( Figure 3A , B ) , suggesting that activating PKC—but not protein kinase A—inhibits GJC , which is in general consistent with previous reports ( Lampe , 1994; Sirnes et al . , 2009 ) . Mutations in GJA1 , the gene encoding Cx43 , have been linked to a number of diseases such as the inherited oculodentodigital dysplasia ( Paznekas et al . , 2009 ) . We therefore asked whether PARIS could be used to probe the function of Cxs encoded by mutated Cx genes . We performed PARIS in HeLa cells , which do not express measurable levels of endogenous Cxs ( Elfgang et al . , 1995 ) . As expected , no PARIS signal was elicited in receiver HeLa cells upon photoactivating the actuator cell; while in HeLa cells expressing GJA1 , photoactivating the actuator cell elicited a robust fluorescence increase in the adjacent receiver cell ( Figure 3C , D ) . Interestingly , expressing a Cx43 protein with either the R202H or R76H mutation—which affects Cx43 trafficking and gap junction permeability ( Shibayama et al . , 2005 ) —caused a significant reduction in the PARIS signal compared to cells expressing wild-type Cx43 ( Figure 3C , D ) . These data indicate that PARIS can be used to probe the effects of clinically relevant mutations in gap junction proteins . Next , we examined whether PARIS can be used to study gap junctions in a physiologically relevant system , namely cardiomyocytes ( CMs ) . Gap junctions formed by Cx40 , Cx43 , and Cx45 play an important role in CMs by synchronizing their contractions and defects in these connexins have been associated with cardiovascular diseases ( Jongsma and Wilders , 2000 ) . Using CMs cultured from neonatal rats ( Figure 3E ) , we observed that stimulating actuator CMs induced a robust fluorescence increase in receiver CMs with half-rise and half-decay times of approximately 14 and 21 s respectively , and the responses were reversibly blocked by the gap junction blocker heptanol ( Garcia-Dorado et al . , 1997 ) ( Figure 3F–H ) . Neither the rate of spontaneous Ca2+ transients in CMs nor the rate of cellular beating was altered by the expression or stimulation of the actuator protein ( Figure 3—figure supplement 1 ) , supporting the notion that expressing and activating PARIS does not affect cellular functions . We then examined whether PARIS can be used to measure gap junction activity ( i . e . electrical synapses ) between genetically defined cell types in the brain . Using the Drosophila olfactory system as a model system , we first confirmed that ArchT and pHluorin can produce cell-autonomous signals in an ex vivo preparation ( Figure 4—figure supplement 1 ) . We expressed both the actuator and the receiver with dual binary expression systems ( GH146-QF > QUAS ArchT , GH146-Gal4 > UAS pHluorinCAAX ) in excitatory projection neurons ( ePNs ) in the fly olfactory pathway ( Stocker et al . , 1997 ) and measured cell autonomous PARIS signal from the antenna lobe ( AL ) in the isolated fly brain ( i . e . in the cis configuration , Figure 4—figure supplement 1A–G ) . The ePN autonomous signal could be elicited repeatedly in the same sample for up to 2 hr , with no obvious loss of signal strength ( Figure 4—figure supplement 1H , I ) , indicating that PARIS is stable in intact living tissue . We then used PARIS to measure electrical synapses formed between excitatory projection neurons ( ePNs ) and excitatory local neurons ( eLNs ) , both of which have dendritic arborizations in the antennal lobe ( AL ) ( Shang et al . , 2007 ) ( Figure 4—figure supplement 2 , first row ) . We generated a transgenic Drosophila line expressing the actuator ( GH146-QF > QUAS ArchT ) selectively in ePNs and the receiver ( Kras-Gal4 >UAS pHluorinCAAX ) selectively in eLNs ( Figure 4A ) . Stimulating a 20 μm diameter region in the AL elicited a rapid increase in pHluorinCAAX fluorescence , with half-rise and half-decay times of approximately 12 s and 29 s , respectively ( Figure 4B , C , F , G ) , consistent with previously reported electrophysiological data indicating that ePNs and eLNs are electrically coupled ( Huang et al . , 2010; Yaksi and Wilson , 2010 ) . Importantly , no response was elicited in the brain when the transgenic flies were pretreated with CBX or in the brain of ShakB2 flies , which have a mutation in their gap junction proteins ( Zhang et al . , 1999 ) , confirming that the signal measured in the receiver neurons is indeed mediated by gap junctions ( Figure 4B , C , F , G ) . Next , we divided the AL into four regions based on orientation and then scanned each region , revealing that laser illumination can induce a fluorescence increase in each region ( Figure 4D , E , H ) , indicating that electrical coupling is a general property between ePNs and eLNs in the AL . In addition , we examined gap junction activity between ePNs and other cell types by pairing ePNs as the receiver cells with various actuator cells that have anatomical overlap with ePNs , including inhibitory local neurons ( iLNs ) , glial cells , and Keyon cells ( Figure 4—figure supplement 2 , second-bottom rows ) . However , when activated , none of these three cell types caused a measurable PARIS signal in the receiver cells ( Figure 4—figure supplement 3 ) , suggesting that ePNs may not form functional gap junction connections with iLNs , glial cells , or Keyon cells . Electrical coupling between neurons could happen through dendritic networks , axon-axonal connections or somatic contacts , which contribute the signal integration and decision of neuronal firing ( Belousov and Fontes , 2013; Yaksi and Wilson , 2010 ) . Capitalizing on the entire optical nature of PARIS , we further examined whether PARIS can be used to measure functional GJC in subcellular compartments , thereby providing spatial resolution that is not accessible to traditional methods such as electrophysiological recording or dye injection . In the Drosophila olfactory system , ventrally localized inhibitory projection neurons ( iPNs ) and ePNs form gap junctions participating odor information processing ( Wang et al . , 2014 ) . Anatomically , both ePNs and iPNs have dendrites in the AL and axons that project to the lateral horn ( LH ) ( Figure 6A , B; see also Liang et al . ( 2013 ) ; Parnas et al . , 2013 ) . Thus , whether gap junctions are formed between iPNs and ePNs in the AL , LH , or both is currently unknown . To answer this question , we expressed the actuator ( Mz699-Gal4 > UAS ArchT ) in iPNs and the receiver ( GH146-QF > QUAS pHluorinCAAX ) in ePNs . We then separately illuminated ePNs-iPNs overlapping regions in the AL or LH to test the presence of functional GJC . Interestingly , stimulating the AL—but not the LH—elicited a significant increase in pHluorinCAAX fluorescence , and this response was eliminated in the presence of CBX ( Figure 5B–D ) . As a control , we confirmed that CBX had no significant influence on the autonomous signal measured at either the AL or LH when both the actuator and receiver were co-expressed in iPNs ( i . e . in the cis configuration; Mz699 >ArchT/pHluorin , Figure 5B , C , F ) . These results support the notion that iPNs and ePNs form functional gap junctions at the AL ( i . e . via dendrite-dendrite contacts ) , but not at the LH ( i . e . via axon-axon contacts ) . To further improve PARIS’s performance for in vivo GJC detecting , we explored the light sensitivity of proton pumps with high-sequence homology to ArchT cloned from fungi , algae , bacteria to proteobacteria ( Figure 6A , up right ) . By measuring the cell-autonomous pHluorin fluorescence increase in response to green-yellow light , we found that six candidates exhibited larger ΔF/F0 than ArchT with a fungi rhodopsin that we named Lari showed the best membrane trafficking performance and 26-times the light-sensitivity of ArchT ( Figure 6A–C , Figure 6—figure supplement 1 ) . Therefore , Lari provides a more powerful actuator for in vivo application of PARIS in the future .
We initially screened three pairs of actuators/receivers , namely ArchT/pHluorin , BeCylOp/FlincG3 and CsChrimson/GCaMP6s . The latter two pairs failed to function in cis to generate receiver responses by activating the actuator ( Figure 1—figure supplement 1 ) . For the cGMP based pair , we have also performed the cis experiments in the presence of PDE inhibitor IBMX that prevented the cGMP hydrolyzation and still observed no signal; meanwhile FlincG did response to exogenous application of cGMP ( data not shown ) . Thus , one possible explanation for the absence of the autonomous signal is that light activation of BeCylOp generated limited cGMP that could not induce FlincG3 ( EC50 = 0 . 89 μM ) ( Bhargava et al . , 2013 ) response . For the pair with CsChrimson , a non-selective cation channel allows not only Ca2+ but also other cations to pass the channel ( Klapoetke et al . , 2014 ) , we deduce the photoactivation induced Ca2+ influx in the CsChrimson expressing HEK293T cells was still under the detection limit of GCaMP6s . Indeed , we found CsChrimson/GCaMP6s could function in cis to generate cell autonomous signals in cultured hippocampus neurons that endogenously express voltage-gated Ca2+ channel to allow further Ca2+ influx ( data not shown ) . First , PARIS relies solely on light and therefore is virtually non-invasive compared with existing methods including paired-recording ( Bennett et al . , 1963; Furshpan and Potter , 1959 ) , dye microinjection ( Stewart , 1978 ) and scrape loading ( el-Fouly et al . , 1987 ) . In addition , given that the activation of the actuator can be specific to subcellular resolution , PARIS can provide spatial information of the functional gap junctions , as shown by our ability to functionally map gap junctions formed at dendrite-dendrite contacts in AL but not at axon-axon contacts in LH between ePNs and iPNs in the Drosophila olfactory system ( Figure 5 ) , while such resolution cannot be easily achieved by any of the previously existed method . With respect to those relatively non-invasive methods which rely on the diffusion of small fluorescent dyes across gap junctions such as FRAP ( Wade et al . , 1986 ) and LAMP ( Dakin et al . , 2005 ) , a significant advantage of PARIS is that it is fully reversible and does not require the delivery of exogenous dyes . PARIS may serve as a robust tool for screening existing and newly developed gap junction blockers and/or modulators , including clinically relevant compounds such as inhibitors of PKC signaling . In addition , PARIS could possibly be applied to study the dynamic regulation of gap junctions in vivo , such as the formation and break of gap junction connections during brain development ( Arumugam et al . , 2005 ) . Moreover , the actuator and receiver proteins in PARIS are both genetically encoded . Recently , several innovative hybrid approaches were developed to monitor gap junctions ( Kang and Baker , 2016; Qiao and Sanes , 2015; Tian et al . , 2012 ) . Using a channel/transporter/enzyme-dependent step for the transfer of small molecules , these approaches can in principle achieve genetic specificity in terms of targeting defined cells . In practice , however , these methods require the addition of an exogenous substrate ( Qiao and Sanes , 2015; Tian et al . , 2012 ) or a patch electrode to establish an electrochemical gradient between connected cells ( Kang and Baker , 2016 ) , thereby limiting their application , particularly in vivo . In contrast , all the components in PARIS are genetically encoded by relatively small genes , vastly increasing the range of cell types in which they can be selectively expressed . For example , we show that the PARIS proteins can be introduced using transfection ( Figures 1–3 ) , virus-mediated expression ( Figure 3 ) , and ex vivo transgenic labeling ( Figure 4 and 5 ) . Given that similar transgenic tools are available for higher organisms , including mammals , PARIS can easily be adapted to other preparations and animal models , including the highly complex mammalian nervous system . In mammalian systems and model organisms in which transgenic techniques are currently unavailable or impractical , recombinant viruses such as lentiviruses , retroviruses , and adeno-associated viruses can be conveniently packaged with these genetic components and used to infect the appropriate target cells . In most animal cells , intracellular pH is believed to be tightly regulated for optimal biochemical reactions; thus , even a small change in intracellular pH is rapidly restored to a set point by buffers and proton exchangers/transporters located on the membrane of cells or intracellular organelles ( Hoffmann and Simonsen , 1989 ) . This robust system for maintaining pH homeostasis enabled us to repeatedly elicit a PARIS signal in both cultured cells and transgenic animals . One caveat to our approach may be the ability of pH to regulate gap junction activity during PARIS . The uncoupling effect of acidic intracellular pH on GJC has long been described across different Cx-consisted gap junctions in vertebrates ( Peracchia , 2004; Turin and Warner , 1977 ) and different innexin-consisted gap junctions in invertebrates ( Giaume et al . , 1980; Obaid et al . , 1983; Stergiopoulos et al . , 1999 ) , while alkalization was reported to increase the junctional conductance and the total number of opened channels in gap junction plagues ( Palacios-Prado et al . , 2010 ) . For the mostly wide expressed Cx43 channels , it has a pKa of ~6 . 7 in oocytes and fully closed when pH is under 6 . 4 while fully open when pH is above 7 . 2 ( Stergiopoulos et al . , 1999 ) , which enables PARIS measurement to reveal the GJC mediated by Cx43 . However , there is one type of gap junctions that has been reported to be sensitive to alkalization—Cx36 consisted gap junctions ( González-Nieto et al . , 2008 ) . Based on the reported pH-conductance curve , the junctional conductance decreased to 50% when pH increased 0 . 8 unit from 7 . 2 to 8 . As shown in Figure 1—figure supplements 4 and 0 . 1-unit pH increase from 7 . 35 to 7 . 45 was enough to induce PARIS signal . So PARIS is still possible in reporting Cx36 consisted GJC under proper activation of the actuator . Even though we conclude that PARIS induced pH fluctuation is controllable and transient , one should still be cautious to the possible modification towards gap junctions as well as cell physiology . For the long-time measurements , either reduce the power or shorten the time of laser illumination , meanwhile increase the interval between each measurement should be helpful to decrease the pH influence . An even more sensitive pH indicator could help to minimize the pH influence as well . As we have demonstrated , PARIS is powerful as a genetic tool to map gap junction connections between targeted neurons in the complex central nervous system ( Figure 4 and 5 ) . To map unknown gap junctions , firstly we need anatomic information about the two group of cells that we concern to make sure they are spatially contacted , which could be achieved by immunostaining or EM . As the intensive efforts have been or being made to create whole brain connectomes from C . elegans ( Jarrell et al . , 2012 ) , Drosophila ( FlyEM project , Janelia Campus Research ) to zebra fish ( Hildebrand et al . , 2017 ) and mice ( Allen Mouse Brain Connectivity Atlas ) , PARIS could utilize these information and databases and possibly help creating electrical connectome . PARIS requires the exclusive expression of the actuator and receiver in different cells . Such genetic tools specifically labeling two distinct groups of cells from different cell types or two subgroups from one cell type , might not be accessible in mammals . However , this limitation can be overcome by the intersectional non-overlapping expression of the actuator and receiver , for example using flp-out system . By fusing the receiver with a flippase , the frt sequence flanked actuator would not express in the presence of the receiver . Meanwhile the receiver and the actuator can both designed to be turned on in a cre-dependent manner . This design could make PARIS more versatile in detecting GJC between specific cells labeled by cre-lines without the contamination of the autonomous signal . Lastly , to protect against false negatives of PARIS , a control experiment in the cis configuration is recommended in different customized preparations and context to ensure the function of the actuators and help to optimize the expression level of actuators/receivers as well as the photostimulation parameters accordingly; Meanwhile , PARIS signals alone from cells connected by potential unknown gap junctions should not be interpreted as definitive without confirmation from pharmacology , genetic interventions or a complementary method . Future refinements to PARIS include the use of the new actuators we have screened combining a receiver with higher pH sensitivity , thereby increasing both the signal-to-noise ratio and temporal resolution , allowing for an even wider range of in vitro and in vivo applications . Finally , the use of additional spectrally non-overlapping pairs of proton-related actuators and receivers , as well as developing actuator-receiver pairs that transport and detect other gap junction‒permeable small molecules , may provide the opportunity to detect gap junctions between multiple cell types and/or structures .
All plasmids were constructed using the Gibson assembly ( Gibson et al . , 2009 ) method . In brief , plasmid vectors and inserts were amplified by PCR using ~30 bp overlapping primers . The fragments were assembled using T5 exonuclease , Phusion DNA polymerase , and Taq DNA ligase ( New England Biolabs ) . All sequences were verified using Sanger sequencing in our in-house facility ( sequencing platform in the School of Life Sciences of the Peking University ) . For cultured cell expression experiments , genes were cloned into the pcDNA3 . 1 vector unless otherwise noted . ArchT was amplified from pAAV-CAG-ArchT-GFP ( Han et al . , 2011 ) , codon-optimized rhodopsin genes from different species were synthesized by Qinglan Biotec then fused at the C-terminus with BFP2 , a trafficking sequence ( TS ) , and an ER export sequence ( ERex ) , producing Actuator-BFP2 . In addition , ArchT was linked directly with a TS and ERex and then fused with mRuby3 , yielding pcDNA3 . 1-ArchT-P2A-mRuby3 . mTagBFP was fused to the N-terminus of pHluorinCAAX to generate pcDNA3 . 1-mTagBFP-pHluorinCAAX . The GJA1 gene was amplified from a cDNA library ( hORFeome database 8 . 1 ) and fused with RFP ( pHuji ) ( Shen et al . , 2014 ) via a P2A linker to generate pHuji-P2A-GJA1 , which was then cloned into the N3 expression vector . Mutations in ArchT and GJA1 were introduced using overlapping PCR with primers containing the mutations of interest . To generate transgenic Drosophila lines , the following four plasmids were used: ArchT ( linked with mRuby3CAAX , HA tag and Flag tag: ArchT-HA-TS-ERex-P2A-mRuby3-Flag-CAAX ) and pHluorinCAAX were cloned into pJFRC28-10xUAS vector ( Pfeiffer et al . , 2012 ) /pJFRC28-5xQUAS vector ( made by replacing 10xUAS with 5XQUAS in pJFRC28 ) respectively , yielding UAS/QUAS-ArchT and UAS/QUAS-pHluorin transgenic flies . HEK293T cells and HeLa cells were purchased and certified from ATCC ( ATCC , Gaithersburg , MD ) . The cells were negative for mycoplasma . HEK293T cells and HeLa cells were cultured in DMEM containing 10% ( v/v ) FBS and 1% penicillin-streptomycin ( all from Gibco ) at 37°C in humidified air containing 5% CO2 . For transfection , cells were plated at 50% confluence on 12 mm glass coverslips in 24-well plates; 12–24 hr after plating , the cells were transfected using polyethylenimine ( PEI ) , with a typical ratio of 1 μg DNA: 3 μg PEI per well; 6 hr after transfection ( or 2 hr after transfection for electrophysiological experiments ) , the culture medium was replaced with fresh medium and the cells were incubated for an additional 18–24 hr prior to imaging or electrophysiological recording . For PARIS transfection , the ArchT-BFP2 and pHluorinCAAX constructs were transfected in separate wells; 6 hr after transfection , the cells were dissociated , mixed by pipette , combined into a single well , and incubated for 24 hr prior to imaging . Alternatively , in some experiments , we used sequential transfection , in which the cells were first transfected with the pHluorinCAAX construct; 6 hr later , the medium was changed and the cells were transfected with the ArchT-BFP2 construct . The medium was changed 6 hr later , and the cells were incubated for 24 hr prior to imaging . For PARIS measurements in HeLa cells , 0 . 5 μg pHluorinCAAX was mixed with 0 . 5 μg pHuji-P2A-GJA1 ( or the R202H or R76H mutant version ) and transfected into the cells; 10 hr later , the medium was replaced with new medium and the cells were transfected with 0 . 5 μg ArchT-BFP2 mixed with 0 . 5 μg pHuji-P2A-GJA1 ( or the R202H or R76H mutant version ) . The medium was replaced 10 hr later , and the cells were incubated for an additional 24 hr prior to imaging . pH calibration in HEK293T cells . HEK293T cells were co-transfected with the mTagBFP-pHluorinCAAX and ArchT-P2A-mRuby3 plasmids ( each at 0 . 5 μg per well in a DNA:PEI ratio of 1:3 ) or mTagBFP-pHluorinCAAX alone ( 1 μg per well ) 24 hr prior to imaging . Cover slips with cells attached were immersed in Tyrode’s solution containing ( in mM ) : 150 NaCl , 4 KCl , 2 MgCl2 , 2 CaCl2 , 10 HEPES , and 10 glucose ( pH 7 . 4 ) and then pre-treated with 10 μM Nigericin for 5 min , after which the calibration buffers were perfused into the chamber . After Nigericin was added , GFP and BFP channels were recorded simultaneously at 5 s intervals ( 500 ms/frame , 512 × 512 pixels ) using a Nikon A1 confocal microscope . Calibration buffers ( containing 10 μM Nigericin ) at pH 7 , 7 . 5 , 8 , 8 . 5 , and 9 . 5 contained ( in mM ) 120 KCl , 30 NaCl , 2 MgCl2 , 10 Glucose , and 10 HEPES; for pH 5 . 5 and 6 . 5 buffers , HEPES was replaced with 10 mM MES . Patch-clamp recordings were performed using an Olympus IX81 upright microscope equipped with a 40x/0 . 95 NA objective; images were acquired using Micro-Manager ( https://micro-manager . org/ ) . Laser light was delivered via a Sutter DG-4 equipped with a xenon lamp . Cultured HEK293T cells were bathed in Tyrode’s solution . Actuator HEK293T cells were identified by blue fluorescence ( excitation filter: 350/50 nm; emission filter: 448/60 nm ) . ArchT was activated using the same light source filtered through a 560/25 nm Sutter VF5 filter . Light intensity was adjusted using the Sutter DG-4 and was calculated by measuring the light power transmitted through the objective using a SANWA laser power meter LP1 ) . Recording electrodes ( with a tip resistance of 3–6 MΩ when filled with internal solution ) were fabricated using a Sutter P-97 electrode puller and controlled using Sutter MP-225 micromanipulators . The recording electrodes were filled with an internal solution containing ( in mM ) : 130 K-gluconate , 10 KCl , 2 MgCl2 , 2 . 5 Mg-ATP , 0 . 25 Na-GTP , 10 HEPES , and 0 . 4 EGTA ( pH adjusted to 7 . 4 with KOH; osmolarity adjusted to 300 mOsm with sucrose ) ; where indicated , 100 μM carbenoxolone ( CBX ) ( Sigma-Aldrich ) was applied by perfusion . The recording signal was amplified and digitized using a HEKA EPC10 double patch-clamp amplifier and collected using Patch Master . Currents were smoothed using a 20 ms moving average in order to minimize 50 Hz AC noise . For simultaneous optical and electrophysiology recordings , the Sutter DG-4 light source was triggered by the HEKA EPC10 amplifier in order to synchronize the electrophysiological recording with the light simulation . All recordings were performed at room temperature . A 1 mM stock solution of Calcein-AM was added to the culture medium to a final concentration of 1 μM . The cells were then incubated for 10 min before washing 3 times with 1 ml Tyrode’s solution . The coverslips containing the attached cells were then placed in a chamber containing Tyrode’s solution and imaged using a Nikon A1 confocal microscope . A typical FRAP experiment was performed using 488 nm imaging for measuring the baseline ( 1 s/frame , 5 s interval , five frames , 512 × 512 pixels ) , 405 nm bleaching ( 12 s , ROI ~5 μm in diameter within the cell ) , and 488 nm imaging for 2 min in order to track Calcein recovery . Cardiomyocytes ( CMs ) were enzymatically dissociated from the ventricles of neonatal ( P0 ) Sprague-Dawley rats , and 0 . 5–1 × 105 CMs per well were seeded on 12 mm glass coverslips coated with poly-D-lysine ( Sigma-Aldrich ) in 24-well plates and grown in DMEM containing 10% FBS and 1% penicillin-streptomycin ( all from Gibco ) at 37°C in humidified air containing 5% CO2 . Forty-eight hours after plating , CMs were simultaneously transfected with the ArchT-BFP2 plasmids using Lipofectamine 3000 ( Invitrogen ) and infected with an adenovirus carrying pHluorinCAAX under CMV promoter ( Vigene Biosciences ) . In brief , before transfection/infection , the medium in each well was replaced with 500 μl Opti-MEM . DNA ( 1 μg ) and Lipofectamine 3000 ( 1 . 5 μl ) were diluted in 100 μl Opti-MEM and incubated for 15 min at room temperature before additional to each well . At the same time , 1 μl of adeno-associated virus carrying the pHluorinCAAX ( 3 × 1010 pfu/ml ) was added into the same wells . Eight hours later , the Opti-MEM was replaced with standard DMEM medium , and the CMs were incubated for an additional 24 hr prior to imaging . CMs were transfected with ArchT-BFP2 plasmids using Lipofectamine 3000 as described above; 24 hr after transfection , CMs expressing ArchT-BFP2 were loaded with 1 μM Fluo-8 AM ( AAT Bioquest ) for 20 min , washed with 1 ml Tyrode’s solution for three times . CMs were imaged with 488 nm excitation in Tyrode’s solution at room temperature using a Nikon A1 confocal microscope for 1 min to record Ca2+ transients , followed by stimulation with 561 nm for 20 s ( 5 trials at 2 min intervals ) . After stimulation , another 1 min of time-lapse imaging was performed using 488 nm excitation to record Ca2+ transients . Ca2+ transients recorded before and after light stimulation were counted using ImageJ analysis of the green channel . All images were generated at a rate of 100 ms/frame . Beating rate was measured using ImageJ analysis of the white-field images . Drosophila stocks were raised at 25°C on standard corn meal-agar-molasses medium . GH298-Gal4 , Mz699-Gal4 ( III ) ( Ito et al . , 1997 ) , and GH146-Gal4 ( II ) ( Stocker et al . , 1997 ) and GH146-QF ( Potter et al . , 2010 ) were kindly provided by Dr . Liqun Luo . Krasavietz-Gal4 ( Dubnau et al . , 2003 ) and ShakB2 ( Zhang et al . , 1999 ) strains were gifts from Dr . Donggen Luo which has been verified by genotyping and sequencing the mutated site . Repo-Gal4 and MB247-Gal4 strains were gifts from Dr . Yi Rao . UAS/QUAS-ArchT and UAS/QUAS-pHluorinCAAX transgenic flies were generated at the Core Facility of Drosophila Resource and Technology , Institute of Biochemistry and Cell Biology , Chinese Academy of Sciences . All the transgenic flies have been genotype verified by sequencing in our in-house facility ( sequencing platform in the School of Life Sciences of the Peking University ) . The entire brain from adult flies within 2 weeks after eclosion ( no gender preference ) was dissected using fine forceps into Ca2+-free adult-like hemolymph ( ALH ) at RT containing ( in mM ) : 108 NaCl , 5 KCl , 5 HEPES , five trehalose , five sucrose , 26 NaHCO3 , 1 NaH2PO4 , 2 CaCl2 , and 1–2 MgCl2 ( 275 mOsm ) ( Wang et al . , 2003 ) . The brain was then transferred to a glass-bottom chamber containing ALH for confocal imaging . The brain was held in place using a custom-made platinum frame and positioned with the anterior surface of the brain toward the objective ( for imaging and stimulation of the antennal lobe ) , or with the posterior surface toward the objective ( for imaging and stimulation of the lateral horn ) . Imaging and light stimulation were performed at RT using an inverted Nikon A1 confocal microscope equipped with a Nikon sapphire laser and either a 40x/1 . 35 NA oil objective ( for HEK293T and HeLa cells ) or a 20x/0 . 75 NA air objective ( for CMs and fly brains ) . During imaging , the cells ( HEK293T cells , HeLa cells , or CMs ) were immersed in Tyrode’s solution . Cells expressing the actuator ( ArchT-BFP2 ) or receiver ( pHluorinCAAX ) were identified by the presence of blue or green fluorescence respectively . ArchT-BFP2 was excited at 405 nm and visualized using a 450/50 nm filter; pHluorin was excited at 488 nm and visualized using a GaSP photomultiplier after passing through a 525/50 nm filter . Blue fluorescent cells in adjacent to green cells were selected to conduct photostimulation with a ROI of 10–20 μm in diameter . ArchT was photostimulated using a 561 nm scanning laser at 0 . 5 mW for experiments using cell lines ( except for the power-dependent measurements ) and 0 . 1 mW for experiments using CMs . Typically , cells were first imaged using 488 nm excitation ( 256 × 256 pixels , 500 ms/frame , 2–5 s intervals ) to obtain a baseline fluorescence measurement . After obtaining a baseline , pulses of 561 nm laser light were applied to activate the ArchT ( 4 s/pulse , 5–10 pulses with no delay ) , intertwined with 488 nm imaging of the receiver . The 488 nm imaging was continued ( 2–5 s intervals ) for 1–2 min after the 561 nm stimulation in order to record the fluorescence recovery of the receiver . For experiments using gap junction blockers and modulators , 100 μM CBX ( Sigma-Aldrich ) , 2 mM heptanol ( J and K Scientific ) , 500 μM 8-Br-cAMP ( Sigma-Aldrich ) , 500 μM forskolin ( TargetMol ) , or 340 nM TPA ( Sigma-Aldrich ) was applied by adding a 1000x stock solution to the chamber . For CM experiments , heptanol was washed out of the chamber for 3 min by perfusion with Tyrode’s solution . Fly brains were stimulated and imaged in ALH using the same laser configuration described above at RT . Genotypes of samples were verified by both the presence and the pattern of green ( pHluorin ) or red ( ArchT ) fluorescence . The antennal lobe ( AL ) and lateral horn ( LH ) were identified by the green fluorescence of pHluorin and were stimulated with 0 . 5 mW laser intensity using an ROI 20–30 μm in diameter . Ctrl brain and shakeB mutant brain were photostimulated and imaged in ALH after dissection , brain in CBX group were immersed in ALH containing 100 μM CBX for 15 min before photostimulation and imaging . Time series images were analyzed using Nikon NIS and ImageJ with a stack stablizing plugin ( http://www . cs . cmu . edu/~kangli/code/Image_Stabilizer . html ) . A mean background value obtained from regions away from the pHluorin was subtracted in order to correct for fluorescence intensity ( F ) . For each pixel , we calculated the normalized change in fluorescence intensity ( ΔF/F0 ) , where F0 is the baseline fluorescence averaged from five frames obtained prior to light stimulation . ΔF/F0 over time was further processed using Origin 9 . 1 ( OriginLab ) . ΔF/F0 images were processed with MATLAB ( MathWorks ) using custom-written scripts in order to produce pseudocolor images which is provided as Source code 1 . All summary data are reported as the mean ±s . e . m . The raw data of each cell or brain sample are presented in the graphs and the sample size are indicated in the legends . All data analyses were performed using Origin 9 . 1 ( OriginLab ) . Differences between groups were tested using the Student’s unpaired or paired t-test , and differences with p<0 . 05 ( two-sided ) were considered statistically significant . For Figure 4H , a two-way ANOVA with Tukey post-hoc test was used . | For the tissues and organs of our bodies to work properly , the cells within them need to communicate with each other . One important part of cellular communication is the movement of signals – usually small molecules or ions – directly from one cell to another . This happens via structures called gap junctions , a type of sealed ‘channel’ that connects two cells . Gap junctions are found throughout the body , but investigating their precise roles in health and disease has been difficult . This is due to problems with the tools available to detect and monitor gap junctions . Some are simply harmful to cells , while others cannot be restricted to specific cell populations within a tissue . This lack of specificity makes it difficult to study gap junctions in the brain , where it is important to understand the connectivity patterns between distinct types of nerve cells . Wu et al . wanted to develop a new , non-harmful method to track gap junctions in distinct groups of cells within living tissues . To do this , Wu et al . devised PARIS , a two-part , genetically encoded system . The first part comprises a light-sensitive molecular ‘pump’ , which can only be turned on by shining a laser onto the cell of interest . When the pump is active , it transports hydrogen ions out of the cell . The second part of the system is a fluorescent sensor , present inside ‘receiving’ cells , which responds to the outcoming hydrogen ions ( small enough to pass through gap junctions ) . If an illuminated ‘signaling’ cell is connected via gap junctions to cells containing the fluorescent sensor , they will light up within seconds , but other cells not connected through gap junctions will not . The researchers first tested PARIS in cultured human and rat cells that had been genetically engineered to produce both components of the system . The experiments confirmed that PARIS could both detect networks of gap junctions in healthy cells and reveal when these networks had been disrupted , for instance by drugs or genetic mutations . Experiments using fruit flies demonstrated that PARIS was stable in living tissue and could also map the gap junctions connecting specific groups of nerve cells . PARIS is a valuable addition to the toolbox available to study cell communication . In the future , it could help increase our understanding of diseases characterized by defective gap junctions , such as seizures , cardiac irregularities , and even some cancers . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2019 | PARIS, an optogenetic method for functionally mapping gap junctions |
Inhibitions and antagonists of L-type Ca2+ channels are important to both research and therapeutics . Here , we report C-terminus mediated inhibition ( CMI ) for CaV1 . 3 that multiple motifs coordinate to tune down Ca2+ current and Ca2+ influx toward the lower limits determined by end-stage CDI ( Ca2+-dependent inactivation ) . Among IQV ( preIQ3-IQ domain ) , PCRD and DCRD ( proximal or distal C-terminal regulatory domain ) , spatial closeness of any two modules , e . g . , by constitutive fusion , facilitates the trio to form the complex , compete against calmodulin , and alter the gating . Acute CMI by rapamycin-inducible heterodimerization helps reconcile the concurrent activation/inactivation attenuations to ensure Ca2+ influx is reduced , in that Ca2+ current activated by depolarization is potently ( ~65% ) inhibited at the peak ( full activation ) , but not later on ( end-stage inactivation , ~300 ms ) . Meanwhile , CMI provides a new paradigm to develop CaV1 inhibitors , the therapeutic potential of which is implied by computational modeling of CaV1 . 3 dysregulations related to Parkinson’s disease .
L-type Ca2+ channels ( LTCC or CaV1 channels ) play pivotal roles in numerous physiological functions by mediating Ca2+ influx and membrane excitability ( Striessnig et al . , 2014 ) . Among four isoforms of CaV1 . 1–CaV1 . 4 in LTCC family , CaV1 . 3 channels exhibit unique biophysical properties ( Lieb et al . , 2014; Xu and Lipscombe , 2001 ) involving diverse regulatory mechanisms ( Ben-Johny and Yue , 2014; Christel and Lee , 2012; Huang et al . , 2012; Tan et al . , 2011 ) . Widely distributed in various excitable tissues , including both cardiovascular and nervous systems ( Brandt et al . , 2003; Mangoni et al . , 2003; Nemzou et al . , 2006; Striessnig , 2007 ) , CaV1 . 3 is also involved in a wide spectrum of pathology , including multiple inheritable diseases ( Pinggera and Striessnig , 2016; Platzer et al . , 2000 ) . As one prominent exemplar of its pathophysiological linkage , CaV1 . 3 expressed in substantia nigra pars compacta ( SNc ) neurons tightly controls the autonomous subthreshold Ca2+ oscillations which have been considered to underlie the pathophysiology of Parkinson’s disease ( PD ) ( Chan et al . , 2007; Guzman et al . , 2009 , 2010 ) . CaV1 . 3 is tuned by its own distal carboxyl tail ( DCT ) to compete with apoCaM ( Ca2+-free calmodulin ) , which is pre-associated with the carboxyl terminus of the channel at preIQ3-IQ domain ( denoted as IQV ) . Closely involved in both apoCaM and Ca2+/CaM binding ( Jurado et al . , 1999 ) , IQV plays important roles in channel functions ( Ben Johny et al . , 2013; Liu et al . , 2010; Singh et al . , 2006; Wahl-Schott et al . , 2006 ) . The competitive tuning is highly regulated by fluctuations of [apoCaM] , the strength of particular DCT isoforms , or the apoCaM affinity with IQV ( Adams et al . , 2014; Bazzazi et al . , 2013; Liu et al . , 2010 ) . DCT consists of two putative α-helical domains—a proximal and a distal C-terminal regulatory domain ( termed PCRD and DCRD , respectively ) ( Hulme et al . , 2006 ) . Positively charged PCRD could coordinate with negatively charged DCRD , in the context of CaV1 . 3 ( Liu et al . , 2010 ) , to regulate effective CaM affinity to the IQV region of the channel . Besides the reported inhibition on Ca2+-dependent inactivation ( CDI ) ( Liu et al . , 2010; Singh et al . , 2006; Wahl-Schott et al . , 2006 ) , several studies suggest that DCT also concurrently attenuates voltage-gated activation ( VGA ) , reducing the maximum open probability and positively shifting the voltage dependence ( Hulme et al . , 2006; Lieb et al . , 2014; Liu et al . , 2016; Scharinger et al . , 2015; Singh et al . , 2008; Wahl-Schott et al . , 2006 ) . Altogether , this competitive tuning of CaV1 gating ( activation/inactivation ) emerges as a new modality distinct from conventional inhibitions such as intensively-studied Ca2+ channel blockers ( CCBs ) that normally only reduce VGA ( Hockerman et al . , 1997 ) , or Ca2+/CaM-triggered conformational changes that induce CDI ( Ben-Johny and Yue , 2014 ) . However , several key matters still remain to be clarified before such new CMI ( C-terminus Mediated Inhibition ) could be fully established . First , PCRD , DCRD and IQV seemingly work cooperatively to mediate CMI as suggested by aforementioned analyses , but the exact interrelationships among these three motifs are yet to be elucidated . Second , CMI is supposed to act on channels in an acute manner similar to widely-applied CCBs or like interventions , but direct evidence is still lacking . Third , CMI is expected to reduce the overall Ca2+ influx; however , it is still unclear whether and how CMI is able to ensure the actual inhibition when both VGA and CDI are attenuated , apparently leading to contradictory effects on Ca2+ influx . Embarking on these frontiers regarding CMI , we unveiled the principle of cooperation among the three key C-terminal motifs , with which we developed chemical-inducible CMI for CaV1 . 3 channels . We then resolved all the aforementioned questions crucial to establish CMI , and gained new insights into CaV1 . 3 gating by comparing CMI and CDI . In addition , with computational modeling of Ca2+ oscillations and pacemaking behaviors under the control of CaV1 . 3 channels in SNc neurons , we explored the potentials of CMI-based inhibitors as therapeutic interventions for CaV1 . 3-related PD .
Alternative splicing at the carboxyl terminus of α1D results into two important native variants of long and short forms: α1DL ( exon 42 ) and α1DS ( exon 42A ) . Short variant of α1DS terminates shortly after the IQV motif and lacks almost entire DCT domain , and its Ca2+ current ( ICa ) exhibited strong CDI and VGA ( Figure 1A ) . Regarding CDI , α1DS still contains all the core elements including the IQV domain for apoCaM pre-association and the EF-hand motifs and N-terminal Ca2+/CaM binding NSCaTE motif ( Dick et al . , 2008 ) for CDI transduction . ICa upon depolarization rapidly decayed ( second row , red trace ) , indicative of strong CDI . In contrast , Ba2+ current ( IBa ) decayed rather slowly ( trace in grey ) representing the background voltage-dependent inactivation ( VDI ) . Thus , the fraction of peak current ( Ipeak ) remaining after depolarization for 50 ms ( third row , r50 ) is closely correlated with inactivation , with the difference between the r50 profiles in Ba2+ and Ca2+ serving as the ideal index of CDI . In practice , due to rather weak VDI of CaV1 . 3 at 50 ms ( r50 , Ba close to 1 ) ( Tadross et al . , 2010 ) , thus for simplicity the value of 1−r50 , Ca at −10 mV was taken as the index of CDI strength ( SCa ) . VGA was evaluated by peak current density Jpeak ( pA/pF ) , measured at different voltages while excluding potential artifacts due to cell-size factor ( capacitance ) . Throughout this study , SCa and JCa ( Jpeak at −10 mV ) served as the quantitative indices for CDI and VGA respectively . Thus , by way of DCT competition against apoCaM , the provisional CMI was expected to attenuate both SCa and JCa , as suggested by prior studies ( Liu et al . , 2010; Singh et al . , 2008 , 2006; Tan et al . , 2011 , 2012; Wahl-Schott et al . , 2006 ) . However , most of these reports focused on either CDI or VGA only; or even when both were studied for whole-cell ICa , CDI and VGA were separately evaluated with different experimental groups . Here , we conducted a side-by-side analysis for each variant or condition , not only more confirmative since concurrent changes in CDI and VGA would be mutually supportive , but also critical to later insights into CMI mechanisms and significance originated from such concurrency . Technically , compared to VGA and JCa , CDI and SCa are more robust so more favorable since CDI of CaV1 . 3 is independent of global Ca2+ so insensitive to Ca2+-buffer capacities and variations in channel expressions partly due to transient transfections ( Dick et al . , 2008; Tadross et al . , 2008 ) . In general , for data and analyses we provided for both CDI and VGA , CDI ( SCa ) was considered as the major index for CMI evaluations and VGA as the supplement . 10 . 7554/eLife . 21989 . 003Figure 1 . Inhibition of CaV1 . 3 gating by carboxyl terminal motifs . ( A ) Parameters and indices illustrated by the control group of short CaV1 . 3 channels . Representative current exemplars ( Ca2+ current ICa in red , with the scale bar in red; Ba2+ current IBa in gray , rescaled ) were shown for α1DS at the membrane potential ( V ) of −10 mV , with the amplitudes measured at the time of peak ( Ipeak ) and 50 ms ( I50 ) . Inactivation profiles across the full range of V for IBa and ICa were quantified by the remaining current at 50 ms ( r50 ) , in ratio between I50 and Ipeak . The CDI strength was quantified by 1−r50 , Ca or SCa , serving as one of the major indices . Based on Ca2+ current density normlized to the cell capacitance ( Cm ) , VGA in Ca2+ was profiled by Jpeak ( in pA/pF ) , with the Jpeak amplitude at −10 mV or JCa as the other major index . ( B ) In contrast to α1DS channels with pronounced CDI ( SCa ) and strong VGA ( JCa ) , α1DS-PCRD-DCRD incorporating all the three motifs of IQV , PCRD and DCRD exhibited strong inhibitions on both CDI and VGA ( less pronounced U-shape or V-shape ) , indexed with SCa and JCa ( smaller values ) respectively . Thick semi-transparent lines in red depict the CDI and VGA profiles of the α1DS control group ( A ) ; and the differences in profiles ( green areas as visual cues ) illustrate the potency of CMI . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 00310 . 7554/eLife . 21989 . 004Figure 1—figure supplement 1 . Sequence alignment of carboxyl termini for L-type CaV1 channels . The sequences of α1 subunits were aligned for CaV1 . 1 ( α1S , XM_983862 . 1 ) , CaV1 . 2 ( α1C , NM_199460 . 3 ) , CaV1 . 3 ( α1D , NM_000720 ) and CaV1 . 4 ( α1F , NP005174 ) , with GenBank accession numbers in parentheses . Key domains of pre-IQ3 , IQ , IQV , PCRD , and DCRD were underlined with different colors , and the conjunction domain ( … ) between PCRD and DCRD was skipped for clarity . Identical ( yellow and blue ) , similar ( green ) and gapped ( − ) sequences were indicated . Key residues for potential interactions were marked by red dots , including valine ( V ) on DCRD , and residues of positively charged arginine ( R ) or negatively charged aspartic acid ( D ) or glutamine ( E ) on PCRD or DCRD , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 004 To establish CMI , several key issues as introduced earlier regarding its mechanisms of action and potential applications need to be clarified . We decided to employ homologous DCTF from CaV1 . 4 ( α1F ) with the strongest DCT tuning effects ( stronger than DCTD ) among the CaV1 members ( Figure 1—figure supplement 1 ) , to construct the channel variant by fusing PCRDF-DCRDF onto α1DS right after IQV motif; meanwhile , since the potency of DCT effects is mainly determined by DCRD ( Liu et al . , 2010 ) , PCRD from either CaV1 . 3 or CaV1 . 4 supposedly makes no or little difference and thus we did not distinguish these two PCRD isoforms in this work . Such channel variant was simply denoted as α1DS-PCRD-DCRD , which represents CMI intrinsic to the long variant of CaV1 . 3 ( α1DL ) but with higher potency since strong DCRDF replaces endogenous DCRDD . Compared with α1DS control , SCa and JCa from α1DS-PCRD-DCRD were substantially attenuated , which provided ample dynamic ranges ( highlighted by green areas ) to explore CMI effects on CDI/VGA and its mechanisms ( Figure 1B ) . With such baseline CMI effects of PCRD-DCRD , we went further to examine whether both PCRD and DCRD are required , by fusing only PCRD or DCRD onto α1DS . Since DCRD might need some structural freedom as in DCT or PCRD-DCRD , a glycine linker of different length ( GX ) was inserted between DCRD and IQV to construct α1DS-GX-DCRD , mimicking the configurations that permit potent effects . From all these channel variants , we concluded no alteration in CDI in comparison with the α1DS control , similarly from VGA results and analyses ( Figure 2A and Figure 2—figure supplement 1A ) . In general , current densities should be more carefully scrutinized before any conclusive claim on VGA; and effects on Jpeak profiles and JCa values normally need to be backed up by evaluations on concurrent CDI . For instance , current densities of α1DS-GX-DCRD ( Figure 2—figure supplement 1A , the three columns on right ) seemed different from the control , potentially interpreted as mild CMI; however , considering that the changes in CDI of these channels were rather small , the mild changes in VGA should be considered as variations instead of conclusive attenuations . Statistical test of significance ( Student’s t-test ) was performed subsequently to help identify significant CMI effects . Based on the fact that all the configurations under test failed to mediate SCa and JCa attenuations , all the three motifs of PCRD , DCRD and IQV might be required for effective CMI , consistent with the preliminary analyses on DCT subsegments from CaV1 . 3 and CaV1 . 4 , in which the segment between PCRD and DCRD ( ‘B’ region ) was dispensable but neither PCRD nor DCRD could be excluded ( Liu et al . , 2010 ) . 10 . 7554/eLife . 21989 . 005Figure 2 . DCRD is indispensable and sufficient to induce CMI in low [apoCaM] . ( A ) All the channel variants including α1DS , α1DS-PCRD and α1DS-G24-DCRD exhibited strong CDI , indexed with SCa . ( B ) Schematic illustration for the strategy to explore the minimum requirement of CMI . For α1DS-GX-DCRD containing glycine linkers ( GX ) of different length ( 0 , 6 or 24 ) , BSCaMIQ that binds apoCaM was overexpressed to downregulate endogenous [apoCaM] , in hope to promote the binding of DCRD with the channel ( IQV ) . ( C ) With low [apoCaM] by overexpressing BSCaMIQ , α1DS-G24-DCRD channels exhibited much attenuated CDI , evidenced by ICa trace , SCa value and r50 profile ( green area indicating the potency ) , in contrast to ultrastrong CDI of α1DS or α1DS-PCRD , both lacking DCRD . ( D ) Statistical summary of CDI ( SCa ) and VGA ( JCa ) in endogenous ( control ) or low [apoCaM] for all channel variants of α1DS , α1DS-PCRD and α1DS-GX-DCRD , with additional information in Figure 2—figure supplement 1 . Notably , for the three α1DS-GX-DCRD variants , both CDI and VGA were concurrently and significantly attenuated . Statistical significance was evaluated and indicated ( p<0 . 001 , ***; p<0 . 05 , * ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 00510 . 7554/eLife . 21989 . 006Figure 2—figure supplement 1 . Full CDI and VGA profiles of different channel variants in endogenous and low [apoCaM] . ( A ) Under normal [apoCaM] , channel variants of α1DS , α1DS-PCRD and α1DS-GX-DCRD ( the number of glycine X equals 0 , 6 and 24 , respectively ) exhibited no significant changes in exemplar ICa traces ( top row ) , CDI ( middle row ) and VGA ( bottom row ) profiles with respective indices of SCa and JCa . Small changes in VGA profiles for α1DS-GX-DCRD were within the experimental variations at least partially due to expression fluctuations in transient transfections; meanwhile , CDI , supposedly concurrently changing with VGA , was very close to the α1DS control ( thick semitransparent red lines ) . ( B ) When [apoCaM] was substantially lower down by overexpressing the apoCaM buffer of BSCaMIQ , α1DS-GX-DCRD exhibited potent attenuations on both CDI and VGA ( green areas indicating the differences from the control group ) , but not for α1DS or α1DS-PCRD containing no DCRD . Thick red lines ( semitransparent ) represent the r50 and Jpeak profiles of the control group ( α1DS under low [apoCaM] , left ) for direct comparison . Dotted lines in the control group ( left column ) were replicated from α1DS in normal [apoCaM] ( A ) , essentially no difference . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 006 Notably , these channel variants expressed in HEK cells were quantified under normal concentrations of Ca2+-free CaM ( [apoCaM] ) , in which the three-motif complex of IQV/PCRD/DCRD as the core of CMI was perturbed by apoCaM that tends to pre-associate with IQV of the channel . To alleviate the interference with the potential cooperation in forming the threesome complex , we managed to substantially reduce [apoCaM] in the cells by overexpressing BSCaMIQ , a strong apoCaM buffer ( Black et al . , 2006 ) , to help identify the minimum requirement for CMI ( Figure 2B ) . In contrast to strong CDI in normal [apoCaM] evidenced from different channel variants ( Figure 2A ) , with [apoCaM] being substantially reduced ( but not lower than a realistic limit ) , surprisingly , α1DS-G24-DCRD and all the other DCRD-containing variants exhibited much attenuated CDI indicative of effective CMI ( Figure 2C , D and Figure 2—figure supplement 1 ) . Meanwhile , the stereotyped ultrastrong CDI normally observed from α1DS or α1DS-PCRD was nearly unaltered by apoCaM buffers , in that under the residue [apoCaM] these channels were still able to be pre-bound with apoCaM considering the effective affinity between apoCaM and the channel ( IQV or IQV-PCRD ) was reasonably high when there was no interference . Hence , mechanistically only DCRD and IQV are required by CMI so that PCRD becomes dispensable if [apoCaM] is low , unveiling the differential roles of PCRD and DCRD in the potential cooperation for CMI induction . The data thus far suggest that CMI mainly relies on DCRD/IQV binding to overcome free [apoCaM] of substantial level in normal cell conditions; and PCRD could facilitate CMI potentially by enhancing the binding of DCRD/IQV and thus the competition against apoCaM . However , the trio failed to attenuate CDI when they were separately expressed as three individual peptides ( Figure 3A–C ) , even when PCRD , DCRD and IQV ( contained in α1DS ) were all well expressed in the same cell as confirmed with fluorescent tags ( Figure 3—figure supplement 1 ) . By examining VGA ( JCa ) with aforementioned precautions , we also concluded that no CMI effect on α1DS was detectable with all the three groups of peptides: PCRD and DCRD , PCRD only , and DCRD only . When [apoCaM] was reduced with chelator BSCaMIQ as in Figure 2 , CDI remained ultrastrong and indistinguishable from α1DS control for all the three test groups ( Figure 3D–F ) , and VGA analyses came to the same conclusions consistently ( Figure 3—figure supplement 2D–F ) . 10 . 7554/eLife . 21989 . 007Figure 3 . Individual DCRD peptides expressed separately from IQV are unable to induce CMI . ( A ) PCRD and DCRD tagged with fluorescent proteins were coexpressed with α1DS as separate peptides . The presence of both YFP-PCRD and CFP-DCRD peptides in the same cell was confirmed under a fluorescence microscope ( Figure 3—figure supplement 1 ) . Under normal [apoCaM] in cells , no CMI effect was observed from ICa trace exhibiting CDI similarly as α1DS control , confirmed by comparable SCa values and indistinguishable r50 profiles . ( B and C ) Experiments and analyses were performed in normal [apoCaM] similarly as ( A ) , except that only YFP-PCRD ( B ) or only YFP-DCRD ( C ) was expressed with α1DS channels . Both resulted into strong CDI , with SCa and r50 indistinguishable from α1DS control . ( D−F ) When free [apoCaM] was substantially reduced by overexpressing BSCaMIQ ( apoCaM buffers ) , the above three cases in ( A−C ) were re-examined . Exemplar ICa traces exhibited strong CDI as quantified by SCa values and r50 profiles , similarly as the control group of α1DS in low [apoCaM] ( semitransparent line in red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 00710 . 7554/eLife . 21989 . 008Figure 3—figure supplement 1 . To confirm the presence of both PCRD and DCRD peptides in single cells . ( A ) YFP-PCRD and CFP-DCRD were coexpressed with α1DS ( left ) . For one single cell under patch recording ( right image ) , the epi-fluorescence of high intensity via CFP ( left image ) and YFP ( middle image ) channels was observed to confirm that both peptides were well expressed . ( B ) The exemplar recording of ICa from that particular cell in ( A ) exhibited strong CDI ( SCa ) , within the range of standard SCa for α1DS ( 0 . 76 ± 0 . 02 , n = 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 00810 . 7554/eLife . 21989 . 009Figure 3—figure supplement 2 . PCRD and DCRD as separate peptides are unable to inhibit VGA . ( A–C ) In normal [apoCaM] , when both PCRD and DCRD were expressed but as separate peptides ( A ) , the VGA profile of α1DS ( control , red thick semitransparent lines ) was unaltered as indexed by JCa . Similarly , no inhibition of VGA was observed when only PCRD ( B ) or only DCRD ( C ) was transfected . ( D−F ) When [apoCaM] was reduced by overexpressing BSCaMIQ , for all the three cases in ( A−C ) , their VGA profiles were indistinguishable from the control group ( red thick semitransparent lines ) , also confirmed by similar JCa values across different groups . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 009 Up to here , we established two extreme cases for CMI , i . e . , either all the three motifs were fused together ( strong CMI ) , or the three were totally isolated to each other ( no CMI ) . To explore intermediate conditions between the two extremes , we physically linked every two modules out of the three to coexpress with the third one , yielding three possible combinations in total . And they were , PCRD-DCRD peptides to coexpress with IQV ( contained in α1DS ) ( Figure 4A , combination I ) , channels containing IQV-PCRD to coexpress with DCRD ( Figure 4B , combination II ) , and channels containing linked IQV and DCRD to coexpress with PCRD ( Figure 4C , combination III ) . Interestingly , all the three combinations exhibited CMI , strongly attenuating both CDI and VGA as illustrated by green areas indicating the altered r50 and Jpeak profiles compared to α1DS control ( Figure 4A—C ) . Combination I also provided a great chance to revisit the assumption we made earlier regarding interchangeable PCRDD and PCRDF from CaV1 . 3 and CaV1 . 4 , respectively . Instead of dealing with two different channel variants , here the same α1DS served as the control to reliably evaluate PCRDD-DCRD vs PCRDF-DCRD peptides . And our assumption was validated by the negligible differences in VGA and CDI between the two PCRD-DCRD peptides ( Figure 4—figure supplement 1 ) . In previous reports , combinations of I and II were already suggested when studying endogenous DCT in CaV1 . 3 ( α1DL ) and CaV1 . 4 ( Singh et al . , 2008 , 2006 ) . In addition , PCRD ( combination III ) was able to induce pronounced CMI effects as we first observed in this work , which also completed all the three possible combinations of peptide-mediated CMI . 10 . 7554/eLife . 21989 . 010Figure 4 . Cooperation by PCRD , DCRD and IQV to induce CMI . ( A ) Both CDI and VGA of α1DS channels were attenuated by pre-linked PCRD-DCRD illustrated in the scheme ( top ) , shown in the exemplar traces ( middle ) and voltage-dependent r50 and Jpeak profiles ( lower two panels ) . The potency of CMI was indexed by SCa and JCa values and also illustrated by the green areas . Profiles of α1DS control were indicated by thick semitransparent lines in red ( lower two panels ) . ( B ) Both CDI and VGA of α1DS-PCRD were strongly prohibited by DCRD . ( C ) Both CDI and VGA of α1DS-G24-DCRD were attenuated by PCRD . ( D ) FRET 2-hybrid assays demonstrated that pre-linked IQV-PCRD motif ( YFP tagged ) exhibited strong binding with CFP-DCRD , with higher binding affinity ( Kd = 1135 , units in donor-cube fluorescence intensity ) than the binding affinity ( Kd = 4700 ) between CFP-DCRD and YFP-IQV itself ( without PCRD being fused ) . Additional PCRD peptides did not make any appreciable change ( Kd = 4624 ) for the binding between CFP-DCRD and YFP-IQV , unable to rescue the low affinity back to the high level that the constitutive PCRD fusion ( YFP-IQV-PCRD ) could achieve . FRmax values for all these binding curves were similar ( 3 . 50—3 . 87 ) . Each data point represented the FR ( y-axis , FRET ratio ) and Dfree ( x-axis , free donor concentration ) values averaged over five adjacent individual cells sorted by Dfree . ( E ) Working model to illustrate the collaboration among three components of PCRD , DCRD and IQV ( embedded within α1DS ) for CMI induction . Grey circles represent the effective combinations , which require that at least two ( out of three ) components are closely engaged ( e . g . , fusion ) to form the trio complex incorporating the third ( separate ) component . The three key combinations are I ( IQV+PCRD-DCRD ) , II ( IQV-PCRD+DCRD ) and III ( IQV-DCRD+PCRD ) , where ‘–’ denotes fusion or spatial closeness and ‘+’ indicates the separate peptide to be coexpressed . In addition , the positive control ( IQV-PCRD-DCRD ) also represents the native long variant α1DL; and the three components ( dotted circles in green , pink and blue ) completely separate to each other ( IQV+PCRD+DCRD ) produce no CMI effect , serving as one negative control . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01010 . 7554/eLife . 21989 . 011Figure 4—figure supplement 1 . Functional similarities between PCRDF and PCRDD . The pronounced CDI ( SCa ) and strong VGA ( JCa ) of α1DS channels ( indicated by red thick semitransparent lines in r50 and Jpeak plots ) were similarly attenuated by PCRD-DCRD peptides made from either PCRDF from CaV1 . 4 ( left column ) or PCRDD from CaV1 . 3 ( right column ) . Representative ICa traces ( top ) , CDI and VGA profiles ( panels in the two bottom rows ) across the full voltage range were shown for the two types of peptides , with no significant difference . Green areas represent the potency of attenuation , supporting that PCRDF and PCRDD were interchangeable for CMI in this study as we assumed . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01110 . 7554/eLife . 21989 . 012Figure 4—figure supplement 2 . Cooperative perturbation of apoCaM/IQV binding by DCRD and PCRD . ( A ) By 2-hybrid FRET with the CFP- or YFP-labeled constructs illustrated on top , IQV-PCRD ( YFP-tagged ) and CaM ( CFP-tagged , apo-state ) exhibited strong binding ( blue ) , quantified by the maximum FRET ratio ( FRmax ) and effective equilibrium dissociation constant ( Kd ) from iterative curve fitting . The values of FRmax and Kd were indicated right above the fitted curves ( in the same color ) when applicable . Unfilled-dots and filled-dots represent individual cells and averaged results ( over five cells ) respectively . The binding between apoCaM and IQV-PCRD were severely perturbed by FKBP-DCRD ( with FKBP tag adopted from subsequent rapamycin-inducible CMI , but without fluorescent tags of CFP/YFP ) , resulted into the data points ( red ) approaching the baseline of CFP/YFP mixture ( cyan ) , indicative of very weak Kd . The perturbation was presumably by way of the close cooperation between FKBP-DCRD and IQV-PCRD to compete against apoCaM . The FRET binding analyses were consistent with the patch-clamp recordings confirming that CDI ( SCa ) of α1DS-PCRD was attenuated in the presence of FKBP-DCRD ( inset ) . ( B ) In the absence of PCRD , the perturbation by FKBP-DCRD was much less effective . Without the presence of PCRD , CFP-CaM and YFP-IQV ( blue ) were still able to bind . However , in contrast to the strong perturbation seen in ( A ) , in the absence of PCRD , inclusion of DCRD peptides made no or very little difference in Kd and FRmaxvalues for CaM and IQV . The binding curve ( dotted line in light blue ) was replicated from CFP-CaM and YFP-IQV-PCRD ( A ) for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 012 Live-cell FRET two-hybrid assays demonstrated that DCRD bound IQV-PCRD with higher affinity ( Kd = 1135 , units in donor-cube fluorescence intensity ) than that between DCRD and IQV ( Kd = 4700 ) , regardless of whether PCRD as the standalone peptide was also present ( Kd = 4624 ) ( Figure 4D ) . Owing to the high affinity between DCRD and IQV-PCRD , DCRD overexpression nearly abolished the binding between apoCaM and IQV-PCRD ( i . e . , apoCaM/channel pre-association ) as demonstrated in the plot of FR-Dfree ( FRET ratio vs free donor concentrations ) ( Figure 4—figure supplement 1A ) ; meanwhile , without the presence of PCRD , apoCaM was still able to bind IQV , but the DCRD perturbation was much less effective ( Figure 4—figure supplement 1B ) . The electrostatic interaction between PCRD and DCRD hinted by CaV1 . 2 ( Hulme et al . , 2006 ) seemed generalizable but was not detected with the FRET pairs based on CaV1 . 4 DCT ( Liu et al . , 2010 ) . Both this putative PCRD/DCRD interaction and the binding between IQV and DCRD should be weaker and secondary compared to the strong binding between IQV-PCRD and DCRD ( Figure 4D ) . This explained why the potential contributions of these secondary interactions may or may not be evidenced , depending on whether the trio complex would come into being in particular settings . Our FRET binding analyses are consistent with the functional data shown earlier , thus providing additional information for the collaborations among CMI motifs . For instance , coexpression of all the three but separate components would not produce CMI ( Figure 3A ) . Also , DCRD strongly attenuated α1DS-PCRD but not α1DS control ( Figure 4B vs Figure 3C ) . Taken together , a cooperative scheme of CMI under physiological [apoCaM] in the cell was unveiled by functional and binding analyses ( Figure 4E ) , which depicted that if any two components have enough spatial closeness or intimacy , e . g . , by prearranged linkage , it would be sufficient for the third component to form the ‘trio’ complex and thus induce effective CMI ( combinations of I , II and III , highlighted in grey ellipses ) . Also , the fusion of all the three components together mimics the intrinsic CMI in α1DL ( positive control , Figure 1B ) ; and expression of three components separate to each other represents the negative control with no CMI ( Figure 3A ) . Based on such cooperation-dependent CMI ( Figure 4E ) and rapamycin-mediated heterodimerization ( Yang et al . , 2013 , 2007 ) , we devised the strategy to implement drug-inducible CMI , illustrated in the design of version 1 ( Figure 5A ) . FKBP ( rapamycin-binding protein ) and FRB fragment ( FKBP and rapamycin binding domain of the kinase mTOR ) as the tags were fused onto DCRD and PCRD , respectively . A small immunosuppressant molecule rapamycin that binds both FRB and FKBP was applied to link one FKBP-DCRD together with one PCRD-FRB , aiming to bind to the IQV domain of α1DS to compete off apoCaM and thus induce CMI according to the scheme of combination I ( Figure 4E ) . This design , if successful , should be directly applicable to native CaV1 . 3 since no modification is needed at the channel side . Attracted by such potentials , we implemented and validated this rapamycin-inducible peptide-mediated CMI in HEK cells with α1DS . In contrast to the stable time-course profiles of the control group ( Figure 5B ) , CDI ( SCa ) and VGA ( Ipeak ) were both rapidly attenuated upon applying 1 μM rapamycin . Within tens of seconds , dimerization between FKBP-DCRD with PCRD-FRB started to attenuate ICa as evidenced by time-dependent decays ( Figure 5C ) . 10 . 7554/eLife . 21989 . 013Figure 5 . Design schemes and experimental implementations of rapamycin-inducible CMI . ( A ) Design principles for chemical-inducible CMI . Rapamycin simultaneously binds one FKBP and one FRB to combine any two FRB/FKBP-tagged components ( PCRD and DCRD , in this version 1 ) selected from the three components of PCRD , DCRD and IQV . Thus , the three components would form the combinations ( combination I , in the version 1 ) to satisfy the requirement of cooperative CMI ( Figure 4E ) . ( B ) One representative implementation of chemical-inducible CMI . According to the design of version one in ( A ) , PCRD-FRB and FKBP-DCRD were constructed and coexpressed with α1DS . Exemplars of Ca2+ current traces demonstrated that acute effects were induced by applying 1 μM rapamycin to the multi-component system of version 1 ( lower panel ) , in contrast to α1DS control with no appreciable changes in ICa ( upper panel ) . ( C ) Statistical summary of rapamycin-induced CMI ( version 1 ) . Averaged values over multiple cells ( number indicated in parentheses ) for the indices SCa ( upper ) or normalized Ipeak ( lower ) demonstrated time-dependent attenuations on both CDI and VGA upon rapamycin application as compared to the control group . ( D ) Rapamycin perfusion induced rapid translocation of YFP-FKBP-DCRD or YFP-FKBP-PCRD ( version 2 ) onto the membrane by linking with FRB-CFP-Ras within 5 min , shown by confocal images via wide-field , CFP , and YFP channels . The local concentrations of YFP-FKBP-DCRD and YFP-FKBP-PCRD were substantially enhanced as suggested by the condensed YFP fluorescence at the membrane ( outlined by CFP fluorescence ) . ( E and F ) According to rapamycin-inducible CMI of version 2 , recombinant channels of α1DS-G12-FRB were coexpressed with cytosolic YFP-FKBP-PCRD and YFP-FKBP-DCRD . To enhance local concentrations of FKBP-tagged peptides near the channels , membrane-localized FRB-CFP-Ras was also overexpressed . Exemplars of Ca2+ current traces ( E ) exhibited strong attenuation ( lower ) , in contrast to stable ICa from α1DS-G12-FRB alone ( upper ) . In contrast to the control group , temporal profiles of rapamycin-inducible CMI ( version 2 ) indicated strong attenuations on CDI ( SCa , upper ) and VGA ( normalized Ipeak , lower ) . Based on totally five trials ( cells ) , SCa changed from 0 . 77 ± 0 . 00 to 0 . 41 ± 0 . 06 and Ipeak was reduced to 35% ± 3% of the basal levels ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01310 . 7554/eLife . 21989 . 014Figure 5—figure supplement 1 . Temporal profiles of fluorescent images for rapamycin-induced membrane-targeting . ( A and B ) Top rows: cartoon depicting the design of the rapamycin-induced translocation of DCRD ( A ) or PCRD ( B ) ( Version 2 of chemical-inducible CMI ) . FRB-CFP-Ras was constitutively anchored to the plasma membrane by incorporating membrane-targeting segment from Ras protein . Lower rows: confocal images acquired via YFP channel showing rapid translocation of YFP-FKBP-DCRD ( A ) or YFP-FKBP-PCRD ( B ) to plasma membrane upon 1 μM rapamycin perfusion ( left ) . At the time of ~120 s or later , YFP fluorescence was substantially condensed to overlap with the stable CFP fluorescence ( FRB-CFP-Ras ) on the membrane ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01410 . 7554/eLife . 21989 . 015Figure 5—figure supplement 2 . Detailed characterizations for rapamycin-induced CMI . ( A and B ) Voltage-dependent profiles of CDI and VGA exhibited no difference when α1DS channels were with ( B ) or without ( A ) rapamycin . Thick lines ( red semitransparent ) represent the CDI and VGA profiles from the α1DS control group . ( C ) Coexpression of CFP-PCRD-FRB and FKBP-DCRD-YFP with α1DS representing the control conditions before rapamycin did not cause any appreciable change in CDI and VGA , confirmed by pronounced SCa and JCa comparable to the values from α1DS control . ( D ) When rapamycin was applied to α1DS coexpressed with CFP-PCRD-FRB and FKBP-DCRD-YFP , SCa and JCa averages over different cells were attenuated with moderate potency ( illustrated by green areas ) presumably by rapamycin-induced linkage between PCRD and DCRD ( version 1 of inducible CMI ) to form the cooperation for effective CMI . ( E ) In rapamycin , with enhanced local concentrations of YFP-FKBP-PCRD and YFP-FKBP-DCRD ( version 2 of rapamycin-induced CMI ) , the potency was much improved , demonstrated by more pronounced attenuations concurrently on both CDI and VGA ( larger green areas in both r50 and Jpeak profiles ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 015 Encouraged by the first prototype of inducible CMI , we then proceeded to enhance the potency from moderate ( as in version 1 ) to ultrastrong attenuation similarly as in Figure 1B . We reasoned that the underperformance here should be mainly due to the low effective concentration of the linked peptides if closely comparing constitutive ( Figure 4A ) and the rapamycin-inducible CMI ( Figure 5C ) . The upgrade version ( version 2 ) was designed to enhance concentration ( local to channels ) and thus the potency of CMI peptides . Membrane-targeting motif Ras ( Yang et al . , 2007 ) was used to construct FRB-CFP-Ras , which would bind DCRD/PCRD tagged with FKBP through rapamycin-inducible heterodimerization . As demonstrated by confocal fluorescence imaging , in about two minutes YFP-FKBP-DCRD and YFP-FKBP-DCRD translocated to the membrane upon rapamycin application ( Figure 5D , Figure 5—figure supplement 1 ) . Meanwhile , FRB was also fused onto the channel to construct α1DS-G12-FRB for inducible dimerization with FKBP-PCRD or FKBP-DCRD , to facilitate peptide-mediated CMI according to the schemes of combination II and III ( Figure 4E , Figure 5D ) . More potently than the earlier design but in a similar time course ( version 1 , Figure 5B , C ) , rapamycin induced strong CMI effects on ICa within 4–5 min ( version 2 , Figure 5E , F ) . Both CDI and VGA were substantially attenuated: SCa ( 0 . 41 ± 0 . 06 , n = 5 ) and Ipeak ( 35% ± 3% , n = 5 ) . Both indices reached the plateau within 4–5 min , which was about the speed of full solution exchange in our recording system , indicating that rapamycin perfusion was the major time-limiting factor . Thus , the acute nature of CMI was strongly supported by the temporal profiles of inducible CMI effects on CDI/VGA ( both version 1 and version 2 ) . To further consolidate the results , direct drug-channel effects were examined for rapamycin as the negative controls; and full CDI and VGA profiles were characterized across the whole voltage range before and after applying rapamycin ( Figure 5—figure supplement 2 ) . One interesting feature we discovered from rapamycin-inducible CMI was that the current amplitude at 300 ms ( I300 ) stayed at the same levels during the whole time course ( Figure 6A , top two rows ) , in contrast to rapidly decayed Ipeak as outlined by blue dotted lines . In this context , CMI was quite unique compared to other conventional inhibitions . In the run-down process ( Kepplinger and Romanin , 2005 ) or the blockage by isradipine ( Anekonda and Quinn , 2011 ) , the number of functional channels was reduced , which resulted into attenuations on ICa during the whole depolarization step including both Ipeak and I300 ( Figure 6A , lower two rows ) . In contrast , SCa remained constant for α1DS undergoing run-down or blockage since intrinsic gating properties ( e . g . , CDI ) of each channel were not altered , whereas SCa was attenuated in CMI ( Figure 6—figure supplement 1A ) . 10 . 7554/eLife . 21989 . 016Figure 6 . Mechanistic insights enlightened by unique features of CMI . ( A ) Temporal profiles between CMI and conventional inhibitions were compared . For rapamycin-inducible CMI ( two versions in the top two rows of panels ) , representative Ca2+ current traces in rapamycin were selected from sequential time-points ( left column ) and superimposed together ( middle column ) for comparison . The time sequence was indicated by color: the first in pink , intermediate in grey , and the last in blue . Ipeak exhibited the trend of inhibition but not for I300 ( ICa at 300 ms ) ( right column ) . In contrast , for run-down process and isradipine blockage ( bottom two rows ) , both Ipeak and I300 exhibited the declining trends indicating substantial inhibitions . ( B ) Ca2+ current density at 300 ms ( J300 , Ca ) remained at the same level ( indicated by the J300 values at −10 mV ) , for various channel variants under test , regardless of whether CMI was in effect . Thick lines ( semitransparent in red ) represent the J300 , Ca profile of α1DS control . ( C ) Functional and the structural insights into the two modes of inhibition . DCT/apoCaM-dependent CMI and Ca2+/CaM-mediated CDI result in indistinguishable gating ( green arrows ) appearing due to similar causes , i . e . , either total or partial loss of the apo-state CaM/IQV complex . That says , in CMI , apoCaM pre-association is totally lost; and in CDI , apoCaM is calcified and dislocated from the pre-association sites . Although the triggers are different for CMI vs CDI , i . e . , DCT competing ( off apoCaM ) vs Ca2+ binding ( onto apoCaM ) , the similarities between the two different inhibitory regulations of CMI and CDI invite the hypothesis that the core gating machinery ( cyan squares ) upon depolarization might step into the same scenario/mode including structural details ( red arrows ) . A series of current traces ( on the right ) indicate CMI with different potency ( enhanced from pink to cyan , upper ) , in comparison with the trace at different stages of CDI ( developed from pink to cyan , lower ) superimposed with the trace from ultrastrong CMI ( dotted trace in cyan ) . These analyses point to one important notion that the lower limit of CMI is determined by end-stage CDI ( the traces or the phases in cyan ) , and thus the dynamic changes for CMI effects on α1DS ( indexed with Ipeak and SCa ) are preset , i . e . , from the pink ( no CMI and ultrastrong CDI ) to the cyan ( ultrastrong CMI and no CDI ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01610 . 7554/eLife . 21989 . 017Figure 6—figure supplement 1 . Detailed comparisons among CMI , run-down and blockage . CMI was compared with conventional inhibitions/inhibitors including channel blockage ( by isradipine ) and run-down process ( reduction of channel numbers ) , indexed with SCa , Ipeak and I300 . Both rapamycin-induced CMI and run-down/blockage exhibited time-dependent decreases in Ipeak , as expected from effective ICa inhibitions ( B ) . In contrast , CDI ( SCa ) did not change in run-down/blockage; whereas for CMI SCa exhibited rapamycin-dependent attenuation , similar to Ipeak ( A ) . Moreover , the plateau of ICa ( I300 ) remained constant in spite of Ipeak attenuation in CMI , whereas I300 changed ( decreased or recovered ) in a similar time-course to that of Ipeak during run-down and blockage ( C ) . Since SCa is in proportion to Ipeak/I300 , the two major types of inhibitions ( CMI vs conventional inhibitions ) were distinct within this context: for CMI , both Ipeak and SCa changed but with I300 being kept constant; in contrast , for conventional inhibitions , both Ipeak and I300 changed , but SCa remained unaltered . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01710 . 7554/eLife . 21989 . 018Figure 6—figure supplement 2 . Indices at different time points of ICa to quantify CMI and end-stage CDI . ( A ) Analyses at both 300 ms and 1000 ms were able to reliably unveil that channels subject to CMI were essentially tuned to approach ICa levels in end-stage CDI . Upper panel , we set I1000 ( ICa at 1000 ms ) of both exemplar traces to the same level for α1DS and α1DS-PCRD-DCRD ( in principle , they should be the same ) ; and the levels of ICa at 300 ms ( I300 ) turned out to be nearly the same , similarly as the analyses with I1000 . At 300 ms , the kinetic difference between IBa ( normalized to the peak of ICa ) and ICa was very little , indicative of end-stage CDI ( upper right ) . Full profiles of Ca2+ current density ( indexed with J300 or J1000 ) were indistinguishable for both groups of α1DS and α1DS-PCRD-DCRD , further supporting that both are qualified as reliable indices for CMI analysis . ( B ) For rapamycin-inducible CMI ( version 2 ) by supplying PCRD , DCRD and α1DS with membrane-targeting and rapamycin-inducible mechanisms , representative ICa traces were selected from sequential time-points ( left ) , and the first ( purple ) and the last ( cyan ) traces were superimposed together for comparison ( right ) . Ipeak exhibited a trend of time-dependent decrease; meanwhile , both I300 and I1000 remained rather constant ( dotted outlines in black , blue and pink respectively ) . ( C ) Statistical summary of ICa amplitudes for rapamycin-induced CMI ( version 2 ) . ICa values were measured at different time points , i . e . , Ipeak , I100 , I300 and I1000 , each of which was either rescaled to Ipeak ( left ) or to its own ( right ) . Ipeak exhibited robust inhibition due to CMI , and a trend of moderate decline ( right ) was noticeable from I100 ( ICa at 100 ms ) ; whereas I300 and I1000 clearly remained constant throughout the full time-course , thus reliably representing the end stage or steady state of CDI . In line with our previous analyses , the current level of I300 or I1000 determined the lower limit of ICa ( measured by Ipeak ) subject to CMI of the maximum potency . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 01810 . 7554/eLife . 21989 . 019Figure 6—figure supplement 3 . The high-affinity binding between DCRD and the channel was not perturbed by Ca2+/CaM . ( A ) Upon applying ionomycin , an ionophore massively raising intracellular Ca2+ levels , endogenous CaM should switch from Ca2+-free state ( apoCaM ) to Ca2+-bound state ( Ca2+/CaM ) . However , similar Kd values were obtained by fitting the FRET binding curves before and after ionomycin application ( pink , before ionomycin; red , in ionomycin ) , indicating that Ca2+/CaM was unable to perturb the strong binding between DCRD and IQV-PCRD . In the FR-Dfree plots , unfilled-dots and filled-dots represent individual cells and averaged results ( over five cells ) respectively . ( B ) Following ionomycin administration , the indices ( Kd and FRmax ) of the binding between CaM and IQV were clearly strengthened consistent with the IQV/CaM interactions previously established . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 019 Would such constant I300 be just a coincidence ? By revisiting different channel variants tested earlier , we concluded that I300 indeed remained at similar amplitudes for them all ( Figure 6B ) , whereas their Ipeak and SCa exhibited broad dynamic ranges . To confirm , depolarization pulses of longer duration ( 1000 ms ) were also utilized to conduct constitutive and rapamycin-inducible CMI ( Figure 6—figure supplement 2 ) . In both modalities of CMI , the behavior of I300 was similar to I1000 and both indices remained constant . In addition , CDI already approached its steady-state at 300 ms ( end-stage CDI ) , and the slight difference between I300 and I1000 was largely due to VDI and negligible in this study . Throughout this work , I300 was employed as the major index to for both CMI and CDI analyses . The constant I300 ( or J300 ) essentially set the lower limit for CMI . That said , while CMI potency was getting higher , Jpeak was tuned down to more closely approach J300 ( e . g . , Figure 6B vs Figure 1B ) . More clear evidence came from rapamycin-inducible CMI: the attenuated Ipeak and constant I300 altogether account for SCa ( CDI ) attenuation; and for ultrastrong CMI , SCa would be nearly abolished ( 0 ) , indicating that Ipeak would reach approximately the level of its lower limit ( I300 ) . These lines of evidence also led us to conclude that a reduction of Ca2+ influx is guaranteed for CMI , despite concurrent VGA/CDI ( activation/inactivation ) attenuations . Such concurrency apparently would cause contradictory effects on Ca2+ influx: inhibition of CDI ( ICDI ) would tend to enhance Ca2+ influx , in opposition to attenuation on VGA ( less Ca2+ influx ) . Even though , we just clarified that CMI attenuation on CDI is actually realized by reducing Ipeak while maintaining the end-stage ICa ( indexed with I300 or I1000 ) , which ensures the overall Ca2+ influx is reduced . Thus , the major uncertainty was relieved for CMI to emerge as a one new modality of Ca2+ channel inhibition . These observations are consistent with the speculation that CDI might be the reversed process to the apoCaM promotion of open probability ( Adams et al . , 2014 ) . High similarities are shared by DCT/apoCaM-dependent CMI and Ca2+/CaM-mediated CDI: functionally CMI ( ultrastrong ) and CDI ( end-stage ) could result into similar gating; and mechanistically they appear to be triggered by similar events: the pre-association between apoCaM and IQV is either totally abolished ( CMI ) or drastically altered ( CDI ) ( Figure 6C ) . These facts suggest that potentially the same set of ‘core machinery’ ( cyan square ) mediates both CMI and CDI with very similar structures and structural changes . To exclude potential complications from ambient Ca2+/CaM , we performed FRET experiments to ensure that the interactions within the trio complex of IQV/PCRD/DCRD were largely unaffected by Ca2+/CaM produced from ionomycin-introduced Ca2+ and endogenous CaM , although Ca2+/CaM indeed exhibited higher binding affinity to IQV than apoCaM ( Figure 6—figure supplement 3 ) . CaV1 . 3 plays a pivotal role in subthreshold oscillation and suprathreshold pacemaking in diverse cell types including neurons ( Chay and Keizer , 1983; Comunanza et al . , 2010 ) , e . g . , dopaminergic neurons in the substantia nigra compacta ( SNc ) ( Chan et al . , 2007 ) . Pathophysiological linkages of PD or other neurodegenerative diseases with CaV1 . 3 functions have been evidenced in multiple lines of studies ( Guzman et al . , 2009; Puopolo et al . , 2007 ) , including CaV1 . 3 antagonists as potential therapeutic interventions ( Anekonda and Quinn , 2011; Pasternak et al . , 2012; Triggle , 2007 ) . To explore the physiological tuning of CMI and its therapeutic potentials , we constructed a customized model for SNc neuron and CaV1 . 3 channels using the software Neuron ( Carnevale and Hines , 2006 ) . Computational analyses were performed to compare the effects between different levels of CMI attenuations , to simulate the tuning of CMI by multiple factors including endogenous DCTs and alternative splicing , apoCaM fluctuations , and apoCaM buffer proteins such as calpacitin ( Gerendasy , 1999; Xia and Storm , 2005 ) . Under CMI of different potency , ICa from α1DS ( no CMI ) , α1DL ( intermediate CMI ) , or α1DS with inducible peptide dimerization ( ultrastrong CMI ) , exhibited distinct Ipeak , SCa and Ca2+ influx ( Figure 7A—C , left ) ( green areas indicating the reduction of Ca2+ influx ) . Meanwhile , the time rates of Ca2+ oscillation and autonomous spiking in the SNc model were also tuned down to different levels in accordance with ICa attenuations or CMI potencies ( Figure 7A—C ) . Experimentally , CMI was compared with other conventional modalities of channel inhibition , for their differences in ICa attenuations ( Figure 6A , Figure 6—figure supplement 1 ) . In parallel , we also simulated the conventional blockage in the CaV1 . 3/SNc model ( Figure 7D ) . Similar attenuations in oscillation and pacemaking were observed from the SNc model when the moderate ( 28% ) blockade clearly reduced overall Ca2+ influx ( green area ) , supporting the current strategy to develop PD therapeutics based on conventional CaV1 . 3 antagonists ( Gudala et al . , 2015; Hurley et al . , 2013; Kang et al . , 2012; Pasternak et al . , 2012 ) . 10 . 7554/eLife . 21989 . 020Figure 7 . Inhibition of SNc oscillation and pacemaking by CMI . ( A and B ) Oscillation and pacemaking of SNc neurons are mediated by CaV1 . 3 , presumably mixed with both short ( α1DS ) and long ( α1DL ) splice variants . Lack of CMI , α1DS ( pink ) exhibited large Ipeak and strong SCa ( A ) , altogether resulted into more Ca2+ influx than α1DL ( blue ) , as indicated by the green area ( B ) . Side-by-side comparison demonstrated that CaV1 . 3-dependent oscillation ( middle column ) and pacemaking ( right column ) were slowed down , presumably by intrinsic CMI of α1DL in comparison to α1DS . ( C ) Effects of ultrastrong CMI ( cyan ) on SNc neurons . Under the CMI of the maximum potency ( left ) , Ipeak was reduced to the lower limit , i . e . , the level of I300 , and SCa was approaching 0 . Accompanied with reduction in Ca2+ influx ( green area ) , oscillation ( middle ) and pacemaking ( right ) rates were also further reduced compared with α1DS ( A ) or α1DL ( B ) . ( D ) Effects of conventional blockage . Conventional inhibition by blockers ( orange ) attenuated ICa ( 28% reduction ) and Ca2+ influx ( green area ) but kept SCa unaltered ( left ) , by which oscillation ( middle ) and pacemaking ( right ) rates were slowed down compared with the α1DL control ( B ) . ( E ) The potential regulatory scheme for CaV1 . 3-mediated pacemaking in SNc neurons . The actual gating of ICa ( similar to B ) could fall into intermediate levels between two extreme conditions of rather weak CMI ( apoCaM-bound and capable of strong CDI ) ( A ) , or ultrastrong CMI ( apoCaM-off and abolishment of CDI ) ( C ) . Hence , both CaV1 . 3 and pacemaking behaviors are potentially bidirectionally regulated , e . g . , apoCaM and apoCaM-binding proteins , to maintain the homeostatic balance of Ca2+ influx . Pathological dysregulations of CaV1 . 3 and the resulted Ca2+ imbalance might underlie PD and other diseases , as the rational to develop therapeutic perturbations based on CMI . DOI: http://dx . doi . org/10 . 7554/eLife . 21989 . 020 Collectively , one common rule was deduced for CMI , CDI and other modalities of inhibition: Ca2+ influx via CaV1 . 3 serves as the major index to determine the downstream of oscillation/pacemaking , i . e . , reduction of Ca2+ influx leads to attenuation of pacemaking whereas enhancement of Ca2+ influx causes augmentation of pacemaking . These analyses on the potential roles of CMI together with those on CMI mechanisms ( Figure 6C ) suggest that multiple molecular processes in CMI , especially those pertaining to apoCaM and the trio complex , are crucial for excitability and signaling of SNc neurons as well as related pathology , such as PD ( Hurley et al . , 2013; Scharinger et al . , 2015 ) ( Figure 7E ) . In this context , CMI , emerging as one key regulatory mechanism distinct from others , essentially reduces Ca2+ influx , local and cytosolic Ca2+ concentration , and oscillation and pacemaking , all of which are potentially subject to bidirectional tuning under pathophysiological conditions .
For LTCCs , gating inhibition by carboxyl-tail motifs ( CMI ) seems to be a universal mechanism preventing excessive Ca2+ influx and intracellular Ca2+ overload . In CaV1 . 1 and CaV1 . 2 channels , C-termini of α1 subunits are subject to proteolytic cleavage and the site of cleavage is located between PCRD and DCRD ( Abele and Yang , 2012; Hulme et al . , 2005 , 2006 ) . Hence , CMI serves as the potential mechanism for endogenous DCRD-containing fragments to attenuate ICa by cooperating with PCRD and IQV . For CaV1 . 3 , the possibility of proteolytic cleavage has not yet been fully explored . Instead , DCT is subject to alternative splicing thus generating α1D isoforms in two major categories: with ( α1DL ) or without ( α1DS or similar variants ) autonomous CMI ( Bock et al . , 2011; Singh et al . , 2008 ) , both broadly expressed in various tissues and organs . CaV1 . 4 channels have profound CMI due to its strong DCT against apoCaM , as evidenced by weak inactivation and small currents in native cells , altogether allowing sustained Ca2+ influx and continuous neural transmission in retinal neurons ( Singh et al . , 2006 ) and immune cells ( Omilusik et al . , 2011 ) . To this end , CMI is a prevalent modality of inhibition across CaV1 family members , awaiting further investigations into the molecular mechanisms and pathophysiology . Based on constitutive and inducible CMI effects , we have gained considerable insights into CaV1 gating . First of all , CMI as an emerging modality of channel inhibition resembles CDI , one of the prominent features of Ca2+ channels ( Ben-Johny and Yue , 2014 ) . Targeting the mechanisms of CDI , most structural efforts have been directed to channels in Ca2+ conditions , evidenced by multiple studies focusing on Ca2+/CaM bound with key channel motifs such as IQ or IQV ( Ben Johny et al . , 2013; Fallon et al . , 2005; Kim et al . , 2008; Mori et al . , 2008 ) and NSCaTE ( Dick et al . , 2008 ) , which have not achieved unequivocal viewpoints toward CDI mechanisms . CMI emphasizes on the importance of apoCaM-bound and apoCaM-off structures in Ca2+-free conditions , which represent the states of ‘Activated’ ( pre-CMI ) and ‘Inhibited’ ( post-CMI ) , promisingly holding the key to understand the core mechanisms that control the gating of Ca2+ channels . Functionally , CMI helps clarify the long-standing complication on the overall effects of CDI on the actual Ca2+ influx into the cell . Historically , CDI is considered as a negative feedback to autonomously reduce Ca2+ entries ( Alseikhan et al . , 2002; Budde et al . , 2002; Liu et al . , 2010 ) . On the other hand , stronger CDI is often accompanied with larger amplitude or facilitated activation of ICa ( Adams et al . , 2014; Liu et al . , 2016; Singh et al . , 2008 ) , altogether raising the question whether Ca2+ influx is reduced or enhanced by CDI . Here we provide direct evidence that attenuation of CDI via endogenous factors and mechanisms is essentially a reduction of Ca2+ influx , assured by the fixed level of Ca2+ current at the late phase of ICa , i . e . , I300 measured in this work , remaining the same regardless of the changes in Ipeak or SCa . In this context , CMI ( in inverse correlation to CDI strength ) is the more direct index of channel inhibition . In contrast , CDI strength itself ( SCa ) could be changed by various mechanisms of action besides [apoCaM] tuning , which may or may not actually reduce Ca2+ influx . For instance , channel mutations in congenital diseases , e . g . , Timothy syndrome ( Splawski et al . , 2004 ) , could cause less inactivation in ICa and thus elevate the Ca2+ entries . Also , small-molecule compounds could produce confounding effects on Ca2+ influx by triggering multiple events opposed to each other ( Hille , 2001 ) , e . g . , roscovintine ( Yarotskyy et al . , 2010 ) reportedly damps activation and meanwhile enhances inactivation , raising uncertainties in its actual effects on Ca2+ influx . CaV1 . 3 acts as the dominant factor underlying subthreshold oscillation and pacemaking activities in SNc dopaminergic neurons , although the detailed mechanisms are not fully elucidated ( Chan et al . , 2007; Durante et al . , 2004; Putzier et al . , 2009 ) . It is currently believed that activity-dependent engagement of CaV1 channels elevates mitochondrial oxidant stress and vulnerability of SNc neurons , contributing to disease progression of PD ( Guzman et al . , 2010 ) . Additional support also comes from clinical evidence that CaV1 blockers for hypertension treatment , e . g . , 1 , 4-dihydropyridines ( DHPs ) appear to reduce the risk of PD ( Pasternak et al . , 2012; Ritz et al . , 2009 ) . However , concerns are raised on developing PD therapeutics based on DHP or like inhibitors . First of all , it has been argued whether CaV1 . 3 is indeed a prerequisite for pacemaking and autonomous oscillation , because of pharmacological complications of DHP , including non-specific side effects and discrepancies in dose dependence ( Guzman et al . , 2009; Putzier et al . , 2009 ) . Viral or transgenic delivery of CMI peptides to CaV1 . 3 channels in neurons could circumvent above problems , in hope to unequivocally confirm the actual role of CaV1 . 3 . Secondly , PD interventions deploying CaV1 . 3 antagonists could be improved in the following aspects: ( 1 ) specificity , for which genetically-encoded CMI-based inhibitors provide the desired specificity intrinsic to CaV1 channels; ( 2 ) assurance of Ca2+ reduction , for which CMI is mechanistically guaranteed to reduce Ca2+ influx; however , certain antagonists may not be able to provide such assurance due to aforementioned reasons . CMI-based interventions and therapeutics could be beneficial to other diseases in addition to PD . A wide spectrum of mental disorders are closely involved in dysregulations of CaV1 channels , in that disease-linked mutations in α1 and β2 subunits result into abnormal gating properties ( Azizan et al . , 2013;Cross-Disorder Group of the Psychiatric Genomics Consortium , 2013;Schizophrenia Working Group of the Psychiatric Genomics Consortium , 2014; Scholl et al . , 2013 ) . Encouraged by the results from SNc modeling that the dysregulated CaV1 . 3 , Ca2+ influx and pacemaking could be corrected back to normal , we expect CMI and its therapeutic potentials to manifest in diverse pathophysiology including CaV1 channelopathies . DCT and DCRD effects have been suggested in a few prior works , e . g . , DCT peptides truncated from long variants of CaV1 . 3 or CaV1 . 4 to coexpress with the truncated channels lacking the DCT motif ( Singh et al . , 2008 , 2006 ) , mainly to demonstrate the ICDI ( inhibition of CDI ) effect . Here , to advance the understanding , we clarify that multiple C-terminal motifs cooperatively and acutely compete apoCaM off the channel; and unveil that CMI functionally resembles CDI such that the maximum potency of CMI is preset by end-stage CDI , which also ensures CMI reduces the overall Ca2+ influx . Moreover , as one major goal beyond basic biophysics , we achieve CMI-based peptides for CaV1 . 3 of native forms , including the short splice variant α1DS in this work and potentially also the long variant α1DL , another major isoform in native tissues ( Yang et al . , 2015 ) . In parallel , one study focusing on CaM , the other side of the competition , speculated on CDI mechanisms as a simple CaM-off process based on the data produced with local enrichment of apoCaM ( Adams et al . , 2014 ) , which , in our view , may have the follow concerns . First , prescreening is needed to control the baseline gating since CaM itself is very well expressed in cells and easily reaches concentrations high enough to diminish the dynamic space to gauge any further change in CDI ( Liu et al . , 2010 ) . Second , even with above practical issues being handled , any claims about CDI still await further proof with direct evidence based on actual inhibition such as CMI . Third , apoCaM , due to its central position and multifaceted roles in cell signaling ( Cheung , 1980; Chin and Means , 2000; Guzman et al . , 2010; Hoeflich and Ikura , 2002 ) , is nearly impossible to develop into molecular tools and therapeutics . To expand CMI-based inhibitors onto in vivo applications , the approach of rapamycin-mediated heterodimerization may not be directly applicable due to potential interferences with normal cell proliferation , growth and survival ( MacMillan , 2013 ) . Nevertheless , the proof-of-concept prototypes in this work lay the foundations for further development and optimization , e . g . , by way of light-inducible dimerization ( Kennedy et al . , 2010; Levskaya et al . , 2009; Yazawa et al . , 2009 ) , to explore the biological ramifications and therapeutic potentials of CMI in diverse settings .
α1D_Short ( denoted as α1DS ) variants were constructed by introducing a unique XbaI site following the IQ domain . PCRD was cloned from either CaV1 . 4 α1F ( NP005174 , GenBank accession number , the same as follows ) or CaV1 . 3 α1D ( NM_000720 ) ; and DCRD was based on CaV1 . 4 α1F ( NP005174 ) , and CFP/YFP-tagged DCRD constructs were made using a similar process as described previously ( Liu et al . , 2010 ) . Then other CFP/YFP-tagged constructs were cloned by replacing DCRD with appropriate PCR amplified segments , via unique NotI and XbaI sites , including YFP-PCRD , CFP-PCRD-DCRD , and CFP-CaM . Peptides of IQV and IQV-PCRD were cloned from CaV1 . 2 α1C ( NM_199460 . 3 ) and CaV1 . 3 α1D ( NM_000720 ) , based on preIQ3-IQ and preIQ3-IQ-PCRD as the templates respectively . YFP-IQV and YFP-IQV-PCRD were cloned similarly as YFP-PCRD . Segments of PCRD-DCRD , PCRD , DCRD with different glycines ( G0 , G6 and G24 ) were PCR-amplified with SpeI and XbaI sites and inserted into aforementioned α1DS . BSCaMIQ , serving as the apoCaM buffer in this study , was kindly provided by Dr . A . Persechini ( Black et al . , 2006 ) . Rapamycin-inducible system consisted of FRB , which was based on 93 a . a . rapamycin binding motif of MTOR; and FKBP , which was 108 a . a . human FKBP-12 ( AAA58472 ) . Constructs of CFP-PCRD-FRB , YFP-FKBP-DCRD , YFP-FKBP-PCRD and FRB-CFP-ras were made by appropriate design . PCRD segment was amplified by PCR with flanking NotI and EcoRI then cloned directionally via these two unique sites into CFP-FRB_pcDNA4_HisMaxC expressing plasmids . DCRD segment was amplified by PCR with flanking BamHI and EcoRI then cloned directionally via these two unique sites into YFP-FKBP_pcDNA4_HisMaxC expressing plasmids . The fusion of YFP-FKBP-PCRD and FRB-CFP-ras were ligated by overlap extension PCR with flanking KpnI and XbaI then cloned into pcDNA3 expressing plasmids via these two unique sites . To make constructs of PCRD-FRB and FKBP-DCRD , segments from CFP-PCRD-FRB and YFP-FKBP-DCRD respectively were amplified by PCR with flanking KpnI and XbaI then cloned directionally into pcDNA4 vector . For chimeric channel α1DS-G12-FRB , a linker containing 12 glycine residues was fused with FRB by overlap PCR with flanking SpeI and XbaI and cloned into α1DS containing engineered cloning sites before the stop codon . For whole-cell electrophysiology and confocal fluorescence imaging , HEK293 cells were cultured in 60 mm dishes or 35 mm No . 0 glass-bottom dishes , and constructs were transiently transfected according to an established calcium phosphate protocol ( Liu et al . , 2010 ) . HEK293 cell line was generously provided by Dr . Zhijie Chang ( Tsinghua University ) . The cell line was free of mycoplasma contamination , checked by PCR with primers 5’- GGCGAATGGGTGAGTAACACG −3’ and 5’- CGGATAACGCTTGCGACCTATG −3’ . 5 μg of cDNA encoding the desired α1D subunit , along with 4 μg of rat brain β2a ( M80545 ) and 4 μg of rat brain α2δ ( NM012919 . 2 ) subunits were applied . All of the above cDNA constructs contained a cytomegalovirus promoter . To enhance expression , cDNA for simian virus 40 T antigen ( 1–2 μg ) was also co-transfected with channel constructs . For all the peptides co-transfected , 2 μg of plasmids were added together with channels . Cells were washed with PBS 6–8 hr after transfection and maintained in supplemented DMEM , then incubated for at least 48 hr in a water-saturated 5% CO2 incubator at 37°C before electrophysiology experiments and confocal microscopy experiments . For FRET optical imaging , HEK293 cells cultured on 35 mm No . 0 glass-bottom dishes were transfected with cDNAs ( 1–5 μg each ) by Lipofectamine 2000 ( Invitrogen , Waltham , MA ) . Cells were washed with PBS 4–6 hr after transfection and maintained in supplemented DMEM , then incubated for 24–48 hr in a water-saturated 5% CO2 incubator at 37°C before FRET 2-hybrid experiments . Whole-cell recordings of transfected HEK293 cells were performed at room temperature ( 25°C ) using an Axopatch 200B amplifier ( Axon Instruments , Sunnyvale , CA ) . Electrodes were pulled with borosilicate glass capillaries by a programmable puller ( P-1000 , Sutter Instruments , Novato , CA ) and heat-polished by a microforge ( MF-830 , Narishige , Japan ) , resulting in 1–3 MΩ resistances , before series resistance compensation of about 70% . The internal solutions contained ( in mM ) : CsMeSO3 , 135; CsCl2 , 5; MgCl2 , 1; MgATP , 4; HEPES , 5; and EGTA , 5 , adjusted to 290 mOsm with glucose and pH 7 . 3 with CsOH . The extracellular solutions contained ( in mM ) : TEA-MeSO3 , 140; HEPES , 10; CaCl2 or BaCl2 , 10 , adjusted to 300 mOsm with glucose and pH 7 . 3 with TEAOH , all according to the previous reports ( Liu et al . , 2016 , 2010 ) . For run-down experiments , the content of MgATP ( Sigma-Aldrich , St . Louis , MO ) in the internal solutions was intentionally reduced . Chemical reagents used for blockage experiments ( isradipine , Sigma-Aldrich ) and drug-inducible experiments ( rapamycin , Fisher Scientific , Waltham , MA ) were dissolved in DMSO as 10 mM or 1 mM stock solution , stored at −20°C , and then diluted to 10 μM or 1 μM using extracellular Ca2+ solution right before experiments . Whole-cell currents were generated from a family of step depolarization ( −70 to +50 mV from a holding potential of −70 mV ) or a series of repeated step depolarization ( −10 mV from a holding potential of −70 mV ) . Currents were recorded at 2 kHz low-pass filtering of the instrument . Traces were acquired at a minimum repetition interval of 30 s . P/8 leak subtraction was used throughout . FRET 2-hybrid imaging experiments were performed with an inverted microscope ( Ti-U , Nikon , Japan ) with Neo sCMOS camera ( Andor Technology , UK ) . The light source was from the mercury lamp filtered at the appropriate wavelengths for CFP and YFP by the optical filters mounted at the computer-controlled filter wheel ( Sutter Instrument ) for excitation , subsequently passing the dichroic mirror and the emission filters . Operations and measurements were controlled by the iQ software ( Andor Technology ) . FRET data were acquired and analyzed by an intensity-based two-hybrid assay ( 33-FRET ) as described ( Butz et al . , 2016; Erickson et al . , 2001; Liu et al . , 2010 ) , based onFR=1+FRmax−11+KdDfree where FRmaxrepresents the maximum FRET ratio FRpertaining to the receptor ( YFP ) , and Dfreedenotes the equivalent free donor ( CFP-tagged ) concentration . By fitting the curve of FR−Dfreewith a set of customized Matlab ( Mathworks , Natick , MA ) codes to iteratively estimate FRmaxand Dfree , effective dissociation equilibrium constant ( Kd ) can be achieved for each binding pair to evaluate the ( relative ) affinity . During imaging , the bath solution was Tyrode’s buffer , containing ( in mM ) : NaCl , 129; KCl , 5; CaCl2 , 2; MgCl2 , 1; HEPES , 25; glucose , 30 , 300 mOsm , adjusted with glucose and pH 7 . 3 , adjusted with NaOH . Ionomycin ( Sigma-Aldrich ) was dissolved in DMSO as 1 mM stock solution , stored at −20°C , and diluted to 1 μM using Tyrode’s buffer immediately before applications . Fluorescence images were achieved in HEK293 cells transfected with membrane-localized CFP-tagged FRB and cytosolic FKBP-PCRD/DCRD tagged with YFP . Dynamic translocations were observed with a ZEISS ( Germany ) Laser Scanning Confocal Microscope ( LSM710 ) through a 100X oil objective and analyzed with ZEN 2009 Light Edition software and Adobe Photoshop CS5 software ( Adobe Systems , San Jose , CA ) . Simulations were performed with NEURON ( Carnevale and Hines , 2006 ) , version 7 . 1 . In light of the evidence that both α1DS and α1DL could express in SNc neurons , we simulated CaV1 . 3 of two extreme settings with distinct inactivation kinetics: one is lack of CDI ( ultrastrong CMI ) and the other exhibits pronounced CDI representing α1DS ( without CMI ) . CMI effects including the endogenous CMI as in α1DL could then be simulated by adjusting the relative weights of the above two states , resulting into varying ( intermediate ) levels of SCa and JCa but with constant I300 throughout . Subsequently , we implemented this new CaV1 . 3 model and substituted the original L-type Ca2+ current mechanism in a published Neuron model of SNc ( Chan et al . , 2007 ) , where amendments were incorporated to appropriately reproduce oscillation and pacemaking ( Guzman et al . , 2009; Putzier et al . , 2009 ) . To simulate CCB effects , ICa and its maximum conductance were decreased by ~30% . Data were analyzed in Matlab and Origin ( Origin Software , San Clemente , CA ) . The standard error of the mean ( S . E . M . ) and Student’s t-test ( unpaired; two-tailed with criteria of significance: *p<0 . 05; **p<0 . 01; ***p<0 . 001 ) were calculated when applicable . | All cells need calcium ions to stay healthy , but having too many calcium ions can interfere with important processes in the cell and cause severe problems . Proteins known as calcium channels on the cell surface allow calcium ions to flow into the cell from the surrounding environment . Cells carefully control the opening and closing of these channels to prevent too many calcium ions entering the cell at once . CaV1 . 3 channels are a type of calcium channel that are important for the heart and brain to work properly . Defects in CaV1 . 3 channels can lead to irregular heart rhythms and neurodegenerative diseases such as Parkinson’s disease . Studies have shown that part of the CaV1 . 3 channel that sits inside the cell – known as the “tail” – responds to increases in the levels of calcium ions inside the cell by closing the channel . The tail region of CaV1 . 3 contains three modules , but how these modules work together to regulate channel activity is not clear . Liu , Yang et al . investigated whether the three modules need to be physically connected to each other in the channel protein . For the experiments , several versions of the protein were constructed with different combinations of tail modules being directly linked as part of the same molecule or present as separate molecules . When any two modules were directly linked , the third module could bind to them and this was enough to close the CaV1 . 3 channel . However , the channel did not close if the modules were totally isolated from each other as three separate molecules . Certain types of neurons in the brain produce electrical signals in a rhythmic fashion that depends on CaV1 . 3 channels . In Parkinson’s disease , increased movement of calcium ions into these neurons via CaV1 . 3 channels interferes with the rhythms of the signals and can cause these cells to die . Liu , Yang et al . performed computer simulations to analyse the effects of closing CaV1 . 3 channels in these neurons . The results suggest that this can restore normal rhythms of electrical activity and prevent these cells from dying . The next step is to understand the molecular details of how the tail region closes CaV1 . 3 channels and its role in healthy and diseased cells . This may lead to new ways to block CaV1 . 3 channels in different types of diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"neuroscience"
] | 2017 | Cooperative and acute inhibition by multiple C-terminal motifs of L-type Ca2+ channels |
Anthracycline-induced cardiotoxicity ( ACT ) is a key limiting factor in setting optimal chemotherapy regimes , with almost half of patients expected to develop congestive heart failure given high doses . However , the genetic basis of sensitivity to anthracyclines remains unclear . We created a panel of iPSC-derived cardiomyocytes from 45 individuals and performed RNA-seq after 24 hr exposure to varying doxorubicin dosages . The transcriptomic response is substantial: the majority of genes are differentially expressed and over 6000 genes show evidence of differential splicing , the later driven by reduced splicing fidelity in the presence of doxorubicin . We show that inter-individual variation in transcriptional response is predictive of in vitro cell damage , which in turn is associated with in vivo ACT risk . We detect 447 response-expression quantitative trait loci ( QTLs ) and 42 response-splicing QTLs , which are enriched in lower ACT GWAS p-values , supporting the in vivo relevance of our map of genetic regulation of cellular response to anthracyclines .
Anthracyclines , including the prototypical doxorubicin , continue to be used as chemotherapeutic agents treating a wide range of cancers , particularly leukemia , lymphoma , multiple myeloma , breast cancer , and sarcoma . A well-known side-effect of doxorubicin treatment is anthracycline-induced cardiotoxicity ( ACT ) . For some patients ACT manifests as an asymptomatic reduction in cardiac function , as measured by left ventricular ejection fraction ( LVEF ) , but in more extreme cases ACT can lead to congestive heart failure ( CHF ) . The risk of CHF is dosage-dependent: an early study ( Von Hoff et al . , 1979 ) estimated 3% of patients at 400 mg/m2 , 7% of patients at 550 mg/m2 , and 18% of patients at 700 mg/m2 develop CHF , where a more recent study puts these numbers at 5% , 26% and 48% respectively ( Swain et al . , 2003 ) . Reduced LVEF shows a similar dosage-dependent pattern , but is not fully predictive of CHF . Perhaps most daunting for patients is that CHF can occur years after treatment: out of 1807 cancer survivors followed for 7 years in a recent survey a third died of heart diseases compared to 51% of cancer recurrence ( Vejpongsa and Yeh , 2014 ) . Various candidate gene studies have attempted to find genetic determinants of ACT , but are plagued by small sample sizes and unclear endpoint definitions , resulting in limited replication between studies . Two ACT genome-wide association studies ( GWAS ) have been published ( Aminkeng et al . , 2015; Schneider et al . , 2017 ) . While neither found genome-wide significant associations using their discovery cohorts , both found one variant that they were able to replicate in independent cohorts . A nonsynonymous coding variant , rs2229774 , in RARG ( retinoic acid receptor γ ) was found to be associated with pediatric ACT using a Canadian European discovery cohort of 280 patients ( Aminkeng et al . , 2015 ) , and replicated in both a European ( p=0 . 004 ) and non-European cohort ( p=1×10−4 ) . Modest signal ( p=0 . 076 ) supporting rs2229774’s association with ACT was also reported in a recent study primarily focused on trastuzumab-related cardiotoxicity ( Serie et al . , 2017 ) . RARG negative cell lines have reduced retinoic acid response element ( RAREs ) activity and reduced suppression of Top2b ( Aminkeng et al . , 2015 ) , which has been proposed as a mediator of ACT . In a different study , a GWAS in 845 patients with European-ancestry from a large adjuvant breast cancer clinical trial , 51 of whom developed CHF , found no variants at genome-wide significance levels ( Schneider et al . , 2017 ) . However , one of the most promising variants , rs28714259 ( p=9×10−6 in discovery cohort ) , was genotyped in two further cohorts and showed modest replication ( p=0 . 04 , 0 . 018 ) . rs28714259 falls in a glucocorticoid receptor protein binding peak , which may play a role in cardiac development . An exciting approach to studying complex phenotypes , including disease , in human is to use induced pluripotent stem cells ( iPSC ) and derived differentiated cells as in vitro model systems . Work by us and others has demonstrated that iPSCs and iPSC-derived cell-types are powerful model systems for understanding cell-type specific genetic regulation of transcription ( Thomas et al . , 2015; Burrows et al . , 2016; Banovich et al . , 2018; Kilpinen et al . , 2017; Alasoo et al . , 2017 ) , but it is less established whether these systems can be used to model the interplay of genetic and environmental factors in disease progression . Encouragingly , the response of iPSC-derived cardiomyocytes ( ICs ) to doxorubicin was recently extensively characterized ( Burridge et al . , 2016 ) . ICs derived from four individuals who developed ACT after doxorubicin treatment ( ‘DOXTOX’ group ) and four who did not ( ‘DOX’ group ) , showed clear differences in viability ( via apoptosis ) , metabolism , DNA damage , oxidative stress and mitochondrial function when exposed to doxorubicin . These observations suggest that ICs recapitulate in vivo inter-individual differences in doxorubicin sensitivity . Gene expression response differences between the DOX and DOXTOX groups were found using RNA-sequencing data , but the sample size was insufficient ( RNA-seq was generated for only three individuals in each group ) to attempt mapping of genetic variants that might explain the observed functional differences between individuals . Here we used a panel of iPSC-derived cardiomyocytes from 45 individuals , exposed to five different drug concentrations , to map the genetic basis of inter-individual differences in doxorubicin-sensitivity . We find hundreds of genetics variants that modulate the transcriptomic response , including 42 that act on alternative splicing . We show that the IC transcriptomic response predicts cardiac troponin levels in culture ( indicative of cell lysis ) in these cell-lines , and that troponin level is itself predictive of ACT . Finally we demonstrate that the mapped genetic variants show significant enrichment in lower ACT GWAS p-values .
We generated iPSC-derived cardiomyocytes ( ICs ) for 45 Hutterite individuals ( Figure 1a ) . iPSC quality was confirmed using qPCR ( Figure 1—figure supplement 1 ) , global gene expression profiling ( Figure 1—figure supplement 2 ) , the embryoid body test ( Supplementary Data ) , and EBV integration analysis ( Figure 1—figure supplement 3 , Figure 1—figure supplement 4 ) . Cardiomyocyte identity was confirmed by FACS for cardiac troponin I and T , with mean purity ( 72 ±12 ) % ( Figure 1—figure supplement 5 ) . We exposed all 45 IC lines to doxorubicin at five different concentrations for 24 hr , after which samples were processed for RNA-sequencing . We obtained sufficient read depth ( 10M exonic reads ) for downstream analysis for 217 of the 5×45=225 individual-concentration pairs , and confirmed sample identity by calling exonic SNPs ( see Methods ) . We observed a strong gene regulatory response to doxorubicin across all concentrations , with 98% ( 12038/12317 ) of quantifiable genes ( 5% FDR ) showing differential expression across the different treatment concentrations . Our data shows excellent concordance with the smaller RNA-seq dataset of ( Burridge et al . , 2016 ) ( Figure 1—figure supplement 6 ) . Principal component analysis ( PCA , Figure 1b ) confirms that the main variation in the data is driven by doxorubicin concentration and that the effect of concentration on gene expression is nonlinear . For some individuals the expression data following doxorubicin treatment with 1 . 25μM is closer to the data from treatment with 0 . 625μM , whereas for others it is closer to data from treatment with 2 . 5μM . This general pattern provides the first indication in our data that that there is systematic variation in how different individuals respond to doxorubicin exposure . Since the majority of genes appear responsive to doxorubicin we clustered genes into six distinct response patterns using a mixture model approach ( Figure 1c , see Materials and methods ) . From largest to smallest , these clusters represent genes that , through the gradient from low to high concentration treatments , are ( 1 ) down regulated ( 2 ) initially up-regulated , then further down-regulated ( 3 ) up-regulated ( 4 ) down-regulated only at lower dosages ( 5 ) up-regulated only at lower dosages ( 6 ) down-regulated then partially recover ( Supplementary file 1 ) . Gene set enrichments ( Figure 1—figure supplement 7 , Supplementary file 2 ) for the up-regulated cluster include metabolic , mitochrondrial and extracellular processes , as well as known doxorubicin response genes in breast cancer cell lines from ( Graessmann et al . , 2007 ) ( 647 overlapping genes of 1090 in term , hypergeometric p=2×10−26 ) . The down-regulated cluster shares genes with those down-regulated in response to UV light , which , like doxorubicin , causes DNA-damage ( 413 overlapping genes of 470 in term , hypergeometric p=3×10−48 ) . Targets of p53 , a transcription factor that responds to DNA damage , are overrepresented in clusters 2 and 5; these clusters involve up-regulation at low concentrations ( 0 . 625μM ) but down-regulation at higher concentrations ( 486 overlapping genes out of 1057 in term , hypergeometric p=2×10−39 ) . Promoter analysis ( Figure 1—figure supplement 8 , Supplementary file 3 ) revealed 21 , 45 , and 6 significantly enriched transcription factor ( TF ) binding motifs for clusters 1 , 2 and 3 respectively ( and none for cluster 4–6 ) . Examples include binding sites for ZNF143 , a TF that promotes GPX1 activity and protects cells from oxidative damage during mitochondrial respiratory dysfunction ( Lu et al . , 2012 ) , which is enriched in cluster 1 ( down regulation w/dox , 318 overlapping genes out of 3555 ZNF143 targets , hypergeometric p=10−8 ) ; RONIN , a regulator of mitochrondrial development and function ( Poché et al . , 2016 ) , which is enriched in clusters 1 and 2 ( 217 and 210 overlapping genes out of 2295 targets , p=10−7 and 10−4 respectively ) ; and MEF2 , myocyte enhancer factor 2 , involved in regulating muscle development , stress-response and p38-mediated apoptosis ( Zarubin and Han , 2005 ) , enriched in cluster 4 ( 32 overlapping genes out of 741 targets , hypergeometric p=10−3 ) . We next sought to map single nucleotide polymorphisms ( SNPs ) that modulate the observed inter-individual transcriptomic response to doxorubicin , leveraging available genetic variation across the 45 individuals ( Livne et al . , 2015 ) . We developed a linear mixed model approach , called suez , that extends the PANAMA framework ( Fusi et al . , 2012 ) to account for relatedness amongst individuals , repeat measurements , multiple conditions and latent confounding . Testing SNPs within 1 Mb of the transcription start site ( TSS ) , 518 genes have a variant with a detectable marginal effect on expression ( 5% FDR , Supplementary file 4 ) . Using a mutual information approach ( see Methods ) which , unlike a naive replication analysis ( Figure 2—figure supplement 1 ) , controls for differential power across GTEx tissues , we find our expression quantitative trait loci ( eQTLs ) show stronger overlap with the two heart tissues than any other GTEx tissue ( Figure 2a , Figure 2—figure supplement 2 ) . Remarkably , even with our moderate number of individuals , we are able to detect many response-eQTLs ( reQTLs ) , i . e . variants that modulate ( directly or indirectly ) transcriptomic response to doxorubicin . We found reQTLs for 376 genes at a nominal 5% FDR ( Supplementary file 5 ) , which we estimate using a parametric bootstrap corresponds to a true FDR of 8 . 5% ( Figure 2b ) . We explored leveraging allele specific expression ( ASE ) extending our previous work ( Knowles et al . , 2017; van de Geijn et al . , 2015 ) . We fit a beta-binomial generalized linear model ( GLM ) where the response variable corresponds to alternative vs reference read counts and the independent variable is the heterozygosity of the test regulatory eSNP . We found it impractical to directly relate effect sizes in the total expression and ASE models so we instead combined likelihood ratios from the beta-binomial GLM and suez likelihood into a single test statistic . This approach yielded 447 reQTLs at 5% FDR ( Supplementary file 6 ) , an increase of 19% over using total expression alone . We hypothesize that this relatively modest increase in power is due to a ) suez already being reasonably well powered in this direct perturbation setting and b ) the somewhat low sequencing depth of our samples . To characterize the detected reQTLs we assigned the response of the major and minor allele to one of the six clusters previously learned ( Figure 1c ) , with heterozygotes expected to display the average of the two homozygous responses . 172 ( 46% ) of reQTLs result in a qualitatively distinct response as determined by the two alleles being assigned to different clusters . The most common transition , occurring for 33 reQTLs , is that the major allele is associated with simple down-regulation ( cluster 1 ) in response to doxorubicin , whereas the minor allele shows up-regulation at low concentration followed by down-regulation at higher concentration ( cluster 2 ) . We further broke-down the significant reQTLs by considering the effect of genotype on expression at each concentration ( βc in Equation 6 ) . We normalized the effect sizes relative to the βc with the largest absolute value , i . e . we consider βc/βargmax|βc′| , so that the largest genotype effect always corresponds to a normalized value of 1 . The resulting normalized effect profiles were split into nine clusters using k-means clustering ( Figure 2d ) . The largest cluster ( cluster 1 , 85 reQTLs ) represents reQTLs with a modest effect size at low concentrations ( 0 , 0 . 625μM ) which is amplified at higher concentrations ( Figure 2e shows a highly significant example ) . Cluster two corresponds to reQTLs whose effect size is attenuated at the 0 . 625μM treatment: examples of reQTLs in this cluster tend to be associated with higher expression level at the 0 . 625μM treatment ( e . g . rs16853200’s association with ABCA12 response , Figure 2—figure supplement 3 ) . A non-synonymous coding variant in RARG , rs2229774 , was previously associated with ACT ( Aminkeng et al . , 2015 ) . Since RARG codes for a transcription factor we searched transcriptome-wide for rs2229774 trans-eQTLs: genes where the expression response to doxorubicin appears to be different for different RARG alleles . Only two of the individuals in our panel carry the alternative A allele ( as heterozygotes ) with the rest being homozygous reference ( GG ) . While this limits statistical power , suez detects one marginal effect ( RECQL ) and five response trans-eQTLs ( NMRK1 , VMA21 , PAQR3 , SGIP1 and LRRC2 ) at 5% FDR ( Figure 2—figure supplement 4 ) . Interestingly PAQR3 , a membrane protein localized to the Golgi apparatus , is a negative regulator of antioxidant response through the Nrf2-Keap1 pathway ( Zhang et al . , 2016 ) . LRRC2 is a mitochondrial protein whose RNA expression level has been previous linked with heart failure ( McDermott-Roe et al . , 2017 ) . Oxidative stress , a major downstream consequence of doxorubicin exposure , disrupts splicing of individual genes including HPRT , POLB ( Disher and Skandalis , 2007 ) , and SMA ( Seo et al . , 2016 ) . We queried the extent to which doxorubicin exposure disrupts splicing patterns across the transcriptome using LeafCutter ( Li et al . , 2017 ) . Across all samples LeafCutter detected 27769 alternative splicing ‘clusters’ ( referred to here as ‘ASCs’ to avoid confusion with k-means clusters ) , which correspond approximately to splicing events , with a median of 3 . 0 splice junctions per ASC . Of 17755 ASCs with sufficient coverage to test , 10430 ( 59% ) , corresponding to 6398 unique genes , showed an effect of doxorubicin exposure on splicing outcomes ( 5% FDR , Supplementary files 7–8 ) . To characterize these changes we calculated the entropy of the splicing choices made for each significant ASC at each concentration and used k-means clusters patterns of change in entropy ( Figure 3a ) . The largest cluster has 6166 ASCs ( 59% ) , and corresponds to the null of no clear change in entropy across concentrations . Clusters 2 ( n=1136 ) and 5 ( n=475 ) correspond to increasing entropy with concentration , and clusters 3 , 4 , 6 , 8 and 9 correspond to the maximum entropy being at different concentrations and reaching different maximum levels . Interestingly , only the relatively small cluster 8 ( n=304 , 3% of ACSs ) corresponds to a reduction in entropy at higher concentrations , suggesting the dominant behavior is reduced splicing fidelity and increased alternative splicing in response to doxorubicin . We further tested the hypothesis that splicing fidelity decreases in the presence of doxorubicin by comparing patterns of intronic percent excised ( Ψ ) with canonical vs cryptic ( unannotated ) splice site usage . We clustered the 7792 introns in significantly differentially spliced ASC , that have a change in percent excised ΔΨ>0 . 1 for some pair of concentrations , into eight response patterns based on their relative excision proportions across concentrations . For each cluster we calculated the proportion of member introns with neither end annotated , one end unannotated , or both ends annotated ( Figure 3b ) . The clusters representing increased Ψ with concentration ( clusters 2 , 4 , 6 and 7 ) all show enrichment for cryptic splice site usage . The two most populous clusters ( 1 and 2 ) correspond to Ψ decreasing and increasingly continuously with doxorubicin concentration , respectively , and the difference in levels of cryptic splicing is extremely apparent ( hypergeometric p<2×10−16 , odds ratio for one annotated end vs two is 28 . 0 ) . We additionally used LeafCutter quantification of percentage spliced in ( PSI ) for each splice junction to map splicing QTLs ( sQTL ) and response-splicing QTLs ( rsQTL ) using suez . We tested SNPs within 100 kb of either end of the splice junction . At 5% FDR we found 467 ASCs with a marginal effect sQTL ( Supplementary file 9 ) and 42 with a rsQTL ( Supplementary file 10 ) . An example rsQTL is rs72922482’s association with inclusion of exon 2 of APAF1 Interacting Protein ( APIP ) . Under the major T allele exon skipping is extremely rare: the LeafCutter PSI for the spanning junction ranges from 0 . 00059 to 0 . 0049 across concentrations ( Figure 3c ) . In rs72922482 heterozygotes , however , the exon is skipped in a significant proportion of transcripts , and this effect is most pronounced in the data collected after treatment at 1 . 25μM , with approximately 50% exon inclusion , suggesting the minor C allele results in very low inclusion of the cassette exon . Another interesting example is NDUFAF6 , another mitochrondrial Complex I protein , where doxorubicin exposure ( particularly at 0 . 625μM ) results in increased use of an alternative downstream transcription start site ( TSS ) which unmasks the influence of rs896853 on a cassette exon between the two alternative TSS ( Figure 3—figure supplement 1 ) . We used the level of cardiac troponin released into the culture media by lysed cardiomyocytes ( see Methods , Supplementary file 11 ) to estimate damage occurring as a result of doxorubicin exposure at different concentrations . We observed significant variation in measurable damage caused by doxorubicin across individuals , with 13 of 45 cell lines having a significant correlation between doxorubicin dose and troponin measurement ( Figure 4a ) . We first sought to determine whether the inter-individual variation in troponin in culture could be explained by variation in the overall gene expression response . Since we are interested in this case in inter-individual differences rather than differences between concentrations we normalized the troponin measurements to have 0 mean and variance of 1 across samples at each doxorubicin treatment . We found 96 . 1% ( 95% credible interval 91 . 5%−98 . 6% ) or 91 . 5% of the variance in this normalized troponin level could be explained using gene expression levels ( we excluded the troponin genes TNNT1-3 and TNNI1-3 from the analysis ) at the corresponding doxorubicin concentrations , using a GREML-analysis ( Yang et al . , 2010 ) or leave-out-one cross validated ( LOOCV ) lasso ( Tibshirani , 1996 ) respectively . The optimal lasso model included 118 genes ( Supplementary file 12 ) . To test whether gene expression mediates a link from genotype to troponin level we performed a transcriptome-wide association study ( TWAS , Gamazon et al . , 2015 ) . For each gene we built an elastic-net predictor of expression at each doxorubicin concentration using SNPs within 100 kb , with 10-fold cross-validation to choose the regularization parameters . The fitted predictions ( the ‘pre-validation’ values ) represent the genetically-determined component of expression . We used the 3840 genes with a statistically significant genetic component ( at 1% FDR ) to predict troponin level using LOOCV lasso regression . 89% of the variance in normalized troponin level can be explained by the genetic component of 102 genes ( Supplementary file 13 ) . This analysis is analogous to two-stage least squares Mendelian randomization ( Angrist and Imbens , 1995 ) analysis and therefore suggests the existence of a causal link from genotype through gene expression to troponin level , and highlights potential mediating genes . However , further assumptions — in particular that the SNPs and troponin level are independent conditional on gene expression — would be required to formally establish a causal connection . To further explore the relationship between transcriptomic response and troponin presence in culture , we analyzed differential expression ( DE ) with respect to troponin measurement at each doxorubicin concentration separately . We found 0 , 7 , 78 , 2984 and 2863 differentially expressed genes ( 5% FDR , Supplementary file 14 ) at the five concentrations respectively ( Figure 4b ) . The most strongly DE gene ( with respect to effect size ) at the 5μM treatment is DUSP13 , a known regulator of ASK1-mediated apoptosis ( Park et al . , 2010 ) . The large number of DE genes at the 2 . 5μM and 5 . 0μM treatments are broadly shared ( nominal replication rate 82 to 85% ) , and DE genes at the 1 . 25μM treatment generally represent the most strongly DE genes at the higher concentrations ( Figure 4c ) . To compare troponin measurements to transcriptomic response we determined an overall per-individual level of transcriptomic response with respect to doxorubicin concentration . To this end we fit a principal curve ( Hastie and Stuetzle , 1989 ) through all gene expression samples , initializing the curve to pass sequentially through the successive doxorubicin concentrations ( Figure 4d ) . Projecting every sample on the principal curve gives a single measure of ‘progression’ through response to doxorubicin at increasing concentrations . We then regressed these values against concentration for each individual to obtain a progression rate . We found the troponin measurement slope is significantly negatively correlated ( Spearman ρ=−0 . 42 , p=0 . 004 , Figure 4e ) with the transcriptomic response rate , suggesting that much of the gene expression program being activated in response to doxorubicin is in fact protective against cardiac damage . Using previously published data ( Burridge et al . , 2016 ) , we built a predictive model of ACT risk trained on RNA-seq of ICs exposed to 1μM doxorubicin from doxorubicin-treated patients who did ( ‘DOXTOX’ , n=3 ) or did not ( ‘DOX’ , n=3 ) develop ACT . Using lasso with fixed λ=10−5 the optimal model included 17 genes as features ( Supplementary file 15 ) . We applied this model to our expression data from the 0 . 625μM treatment ( since this concentration shows excellent concordance with the 1μM data of Burridge et al . , see Figure 1—figure supplement 6 ) to obtain predicted log-odds of ACT . While these log-odds are unlikely to be well-calibrated due to differences in the training and test datasets , they may still accurately represent relative risk of ACT across our 45 individuals . Indeed , the log-odds correlated significantly with the troponin measurement slope ( Spearman correlation p=0 . 01 , Figure 4f ) , suggesting our troponin measurements , and by extension our expression response data , recapitulate in vivo cellular response to doxorubicin . To determine the disease-relevance of our molecular QTLs we obtained summary statistics for the largest ACT GWAS to date ( Schneider et al . , 2017 ) . While this GWAS was not sufficiently powered to find genome-wide significant associations , 11 variants representing nine independent loci have p<10−5 , with the most significant ( rs2184559 ) at p=2 . 8×10−6 . Of the 8 GWAS variants with p<10−5 either tested in our eQTL mapping , or in high LD ( R2>0 . 8 ) with a tested SNP , seven have a nominally significant marginal eQTL ( p<0 . 05 , the 8th has p=0 . 07 ) and four have a reQTL with p<0 . 1 . The one replicated variant in this GWAS , rs28714259 , was not genotyped in our data but is in high LD ( R2=0 . 98 ) with rs11855704 which is a nominally significant marginal eQTL for tubulin gamma complex associated protein 5 ( TUBGCP5 , Figure 3—figure supplement 2 ) . rs4058287 ( GWAS p-value 9 . 68×10−6 ) has a marginal effect on Alpha-Protein Kinase 2 ( ALPK2 , also known as ‘Heart Alpha-Protein Kinase’ since it was discovered in mouse heart ( Ryazanov et al . , 1999 ) and is expressed in few other tissues ( Melé et al . , 2015 ) ) expression ( p=0 . 0016 ) as well as a weak interaction effect ( p=0 . 06 , see Figure 5a ) . Interestingly , ALPK2 has been shown to upregulate DNA repair genes and to enable caspase-3 cleavage and apoptosis in a colorectal cancer model ( Yoshida et al . , 2012 ) . The replicating variant from Aminkeng et al . , 2015 , rs2229774 only occurs in two individuals in our cohort ( who are heterozygous ) making eQTL mapping infeasible . Additionally we find a marginal effect eQTL ( p=0 . 0017 , Figure 5—figure supplement 1 ) on SLC28A3 for rs885004 , which has previously been associated with ACT in a candidate gene study ( Visscher et al . , 2013 ) . rs885004 is intronic , is in LD ( R2=0 . 98 ) with another ACT implicated variant , rs7853758 ( Visscher et al . , 2012 ) , and falls in a DNase I hypersensitivity and H3K27ac peak present in numerous ENCODE cell lines ( and is open in our ICs according to ATAC-seq data , see Figure 5—figure supplement 2 ) . To determine whether our molecular QTLs are more useful than published QTLs for interpreting ACT risk variants we first sought to obtain the best powered GWAS data possible . Since the Schneider et al . GWAS was overall underpowered , we obtained additional ACT GWAS summary statistics from a more recent study ( Serie et al . , 2017 ) and performed a meta-analysis with Schneider et al . We used this data to assess whether there was detectable enrichment of low GWAS p-values for our regulatory QTLs . When considering eQTL with nominal p<10−5 ( corresponding approximately to 5% FDR ) we found no enrichment for GWAS p<0 . 05 for three GTEx tissues ( heart , brain and lymphoblastoid cell lines—LCLs ) , our marginal effect eQTLs or baseline ( no doxorubicin ) only eQTLs ( Figure 5—figure supplement 3 ) . However , considering SNPs that are either main effect or response eQTL we see significant enrichment ( one-sided hypergeometric p=6 × 10−20 , OR = 1 . 40 ) . Similarly for ‘combined’ eQTL where we explicitly test for any effect of genotype ( main or interaction effect , see Methods ) we see enrichment ( p=5 × 10−12 , OR = 1 . 29 ) . Furthermore , focusing on response eQTL we see a stronger enrichment ( p=2 × 10−37 , OR = 1 . 95 , Figure 5b ) , suggesting that the enrichment in combined eQTL is driven by this signal . Response eQTLs mapped using allelic-specific expression as well as total expression show the strongest enrichment ( p=6 × 10−60 , OR = 2 . 22 ) . When considering splicing QTLs ( Figure 5—figure supplement 4 ) we found no enrichment for marginal sQTLs mapped in LCLs ( Li et al . , 2016 ) . Interestingly , in contrast to the total expression QTLs we found a significant enrichment ( p=4 × 10−17 , OR = 1 . 36 , Figure 5c ) for IC marginal sQTLs although the enrichment in response sQTLs was still higher in absolute terms ( p=2 × 10−5 , OR = 1 . 57 ) . These findings are indicative that molecular response QTL mapping has potential for understanding the molecular basis of environmentally-dependent human disease . Finally , we attempted colocalization analysis for our response eQTLs and meta-analyzed GWAS using coloc ( Giambartolomei et al . , 2014 ) , a Bayesian method based on summary statistics . For each region ( gene in our case ) coloc infers a posterior probability for each of five possibilities: ( H0 ) no association , ( H1 ) association for the eQTL only , ( H2 ) association for the GWAS only , ( H3 ) independent variants , or ( H4 ) colocalization to one variant . Out of 43 genes with a SNP with a reQTL at p<10−5 and GWAS SNP at p<0 . 05 coloc gave maximum posterior probability to the null hypothesis ( H0 , no association ) for 32 genes , association only for reQTL ( H1 ) for 9 , and colocalization ( H4 ) for one gene , NOL10 ( posterior probability of colocalization 0:54 , Figure 5—figure supplement 5 , Supplementary file 16 ) . While these results suggest our data is not sufficiently well-powered for colocalization analysis , we note that the posterior probability of colocalization ( H4 ) is higher than that for independent signal ( H3 ) in 40/43 tested genes .
Human iPSC-derived somatic cells provide a powerful , renewable and reproducible tool for modeling cellular responses to external perturbation in vitro , especially for non-blood cell-types such as cardiomyocytes which are extremely challenging to collect and even then are typically only available post-mortem . We established a sufficiently large iPSC panel to effectively query the transcriptomic response of differentiated cardiomyocytes to doxorubicin . We were also able to characterize the role of genetic variation in modulating this response , both in terms of total expression and alternative splicing . There are , of course , caveats associated with using an in vitro system , which may not accurately represent certain aspects cardiac response to doxorubicin in vivo . That said , the replication of GTEx heart eQTLs , association of troponin levels with predicted ACT-risk ( Burridge et al . , 2016 ) , and the observed GWAS enrichment , all support the notion that the IC system recapitulates substantial elements of in vivo biology . It is challenging to quantify this agreement , and there are in vivo factors that are certainly not represented . For example , excessive fibrosis may contribute to ACT ( Cascales et al . , 2013; Zhan et al . , 2016; Farhad et al . , 2016; Heck et al . , 2017 ) , although is unclear how substantial this contribution is as well as whether fibroblasts are directly activated by doxorubicin exposure or simply respond indirectly to cardiomyocyte damage . While our FACS analysis shows cardiomyocytes are the dominant cell type in our cultures , heterogeneity remains and other cell types could be mediating some of the observed changes . For many diseases such as ACT which involve an environmental perturbation it is reasonable to suppose that eQTLs detected at steady-state are only tangentially relevant when attempting to interpret disease variants . Such concerns motivated us to focus on response eQTLs , that is , variants that that have functional consequences under specific cellular conditions because they interact , directly or indirectly , with the treatment . We used a statistical definition of reQTLs corresponding to cases where gene expression levels are significantly better explained using a model including an interaction term between genotype and treatment ( represented as a categorical variable ) , compared to a model with only additive effects for genotype and treatment . Our characterization of the detected reQTL demonstrates that these variants are indeed candidate drivers of differences in individual transcriptomic response to doxorubicin . The strongest reQTL effects correspond to completely different response patterns for the major and minor alleles , while weaker effects correspond to more subtle modulation of the same response pattern . We note that it is not necessarily the case that such reQTLs are the only functionally relevant eQTLs . eSNPs with a marginal ( additive ) effect on expression of a gene responsive to doxorubicin ( as most genes are ) could still be important if the relationship between expression and ACT-risk is nonlinear , for example involving thresholding effects . We observed a statistical enrichment of expression and ( to a lesser extent ) splicing QTLs in ACT GWAS . However , with no reproducible genome-wide significant associations available , fine-mapping of causal variants remains fraught . We anticipate our findings will be increasingly valuable as larger-scale ACT GWAS become available . We derived ICs from healthy individuals so we do not known which individuals would develop ACT if they required anthracycline treatment . Mapping molecular response QTLs in larger panels of ICs from patients treated with anthracyclines who do or do not develop ACT symptoms would allow stronger conclusions to be drawn about the contribution of the detected ( r ) eQTLs to disease etiology . We used a panel of Hutterites individual since this homogeneous population offers unique advantages for mapping genetic traits: exposure to a fairly uniform environment and less variable genetic background , despite still representing much of European diversity ( Newman et al . , 2004 ) . However , the genetic basis of ACT susceptibility is likely complex and some relevant genetic variation may not be well represented in this cohort . Finally , an interesting observation in our study is that splicing fidelity is reduced upon doxorubicin exposure . This is not completely unexpected since a key downstream side-effect of doxorubicin is increased oxidative stress , which has been previously associated with dysregulated splicing of specific genes ( Disher and Skandalis , 2007; Seo et al . , 2016 ) . Our finding that this effect is prevalent across the transcriptome poses further questions about what known effects of doxorubicin might , in fact , be mediated by changes in RNA splicing .
Generation of lymphoblastoid cell lines ( LCLs ) and genome-wide genotyping of many individuals from a multi-generational pedigree were performed previously . Briefly , lymphocytes were isolated from whole blood samples using Ficoll-Paque and immortalized using Epstein Barr Virus ( EBV ) ( Cusanovich et al . , 2012; Cusanovich et al . , 2016 ) . Phased genotypes were obtained by combining pedigree information , genotypes from SNP arrays , and genotypes from whole genome sequencing of related individuals ( Livne et al . , 2015 ) . We reprogrammed 75 LCLs to iPSCs using episomal plasmid vectors , containing OCT3/4 , p53 shRNA , SOX2 , KLF4 , L-MYC , and LIN28 which avoids integrating additional transgenes ( Okita et al . , 2011 ) . Initially , the lines were generated on mouse embryonic fibroblasts ( MEF ) , which coated the well and served as feeder cells to create an environment supportive of pluripotent stem cells . The colony was then mechanically passaged on MEF and tested for expression of pluripotency-associated markers by immunofluorescence staining and RT-PCR . The lines were passaged for at least 10 weeks on MEF to ensure lines had stabilized . All iPSC lines were characterized as described previously ( Gallego Romero et al . , 2015 ) . Briefly , we initially performed qPCR using 1μg of total RNA , converted to cDNA , from all samples to confirm the endogenous expression of pluripotency genes: OCT3/4 , NANOG , and SOX2 ( Figure 1—figure supplement 1 ) . We next confirmed pluripotency using PluriTest ( Müller et al . , 2011 ) . All samples were classified as pluripotent and had a low novelty score ( Figure 1—figure supplement 2 ) . Additionally , we confirmed the ability of all iPSC lines to differentiate into the three main germ layers using the embryoid body ( EB ) assay ( Supplementary Data ) . Finally , we tested for the presence and expression of the EBV gene EBNA-1 using PCR ( Figure 1—figure supplement 3 , Figure 1—figure supplement 4 ) . We tested all samples for both genomic integrations and vector-based EBV . If the cells were positive ( two positive and one indeterminate case was identified ) , we further tested the origin of the EBV ( genomic or episomal ) using primers specific to the LMP-2A gene found in EBV or part of the sequence specific to the episomal plasmid ( Figure 1—figure supplement 3 ) . We concluded that two lines still had EBV present in the genome , this was also reflected in EBNA-1 gene expression for these individuals ( Figure 1—figure supplement 4 ) . We retained these individuals because they passed all quality control metrics and were not outliers based on genome-wide gene expression . It should also be noted that gene expression levels are extremely similar between iPSC lines . This relative homogeneity further demonstrates the quality of our iPSC lines . In summary , all iPSC lines showed expression of pluripotent genes quanti1ed by qPCR , generated EBs for all three germ layers , and were classified as pluripotent based on PluriTest . iPSC lines were transitioned to feeder-free conditions , which was necessary to prime the iPSCs for differentiation . Next we differentiated the iPSCs to cardiomyocytes ( Lian et al . , 2013; Burridge et al . , 2014 ) . iPSC lines were covered with a 1:60 dilution matrigel overlay for 24 hr . On day 0 iPSC lines were treated with 12μM of the GSK3 inhibitor , CHIR99021 , in RPMI+ B27 medium ( RPMI1640 , 2 nM L-glutamine and 1x B27 supplement minus insulin ) for 24 hr at which time media was replaced with fresh RPMI + B27 . 72 hr after the addition of CHIR99021 ( Day 3 ) , 2μM of the Wnt inhibitor Wnt C-59 was added for 48 hr . Fresh RPMI + B27 was added on Days 5 , 7 and 10 . Beating cells appeared between Days 8–10 . These cardiomyocytes consisted of ventricular , atrial and pacemaker-like cells . The cells formed thick layers and contract throughout the well . Metabolic selection was used to purify the cardiomyocytes ( Tohyama et al . , 2013 ) from Day 14 to Day 20 when glucose-free RPMI media supplemented with the components essential for cardiomyocyte differentiation ( Burridge et al . , 2014 ) , ascorbic acid and human serum albumin , together with lactate , a substrate uniquely metabolized by cardiomyocytes , was added to cells . Because this lactate media can only be metabolized by cardiomyocytes , the non-cardiomyocytes in the culture were removed over the 6 day treatment . On day 20 the cardiomyocytes , now at a high cTnT purity , were replated for experiments in media that contains only galactose and fatty acids as an energy source . This galactose media forces the cardiomyocytes to undergo aerobic respiration , rather than anaerobic glycolysis common in cultured cells . At day 20 when ICs were plated for doxorubicin exposure , a portion of cells were collected to assess purity . Cells were harvested from plates by incubating with TrypLE Express ( Thermo Fisher Scientific , Cat . No 12604013 ) for 5 min at 37°C . Once removed , cells were manually dissociated further by passing through a 100μm and then 40μm strainer to create a single cell suspension . Cells were then resuspended , fixed , and permeabilized ( Foxp3/Transcription Factor Staining Buffer Set; eBioscience/Affymetrix , Cat . No 00-5523-00 ) according to the manufacturer’s instructions . Cells were stained with directly conjugated antibodies to cTnI ( Alexa Fluor647 Mouse Anti-Cardiac Troponin I Clone C5; BD Biosciences , Cat . No 564409 ) and cTnT ( PE Mouse Anti-Cardiac Troponin T Clone 13–11; BD Biosciences , Cat . No 564767 ) . The Zombie VioletTM Fixable Viability Kit ( BioLegend , Cat . No 423113 ) was used to assess cell viability at the time of fixation . The following isotype controls were used: Alexa Fluor647 Mouse IgG2b , κ Isotype Control Clone 27–35 ( BD Biosciences Catalog No . 558713 ) and PE Mouse IgG1 , κ Isotype Control Clone MOPC-21 ( BD Biosciences Catalog No . 554680 ) . Cells were analyzed using a FACS Canto or LSR-II flow cytometers ( BD Biosciences ) , and the data were analyzed with FlowJo software ( v10 . 0 . 7 , Tree Star ) . All gates were established such that <2% of cells stained with isotype controls were positive and dead cells were excluded . We incubated the cardiomyocytes in 0 , 0 . 625 , 1 . 25 , 2 . 5 , or 5 μM doxorubicin . After 24 hr , we collected the serum and cells from each condition . From the serum , we measured cardiac Troponin T levels using the ABNOVA Troponin I ( Human ) ELISA kit ( cat . no . KA0233 ) . From the cells , we extracted RNA for sequencing . Cells from each individual were treated separately , but batches of experiments were performed on different days . Each treatment batch contained 1 to 4 individuals . RNA quality was assessed with the Agilent Bioanalyzer . We prepared libraries using the Illumina TruSeq Library Kit and generated 50 bp single-end reads on a HiSeq 4000 at the University of Chicago Functional Genomics Facility . We confirmed sequencing quality using FastQC and MultiQC ( Ewels et al . , 2016 ) . We confirmed sample identity by ( 1 ) comparing allelic counts ( quantified using samtools mpileup [Li et al . , 2009] ) of exonic SNPs to the known genotypes and ( 2 ) running verifyBamID ( Jun et al . , 2012 ) . We aligned RNA-seq reads using STAR version 2 . 5 . 2a ( Dobin et al . , 2013 ) to GRCh38/GENCODE release 24 . We counted reads using feature Counts ( Liao et al . , 2014 ) and calculated counts per million reads ( cpm ) using `cpm` from the `edgeR` ‘R package ( version 3 . 18 . 1 ) ( Robinson et al . , 2010 ) . We discarded samples with <107 exonic reads and genes with median log2 ( cpm ) less than 00 . We performed differential expression ( DE ) analysis across all five doxorubicin concentrations jointly , using either a linear model on quantile normalized cpm value or Spearman correlation , followed by Benjamini-Hochberg False Discovery Rate ( FDR ) control . Since the vast majority of genes showed differential expression we did not investigate better powered DE methods such as DESeq2 . We clustered genes into ‘response patterns’ using a K-component mixture model ( 1 ) π∼Dir ( 1/K , ⋯ , 1/K ) zg|π∼Discrete ( π ) yngc|zg=k , θ∼N ( θck , σ2 ) where π is a prior probability vector over cluster assignments , Dir is the Dirichlet distribution , zg is cluster from which gene g is generated , yngc is the expression of gene g in individual n at concentration c , θck is the mixture parameter ( mean ) across concentrations for cluster k , and σ2 is a shared noise variance . We marginalize ( sum ) over zg and optimize with respect to π , θ , σ using the rstan R package ( Carpenter et al . , 2016 ) ( version 2 . 16 . 2 ) . The hyperparameters of the Dirichlet distribution are set such that in the limit of large K the model approximates a Dirichlet process mixture ( MacEachern and Müller , 1998 ) which automatically learns of an appropriate number of mixture components to use from data . Gene set and promoter motif enrichment were performed using HOMER v4 . 9 . 1 ( Heinz et al . , 2010 ) using default parameters and without de novo motif search . We developed an extension of the PANAMA ( Fusi et al . , 2012 ) linear mixed model ( LMM ) framework to map eQTLs and response eQTLs while accounting for latent confounding , which we call suez . suez entails a two step procedure . Step one is used to learn latent factors from all genes , using the modelyncg=∑kWkgxnck+ung+vcg+ξncg+ϵncgWkg∼N ( 0 , σk2 ) factor loadings/coefficientsung∼N ( 0 , σu2 ) individual random effectsξ∼MVN ( 0 , σξ2Σ ) kinship random effectϵ∼MVN ( 0 , diag ( σϵ2 ) ) noisewhere xnck are latent factors , vcg are per gene , per concentration fixed effects . We integrate over W , u , ξ and ϵ , which results in a per gene multivariate normal , ( 2 ) y:g∼MVN ( Vv:g , ∑k σk2x:kxk:T+σu2U+σξ2Σ+σe2I ) , where y:g refers to the vector of expression for gene g across all individuals and concentrations ( i . e . all ‘samples’ where a sample is an individual-concentration pair ) , V is a matrix mapping concentrations to samples ( i . e . Vsc=1 iff sample s is at concentration c ) and U is a matrix of which samples are for the same individual ( i . e . Uss′=1 if sample s and sample s′ come from the same individual ) . We optimize x , v and the variances {σu2 , σk2 , σξ2 , σϵ2} jointly across all genes g . In step 2 we test individual gene-SNP pairs while accounting for confounding using the covariance matrix ( 3 ) Σπ=∑k σk2x:kxk:T+σu2U+σξ2Σwhich includes both latent confounding , individual random effects and similarity due to kinship . We consider three LMMs , all with the same parameterization of the covariance σπ2Σπ+σe2I where σπ2 and σe2 are optimized along with the fixed effects to allow the extent to which each gene follows the global covariance pattern to be adapted . The simple structure of this covariance also allows pre-computation of the eigen-decomposition of Σπ which enables linear ( rather than cubic ) time evaluation of the likelihood and its gradient . Model 0 involves no effect of the SNP ( and can therefore be fit once for a gene ) and a fixed effect for concentration . Model 1 adds a marginal effect of the SNP genotype dosage d . Finally model 2 adds an interaction effect between concentration and genotype , which is equivalent to a concentration-specific genotype effect . In summary: ( 4 ) Model 0: E[yncg]=vcg ( 5 ) Model 1: E[yncg]=vcg+βdn ( 6 ) Model 2: E[yncg]=vcg+βcdn We optimize σπ2 , σe2 and the regression coefficients for each of the three models separately , and use likelihood ratio tests ( LRT ) to compare the models . Comparing Model 1 vs 0 ( one degree of freedom ) tests whether there is a marginal effect of the variant . Comparing Model 2 vs 1 ( C−1=4 degrees of freedom , where C is the number of conditions/concentrations ) tests whether there is an interaction effect , i . e . whether the genetic effect on expression is different at different concentrations ( or equivalently whether the response to doxorubicin is different for different genotypes ) . Finally Model 2 vs 0 ( C=5 degrees of freedom ) tests whether there is any effect of genotype on expression , either in terms of a marginal or concentration-specific effect ( we refer to these as ‘combined’ eQTL ) . We use the conservative approach of using Bonferroni correction across SNPs for a gene , followed by Benjamini-Hochberg FDR control . We quantile normalize the expression levels across all samples for each gene to a standard normal distribution so that the distributional assumptions of our linear mixed model are reasonable . However , optimizing the variance parameters σπ2 and σe2 means that the χ2 distribution for the LRT will only hold asymptotically and p-values for finite sample sizes will tend to be somewhat anti-conservative . To account for this for response-eQTLs , we use a parametric bootstrap since there is no fully valid permutation strategy for testing interaction effects . This involves first fitting Model 1 and then simulating new expression data under the fitted model . Models 1 and 2 are then ( re ) fit to this data and compared using an LRT . We then perform Bonferroni correction across SNPs for each gene to obtain an empirical null distribution of per gene p-values which we use to estimate the true FDR for our response-eQTL results . For significant reQTLs we assigned the response of the minor allele and major allele to the previously determined clusters using the modelync|zA , za , θ∼N ( 12dnθczA+12 ( 2−dn ) θcza , σ2 ) , where ync is the expression for individual n at concentration c , zA and za are the cluster assignments for the major and minor allele respectively , dn∈{0 , 1 , 2} is the genotype dosage , and θ and σ2 are fixed at the values learned in Equation 1 . For each reQTL separately we calculate the likelihood of y given all possible pairs of assignments ( zA , za ) and choose the maximum likelihood solution . As for all k-means clustering in the paper , we used KMeans_rcpp function of the R package ClusterR v1 . 0 . 6 , taking the best of 10 initializations using the k-means++ option , to cluster the normalized genotype effect profiles of the significant associations . The choice of 9 clusters was determined manually . We ( Knowles et al . , 2017; van de Geijn et al . , 2015 ) and others ( Kumasaka et al . , 2016 ) have demonstrated that modeling allele-specific expression can improve power to detect both cis eQTLs ( van de Geijn et al . , 2015; Kumasaka et al . , 2016 ) and reQTLs ( Knowles et al . , 2017 ) . Here we employ a combination of ideas from these methods: Under the hypothesis that the allelic effect of the test regulatory SNP varies across concentrations ( analogous to Model 2 in Equation 6 ) , we have ( 7 ) ynkc|rnkc , ϕnk∼BB ( rnkc , σ ( μ+ϕnkβc ) , γ ) where there are K exonic SNPs in a gene , with alternative allele counts ynkc and read coverage rnkc . σ is the logistic function . μ is an intercept term to account for reference mapping bias , and γ is a per-gene concentration ( reciprocal of over dispersion ) parameter . We regularize γ using a Γ ( 1 . 001 , 0 . 001 ) prior . ϕnk∈{−1 , 0 , +1} is the phased heterozygosity of the test regulatory SNP in individual n , with ϕnk=0 if the regulatory SNP is homozygous ( these individuals are included to help estimate μ and c ) and ϕnk=1 or −1 if the regulatory SNP is heterozygous and in phase ( 1 ) or opposite phase ( -1 ) with SNP k . βc is the effect size in concentration c . The null ( no interaction effect , corresponding to Model 1 in Equation 6 ) is that βc=β for all all c . To integrate evidence from total and allelic specific expression it is valid to add the log likelihood ratios , which can be seen as either fitting one model with likelihood terms for the two components , or as a result of χ2 random variables being closed under addition . This approach has substantially better power than applying Fisher’s combined probability test to p-values from testing the total and allelic expression components separately . Twice the summed log likelihood ratio is asymptotically χ2 with degrees of freedom being simply the sum of the degrees of freedom of the two components ( so usually 4+4=8 in our case ) . In practice we only fit the total expression model if there are at least 5 alternative alleles observed for the test regulatory SNP , and only fit the allele-specific model if there are at least 2000 supporting allelic reads for the gene , so some regulatory SNPs are only tested using one component or the other . In addition , for some genes whose expression is very low for specific concentrations there may be no allelic reads for a concentration , in which case the degrees of freedom for the allele-specific component will be reduced since no βc is learned for that concentration . We initially compared our eQTLs to GTEx eQTLs by estimating Storey’s π1 ( Storey and Tibshirani , 2003 ) , using the q value R package v2 . 8 . 0 , for GTEx nominal p-values for our significant eQTLs ( at a nominal p<10−5 ) . While the GTEx heart tissues show higher replication than most tissues , surprising tissues ranked higher ( Figure 2—figure supplement 1 ) . We reasoned that differential power across the GTEx tissues due to differing sample size and noise levels confound this simple approach . We therefore used an extension of Storey’s π1 to test for overlap between two sets of p-values . For each GTEx tissue we fit ( using LBFGS in Stan ) a mixture model ( 8 ) ( pi1 , pi2 ) ∼π00U ( pi1 ) U ( pi2 ) +π10B ( pi1|a1 , b1 ) U ( pi2 ) +π01U ( pi1 ) B ( pi2|a2 , b2 ) +π11B ( pi1|a1 , b1 ) B ( pi2|a2 , b2 ) where pij is the p-value for SNP-gene pair i in tissue j , U is the uniform distribution on [0 , 1] ( corresponding to p-values coming from the null ) and B is the beta distribution ( corresponding to non-null p-values ) . π00 , π01 , π10 , π11 correspond to mixture weights are estimates of the proportion of SNP-gene pairs that are ( a ) null for both tissues , ( b ) null for tissue one and non-null for tissue 2 , ( c ) non-null for tissue one and null for tissue 2 , ( d ) non-null for both tissues . Note that π sums to 1 . We constrain the hyperparameters aj∈[0 , 1] and bj≥1 to encode the assumption that non-null SNP-gene pairs should have low p-values . Due to the large number of SNP-gene pairs tested , in practice we bin p-values on a regular 100×100 grid and use the bin counts to weight the likelihood . Finally we estimate the mutual information between the pair of tissues as: ( 9 ) MI=∑k={0 , 1} ∑j={0 , 1} πkjlogπkjπkπj′where πk=∑j={0 , 1} πkj and π′j=∑k={0 , 1} πkj are marginal probabilities . This approach explicitly estimates the proportion of null tests in tissue 1 ( π0 ) and tissue 2 ( π′0 ) as well as the proportion of tests that are non-null in both ( π11 ) . This approach both controls for power in both tissues and negates the need to choose arbitrary significance thresholds . We ran LeafCutter v0 . 2 . 6_dev ( using default settings ) which allows joint differential intron excision testing across more than two conditions . For each Alternative Splicing Cluster ( ASC ) LeafCutter fits a set of PercentSplicedIn probability vectors ψc , across detected splice junctions i , at each concentration c . For ASCs determined to be significantly ( 5% FDR ) differential spliced across concentrations , we calculated the entropy hc=−∑i ψcilogψci at each concentration c . We normalized these profiles as hc~=hc/hc¯ and clustered these profiles , using KMeans_rcpp as above . To investigate the relative usage of cryptic splice sites we first determined the set of 7792 splice junctions that ( a ) fell in ASCs determined to be significantly differentially spliced ( 5% FDR ) and ( b ) had maxcψci−mincψci>0 . 1 . We obtained normalized intron excision rates by subtracting the per intron mean and dividing by the per intron standard deviation . These ψ profiles were clustered using KMeans_rcpp . Cryptic splice site usage was determined by considering all exons in Gencode v26 and ignoring transcript structure ( i . e . a junction spanning two splice sites used but only in different transcripts would still be considered ‘annotated’ ) . For ( response ) splicing QTL we calculated within ASC intron excision ψ with pseudocount of 0 . 5 , and set entries with 0 denominator ( no reads for that ASC in that sample ) to the mean across all other samples . These values were then ( 1 ) z-score normalized across samples and ( 2 ) quantile normalized to a normal across introns . QTL mapping was then performed using suez considering each intron as a ‘gene’ . We assessed the proportion of variance in cardiac troponin explained by gene expression response . Let yci represent the troponin level measured in individual i at doxorubicin concentration c , normalized to have 0 mean and variance 1 across individuals at each concentration . Let xcig be the expression of gene g ( in individual i at concentration c ) , z-score normalized across samples . We consider the linear model ( 10 ) yci=∑g βgxcig+ϵciwhere ϵci∼N ( 0 , σϵ2 ) is noise and the coefficients βg are given a prior N ( 0 , σβ2/G ) where G=12 , 317 is the number of genes in the analysis . Integrating over βg we have ( 11 ) y:∼N ( 0 , σβ21G∑gx:gx:gT+σϵ2I ) We optimize this model wrt σβ and σϵ to obtain an estimate ρ=σβ2/ ( σβ2+σϵ2 ) of the percent variance of y explained by x . A Bayesian credible interval for ρ is obtained under this model using 8000 iterations of Hamiltonian Monte Carlo ( with the first 4000 discarded as burnin ) implemented using RStan ( Carpenter et al . , 2016 ) ( v2 . 16 . 2 ) . For the transcriptome-wide association study for cardiac troponin levels we use the R package glmnet ( v2 . 0–13 ) to build elastic-net predictors of gene expression for each gene , using 10-fold cross-validation to choose λ and α=0 . 5 which we found gave comparable performance to higher values . A single model was learnt per gene jointly across concentrations , including main effects for all SNPs within 100 kb of the gene TSS , main effects for doxorubicin dosage ( encoded categorically ) and interaction terms between each SNP and the dosage factor . The fitted values on the test-folds from the cross-validation are known as the ‘prevalidation’ response . To test which genes have a significant genetic component we tested ( using analysis-of-variance ) whether the observed expression was better predicted under a linear model including the prevalidation values and the dosage variable than by dosage alone . The prevalidated response for the 3840/12317 genes ( 1% FDR ) genes that are predictable from genotype are then used to predict troponin level , normalized as for Equation 10 , using leave-out-one-cross-validated lasso regression . All the custom analysis scripts used for this project are available at https://github . com/davidaknowles/dox ( Knowles and Blischak , 2017; copy archived at https://github . com/elifesciences-publications/dox ) . The suez response eQTL mapping R package is available at https://github . com/davidaknowles/suez ( Knowles , 2017; copy archived at https://github . com/elifesciences-publications/suez ) . The following data are available as Supplementary Data: ( 1 ) differential expression cluster assignments , ( 2 ) significant ( 5% FDR ) eQTLs and sQTLs , ( 3 ) differential splicing results , ( 4 ) levels of cardiac troponin and the predicted transcriptomic response . In addition to the Supplementary Data included with this paper additional results are hosted at http://web . stanford . edu/~dak33/dox/ and Dryad ( doi:10 . 5061/dryad . r5t8d04 ) including ( 1 ) gene-by-sample matrix of RNA-seq quantification ( log counts per million ) , ( 2 ) LeafCutter intron excision quantification ( 3 ) p-values for all tested eQTLs , reQTLs , sQTLs , and rsQTLs , ( 4 ) RARG variant response and marginal trans-eQTLs , ( 5 ) RIN , RNA concentration and other technical covariates , ( 6 ) embryoid body imaging for all iPSC lines . The RNA-seq FASTQ files will be added to the dbGaP database ( Tryka et al . , 2014 ) under dbGaP accession phs000185 ( https://www . ncbi . nlm . nih . gov/projects/gap/cgi-bin/study . cgi ? study_id=phs000185 ) . The genotype data files cannot be shared because releasing genotype data from a subset of individuals in the pedigree would enable the reconstruction of genotypes of other members of the pedigree , which would violate the original protocol approved by the research ethics board ( Livne et al . , 2015 ) . The summary statistics for the ACT GWAS were given to us by the authors of the studies ( Schneider et al . , 2017; Serie et al . , 2017 ) . | Many cancers , including leukaemia , lymphoma and breast cancer , are treated with potent chemotherapy drugs such as anthracyclines . However , anthracyclines have strong side effects known as anthracycline cardiotoxicity , which affect the health of the heart . Almost half of the patients given high doses of anthracyclines develop chronic heart failure . While anthracycline cardiotoxicity is very common , people’s genes may contribute to how sensitive they are to these drugs but it is not understood which genes can cause this effect . Previous studies using only a small number of participants have not been able to pin down the genetic factors that make some patients respond well to anthracyclines , and others prone to developing heart failure when taking these drugs . To find out which genes affect anthracycline cardiotoxicity , Knowles , Burrows et al . transformed blood cells from 45 individuals into stem cells , which were then developed into heart muscle cells . Then , the activity of genes was analyzed by measuring the amount of RNA ( the template molecules used to make proteins ) produced by those genes . After the cells had been exposed for 24 hours to the anthracycline drug doxorubicin , hundreds of gene activity differences could be found in the heart muscle cells between individuals . Some of these differences were linked to poorer health of the cells after treatment with the drug . As a result , a number of genetic variants that could predispose patients to the side effects of doxorubicin were discovered . The experiments also revealed how doxorubicin disrupts an important process that separates ‘junk’ parts of the RNA from the parts that are used as a template for proteins . Being able to predict who is likely to be sensitive to drugs such as doxorubicin could help doctors to tailor chemotherapy treatments more effectively , minimising the risk of heart failure . In future , larger studies could lead to accurate predictions of a patient’s response to a particular chemotherapy drug to personalize their cancer treatment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2018 | Determining the genetic basis of anthracycline-cardiotoxicity by molecular response QTL mapping in induced cardiomyocytes |
The coordination of activity across neocortical areas is essential for mammalian brain function . Understanding this process requires simultaneous functional measurements across the cortex . In order to dissociate direct cortico-cortical interactions from other sources of neuronal correlations , it is furthermore desirable to target cross-areal recordings to neuronal subpopulations that anatomically project between areas . Here , we combined anatomical tracers with a novel multi-area two-photon microscope to perform simultaneous calcium imaging across mouse primary ( S1 ) and secondary ( S2 ) somatosensory whisker cortex during texture discrimination behavior , specifically identifying feedforward and feedback neurons . We find that coordination of S1-S2 activity increases during motor behaviors such as goal-directed whisking and licking . This effect was not specific to identified feedforward and feedback neurons . However , these mutually projecting neurons especially participated in inter-areal coordination when motor behavior was paired with whisker-texture touches , suggesting that direct S1-S2 interactions are sensory-dependent . Our results demonstrate specific functional coordination of anatomically-identified projection neurons across sensory cortices .
Sensory perception , fine voluntary motor control , and higher cognitive functions depend on neural dynamics in the mammalian neocortex , which itself relies on the exchange of information between cortical areas through both bottom-up ( feedforward ) and top-down ( feedback ) neuronal pathways across the cortical hierarchy ( Bressler and Menon , 2010; Buschman and Miller , 2007 ) . Cortico-cortical connections are formed between columnar microcircuits via long-range axons of pyramidal neurons in superficial layer 2/3 ( L2/3 ) and deeper layer 5 . A given cortical area typically establishes connectivity patterns not only with one particular area but with multiple target areas in a distributed and often reciprocal fashion ( Markov et al . , 2013; Oh et al . , 2014; Zingg et al . , 2014 ) . Thus , in order to fully understand the cortical interactions underlying behavior , it is necessary to disentangle how neuronal subpopulations defined by both their functional properties and their specific anatomical projections contribute to local computation and long-range communication . Such an understanding has been limited by the difficulty in measuring population activity across areas with sufficient spatial and temporal resolution . Present methods to study large-scale cortical dynamics either lack cellular resolution and sensitivity to low numbers of action potentials ( e . g . , human fMRI; Hutchison et al . , 2013; or wide-field functional imaging in mice , Ferezou et al . , 2007; Lim et al . , 2013; Minderer et al . , 2012 ) or they are restricted to poorly defined neuronal subsets as for extracellular recordings ( Melzer et al . , 2006 ) . The main limitation for these recording approaches is the reliance on correlated activity patterns to infer information flow without the additional ability to attribute such activity to anatomically-defined neuronal subsets . Consequently , it has not been possible to definitively determine whether the underlying measured inter-areal dynamics could reflect: i ) direct cortico-cortical interactions; ii ) indirect cortico-thalamocortical pathways; iii ) or synaptic drive from common input areas . To dissect these possibilities new technologies are needed to monitor inter-areal dynamics with cellular resolution while at the same time identifying subsets of neurons that project across areas . Two-photon microscopy is well suited to monitor action potential firing across neuronal populations , mainly using calcium imaging , as well as to optically identify molecularly or anatomically-defined cell types ( Chen et al . , 2013a ) . So far , standard two-photon microscopes have been limited to imaging long-range activity within one cortical area ( Chen et al . , 2013b; Glickfeld et al . , 2013; Jarosiewicz et al . , 2012; Petreanu et al . , 2012; Sato and Svoboda , 2010 ) . New systems have recently been developed that enable simultaneous imaging of neuronal populations across cortical areas across increasingly larger fields of view ( Lecoq et al . , 2014; Stirman et al . , 2014; Tsai et al . , 2015 ) . Here , we present a novel 'multi-area' two-photon microscope for simultaneous measurements across primary and higher sensory areas of mouse neocortex . We have combined this system with anatomical labeling strategies to identify feedforward and feedback projection neurons between reciprocally connected cortical areas to image their functional interactions . In order to investigate the role of direct cortico-cortical interactions among other potential sources of correlated activity , we have applied this approach in the whisker primary ( S1 ) and secondary ( S2 ) somatosensory cortices , two areas that are anatomically coupled through reciprocal connections , cortico-thalamocortical pathways , and other common inputs ( Deschenes et al . , 1998; Suter and Shepherd , 2015; Theyel et al . , 2010 ) . Expanding our recent work on the activity of divergent projection pathways originating in S1 during a texture discrimination task ( Chen et al . , 2013b; 2015 ) , we sought here to examine how population activity in S1 and S2 evolves over time during such tactile whisker-based behavior . Whisking behavior spans a range of time scales , from individual whisk cycles of about 100-ms duration , to bouts of whisking over a second , and to prolonged whisking , for example during locomotion ( Kleinfeld and Deschenes , 2011 ) . Our multi-area imaging approach enabled us to analyze the slower aspects of whisking envelope changes and whisker-touch contacts whereas analysis of neuronal dynamics on the rapid time scale of tens of milliseconds was precluded by our limited temporal resolution . Our main goal was , however , to take advantage of the ability to simultaneously image in S1 and S2 and to investigate how the subsets of reciprocally projecting neurons contribute to the coordination of activity across these areas and to the coding of sensory and behavior information .
We built a two-photon microscope capable of simultaneous scanning of two sub-areas within a relatively large field of view ( FOV ) , enabling one to freely and independently position the sub-areas in order to select appropriate imaging spots . To achieve this goal we coupled two laser beams through a galvanometric scanner system into a low-magnification , high-NA objective ( Figure 1A–D and Materials and methods ) . Specifically , we chose a 16x water-immersion Nikon objective ( NA 0 . 8 ) as core element , which supports imaging in a FOV of 1 . 8-mm maximum side length with cellular resolution ( Figure 1E , Figure 1—figure supplement 1 and Video 1 ) . We split laser light from a Ti:sapphire laser ( 80 MHz pulse repetition rate ) into two excitation beams using a 50:50 beam splitter and delayed the laser pulse train of one beam by 6 . 25 ns , half of the inter-pulse interval , to interlace the two pulse trains so that the two sub-areas receive alternating laser excitation pulses . For disambiguating the fluorescence signal generated by the two laser foci , we adopted a rapid de-multiplexing approach ( Cheng et al . , 2011 ) . For typical 2–4 ns fluorescence lifetimes of fluorescent proteins ( Akerboom et al . , 2013 ) , the 6 . 25-ns time windows are sufficiently long to capture mainly fluorescence photons generated by the last excitation pulse . Some crosstalk between areas may remain but can be corrected for post hoc using spatial linear unmixing ( Materials and methods and Figure 1—figure supplement 2 ) ( Cheng et al . , 2011 ) . 10 . 7554/eLife . 14679 . 003Figure 1 . Multi-area two-photon microscope for flexible simultaneous imaging of sub-areas within a large field-of-view . ( A ) Schematic of multi-area two-photon microscope . Light from a Ti:sapphire laser is split into two beams and one beam sent to a delay line . Each beam then enters a focal plane unit ( FPU ) , which allows axial focusing with an electrically tunable lens ( ETL ) . Both beams are scanned in parallel by a pair of galvo mirrors . ( B ) Schematic of FPU . ( C ) Imaging modes include scanning of a single large FOV ( with one beam switched off ) and parallel scanning of two sub-areas . ( D ) Principle of spatiotemporal multiplexing: The detected fluorescence photons can be attributed to the correct area of origin by rapid demultiplexing synchronized to the laser pulse train . ( E ) Example two-photon image ( 1 . 7 mm FOV ) at 160–180 µm depth in a YCX2 . 60-expressing transgenic mouse in L2/3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 00310 . 7554/eLife . 14679 . 004Figure 1—figure supplement 1 . Variation of the point-spread function over field-of-view position and ETL tuning range . The point-spread function ( PSF ) was measured using 200 nm beads ( Fluoresbrite Plain YG Microspheres , Polysciences Inc ) at 840 nm excitation . Off-axis positions were accessed by translating a focal plane unit ( FPU ) to a position corresponding to a sample-offset of 500 μm and 900 μm . Full-width-half-maxima ( FWHM ) of the PSF are shown for the x- , y- , and z-direction . The PSFs exhibit residual astigmatism and degrade when the focus is moved away from the nominal working distance ( ∆z = 0 ) and the on-axis position . Due to vignetting , the ETL tuning range is limited at 900 μm off-axis . n = 5 beads , error bars: 95% confidence interval . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 00410 . 7554/eLife . 14679 . 005Figure 1—figure supplement 2 . Crosstalk between both sub-areas observed in vivo . Neurons expressing YC-Nano140 were imaged with a single beam exciting fluorescence either in sub-area 1 or 2 ( Average of n = 49 frames with motion correction ) . The detected signal in the non-illuminated sub-area is due to crosstalk caused by the exponential fluorescence decay ( shown amplified by a factor of 10 for better visibility ) . The crosstalk can be quantified by the ratio of both images after background subtraction ( the background was estimated from a ROI placed in the dark blood vessels ) . The percent crosstalk estimates are averages over 10000 pixels excluding blood vessels with standard deviation . As CFP is quenched by YFP in the Förster resonance energy transfer ( FRET ) interaction between both fluorophores , the lifetime of the CFP emission is shortened , which leads to a lower crosstalk . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 00510 . 7554/eLife . 14679 . 006Video 1 . In vivo z-stack of YC-Nano140 expressing neurons . Single area images from the multi-area two-photon microscope of L2/3 neurons were taken from 70-210 µm below the pial surface at 1 µm z-step resolution . Sub-area excitation beam was delivered through the ETL , positioned either on-axis ( left ) or 900 µm off-axis ( right ) , and focusing was achieved through translation of the objective by the z-stage . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 006 Each beam enters a movable coupling unit , named 'focal plane unit' ( FPU ) , which enables independent positioning and focusing of its respective imaging area below the objective ( Figure 1A–C ) . Independent positioning is achieved by coupling the FPU output beams to the scanner unit via small fold mirrors that sit at the end of cantilever arms . Lateral x/y-movement of each FPU introduces an offset of the respective beam from the optical axis of the first scan lens , which converts this offset into a pivoting angle of the beam around the scan mirrors . In the remaining optical path , this pivoting angle is translated into lateral movement of the corresponding imaging sub-area below the objective . Independent focusing is achieved with electrically tunable lenses ( ETLs ) in the FPUs ( Grewe et al . , 2011 ) . Each ETL is combined with an offset lens to allow tuning the beam from divergent to convergent . These divergence changes translate into axial shifts of the intermediate foci at the FPU output and in between scan and tube lens , corresponding to down- and upward shifts of the excitation focus along the optical z-axis below the objective . In combination with a 6-mm pair of scan mirrors , ETL focusing provides a z-range of up to 600 µm . Mice can actively sense the environment by moving their whiskers to gather information regarding the location , shape , size , and texture of an object ( Diamond et al . , 2008 ) . Processing of tactile information at the cortical level is thought to occur through interactions between S1 , S2 , as well as primary motor cortex ( M1 ) ( Aronoff et al . , 2010; Bosman et al . , 2011 ) . In order to investigate direct interactions between S1 and S2 , we sought to apply the multi-area two-photon microscope to simultaneously monitor activity in feedforward neurons in S1 projecting to S2 ( S1S2 ) and feedback neurons in S2 projecting to S1 ( S2S1 ) in wild-type adult mice during tactile whisker behavior . To distinctly label these projection neurons in a mutually exclusive manner across the reciprocally connected areas we employed a viral strategy making use of orthogonal recombinase systems . To label S2S1 neurons , we delivered a retrogradely-infecting AAV6 expressing Cre recombinase ( AAV6-pgk-Cre ) into S1 along with S2-injection of an AAV1 expressing Cre-dependent nuclear tdTomato ( AAV1-EF1α-dio-NLStdTomato; Figure 2A ) . S1S2 neurons were labeled by delivering an AAV6 expressing Flpe ( AAV6-syn-Flpe ) into S2 along with S1-injection of an AAV1 expressing Flpe-dependent nuclear LSSmKate2 ( AAV1-EF1α-fio-H2BLSSmKate2 ) . In addition to these viruses , we broadly expressed the genetically encoded calcium indicator YC-Nano140 in S1 and S2 using AAV1-EF1α-YCNano140 ( Chen et al . , 2013b; Horikawa et al . , 2010 ) . For targeting viral injections as well as for selecting regions for later two-photon imaging , we employed optical intrinsic signal imaging to identify areas within S1 and S2 corresponding to the same principal whisker ( Figure 2B and Materials and methods ) . Following cranial window implantation , LSSmKate2-positive S1S2 neurons and tdTomato-positive S2S1 neurons in L2/3 were identified in vivo ( Figure 2C ) . YC-Nano140 expressing neurons that did not express LSSmKate2 or tdTomato were classified as S1ND and S2ND neurons , respectively ( target area 'not determined' ) , possibly comprising unlabeled S1S2 and S2S1 neurons as well as projection neurons targeting different brain regions . Animals were habituated to head-fixation and trained to perform a whisker-based go/no-go texture discrimination task ( Figure 2D , E ) ( Chen et al . , 2013b; 2015 ) . On ‘go’ trials , animals were rewarded with a water droplet if they correctly licked ( ‘Hit’ ) when presented with a target texture ( a panel of coarse sandpaper , P100 ) . On ‘no-go’ trials , mice were supposed to withhold licking ( ‘correct rejection’ or ‘CR’ ) when presented with one of two non-rewarded , ‘non-target’ textures of smoother grades ( P280 , P1200 ) . Misses on go trials were not rewarded and false alarms ( ‘FA’ ) on no-go trials were punished with an air puff and a time-out period . Whisker movements were monitored with high-speed videography ( 500 Hz ) and licking behavior was measured with a piezo film attached to the water spout . Whisking and licking recordings were downsampled to match the frame rate of imaging ( 7 Hz ) , allowing analysis of how neuronal activity relates to slow amplitude changes of whisking envelope and to the occurrence of whisker-texture touches ( Materials and methods ) . Since simultaneous imaging in two cortical regions presents unique opportunities to examine the coordination of activity across areas , we sought to increase the number of pairwise imaged S1 and S2 neuronal populations . To this end we used the ETLs to implement a ‘combinatorial plane hopping’ mode , in which two sub-areas are scanned simultaneously but each imaging plane is independently refocused in a combinatorial manner during the inter-trial interval ( Figure 2C , F and Video 2 ) . Using this approach , we imaged in 7 mice ~150 neurons per sub-area ( distributed over three imaging planes at different cortical depths ) across ~1800 trials over 5–6 experimental sessions . Combinatorial hopping between three imaging planes in each area resulted in simultaneous imaging of 9 combinations of planes per animal , for which ~200 trials were acquired per pair of planes , still sufficient for our analysis . In total , 228 S1S2 , 817 S1ND , 193 S2S1 , and 750 S2ND neurons were imaged in 63 pairs of focal planes across S1 and S2 . For comparison with non-task conditions , we additionally imaged the same neuronal populations as measured during texture discrimination behavior for another ~1800 trials over 5–6 sessions , during which mice were passively presented with the same textures . In order to improve statistical analysis of single-trial responses for trial conditions with low trial numbers , calcium traces were denoised using tensor decomposition ( Figure 2—figure supplement 1 and Figure 2—source data 1 , Materials and methods ) . 10 . 7554/eLife . 14679 . 007Figure 2 . Simultaneous calcium imaging of identified feedforward and feedback neurons in S1 and S2 of mouse neocortex during behavior . ( A ) Viral injection scheme for simultaneous labeling of feedforward and feedback neurons and YC-Nano140 expression . ( B ) Functional mapping of S1 and S2 through optical intrinsic signal imaging . Intrinsic signals evoked by stimulation of the C2 whisker ( top left ) and the B2 whisker ( top right ) . In addition to localized intrinsic signals in S1 barrel columns , additional activation spots are visible in S2 . Identified barrel columns ( circles ) are overlaid over blood vessel ( bottom left ) and YC-Nano140 expression ( bottom right ) images . ( C ) In vivo 2-photon images of LSSmKate2-positive S1S2 neurons ( blue ) , tdTomato-positive S2S1 neurons ( red ) with non-co-labeled YC-Nano140-expressing neurons ( grey ) in S1 ( S1ND ) and S2 ( S2ND ) . ( D ) Behavior setup for texture discrimination task . ( E ) Trial structure for go/no-go texture discrimination task . ( F ) Example calcium transients for individual neurons in [C] measured episodically during texture discrimination task along with periods of whisker-to-texture touch ( orange area ) , whisking amplitude ( brown trace ) , and reaction time on Hit trials ( green area ) . For each trial the selected plane in each sub-area is indicated on top , illustrating the combinatorial plane hopping . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 00710 . 7554/eLife . 14679 . 008Figure 2—source data 1 . Optimized low tensor rank across animals . Table of optimum column size of each factor matrices related to neurons ( N’ + N’offset ) , time points ( T’ ) , and trial conditions ( C’ ) determined after cross-validation and cost function procedures for each animal used for denoising . Total possible column sizes are also indicated along with number of active neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 00810 . 7554/eLife . 14679 . 009Figure 2—figure supplement 1 . Denoising with tensor decomposition . ( A ) Calcium responses from one animal across multiple sessions are organized into a data tensor . The tensor is decomposed and a low-rank tensor representing denoised calcium responses is generated . ( B ) Denoising procedure: A low-rank tensor for trials and conditions was determined by cross-validation methods . A low-rank tensor for neurons was determined by convolving estimated spike trains from experiment data to generate simulated calcium traces . Noise is added to simulate calcium traces and tensor decomposition is performed to determine an optimum low-rank tensor by comparing simulated vs . simulated denoised traces . Once an optimum low rank tensor is determined for all dimensions , tensor decomposition is applied to raw traces . ( C ) Low tensor rank ( arrow ) computed by cross-validation of training set . ( D ) Contribution of each dimension to low tensor rank in [C] . ( E ) Final low tensor rank offset ( arrow ) computed by cost function of simulated vs . simulated denoised traces . ( F ) Comparison of simulated traces denoised with tensor decomposition vs . Gaussian filter . ( G ) Denoised fits from Gaussian filter vs . tensor decomposition for single neurons from simulated data . ( H ) Example of experimental data before and after denoising . ( I ) Optimum T’ or N’ + N’offset vs . active neurons for each animal . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 00910 . 7554/eLife . 14679 . 010Video 2 . Simultaneous calcium imaging across S1 and S2 . Single trial video of calcium responses during texture discrimination acquired at 7 Hz with the multi-area two-photon microscope ( 1x playback speed ) . YFP ( green ) and CFP ( blue ) fluorescence from YC-Nano140 are shown and overlaid with calculated ∆R/R ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 010 While sensory- and behavior-related responses of S1S2 , S1M1 , and S1ND neurons have been characterized during texture discrimination ( Chen et al . , 2013b; 2015 ) , responses of S2S1 and S2ND neurons have not . We first assessed for each cell class how calcium signals relate to behavioral aspects using a general linear model ( GLM ) against vectors for whisker-touch onset , whisking envelope amplitude , and licking onset ( Figure 3A , B and Figure 3—figure supplement 1 ) ( Miri et al . , 2011; Pinto and Dan , 2015 ) . S1ND and S2S1 neurons showed better overall GLM fits to these behavioral parameters compared to their neuronal counterparts in their respective areas ( Figure 3C and Figure 3—figure supplement 1; S1ND vs . S1S2 , p<0 . 002; S2ND vs . S2S1 , p<0 . 005 , KS-test ) . Further analysis of fits to specific regressors revealed that S2S1 and S1ND neurons showed higher GLM coefficients for whisker-touch onset than their within-area counterparts ( Figure 3D; S1ND vs . S1S2 , p<0 . 05; S2ND vs . S2S1 , p<0 . 005 , one-way ANOVA with repeated measures ) . While no specific differences were observed for cell classes in S1 , S2S1 neurons showed higher GLM coefficients than S2ND neurons for whisking and licking onset ( p<0 . 001 , one-way ANOVA with repeated measures ) . These results suggest that S2S1 neurons exhibit higher whisking- and licking-related activity compared to other neurons in S2 . 10 . 7554/eLife . 14679 . 011Figure 3 . Feedback neurons in S2 exhibit behavior-related responses . ( A ) General linear model ( GLM ) of behavior-related responses . Example of GLM fit for one neuron of calcium responses against touch , licking , and whisking as behavior events . Single-trial calcium responses are plotted along with model fit as well as touch periods with onset indicated , individual licks with onset indicated , whisking envelope amplitude , and decision . ( B ) GLM coefficients ( B ) for example neuron shown in [A] for regressors for touch onset , whisking envelope amplitude , and licking onset across different delays . Delays are aligned to the onset of each behavioral event . ( C ) Cumulative probability distribution ( cpd ) of overall GLM fit across cell types . ( D ) GLM coefficients for different cell types for regressors for touch onset ( left ) , whisking envelope amplitude ( middle ) , and licking onset ( right ) across different delays . Grey line indicates average GLM coefficient for neurons with non-significant coefficients at that time point . ( E ) Fraction of active neurons able to discriminate Hit vs . CR , FA vs . CR , and P280 vs . P1200 trials above chance determined by single-cell ROC analysis . ( shaded area: s . e . m . error bars: s . d . from bootstrap test; n = 44 S1S2 , 161 S1ND , 59 S2S1 , 198 S2ND neurons ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 01110 . 7554/eLife . 14679 . 012Figure 3—figure supplement 1 . General linear model of whisking- and licking-related calcium responses . ( A–B ) Example of GLM fit for neurons showing prominent whisking [A] and licking [B] related calcium responses against touch , licking , and whisking behavior events . Single-trial calcium responses are plotted along with model fit as well as touch periods with onset indicated , individual licks with onset indicated , and whisking envelope amplitude . ( C–D ) , GLM coefficients ( B ) for neurons in [A] and [B] , respectively , for regressors for touch onset , whisking envelope amplitude , and licking onset across different delays . Delays are aligned to the onset of each behavioral event . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 012 We next analyzed single-neuron responses to different sensory conditions or different behavior conditions by performing single-cell receiver operating characteristic ( ROC ) analysis against different trial conditions ( Green and Swets , 1966 ) . Single-cell ROC analysis of Hit vs . CR trials revealed that a larger fraction of S2S1 neurons compared to other neuronal classes ( 72% ) was able to discriminate these two conditions above chance ( Figure 3E; p<0 . 002 , χ2 test ) . Differences between Hit vs . CR trials could reflect encoding of sensory information , decision , or decision-related actions such as licking . To disambiguate these possibilities , we also performed ROC analysis of FA vs . CR trials , which were previously shown to consist of similar whisking and sensory conditions ( Chen et al . , 2015 ) . Again , a larger fraction of S2S1 neurons ( 50% ) was able to discriminate these two conditions above chance ( p<0 . 05 , χ2 test ) , suggesting that this greater discrimination power of S2S1 neurons represents decision- or action-related information . As an additional control , we assessed sensory-related responses by ROC analysis of P280 vs . P1200 textures on CR trials and found that S2S1 neurons were not more likely to discriminate these trial types compared to other cell types ( Figure 3E ) . Overall , we find that S2S1 were more likely to encode for non-sensory aspects of task-related behavior compared to other neurons in S2 and S1 . In order for task-related information exchange to occur between areas , activity across areas must be 'coordinated' during relevant behavioral conditions and such coordination should be specific to neurons that anatomically project between areas . To investigate how activity is coordinated across S1 and S2 , we first sought a measure of population activity for each area that would capture the diverse response properties of individual neurons and allow us to determine if their dynamics evolve similarly across time . To this end , we characterized population activity in S1 and S2 , respectively , by using linear discriminant analysis ( LDA ) ( Fisher , 1936; Safaai et al . , 2013 ) . For n neurons in an imaging area , LDA finds for each time point an axis in n-dimensional space so that the distributions of population responses for two chosen trial conditions – projected onto this axis – are best separated ( Materials and methods ) . Similar to the ROC analysis , we selected not only Hit vs . CR but also various other pairs of trial conditions that would allow us to disambiguate sensory- and behavior-related dynamics ( Table 1 ) . The dimensionality reduction resulting from this approach effectively extracts time-dependent ‘linear discriminant’ variables LD ( t ) as one-dimensional representations of neuronal population activity with respect to the chosen trial conditions . For illustration purposes , we exemplify this LDA procedure for measurements from only two neurons in Figure 4A , B , whereas typically populations of active neurons within an imaging area were used for analysis . 10 . 7554/eLife . 14679 . 013Table 1 . Axes used for linear discriminant analysis . Summary of trial conditions compared and used for linear discriminant analysis . For each axes , noted are potential differences in texture , licking , and whisking parameters between trial conditions as well as the utility in comparing such trial conditions for isolating sensory or behavior responses . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 013Axes for LDATextureLickingWhiskingUtility in analysisHit vs . CRDifferentDifferent ( Hit ) SameCannot isolate sensory , decision , or action-related responsesFA vs . CRSameDifferent ( FA ) SameIsolate decision and action-related responsesPre- vs . post-touch licking ( FA trials ) SameDifferentSameIsolate licking-related responsesHigh vs . Low Whisking ( CR trials ) SameNoneDifferentIsolate whisking-related responsesP280 vs . P1200 ( CR trials ) DifferentNoneSameIsolate sensory-related responsesTarget vs . Non-target ( Non-task ) DifferentNoneNoneIsolate sensory-related responses10 . 7554/eLife . 14679 . 014Figure 4 . Illustration of extracting population response time courses by linear discriminant analysis . ( A ) While LDA is performed on multiple simultaneously imaged neurons , for demonstration purposes , here calcium transients of two simultaneously imaged neurons within an imaging plane are plotted and sorted according to Hit and CR trials . Dotted line indicates whisker-touch onset . ( B ) Example linear discriminant analysis performed on the two neurons in [A] . Bottom panel shows scatter plot of trial-by-trial responses for each neuron at the indicated time point ( red region in [A] ) rotated along the LD axis for Hit vs . CR trials . Top panel shows distribution of trials for population activity projected along the LD axis along with mean LD response . ( C ) Average S1 or S2 population responses after LDA in Hit and CR trials across the first second prior to and following whisker-touch onset . ( D ) , ROC analysis of S1 or S2 population responses shown in [C] for Hit vs . CR trials under task conditions demonstrating the performance of the LDA . Dotted line indicates touch onset . ( shaded area: s . e . m . ; n = 21 S1 planes , 21 S2 planes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 014 In the initial analysis of population responses , we did not distinguish between neuronal cell types in each area and thus included both S1S2 and S1ND neurons for S1 and S2S1 and S2ND neurons for S2 . We performed LDA at each time point for 1-s periods prior to and following either whisker-touch onset or licking onset , generating mean LD time courses for S1 or S2 by averaging LDA results from all imaging areas in these respective regions . For LDA performed on Hit vs . CR trials , we observed that mean population responses for both S1 and S2 diverged following whisker-touch onset ( Figure 4C ) . ROC analysis using the LD variable as measure of population activity in S1 and S2 , respectively , revealed that discrimination power for both areas increased immediately following whisker-touch onset and through the first second of touch ( Figure 4D ) . S1 and S2 receive common input from several areas including M1 , which controls licking and whisking ( Brecht et al . , 2004; Huber et al . , 2012; Suter and Shepherd , 2015 ) , the posteromedial thalamic nucleus ( POm ) , which relays re-afferent whisking ( Deschenes et al . , 1998; Moore et al . , 2015; Yu et al . , 2006 ) , and the ventral lateral region of the ventral posterior medial thalamic nucleus ( VPMvl ) , which relays whisker touch ( Pierret et al . , 2000 ) . Correlations of activity between S1 and S2 could thus reflect these aspects of behavior . To measure how S1 and S2 activities are coordinated across time , we calculated the trial-by-trial correlation of LDS1 and LDS2 , the LD time courses obtained for active neurons in simultaneously imaged populations in S1 and S2 , respectively ( Figure 5A ) . We termed this cross-areal correlation LDCCS1:S2 , which during task performance increased immediately following whisker-touch onset for both Hit and CR trials . 500 ms after touch onset , however , LDCCS1:S2 remained elevated for Hit trials relative to CR trials ( Figure 5B , p<0 . 05 , one-way ANOVA with repeated measures ) . The time point of this divergence corresponded to the average delay of licking onset from whisker-touch onset ( mean reaction time: 524 ± 5 ms for Hit trials; Figure 5C ) ( Chen et al . , 2013b; 2015 ) . To examine whether LDCCS1:S2 changes indeed relate to the reaction time , LDCCS1:S2 on Hit trials were re-aligned to licking onset ( Figure 5D ) . LDCCS1:S2 increased and peaked at licking onset and remained elevated thereafter , further suggesting that coordination of activity across S1 and S2 could be associated with such behavior . 10 . 7554/eLife . 14679 . 015Figure 5 . Motor behavior is associated with coordinated population activity across S1 and S2 . ( A ) Analysis of coordinated activity across S1 and S2 . Left panel shows example of single-trial population responses for Hit trials projected along Hit vs . CR axis for simultaneously imaged S1 ( LDS1 ) and S2 ( LDS2 ) of S2 sub-areas . Upper right panels shows trial-by-trial correlations ( LDCCS1:S2 ) between LDS1 andLDS2 atindicated time points . Bottom right panel shows calculated LDCCS1:S2 across the trial period . ( B ) LDCC S1:S2 for Hit vs . CR trials . ( C ) Normalized histogram of reaction times across Hit trials . ( D ) LDCC S1:S2 for Hit trials along Hit vs . CR axis aligned to licking onset . ( E ) LDCC S1:S2 for FA vs . CR trials . ( F ) Licking rate for FA trials in which licking onset precedes ( pre-wo ) and follows ( post-wo ) whisker-touch onset . ( G ) LDCC S1:S2 for pre-wo vs . post-wo FA trials . ( H ) High vs . low whisking amplitude CR trials . ( I ) LDCC S1:S2 for high vs . low whisking amplitude CR trials . All time course data are aligned to whisker-touch onset ( black dotted line , x-axis ) except for [D] which is aligned to licking onset ( red dotted line ) . shaded area: s . e . m . ; ( C , D , E , G , I ) n = 63 pairs of S1 and S2 planes in 7 animals; ( C ) n = 7120 trials ( F ) n = 1120 trials ( H ) n = 7 animals , 6960 trials . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 01510 . 7554/eLife . 14679 . 016Figure 5—figure supplement 1 . Linear discriminant analysis across different sensory or behavior axes . Average S1 or S2 population responses across the first second prior to and following whisker touch onset for: ( A ) FA vs . CR trials; ( C ) high- vs . low-amplitude whisking CR trials; ( E ) target vs . non-target textures under non-task conditions; G , P280 vs . P1200 textures for CR trials . ROC analysis of S1 or S2 population response for: ( B ) FA vs . CR trials; ( D ) high- vs . low-amplitude whisking CR trials; ( F ) target vs . non-target textures under non-task conditions; ( H ) P280 vs . P1200 textures for CR trials . Dotted line indicates touch onset . ( shaded area: s . e . m . ; n = 21 S1 planes , 21 S2 planes ) DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 01610 . 7554/eLife . 14679 . 017Figure 5—figure supplement 2 . Coordinated actitvity across S1 and S2 is not stimulus-specific . ( A ) LDCCS1:S2 for target vs . non-target textures under non-task condition trials . ( B ) LDCCS1:S2 for P280 vs . P1200 textures for CR trials . We observed no increased or different LDCCS1:S2 between target and non-target textures under non-task conditions when animals received sensory stimulation but neither whisked nor licked . We also observed no increase or difference in LDCCS1:S2 when analyzing population responses for P280 vs . P1200 textures on CR trials under task conditions . All time course data are aligned to whisker-touch onset ( black dotted line , x-axis ) . shaded area: s . e . m . ; n = 63 pairs of S1 and S2 planes in 7 animals . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 017 To dissociate whether and how cross-areal coordination related to sensory versus motor parameters , we first controlled for sensory input by measuring LDCCS1:S2 for population responses projected along the FA vs . CR axis , where the same non-target textures were presented ( Figure 5E and Figure 5—figure supplement 1A , B ) . We found LDCCS1:S2 was higher for FA compared to CR trials both prior to and following whisker-touch onset , ( p<0 . 05 , one-way ANOVA with repeated measures ) . We asked if this increased LDCCS1:S2 on FA trials could partially be explained by licking behavior . We therefore subdivided FA trials into trials in which licking preceded whisker-touch onset – likely reflecting impulsive licking behavior – and those trials in which licking onset occurred after whisker-touch onset ( 67 . 1% and 32 . 9% of FA trials , respectively; Figure 5F ) . LDCCS1:S2 showed an increased level prior to whisker-touch onset specifically for the subset of trials with early licking ( Figure 5G , p<0 . 05 , one-way ANOVA with repeated measures ) . This suggests that population activity in S1 and S2 can be coordinated during licking behavior both in the presence and absence of sensory stimulation . We next asked whether LDCCS1:S2 is also related to other motor behaviors such as whisking . During texture discrimination , animals adopted a high-amplitude , rhythmic whisking strategy prior to whisker-touch onset in anticipation of the delivered texture , which drives texture-specific kinematics and is absent in non-task sessions ( Chen et al . , 2013b; 2015 ) . During task conditions , we measured LDCCS1:S2 from population responses for CR trials ( i . e . same texture , no licking ) projected along the high- vs . low-amplitude whisking axis ( Figure 5H and Figure 5—figure supplement 1C–D ) . Similar to the results for licking behavior , high-amplitude whisking trials were associated with higher LDCCS1:S2 prior to and after whisker-touch onset when compared to low-amplitude whisking trials ( Figure 5I , p<0 . 02 , one-way ANOVA with repeated measures ) , demonstrating another motor-related component of S1-S2 coordination . By using LDA for other pairs of trial conditions , we found that stimulation with distinct textures did not result in elevated LDCCS1:S2 , suggesting that S1-S2 coordination is not stimulus-specific ( Figure 5—figure supplement 1G–H and Figure 5—figure supplement 2 ) . Taken together , this demonstrates that the coordination of population activity across S1 and S2 can be associated with licking and whisking behavior that is independent of sensory stimulus . Correlated changes in population dynamics across cortical areas can either reflect direct cortico-cortical interactions , indirect interactions through cortico-thalamocortical pathways , or co-activation from another common input source ( Salinas and Sejnowski , 2001 ) . In order for direct cortico-cortical interactions to be present , such correlations should exist in neurons that project between those areas . To understand how S1S2 and S2S1 neurons might contribute to the coordination of population activity in S1 and S2 , we repeated the LDA for S1 or S2 but shuffled the trial-by-trial responses of S1S2 and S2S1 neurons when projecting the population response onto the LD axis ( Figure 6—figure supplement 1A ) . In order to ensure that changes in the population response were specific to these neurons and not merely a result of altering any given subpopulation of neurons , we also computed population responses , in which trials from an equal number of S1ND and S2ND neurons were shuffled ( see details in Materials and methods ) . We observed no significance difference in the trajectory or discrimination power of S1 and S2 population responses when shuffling any of these cell types ( Figure 6—figure supplement 2 ) , suggesting that the average population response within each area was not altered with this analysis . To determine the specific contribution of S1S2 and S2S1 to inter-areal coordination , we measured the change in correlation between the S1 and S2 population responses ( △LDCCS1:S2; relative to unshuffled controls ) that resulted from shuffling trials of these projection neurons and compared it to the result of shuffling trials of S1ND and S2ND neurons ( Figure 6A and Figure 6—figure supplement 1B ) . If S2S1 and S1S2 neurons especially contribute to LDCCS1:S2 , their trial-shuffling should lead to a larger reduction ( more negative △LDCCS1:S2 ) compared to trial-shuffling S1ND and S2ND neurons . Analysis of coordinated activity projected along the Hit vs . CR axis showed no significant difference in △LDCCS1:S2 between S2S1 and S1S2 neurons and S2ND and S1ND neurons when aligned to whisker-touch onset ( Figure 6—figure supplement 3 ) . However , analysis of Hit trials after aligning to licking onset revealed a negative dip in △LDCCS1:S2 when shuffling projection neurons , indicating that S1S2 and S2S1 neurons especially contributed to LDCCS1:S2 upon licking onset ( Figure 6B , p<0 . 0001 , one-way ANOVA with repeated measures ) . 10 . 7554/eLife . 14679 . 018Figure 6 . Projection neurons contribute to coordinated S1 and S2 activity . ( A ) The contribution of specific cell types to coordinated activity across S1 and S2 is measured by trial-shuffling responses for those cell types prior to calculating the LDCCS1:S2 . The resulting LDCCS1:S2 from the shuffled condition is then subtracted by the LDCCS1:S2 from the control condition to obtain ∆LDCCS1:S2 ( see also Figure 6—figure supplement 1 ) . ( B ) , ∆LDCCS1:S2 for Hit trials along the Hit vs . CR axis after aligning to licking onset . ( C ) ∆LDCCS1:S2 for high-amplitude whisking CR trials along the high vs . low whisking amplitude CR trial axis . ( D ) ∆LDCCS1:S2 for FA trials , in which licking onset precedes whisker-touch onset along the FA vs . CR axis . ( E ) ∆LDCCS1:S2 for FA trials , in which licking onset follows whisker-touch onset along the FA vs . CR axis . ( F ) ∆LDCCS1:S2 for CR trials along the FA vs . CR axis . All time course data are aligned to whisker-touch onset ( dotted line , x-axis ) except for [B] which is aligned to licking onset ( red dotted line ) . ( shaded area: s . e . m . ; n = 21 S1 planes , 21 S2 planes , 63 pairs of S1 and S2 planes in 7 animals ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 01810 . 7554/eLife . 14679 . 019Figure 6—figure supplement 1 . Measuring the contribution of specific cell types to coordinated population activity . ( A ) , An example of population response after trial shuffling . Trial responses of neuronal subpopulations were shuffled in order to determine their contribution to the population response . For the two example neurons also shown in [Figure 4A , B] , the contribution of ’cell 2‘ on the population response for Hit trials was examined by shuffling the order of trial responses for’ cell 2‘ . A scatter plot of the trial-by-trial responses of ’cell 1‘ vs . ’cell 2‘ is shown before ( middle ) and after ( bottom ) shuffling . The corresponding trials in control and shuffled conditions are indicated by color . The population responses under shuffled conditions were projected onto the LD axis for Hit vs . CR trials determined under control conditions ( top ) . ( B ) Example analysis of coordinated activity across S1 and S2 after trial shuffling . Single-trial population responses for CR trials projected along the FA vs . CR axis for simultaneously imaged S1 ( LDS1 ) and S2 ( LDS2 ) sub-areas are shown under control conditions ( top left ) and after shuffling trials of S1S2 and S2S1 neurons ( top right ) . Bottom panel shows trial-by-trial correlations ( LDCCS1:S2 ) between LDS1 and LDS2 at indicated time point for control and shuffled conditions . See [Figure 6A , B] for subsequent analysis of the same sub-areas to measure △LDCCS1:S2 across the trial period . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 01910 . 7554/eLife . 14679 . 020Figure 6—figure supplement 2 . Projection of shuffled trials does not alter average population response . ( A ) Cumulative distribution of average peak calcium responses for each cell type . ( B ) Average S1 ( top ) or S2 ( bottom ) population responses projected along the Hit vs . CR axis across the first second prior to and following whisker-touch onset in data in which no trials are shuffled ( control ) , trials of S1S2 and S2S1 neurons are shuffled ( S1S2S2S1 ) , and S1ND and S2ND neurons are shuffled ( S1NDS2ND ) . ( C ) ROC analysis of S1 ( top ) or S2 ( bottom ) population responses , in which no trials are shuffled ( control ) , trials of S1S2 and S2S1 neurons are shuffled ( S1S2S2S1 ) , and S1ND and S2ND neurons are shuffled ( S1NDS2ND ) vectors for Hit vs . CR trials under task conditions . Dotted line indicates touch onset . ( shaded area: s . e . m . ; n = 21 S1 planes , 21 S2 planes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 02010 . 7554/eLife . 14679 . 021Figure 6—figure supplement 3 . Contribution of S1S2 and S2S1 neurons to Hit and CR trials relative to whisker-touch onset . ( A ) Contribution of S1S2 and S2S1 neurons or S1ND and S2ND neurons to coordinated S1 and S2 activity for Hit trials along the Hit vs . CR axis after aligning to whisker-touch onset ( dotted line ) . ( B ) Contribution of S1S2 and S2S1 neurons or S1ND and S2ND neurons to coordinated S1 and S2 activity for Hit trials along the Hit vs . CR axis after aligning to whisker-touch onset . ( shaded area: s . e . m . ; n = 63 pairs of S1 and S2 planes in 7 animals ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 021 We further assessed the contribution of S2S1 and S1S2 neurons to inter-areal coordination along motor conditions by analyzing whisking- and licking-related LDCCS1:S2 . We first measured △LDCCS1:S2 from population responses projected onto the high- vs . low-whisking amplitude axis ( for CR trials ) and found that S1S2 and S2S1 neurons significantly contributed to LDCCS1:S2 in high-amplitude whisking trials following but not preceding whisker-touch onset ( Figure 6C , p<0 . 02 , one-way ANOVA with repeated measures ) . Similarly , analysis of FA vs . CR trials showed that S2S1 and S1S2 neurons did not specially contribute to LDCCS1:S2 for FA trials in which licking onset preceded whisker-touch onset ( Figure 6D ) , but they did so for trials in which licking onset followed whisker-touch onset ( Figure 6E , p<0 . 02 , one-way ANOVA with repeated measures ) . These findings indicate that while licking and whisking behavior is associated with correlations in population responses across S1 and S2 , any special contribution of S1S2 and S2S1 neurons to this coordination depends on the presence of sensory stimulus , thus occurring after the whisker-touch onset . Hence , the specific contribution of projection neurons mutually connecting S1 and S2 could reflect sensory- rather than motor-related activity . In line with this notion , further analysis showed that △LDCCS1:S2 decreased when trial-shuffling S1S2 and S2S1 neurons compared to S1ND and S2ND neurons following whisker-touch onset on CR trials , when licking behavior is absent ( Figure 6F , p<0 . 02 , one-way ANOVA with repeated measures ) . These results demonstrate that direct cortico-cortical interactions through S1S2 and S2S1 neurons reflect exchange of sensory or decision information rather than motor information .
In summary , we have demonstrated simultaneous measurement of calcium signals in identified feedforward and feedback neurons across S1 and S2 in the awake behaving mouse using a multi-area two-photon microscope in combination with viral-mediated labeling of long-range projection neurons . We have used this approach to investigate the contribution of cortico-cortical projection neurons to the coordinated activity patterns across these areas . While the acquisition rate of the imaging system and the kinetics of the expressed calcium indicator ( Chen et al . , 2013b ) used in this study precludes our ability to capture the 4–10 ms spike latencies reported across mouse cortical areas ( Ferezou et al . , 2007 ) for examining spike-timing and monosynaptic relationship of long-range cortical dynamics , we nevertheless observe that population activity across S1 and S2 is coordinated during relevant task periods in a behavior-dependent manner . We took a simplified view of the population activity by performing dimensionality reduction with LDA , which is a supervised method to project high-dimensional dynamics onto a single axis . Specifically , through the analysis of correlated population responses across S1 and S2 along multiple LDA axes , we find that inter-areal coordination is associated with both goal-directed licking as well as whisking behavior and that it can occur independent of sensory stimuli . In the absence of tactile stimuli , S1S2 and S2S1 neurons do not show a special contribution to motor-related coordination , suggesting that this coordination does not necessarily reflect direct cortico-cortical interactions . S1 and S2 receive common input from M1 and POm , conveying efferent and re-afferent motor information ( Deschenes et al . , 1998; Suter and Shepherd , 2015 ) , and are additionally coupled by thalamic relays through POm ( Theyel et al . , 2010 ) ( Figure 7 ) . We speculate that motor-related S1 and S2 coordination could be a result of common drive or cortico-thalamocortical pathways through these shared areas . In contrast , we find a special contribution of identified S1S2 and S2S1 neurons to inter-areal coordination occurring during whisker-texture touch , indicating that their participation particularly depends on sensory stimuli . This contribution is most prominent when sensory stimuli and motor behavior are paired , such as upon licking onset on Hit trials , which further suggests that such cortico-cortical interactions could be involved in a form of 'active sensation' . However , we reason that this interaction does not necessarily reflect motor behavior . Given that these neurons also specially contribute to inter-areal coordination during whisker-touch periods in CR trials , we propose that these direct cortico-cortical interactions more likely represent the exchange of sensory- or decision-related activity . 10 . 7554/eLife . 14679 . 022Figure 7 . Model of coordinated activity across S1 and S2 . Our results identify coordinated activity patterns across S1 and S2 that are related to motor behaviors , which could arise from common input from M1 or POm or indirect cortico-thalamocortical ( CTC ) pathways through POm . S1S2 and S2S1 neurons especially participate in inter-areal coordination when motor behavior is paired with sensory stimuli suggesting that such cortico-cortical ( CC ) interactions specifically reflect the exchange of sensory information during active sensation . DOI: http://dx . doi . org/10 . 7554/eLife . 14679 . 022 Our findings provide the first direct evidence for a unique contribution of direct cortico-cortical interactions over other sources of correlated activity across these areas . Such specificity points towards distinct but potentially synergistic roles for how different inputs may be involved in information flow across the cortex . The division between direct and indirect interactions along sensory and motor parameters , respectively , is in line with theories that indirect cortico-thalamocortical pathways are more involved in relaying motor , rather than sensory , signals ( Sherman and Guillery , 2011 ) . Additionally , both frontal cortical areas such as M1 and higher-order thalamic nuclei such as POm have been implicated in gating and coordinating activity across somatosensory areas ( Pais-Vieira et al . , 2013; Theyel et al . , 2010; Zagha et al . , 2013 ) . In the visual system , the pulvinar , a higher-order thalamic nuclei involved in attention , has been identified in coordinating activity between visual areas ( Saalmann et al . , 2012 ) . Our findings support the notion that nuclei that also drive motor-related or attention-related activity might help to coordinate primary and higher sensory areas in a manner that facilitates sensory-related direct cortico-cortical communication upon stimulus presentation . What is the relevance of this sensory-related cortico-cortical interaction ? It has been suggested that feedback inputs from higher sensory areas provide contextual information to help extract relevant sensory features provided by feedforward inputs in primary areas ( Gilbert and Li , 2013 ) . We find that S2S1 neurons exhibit prominent licking and decision-related activity , which is in line with recent evidence that behavior-related activity in S1 can be inherited from S2 ( Yang et al . , 2016 ) . Coordinated activity between S1S2 and S2S1 neurons during texture discrimination may reflect sensory processing involved in several functions . One function may be associated with decision making , as exemplified by correlations increasing upon whisker touch onset and peaking upon licking onset on Hit trials . Another function may be associated with the reinforcement of particular aspects of the sensory signal that might strengthen or stabilize sensory representations through goal-directed learning , exemplified by the persistent coordinated activity after licking onset on Hit trials ( Chen et al . , 2015 ) . Future work to dissect S1S2 and S2S1 neuronal dynamics using projection-targeted multi-area calcium imaging will help to resolve these possibilities . In conclusion , multi-area calcium imaging with anatomical tracers presents new opportunities for overlaying functional measurements with recent comprehensive mapping of the long-range connectivity in mouse neocortex ( Oh et al . , 2014; Zingg et al . , 2014 ) . Several different approaches have been implemented for imaging across brain areas . While the use of multiple miniature objectives does not limit the maximum distance between areas that can be imaged ( Lecoq et al . , 2014 ) , the physical size and working distance of such objectives does limit the proximity between areas and depth that can be imaged noninvasively . The use of multiple beams through a single large-FOV , long-working distance objective thus provides a complementary approach . The multi-area two-photon miscroscope described here shares similar design principles as reported in Stirman et al . ( 2014 ) . The core principles and modularity of these designs readily allows for improvements in temporal resolution through resonant , free-line , or random-access scanning systems ( Bathellier et al . , 2012; Grewe et al . , 2010 ) , increasing the number of simultaneously imaged areas and cortical layers with low repetition rate lasers ( Cheng et al . , 2011; Quirin et al . , 2014 ) , and imaging across larger FOVs with different optical configurations ( Stirman et al . , 2014; Tsai et al . , 2015 ) . In addition to developments in imaging technology , new genetic tools are being developed for combinatorial conditional gene expression to concurrently label increasing number of pathways ( Fenno et al . , 2014 ) and genetically encoded voltage indicators for reporting electrical signals ( Akemann et al . , 2013; Sun et al . , 2013 ) . These developments will expand the range of biological questions that can be addressed in elucidating the relationship between long-range cortical communication and the fine-scale organization and computations occurring within local circuitry .
The microscope consists of several building blocks: The beam preparation stage , which sends two pulse trains with the correct delay and intensity to the focal plane units ( FPUs ) , which in turn allow independent focusing and positioning of each sub-area . The scan system scans the beams directed to each sub-area in parallel and sends them to the objective via the excitation optics . The microscope front end consists of the objective , z-stage and detection system . FPUs , excitation optics and the microscope front end were mounted on an elevated breadboard . The optical design software Zemax ( Zemax LLC , Redmond , USA ) was used for system layout and performance evaluation . For construction of the AAV-EF1α-fio-H2BLSSmKate2 viral construct , the double-inverse oriented FRT ( fio ) sites was synthesized de novo ( GenScript ) with flanking BamH1 and EcoRI restriction sites and an internal AscI and NheI sites and insert into an AAV-EF1α-YC-Nano140 ( Chen et al . , 2013b ) backbone plasmid . The H2B subunit with 5' NheI and 3' AgeI restriction sites was generated by PCR amplification from a pTagRFP-H2B vector ( Evrogen ) and subcloned into an pLSSmKate2-N1 plasmid ( Piatkevich et al . , 2010 ) . Subsequently , H2BLSSmKate2 with 5' NheI and 3' AscI restriction sites was generated by PCR amplification followed by insertion into the AAV-EF1α-fio plasmid . The AAV-syn-Flpe viral construct was generated by restriction enzyme digest of pCAG-Flpe ( Matsuda and Cepko , 2007 ) and insertion into the pAAV-6P-SEWB backbone plasmid . For the AAV-EF1α-dio-NLStdTomato viral construct , NLStdTomato with 5' NheI and 3' AgeI restriction sites was generated by PCR amplification from a pTagRFP-H2B vector followed by insertion into the AAV-EF1α-dio-eYFP plasmid . The AAV-pgk-Cre construct was previously described3 . Recombinant serotype 6 AAV particles were produced by co-transfecting AAV-293 cells with the shuttle plasmid and the pDP6 packaging plasmid . Recombinant serotype 1 AAV particles were produced by co-transfecting AAV-293 cells with the shuttle plasmid and the pDF1 packaging plasmid . Cell lysates were subjected to purification on iodixanol density gradients followed by HPLC with HiTrap Heparin column for AAV2 or by anion exchange HPLC for AAV1 ( GE Healthcare Bio-Sciences AB ) using standard procedures . The viral suspension obtained was concentrated using Centricon centrifugal filter devices with a molecular weight cut-off of 100 kDa ( Millipore , Billerica , MA ) , and the suspension medium replaced with PBS . Vector titres were determined by measuring the number of encapsidated genomes per ml by real-time PCR . Experimental procedures followed the guidelines of the Veterinary Office of Switzerland and were approved by the Cantonal Veterinary Office in Zurich . Stereotaxic viral and tracer injections were performed on young adult ( P35-42 ) male wild type C57Bl6/J mice as previously described ( Chen et al . , 2013b ) . A solution containing AAV1-EF1α-YC-Nano140 , AAV1-EF1α-fio-H2BLSSmKate2 , and AAV6- pgk-Cre ( 200 nl total volume , ~1 x 109 vg/μl per virus , 2:1:1 ratio by volume ) was delivered into S1 , targeting L2/3 and L5 ( ~300 and 500 μm below the pial surface ) . A solution containing AAV1- EF1α-YC-Nano140 , AAV1-EF1α-dio-NLStdTomato , and AAV6-syn-Flpe ( 200 nl total volume , ~1 x 109 vg/μl per virus , 2:1:1 ratio by volume ) was delivered into S2 , targeting L2/3 and L5 ( ~300 μm and 500 μm below the pial surface ) . Injection regions were selected by optical intrinsic signal imaging or stereotaxic coordinates ( 1 . 1 mm posterior to bregma , 3 . 3 mm lateral for S1; 0 . 7 mm posterior to bregma , 4 . 2 mm lateral for S2 ) . To allow long-term in vivo calcium imaging , a cranial window was implanted 24 hr after virus injections over S1 as described33 . A metal post for head fixation was implanted on the skull , contralateral to the cranial window , using dental acrylic . For demonstration of large single-FOV imaging , structural images were acquired from one adult male Rasgrf2-2A-dCre;Camk2atTA;Ai92 ( TITL-YCX2 . 60 ) transgenic mouse ( Madisen et al . , 2015 ) implanted with a cranial window without viral injections . Mice were housed 2–3 per cage in reverse 12 hr light cycle conditions . All handling and behaviour occurred under simulated night time conditions . One week following chronic window implantation , mice were handled daily for 1 week while acclimated to a minimum of 15 min of head fixation . Mice were water restricted and trained to a go/no-go texture discrimination task previously described ( Chen et al . , 2013b ) . Imaging during behaviour began following 3–5 training sessions once animals reached a performance level of d' > 1 . 75 ( 80% correct ) for one session . Imaging under task conditions was performed over the course of 5–6 sessions at a performance level of d’ = 2 . 62 ± 0 . 15 . Once sufficient task-related data was acquired , mice were provided with free access to water and then imaged for an additional 5–6 sessions under non-task conditions , in which textures were presented but no reward or punishment delivered . Sample sizes were chosen based on previous behavioural imaging studies , which comprise 6–10 mice per group ( Chen et al . , 2013b; 2015 ) . Due to their low occurrence ( 6 . 7 ± 0 . 5% of all trials ) , miss trials were excluded from analysis . No statistical methods were used to predetermine sample size . Since animals constitute a single experimental group , no randomization of animals or blinding to experimenter was performed . The S1 and S2 barrel column was identified using intrinsic signal optical imaging under ~1 . 5% isoflurane anaesthesia . The cortical surface was illuminated with 630-nm LED light , single whiskers were stimulated ( 2–4° rostro-caudal deflections at 10 Hz ) , and reflectance images were collected through a 4x objective with a CCD camera ( Toshiba TELI CS3960DCL; 12-bit; 3-pixel binning , 427x347 binned pixels , 8 . 6-µm pixel size , 10-Hz frame rate ) . Intrinsic signal changes were computed as fractional changes in reflectance relative to the pre-stimulus average ( 50 frames; expressed as ∆R/RIOS ) . Barrel column centres for stimulated whiskers were located by averaging intrinsic signals ( 15 trials ) , median-filtering ( 5-pixel radius ) , and thresholding to find signal minima . Reference surface vasculature images were obtained using 546-nm LED illumination and matched to images acquired during 2-photon imaging . Prior to behavior training , all whiskers excluding the principal and first-order surround whiskers corresponding to the expression area were partially trimmed to a length out of reach from texture contact during the task . During whisker trimming , the principal whisker was noted by images taken from the high-speed video camera for re-identification in subsequent imaging sessions for whisker tracking . The whisker field was illuminated with 940-nm infrared LED light and movies were acquired at 500 Hz ( 500 x 500 pixels ) using a high-speed CMOS camera ( A504k; Basler ) . Average whisker angle across all imaged whiskers was measured using automated whisker tracking software ( Knutsen et al . , 2005 ) . Because our limited temporal resolution of imaging ( 7 Hz ) precluded analysis of rapid dynamics within individual whisking cycles , we based our analysis on the envelope amplitude of whisking calculated as the difference in maximum and minimum whisker angle along a sliding window equal to the imaging frame duration ( 142 ms ) . The slower dynamics of the envelope amplitude represents both rhythmic and non-rhythmic forms of whisking behavior . For comparison between high- vs . low-amplitude whisking trials , the mean whisking amplitude during the 1-s period prior to whisker-touch onset was calculated for each animal and high- and low-amplitude trials were identified as those whose amplitude during the same period was greater or less than the mean , respectively . For all trials , the first and last possible time point for whisker to texture contact was quantified manually through visual inspection . For in vivo identification of LSSmKate-positive feedforward and tdTomato-positive feedback neurons , 3D-volume image stacks were taken on a standard custom-built 2-photon microscope controlled by HelioScan34 , equipped with a Ti:sapphire laser system ( ~100-fs laser pulses; Mai Tai HP; Newport Spectra Physics ) , a water-immersion objective ( 40×LUMPlanFl/IR , 0 . 8 NA; Olympus ) , galvanometric scan mirrors ( model 6210; Cambridge Technology ) , and a Pockels Cell ( Conoptics ) for laser intensity modulation . An 800-nm excitation with 610/75 nm emission filter and 840–900 nm excitation with 697/75 nm emission filter was used for tdTomato and LSSmKate2 , respectively . Due to suboptimal in vivo 2-photon excitation of LSSmKate2 , additional H2BLSSmKate2-positive neurons were identified followed by antibody staining of LSSmKate2 for signal amplification . Mice were anesthetized ( ketamine/xylazine; 100/20 mg/kg body weight ) and perfused transcardially with 4% paraformaldehyde in phosphate buffer , pH 7 . 4 . Cortical sections ( 50 µm ) were cut along the imaging plane using a vibratome ( VT100; Leica ) , then blocked in 10% NGS and 1% Triton at room temperature and incubated overnight at 4°C in 5% NGS , 0 . 1% Triton and mKate guinea pig polyclonal antibody ( Cai et al . , 2013 ) ; 1:1 , 000 ) . A guinea pig Alexa647–conjugated goat IgG secondary antibodies ( 1:400; Molecular Probes , Invitrogen ) was applied for 2 hr at room temperature . Images were acquired with a confocal microscope ( Fluoview 1000; Olympus ) , green ( YC-Nano140 ) , red ( tdTomato ) , and infrared ( Alexa647 ) excitation/emission filters . Two-channel , two-area ( CFP/YFP ) calcium imaging data was imported into MATLAB ( Mathworks ) for processing . For each channel , spatial linear unmixing was applied for the two area as described below . Background was subtracted on each area and channel ( bottom 1st percentile fluorescence signal across entire frame ) . For each area , Hidden Markov Model line-by-line motion correction was applied to both data channels . Regions of interests ( ROIs ) corresponding to individual neurons were manually selected from the mean image of a single-trial time series using ImageJ ( National Institute of Health ) . Mean pixel value for each ROI was extracted for both channels . Calcium signals were expressed as relative YFP/CFP ratio change ΔR/R= ( R-R0 ) /R0 . R0 was calculated for each trial as the bottom 8th percentile of the ratio for the trial . Active neurons were identified by two-way ANOVA with repeated measures of the neuronal calcium signal against the neuropil signal ( significance value , p<0 . 05 ) for each imaging session . The neuropil is defined as a region of interested selected from the entire imaging frame representing non-somatic tissue expressing YC-Nano140 but excluding blood vessels . Calcium signals were denoised using tensor decomposition before further analysis ( Figure 2—figure supplement 1 and Figure 2—source data 1 ) ( Cong et al . , 2015; Seely et al . , 2014 ) . Tensor decomposition is a method used for dimensionality reduction , which can be viewed as a generalization of singular value decomposition of data represented as tensors rather than matrices ( Hitchcock , 1927 ) . While calcium imaging recordings are often described as two-dimensional matrices comprised of neurons and time dimensions , it can additionally be described along a third dimension representing trial conditions ( Figure 2—figure supplement 1A ) . For such data , tensor decomposition can be used as a form of single-trial denoising by assuming that calcium signals across neurons , time , and trial conditions are not independent and that multi-linear relationships across dimensions therefore can be exploited . Through tensor decomposition , background noise that does not match the assumed multi-linear structure can be reduced if present . Single-trial denoising of calcium transients is desirable when analyzing conditions with low trial counts such as FA trials ( 7 . 4% of all trials ) in order to improve statistical analysis of such conditions . For each animal , calcium signals were arranged into a data tensor ( Y ) across three dimensions according to the number of trial conditions ( I; i . e . , 6 combinations of decision and texture ) , number of neurons ( J ) , number of time points ( K ) . Using Tucker decomposition , this tensor can be described elementwise as: ( 1 ) yijk = ∑c=1C∑n=1N∑t=1Tgcntmicmjnmktfor i=1 . . . I , j=1 . . . J , k=1 . . . K consisting of a factor matrix related to trial-condition containing elements ( mic ) with column size C , a factor matrix related to neuron containing elements ( mjn ) with column size N , a factor matrix related to time point ( mkt ) with column size T , and a core tensor describing the interactions between the matrix components containing elements ( gcnt ) . From this , a low rank tensor , Y’ , containing the denoised traces can be described elementwise as: ( 2 ) yijk′ = ∑c=1C′∑n=1N′∑t=1T′gcntmicmjnmktfor i=1 . . . I , j=1 . . . J , k=1 . . . K This tensor is obtained by reducing the column size of each factor matrices across each dimension resulting in C’ which is related to the number of trial conditions such that ( C’ ≤ I ) , N’ which is related to the number of neurons such that ( N’ ≤ J ) , and T’ which is related to the number of time points such that ( T’ ≤ K ) . From this , a tensor rank ( Θ' ) for Y’can be expressed as the sum of the reduced column sizes across all dimensions: ( 3 ) Θ′ = C′+N′+T' In order to determine the optimum Θ' , a five fold cross validation procedure was first performed ( Figure 2—figure supplement 1B ) ( Seely et al . , 2014 ) . For each trial condition in each neuron , trials were divided into a training set ( 80% of trials ) and a test set ( 20% of trials ) . Single-trial traces in each tensor element were replaced with average traces from the training set . Denoised traces were obtained for a given Θ' and compared to the average traces of the test set by computing the mean squared errors ( MSE ) ( Figure 2—figure supplement 1C , D ) . The optimum Θ' is identified as Θ' with the minimum MSE . Determining Θ' by five fold cross validation is advantageous in that it is unsupervised and can correct for unknown sources of noise . However , since the error estimation used in this procedure is based on comparing average traces , the Θ' determined is not necessarily optimized for denoising single-trial responses and thus neurons with variable trial-to-trial responses may not be properly denoised . Indeed , while five fold cross validation was sufficient in identifying optimum Θ' for T’ and C’ , better fits for some neurons were observed when manually adjusting N’ ( data not shown ) . In order to improve denoising of single-trial responses , a second-step procedure was implemented to optimize N’ through a supervised approach of performing tensor decomposition on noisy simulated calcium transients in order to determine a rank offset ( N’offset ) resulting in a final tensor rank ( Θfinal' ) such that: ( 4 ) Θfinal' = C'+N'+Noffset'+T' where the denoised transient best reflects the ideal transients . In order to emulate the multi-linear structure across neurons , time , and trial conditions present in our experimental data that is required for tensor decomposition , a peeling algorithm ( Grewe et al . , 2010 ) using previously measured YC-Nano140 parameters ( Chen et al . , 2013a ) ( single-action potential transient: A0 = 4 . 54% , τonset = 0 . 186 s , Apeak = 2 . 3% , τdecay = 0 . 673 s ) was applied to raw traces to extract estimated spike trains for all neurons and trials for a single animal . While the accuracy and precision of the estimated spikes may vary depending on noise in the raw trace ( Lütcke et al . , 2013 ) , the multi-linear relationships across each tensor dimension is preserved . The estimated spike trains are then convolved using YC-Nano140 parameters to produce ideal simulated calcium transients . The degree of noise under experimental conditions is estimated by assuming that any variance in calcium signal present in inactive neurons reflects non-neuronal noise . For each inactive neuron , a normal distribution was fit to raw calcium traces to obtain σ representing the degree of noise for that neuron . Noise was then added neuron-by-neuron to simulated calcium transients that matched the σ’s from all inactive neurons in the data set . The similarity between the ideal and denoised simulated trace was measured by computing the Pearson’s correlation coefficient ( CC ) between the two traces for each neuron and taking the average across neurons . From this , the optimum Θfinal' was determined by calculating a cost function representing the difference between Θfinal' and the CC obtained from Θfinal' , each normalized across the range of tested Θfinal' : ( 5 ) Cost ( Θfinal′ ) =∥ Θfinal′∥ −∥ CC ( Θfinal′ ) ∥ such that the optimum Θfinal' resulted in the minimum Cost ( Θfinal' ) ( Figure 2—figure supplement 1E ) . In comparing denoising of simulated transients with tensor decomposition against temporal smoothing by a 5-point Gaussian filter , we observed that denoising with tensor decomposition better preserves the onset and peak of calcium transients , resulting in better CC of denoised to ideal traces ( Tensor decomposition 0 . 69 ± 0 . 01 , Gaussian filter: 0 . 65 ± 0 . 01 , p<1x10-6 , Student’s t-test , Figure 2—figure supplement 1F , G ) . This suggests that denoising with tensor decomposition is preferred when investigating sub-second temporal dynamics of activity as it preserves high frequency components of the calcium signal . The optimum Θfinal' for each animal was determined for denoising ( Figure 2—figure supplement 1H ) . We asked if the size of the optimum low rank tensor used for denoising was similar across animals ( Figure 2—source data 1 ) . We observed that C’ was largely consistent across animals and reflected a rank near the total possible ranks along the condition dimension . For T’ and N'+Noffset' , we observed that the optimum column size across these dimensions was strongly correlated with the number of identified active neurons ( T’: R = 0 . 82 , p<0 . 05; N'+Noffset': R = 0 . 82 , p<0 . 05 , Pearson’s correlation , Figure 2—figure supplement 1I ) . This suggests that the optimum low rank tensor identified for denoising captures a relevant portion of the original data tensor containing real calcium transient events . Spatial linear unmixing is based on the fact that the total PMT signal recorded at the corresponding pixel for both areas in a given channel is the linear sum of the signal for each area weighted by the cross talk resulting from the fluorescence lifetime of the indicator . For a dual beam system , the contribution of the two detected areas can be represented by the following equations: ( 6 ) J1 ( x , y ) =S1 , 1×I1 ( x , y ) +S1 , 2×I2 ( x , y ) J2 ( x , y ) =S2 , 1×I1 ( x , y ) +S2 , 2×I2 ( x , y ) where J is the total signal per area , I is the fluorophore abundance , and S is the crosstalk . These equations can be expressed as a matrix: ( 7 ) [J]=[S][I] whereby the unmixed image [I] can be calculated using the inverse matrix of [S] ( 8 ) [I]=[S]−1[J] assuming the detected signal in both areas represents the total signal: ( 9 ) S1 , 1+S2 , 1=1S1 , 2+S2 , 2=1 [S] [S] was determined empirically at the beginning of each session using the experimentally prepared mouse expressing YC-Nano140 . The intended FOVs were sequentially scanned with a single excitation beam during dual area acquisition mode . The resulting crosstalk into each area was calculated from the acquired reference images and applied for spatial linear unmixing of subsequent dual beam data using MATLAB . Behavior-related activity was described using a general linear model ( GLM ) ( Miri et al . , 2011; Pinto and Dan , 2015 ) expressed as: ( 10 ) Yt = ∑i=−36BiL Xt−iL+ ∑i=−36BiWXt−iW+ ∑i=−36BiT Xt−iT Z-scored regressors ( Xt+i ) representing touch onset ( T ) , whisking envelope amplitude ( W ) , and licking onset ( L ) with regression coefficients ( Bi ) at different delays ( i ) were used to model the z-scored calcium signal ( Yt ) across time frames t . Regressors for touch onset and whisking amplitude were obtained from the whisker-tracking video while regressors for licking onset were obtained from the lick-sensor data . Each regressor was down sampled to match the calcium imaging frame rate . Touch onset was selected to best reflect touch-related responses given previously reported neuronal adaptation in neuronal firing upon repeated touches ( Musall et al . , 2014; Yamashita et al . , 2013 ) . Whisking envelope amplitude was previously observed to best reflect periods of whisking and non-whisking behavior in order to identify whisking-related neurons ( Chen et al . , 2013b ) . Given the slow kinetics of calcium indicators and given that the imaging rate is well below the Nyquist rate of the natural whisking frequency ( ~10 Hz ) ( Kleinfeld and Deschenes , 2011 ) , whisking-related signals measured here do not reflect whisking frequency . Licking onset was selected due to the observation that licking behavior in task-performing mice typically proceeds in licking bouts . Introducing additional behavioral regressors such as licking offset and touch offset to the GLM did not improve model fit ( data not shown ) . In order to capture a physiologically realistic range of response delays to behavioral events as previously observed ( Chen et al . , 2013b; 2015 ) , regressors for each behavioral parameter were generated across a range of delays from i = −3 ( t = 0 . 43 s before behavior event ) to i = +6 ( t = 0 . 85 s after behavioral event ) . Only delays from i = −2 to i = +5 were included for cell type analysis . GLM was applied to active neurons , where the first 5 s from each trial across active sessions were extracted and concatenated for analysis . To fit the GLM , trials were randomly divided into a training set ( 80% of trials ) and a test set ( 20% of trials ) . Ridge regression was used to minimize the Bi at irrelevant delays . The optimum regularization parameter was determined by performing a five fold cross validation within the training set and selecting the value with the best cross validation performance . Bi was then calculated from the training set and applied to the test set to obtain predictions for Yt . To assess GLM fit , a coefficient of determination ( R2 ) was calculated by comparing the predicted and the original traces . To reduce the effect of the particular choice of test trials on R2 , test trials were bootstrapped 1000 times to obtain a final R2 reflecting GLM fit . To assess the significance of individual Bi , a shuffled distribution for each Bi was obtained by permutation test after shuffling calcium traces for time points within each trial 1000 times . Bi whose value was greater than the 95 percentile of the shuffled distribution was identified as significant . GLM does not require normal distribution of the data set . Comparisons of Bi across cell types was performed using one-way repeated measures ANOVA . The variances of each cell type were tested using the F-test and determined to not be significantly different . The performance of neuronal populations or single neurons in discriminating two trial types was assessed using a receiver operating characteristic ( ROC ) analysis ( Green and Swets , 1966; O'Connor et al . , 2010 ) . For neuronal populations , the discriminability of the population response projected along the LD axis was measured at each time point 1 s prior to and following touch . Each trial was assigned a ‘discrimination variable’ score ( DV ) equal to the similarity to the mean projected population response for trial type X minus the similarity to the mean projected population response for trial type Y . Thus , for trial type X ( 11 ) DVx=Xi ( X¯∀j≠i−Y¯ ) and for trial type Y ( 12 ) DVY=Yi ( X¯−Y¯∀j≠i ) where Xi and Yi are the single-trial population response for the i-th trial . X¯ and Y¯ are the mean population response . Trials were classified as belonging to trial type X or Y if DVX or DVY was greater than a given criterion , respectively . To determine the fraction of trials an ideal observer could correctly classify , an ROC curve was constructed by varying this criterion value across the entire range of DVX or DVY . At each criterion value , the probability that a trial of type X exceeded the criterion value was plotted against the probability that a trial of type Y exceeded the criterion value . The area under the ROC curve ( Aobserved ) was then calculated to represent the single-neuron performance ( ‘fraction correct’ ) as the fraction of trials correctly discriminated by an ideal observer using the DV . We corrected for sampling bias due to the limited number of trials collected , using methods described ( Safaai et al . , 2013 ) . The sampling bias ( Abias ) was determined by calculating the mean area under the ROC curve after randomly shuffling trial type or stimulus labels repeated 1000 times . The corrected area under the ROC curve Acorrected was then calculated as Acorrected = Aobserved − Abias + 0 . 5 . For single neurons , classification of trial type X versus trial type Y was based on the similarity of the calcium transient in each trial to the mean calcium transient for trial type X compared to trial type Y . Only the first second of the calcium signals following initial texture contact was considered since it reflected the minimum touch duration common across trial types ( Chen et al . , 2013b; 2015 ) . DV was equal to the dot-product similarity to the mean calcium transient for trial type X minus the dot-product similarity to the mean for trial type Y . Neurons discriminating above chance were identified using repeated permutations tests where trial type or stimulus labels were randomly shuffled . For each permutation test , a threshold corresponding to the shuffled distribution 95th percentile was calculated . Neurons , whose performance values were above the mean value of this threshold across 1000 permutation tests , were considered to be discriminating above chance . Comparison of discriminative neurons across cell types was performed using a χ2 test . Normal distribution was assumed for statistical comparison but not explicitly tested . We used linear discriminant analysis ( LDA ) for dimensionality reduction of neuronal population responses . Observations consisted of the △R/R values at a given time point for all neurons simultaneously recorded within an imaging field , thus representing the neuronal state space vector at this moment ( with each neuron representing one dimension ) , i . e . , representing a ‘snapshot’ of the state space vector trajectory during the given trial . Observations were considered for all n trials , separated into the N1 and N2 trials for the two chosen trial conditions C1 and C2 , respectively ( e . g . , Hit vs . CR or low- vs . high-amplitude whisking; see Table 1 ) . △R/R values were arranged in a matrix x with neurons as columns and trials as rows . The LDA procedure seeks to find a projection vector w such that the projections of the observations onto this axis , collected in the vector ( 13 ) y=wTx+ w0 , are best separated for the two chosen trial conditions . Maximal separation is defined as the maximal difference of the mean vectors μ1=1N1∑n∈C1xn and μ2=1N2∑n∈C2xn for C1 and C2 , respectively , normalized by the within-class scatter . The solution , known as Fisher’s linear discriminant ( Fisher , 1936; Safaai et al . , 2013 ) , is given by ( 14 ) wT=Sw−1 ( μ1−μ2 ) where SW−1 is the within-class covariance given by ( 15 ) SW−1= ∑n∈C1 ( xn−μ1 ) ( xn−μ1 ) T+ ∑n∈C2 ( xn−μ2 ) ( xn−μ ) T The bias is calculated as ( 16 ) w0=−12 ( wTμ1−wTμ2 ) Intuitively , this procedure finds the hyperplane in the state space ( orthonormal to the projection vector w and encompassing w0 ) that results in best separation according to Fisher’s criterion . To analyze the time courses of neuronal population dynamics during behavior trials , the LDA procedure was applied independently to each time point over 1-s periods before and after whisker-touch onset ( or licking onset in some cases ) . Only neurons identified as active in at least one imaging session were included in the LDA . For each individual trial we thereby obtained a time-dependent ‘linear discriminant’ variable LD ( t ) . The mean value LD¯ by definition is half of the distance between the projections of the mean vectors μµ1and μµ2 ( 17 ) LD¯ ( t ) =12 ( wTμ1+wTμ2 ) For whole-region analysis ( S1 or S2 ) we averaged LD values obtained from all imaging areas/planes . Inter-areal coordination as a function of time , termed LDCCS1:S2 , was determined by calculating the Pearson’s correlation coefficient between the population responses LDS1 and LDS2 for S1 and S2 , respectively , across all simultaneously imaged trials at each time point . To determine the specific contribution of S1S2 or S2S1 neurons to cross-areal coordination , a one-dimensional modified discriminant LD’ ( t ) was obtained for each area by shuffling the trial-by-trial calcium responses of S1S2 or S2S1 neurons , respectively , and then projecting the population vector onto the LDA axis determined from the non-shuffled population response . Cross-correlation of LD’S1 and LD’S2 yielded LDCC’S1:S2 . Shuffling was repeated 1000 times to obtain mean and standard error for LDCC’S1:S2 values . The change in S1-S2 correlation ( ∆LDCCS1:S2 ) was calculated as the mean LDCC’S1:S2 minus the unshuffled LDCCS1:S2 . Reductions in correlation strength thus show up as negative values . To control for trial shuffling of S1S2 or S2S1 neurons , trial shuffling was performed on an equal number of S1ND or S2ND neurons , repeated 1000 times , and ∆LDCCS1:S2 was calculated from the average cross-correlation . Comparisons of LDCCS1:S2 and ∆LDCCS1:S2 across trial conditions were performed using one-way repeated measures ANOVA . The variances of each the trial condition were tested using the F-test and determined to not be significantly different . | Behavior and cognition – the process of thought – emerge from computations that occur within vast networks of neurons in the brain . Within these networks , neurons may communicate with their neighbours in the same brain region as well as with distant counterparts in remote brain regions . Neuroscientists have studied these networks by measuring the activity of neurons within a single region or across the brain as a whole . However , it has not been possible to study long-distance communication between pairs of neurons in different brain regions . This has made it difficult to work out exactly what information brain regions exchange . Chen , Voigt et al . now overcome these challenges by developing a new microscope system that allows researchers to measure the activity of individual neurons in different brain regions at the same time . The system works alongside tracing techniques that map the connections between distant neurons . To demonstrate the new tools , Chen , Voigt et al . measured the activity of neurons in two areas of the mouse brain that monitor the whiskers . Mice brush their whiskers against an object to obtain information on its size , shape , texture and location . Two brain regions , called the primary and secondary areas of the whisker cortex , process this information and exchange messages back and forth . However , it was unclear what information these messages contain . Chen , Voigt et al . therefore trained mice to discriminate between coarse and fine sandpapers using their whiskers , and analysed the activity of the neurons that directly connect the two areas of the whisker cortex . The results revealed that although movement and sensory stimulation activated both the primary and secondary areas of the whisker cortex , the direct connections between these regions mainly exchange sensory information . This approach makes it possible to observe brain networks in an unprecedented level of detail . In the future , this technology will be extended to provide a more comprehensive view of how neurons communicate across brain areas . This will increase our understanding of how multiple areas of the brain all work together to produce the activity patterns that give rise to behavior . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Long-range population dynamics of anatomically defined neocortical networks |
How proteins control the biogenesis of cellular lipid droplets ( LDs ) is poorly understood . Using Drosophila and human cells , we show here that seipin , an ER protein implicated in LD biology , mediates a discrete step in LD formation—the conversion of small , nascent LDs to larger , mature LDs . Seipin forms discrete and dynamic foci in the ER that interact with nascent LDs to enable their growth . In the absence of seipin , numerous small , nascent LDs accumulate near the ER and most often fail to grow . Those that do grow prematurely acquire lipid synthesis enzymes and undergo expansion , eventually leading to the giant LDs characteristic of seipin deficiency . Our studies identify a discrete step of LD formation , namely the conversion of nascent LDs to mature LDs , and define a molecular role for seipin in this process , most likely by acting at ER-LD contact sites to enable lipid transfer to nascent LDs .
Lipid droplets ( LDs ) are cellular organelles that act as reservoirs to store neutral lipids for metabolic energy and cell membrane components ( for reviews , see Gross and Silver , 2014; Hashemi and Goodman , 2015; Pol et al . , 2014; Walther and Farese , 2012; Welte , 2015; Wilfling et al . , 2014a ) . Although most eukaryotic cells make LDs , the mechanism underlying the initial formation of LDs is mostly unknown . Excess storage of neutral lipids , such as triacylglycerols ( TGs ) and sterol esters , in LDs underlies many common metabolic diseases , such as obesity . Additionally , there is considerable interest in increasing cellular TG storage for industrial applications . In essence , LD formation corresponds to establishing an oil-in-water emulsion ( Thiam et al . , 2013b ) . In cells , this appears to be a well-organized process and can be conceptually separated into several distinct biochemical steps that are particularly evident when cells are incubated with exogenous fatty acids to induce them to form TG-containing LDs . In the initial step , enzymes utilize fatty acids and glycerolipids to synthesize neutral lipids , such as TGs , in the endoplasmic reticulum ( ER ) . TGs are subsequently packaged into nascent LDs that grow and are thought to bud from the ER to form initial LDs ( iLDs ) . How cells utilize proteins to harness the principles of emulsion physics to control these initial steps in LD formation is unclear . Much later , a second LD pathway responds to fatty acid loading of cells . A subset of the mature iLDs is converted to expanding LDs ( eLDs ) when specific TG synthesis enzymes [e . g . , glycerol-3-phosphate acyltransferase ( GPAT4 ) and acyl CoA:diacylglycerol O-acyltransferase 2 ( DGAT2 ) ] migrate from the ER to LDs via membrane bridges ( Wilfling et al . , 2013 ) . Targeting of GPAT4 to LDs and formation of these bridges depends on the Arf1/COP-I proteins ( Wilfling et al . , 2014b ) . Another enzyme , CTP:phosphocholine cytidylyltransferase 1 ( CCT1 ) , binds to phosphatidylcholine ( PC ) -poor surfaces of eLDs , where it becomes activated and catalyzes the synthesis of PC to coat eLD surfaces ( Krahmer et al . , 2011 ) . In this manner , eLDs grow dramatically through coordinated synthesis of core and surface lipids . The current study focused on the initial steps in LD formation and on the ER-localized protein seipin , which has been implicated in LD formation with an uncertain role ( for reviews , see Cartwright and Goodman , 2012; Fei et al . , 2011a ) . The seipin gene was initially identified by mutations in rare but severe forms of congenital generalized lipodystrophy ( Magré et al . , 2001 ) . Subsequently , LD morphology screens in Saccharomyces cerevisiae showed that the seipin homologue Fld1 is required for normal LDs; in its absence , cells have many small LDs or a few 'supersized' or giant LDs , depending on growth conditions ( Fei et al . , 2008; Szymanski et al . , 2007 ) . Seipin is an integral membrane protein with two transmembrane domains and a large , evolutionarily conserved ER luminal loop ( Agarwal and Garg , 2004; Lundin et al . , 2006 ) . Seipin forms oligomers ( Binns et al . , 2010; Sim et al . , 2013 ) . In yeast , seipin localizes to ER-LD contact regions ( Grippa et al . , 2015; Szymanski et al . , 2007; Wang et al . , 2014 ) , and yeast cells lacking seipin have abnormal LD formation ( Cartwright et al . , 2015; Grippa et al . , 2015; Wang et al . , 2014 ) , suggesting a role for seipin in organizing this process . Alternatively , seipin might affect LDs by regulating lipid metabolism ( Boutet et al . , 2009; Fei et al . , 2011b , 2008 , 2011c; Sim et al . , 2012; Szymanski et al . , 2007; Tian et al . , 2011; Wolinski et al . , 2015 ) or by causing defects in ER calcium homeostasis ( Bi et al . , 2014 ) . Here , we investigated seipin function in LD formation in Drosophila and mammalian cells . We found that seipin acts at a distinct step of LD biogenesis , after nascent LDs form during iLD formation . Our data suggest that seipin localizes to ER-LD contact sites and enables nascent LDs to acquire more lipids from the ER and grow to form mature iLDs . Without seipin , this process appears to be blocked , resulting in massive accumulation of small nascent LDs . The few LDs that do grow exhibit aberrant targeting of lipid synthesis enzymes , such as GPAT4 , involved in forming eLDs . The latter process likely explains the giant LD phenotype characteristically found in seipin-deficient cells .
As reported ( Fei et al . , 2011b , 2008; Szymanski et al . , 2007; Tian et al . , 2011 ) , we showed that depletion of seipin from Drosophila S2 cells by RNAi ( ~80% knockdown efficiency , Figure 1—figure supplement 1A ) led to formation of giant LDs after prolonged oleic acid treatment to induce LD formation ( Figure 1A , 24 hr ) . To determine the molecular basis of this phenotype , we examined when LD formation first appeared to be abnormal in seipin-deficient cells . Within 1 hr of adding oleic acid to cells , LDs in seipin-depleted cells were larger than those in control cells , although almost all LDs were less than 2 μm in diameter ( Figure 1A and B , top ) . Giant LDs ( diameter ≥ 2 μm ) first appeared in seipin knockdown cells ~5 hr after adding oleic acid and were more prevalent after 8 hr . In contrast , giant LDs were rare in control cells . Seipin-depleted cells also had fewer LDs than control cells , particularly at later times ( Figure 1B , bottom ) . Since the total areas with BODIPY-stained LD signal in optical sections of seipin-depleted cells and control cells at late time points were similar , the LDs likely coalesced in seipin-deficient cells . 10 . 7554/eLife . 16582 . 003Figure 1 . Seipin depletion alters LD morphology without affecting cellular lipid synthesis or composition in Drosophila S2 cells . ( A ) Time course of LD formation in control and seipin knockdown ( KD ) cells . S2 cells were treated with 1 mM oleic acid for the indicated times , and LDs were stained with BODIPY 493/503 . Bar , 10 μm . ( B ) Quantification of LD formation over time . Top , LD diameters; lines show median . Bottom , average LD numbers per cell area . n = 20 . ( C ) Seipin deficiency does not affect cellular glycerolipid synthesis . Cells were pulse-labeled with [14C]-oleic acid ( 100 μCi/μmol ) for indicated times . Phospholipids and neutral lipids were extracted and separated by TLC . The TLC plate was exposed on an imaging screen , and the intensities of bands were quantified with FIJI software . Values are presented as integrated density normalized to protein concentration . n=3 . ( D ) Seipin does not affect the flux and steady-state levels of lipids by lipidomics . Cells were labeled with [13C5]-oleic acid for 3 hr . Lipids were extracted , and lipid classes and species were identified by shotgun mass spectrometry–based lipidomics . n=3 biological replicates and 2 technical replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 00310 . 7554/eLife . 16582 . 004Figure 1—figure supplement 1 . Seipin does not affect cell lipid synthesis or composition . ( A ) Typical efficiency of seipin knockdown with dsRNA in Drosophila S2 cells . Primers for dsRNA and qPCR are described in Supplementary file 1 . n=3 . ***p<0 . 001 . ( B ) Glycerolipid synthesis in the ER is not affected by seipin knockdown in Drosophila S2 cells . Metabolic labeling of cells was as described in Figure 1C . Microsomes were purified , and lipids were extracted , separated with TLC , exposed on an imaging screen , and quantified with FIJI software . n=3 . No significant change was observed . ( C ) Sequence analysis of seipin knockout clone . Seipin knockout clone of SUM159 cell line was generated with CRISPR/Cas9-mediated genome editing . Genomic DNA was extracted and sequenced . The seipin knockout clones contain heterozygous mutations in exon 3 , with an 8-nucleotide deletion on one allele and a 1-nucleotide deletion on the other , leading to frame shift in both alleles . ( D ) Expression of seipin protein in wildtype and seipin knockout SUM159 cells was examined by western blot with antibody against endogenous seipin . No detectable seipin protein was found in the seipin knockout clone . ( E ) Seipin does not affect lipid levels in SUM159 cells . Lipids were extracted from cell homogenate and microsomes of wildtype or seipin knockout SUM159 cells . Lipid classes and species were identified with LC-MS based lipidomics . n=3 biological replicates and 2 technical replicates . PA standard curve shows that PA measurement was linear at a concentration range similar to that in samples , as low as 0 . 75 pmol . ( F ) No apparent PA accumulation at LD formation sites in seipin-depleted cells . SUM159 cells co-transfected with cherry-LiveDrop and GFP-PASS were imaged 30 min after adding oleic acid . The majority of the cherry and GFP signal does not overlap at the LD formation site . As a positive control , addition of PMA to control cells induced GFP-PASS distribution to the plasma membrane ( arrowhead ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 00410 . 7554/eLife . 16582 . 005Figure 1—figure supplement 2 . Seipin knockdown does not affect ER morphology or stress . ( A ) Cells were transfected with an ER marker cherry-sec61 ( Drosophila S2 ) or ss-BFP-KDEL ( SUM159 and human fibroblasts ) and imaged as Z stacks . Bars , 5 μm . Images from three independent experiments , containing ~45 cells for each genotype , from each cell type , were scrambled and blind scored by four independent observers . No difference in ER morphology between control and seipin-depleted cells was reported . X2=0 . 9999 . ( B ) Seipin depletion does not induce ER stress . WT or seipin-knockout SUM159 cells and fibroblasts from healthy controls or seipin loss-of-function patients were treated with or without Thapsigargin ( 1 μM , 6 hr ) and examined for ER stress . Xbp-1 splicing was determined by qPCR ( top ) , and protein levels of ER stress markers , BiP and Ire1 , were determined by western blotting . Representative results are shown . ( C ) Seipin depletion does not aggravate ER stress under fatty acid–loaded conditions . Wildtype or seipin knockout SUM159 cells were treated with oleic acid ( 0 . 5 mM ) , palmitate ( 0 . 1 mM ) , or thapsigargin ( 1 μM ) for the indicated times and examined for ER stress by qPCR ( top ) , and BiP and CHOP protein levels ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 005 The altered LD morphology during formation in seipin-deficient cells could result from changes in lipid synthesis , as suggested by some studies ( Boutet et al . , 2009; Fei et al . , 2011b , 2008 , 2011c; Tian et al . , 2011 ) . To examine this possibility , we used [14C]-oleic acid as a tracer to measure lipid synthesis in seipin-depleted cells . Rates of accumulation of TG , PC , and phosphatidylethanolamine ( PE ) were similar in control and seipin knockdown cells both in cell homogenates ( Figure 1C ) and microsomes ( Figure 1—figure supplement 1B ) , indicating similar rates of glycerolipid synthesis . Steady-state levels and synthesis rates of lipids in seipin-depleted cells showed no differences by high-resolution shotgun lipidomics ( Almeida et al . , 2015; Ejsing et al . , 2009 ) at 3 hr after adding [13C5]-oleic acid ( Figure 1D ) . To ensure the lack of differences in lipid synthesis was not due to residual seipin , we deleted seipin in human mammary carcinoma cells ( SUM159 ) by CRISPR/Cas9-mediated genome editing ( Ran et al . , 2013 ) ( Figure 1—figure supplement 1C ) . In this knockout clone , no seipin was detected ( Figure 1—figure supplement 1D ) . LC-MS/MS lipidomics did not show evidence of altered lipid metabolism ( for instance , levels of PC , PE or TG ) between control and seipin knockout cells without oleic acid ( Figure 1—figure supplement 1E ) . Although others reported an accumulation of phosphatidic acid ( PA ) in seipin-deficient yeast ( Fei et al . , 2011b , 2011c; Sim et al . , 2012; Tian et al . , 2011; Wolinski et al . , 2015 ) , we found no increase in cellular PA levels ( Figure 1D , Figure 1—figure supplement 1E ) . Additionally , expression of a probe , GFP-PASS ( Lu et al . , 2016 ) , that senses the accumulation of PA and possibly other anionic phospholipids ( Horchani et al . , 2014 ) did not accumulate at any specific site in cells during LD formation ( Figure 1—figure supplement 1F ) . Since seipin resides in the ER membrane , changes in lipid composition might occur in the ER but be masked in global analyses by lipids from other membranes . However , the ER morphology appeared normal in seipin-depleted cells ( Figure 1—figure supplement 2A ) , and we did not find evidence of altered ER lipid composition ( Figure 1—figure supplement 1B , E ) . Thus , the abnormal LD phenotype found in seipin depletion does not likely result from defects in whole-cell or global ER lipid metabolism . An alternative model of seipin’s function posits a direct role in organizing TG for LD formation ( Cartwright and Goodman , 2012 ) . LDs are thought to form in several steps . First , TG is synthesized in the ER by lipid synthesis enzymes , such as DGAT1 ( Yen et al . , 2008 ) . When the TG reaches a critical mass , it is thought to undergo phase separation and bud from the ER as iLDs . We hypothesized that seipin might act at one of these steps to facilitate iLD formation . Since LDs were abnormal by 1 hr of formation in seipin knockdown cells , we examined earlier stages . Such studies require a fluorescent probe sensitive enough to detect TG collections during the earliest steps of LD formation . These normally have too little lipid mass to be detected by conventional neutral lipid dyes ( e . g . , BODIPY ) that partition preferentially into the neutral lipid phase . We therefore developed a probe , which we named LiveDrop , that contains the membrane hairpin domain ( amino acids 160–216 ) of the glycerolipid synthesis enzyme GPAT4 fused to a fluorescent protein ( mCherry or eGFP ) ( Wilfling et al . , 2013 ) ( Figure 2—figure supplement 1A ) . Although full-length GPAT4 normally localizes only to eLDs ( Wilfling et al . , 2013 ) , LiveDrop localizes from the ER to all newly synthesized LDs upon their formation ( Figure 2A , C , control ) and identifies TG collections before BODIPY ( as seen in Figure 2C ) . LiveDrop localization to foci that become LDs depends on TG synthesis ( Figure 2F and Figure 2—figure supplement 1D ) , and in in vitro experiments using adhesive emulsion ( Thiam et al . , 2012 ) , we found that LiveDrop partitions preferentially into monolayers over bilayer membranes compared with a control protein Arf1 ( Figure 2—figure supplement 1C ) . Together , these data demonstrate that LiveDrop is a sensitive probe , identifying the earliest steps of LD formation . The finding that LiveDrop localizes to all newly synthesized LDs whereas the full-length GPAT4 localizes only to eLDs suggests a retention mechanism that maintains GPAT4 in the ER during LD formation . 10 . 7554/eLife . 16582 . 006Figure 2 . Aberrant accumulation of BODIPY-negative LiveDrop puncta in seipin-depleted Drosophila S2 cells . ( A ) Control or seipin knockdown cells expressing cherry-LiveDrop were treated with oleic acid immediately before the movie was taken at 30 s intervals for 30 min . Images were deconvolved as described in Materials and methods . Frames at indicated time points are shown . Green , BODIPY; red , cherry-LiveDrop . Arrow , BODIPY positive LDs; arrowhead , BODIPY negative puncta . Bar , 5 μm . ( B ) Quantification of numbers and size of LiveDrop objects that are positive or negative for BODIPY from movies taken at single optical plane . n=4 . ( C ) Accumulation of LiveDrop puncta and BODIPY over time . Graphs show line profiles for each channel at indicated lines . Objects accumulate BODIPY overtime in control but not seipin knockdown cells . ( D ) LiveDrop puncta in the absence of seipin are highly mobile . Cells expressing GFP-LiveDrop and BFP-KDEL were incubated with oleic acid for 30 min , before live-cell images were taken at max speed ( ~0 . 38 s/frame ) for 1 min . Movement of LiveDrop puncta are tracked with FIJI software . Representative tracks are shown . Bars , 5 μm . ( E ) Speed and distance of LiveDrop puncta movement were measured with FIJI . n= 3 cells , 8 puncta per cell . **p<0 . 005; ***p<0 . 001 . ( F ) Presence of LiveDrop puncta in seipin-knockdown cells depends on TG synthesis . Cells expressing cherry-LiveDrop were treated with oleic acid for 30 min in the presence or absence of various TG synthesis inhibitors . D1i: DGAT1 inhibitor; BrPal: bromopalmitate; TrC: triacin C . Bar , 5 μm . Quantification of cells with abnormal accumulations of BODIPY negative , LiveDrop puncta are shown in ( G ) . Representative results from two independent experiments are shown . 40 cells from each condition were quantified . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 00610 . 7554/eLife . 16582 . 007Figure 2—figure supplement 1 . Characterization of LiveDrop as an LD formation marker . ( A ) Illustration of cherry-LiveDrop . LiveDrop contains a hairpin protein that preferentially partitions to monolayer membranes at the interface of neutral lipids . ( B ) Initial LD formation in Drosophila S2 cells captured by lattice light-sheet microscopy . Entire cell volume of cells expressing cherry-LiveDrop were imaged every 4 s using a lattice light-sheet microscope by scanning the light-sheet along with the detection objective at 200 nm step size . Images are presented as middle slice and Z project at indicated times . LiveDrop clearly highlights forming LDs over time . Graph shows the increase in LD number with time . ( C ) LiveDrop preferentially partitions to monolayer but not bilayer membranes . Adhesive emulsion was formed by pushing two inside-out oil droplets together . Cherry-LiveDrop and Arf1-alexa488 were added as describe in Materials and methods . LiveDrop is enriched at oil-water interface , compared to Arf1 . Two examples are shown . Bar , 20 μm . ( D ) LiveDrop puncta accumulation in seipin depletion in Drosophila S2 cells depends on TG synthesis . Expression of seipin or seipin in combination with enzymes along TG synthesis pathways was inhibited with dsRNAs , before cells were transfected with cherry-LiveDrop and treated with oleic acid for 30 min . Green , BODIPY; red , LiveDrop . Arrow , BODIPY positive LDs; arrowhead , BODIPY negative LiveDrop puncta . Bar , 5 μm . Quantification of cells with abnormal accumulations of BODIPY-negative LiveDrop puncta are shown in ( E ) , using the same method described in Figure 2G . Representative result from two independent experiments is shown . 40 cells from each condition were quantified . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 00710 . 7554/eLife . 16582 . 008Figure 2—figure supplement 2 . LiveDrop is present in the LD fraction . SUM159 seipin KO cells transfected with Cherry-LiveDrop were fractionated by density centrifugation and protein enrichment in each fraction was determined by Western blot . Cherry-LiveDrop is enriched in the LD fraction . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 008 We used LiveDrop and live-cell imaging to study LD formation in Drosophila S2 cells . In control cells , when LD formation was induced by oleic acid , LiveDrop , which was initially diffuse within the ER , rapidly accumulated in small puncta that grew larger within minutes and subsequently became detectable with BODIPY ( Figure 2A , B , C , Figure 2—figure supplement 1B , and Videos 1 , 2 ) . In contrast , the ER of seipin-deficient cells had large numbers ( up to an average of 60 per cell , Figure 2B ) of small LiveDrop puncta that mostly did not grow large enough to stain with BODIPY ( Figure 2A , B , C , Figure 2—figure supplement 1B , and Videos 1 , 3 ) . Instead , they stayed small and highly mobile ( 0 . 23 ± 0 . 06 μm/s ) at the ER ( Figure 2D , E and Video 4 ) . However , a few larger , BODIPY-positive LDs formed in many of the seipin-depleted cells , possibly due to TG phase separation in a non-seipin organized process . BODIPY-stained LDs formed in control cells were relatively static ( 0 . 07 ± 0 . 02 μm/s ) , compared with LiveDrop puncta in seipin-depleted cells ( Figure 2D , E and Video 4 ) . To clarify the identity of LiveDrop puncta in seipin-knockdown cells , we inhibited TG synthesis with pharmacological inhibitors ( Figure 2F , G ) or by knocking down TG synthesis enzymes ( Figure 2—figure supplement 1D , E ) . These treatments abolished the LiveDrop puncta , indicating that they depend on TG synthesis and are likely very small TG collections associated with the ER . 10 . 7554/eLife . 16582 . 009Video 1 . Seipin depletion leads to aberrant accumulation of BODIPY-negative , LiveDrop in Drosophila S2 cells ( related to Figure 2A ) . Cells expressing cherry-LiveDrop were treated with oleic acid and imaged with spinning disk confocal microscopy as described in Figure 2A . Green , BODIPY; red , cherry-LiveDrop . Time is presented as min: sec . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 00910 . 7554/eLife . 16582 . 010Video 2 . Use of LiveDrop to visualize LD formation with lattice light-sheet microscopy in Drosophila S2 cells ( related to Figure 2A ) . ( A ) A control cell expressing cherry-LiveDrop were seeded onto 5 mm coverslips , and entire cell volume of cells were imaged with custom-made lattice light-sheet microscopy at 4 s intervals and with the light-sheet and objective scan step size of 200 nm . Raw images were deconvolved and 3-D visualization was done with Amira software . The beginning and end sections of the movie ( cyan ) show the light-sheet and objective scanning through slices of the cell . The middle section of the movie ( yellow ) presents the 3-D reconstitution of the cell volume . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01010 . 7554/eLife . 16582 . 011Video 3 . Visualization of LD formation with lattice light-sheet microscopy in control and seipin knockdown Drosophila S2 cells ( related to Figure 2A ) . Control or seipin knockdown cells expressing cherry-LiveDrop were imaged and processed as above . Video shows the 3-D projection of each cell section containing 1/3 of the cells thickness . Time is presented as min: sec . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01110 . 7554/eLife . 16582 . 012Video 4 . LiveDrop punta in seipin-depleted cells are highly mobile ( related to Figure 2D ) . Drosophila S2 cells expressing GFP-LiveDrop and BFP-KDEL were treated with oleic acid and imaged and deconvolved as described in Figure 2D . Time is presented as sec: msec . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 012 To assess evolutionary conservation of seipin function , we examined two mammalian cell models with seipin deficiency . In addition to the SUM159 seipin knockout cell line , we also studied primary fibroblasts from controls and patients with congenital generalized lipodystrophy type 2 , in which frame-shift mutations in BSCL2 ( mammalian seipin ) [patient 1: p . A212fsX231; patient 2: p . T109Nfs*5 ( Agarwal et al . , 2003 ) & p . P65Gfs*28] co-segregate with the disease phenotype . In each mammalian seipin-defective cell line , LiveDrop formed numerous , BODIPY-negative puncta upon oleate treatment ( Figure 3A , B , and Videos 5 , 6 ) , recapitulating the findings from seipin knockdown S2 cells . After prolonged oleic acid treatment ( 24 hr ) , giant LDs formed in each seipin deficiency model . However , unlike Drosophila S2 cells , there were very few giant LDs and more numerous small puncta that were weakly detected by BODIPY staining ( Figure 3C , D ) . Of note , ER morphology seemed normal in seipin-knockout cells ( Figure 1—figure supplement 2A ) , and there were no changes in ER lipid composition ( Figure 1—figure supplement 1E ) or in the activity of ER stress pathways ( Figure 1—figure supplement 2B , C ) . 10 . 7554/eLife . 16582 . 013Figure 3 . Seipin’s functions in initial LD formation and late LD phenotype are evolutionarily conserved . ( A ) LiveDrop puncta accumulate in SUM159 mammary carcinoma cells lacking seipin . Cells were incubated with oleic acid for 30 min before imaging . Red , cherry-LiveDrop; green , BODIPY . Arrows , BODIPY positive LDs; arrowheads , BODIPY negative LiveDrop puncta . Bar , 5 μm . Graphs show the line profile for each channel at dotted lines . ( B ) LiveDrop puncta accumulate in primary human fibroblasts from two subjects with lipodystrophy due to BSCL2 loss-of-function mutations ( patient 1 , p . A212fsX231; patient 2 , p . T109Nfs*5 & p . P65Gfs*28 ) . Bar , 5 μm . ( C ) LD phenotype in seipin knockout SUM159 cells at late stage of formation . Cells were treated with oleic aicd for 16 hr before imaging . Bar , 5 μm . ( D ) LD phenotype during late stage of formation in primary fibroblasts from healthy controls or lipodystrophy patients . Bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01310 . 7554/eLife . 16582 . 014Video 5 . Accumulation of LiveDrop puncta in seipin knockout SUM159 cells ( related to Figure 3A ) . Wildtype or seipin knockout SUM159 cells expressing cherry-LiveDrop were incubated with oleic acid immediately before imaging with spinning disk confocal microscopy . Images were taken at 10 s intervals for 30 min . Red , cherry-LiveDrop; green , BODIPY . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01410 . 7554/eLife . 16582 . 015Video 6 . Accumulation of LiveDrop puncta in human fibroblasts from a lipodystrophy patient with loss of function mutations in seipin ( related to Figure 3B ) . Fibroblasts from an apparently healthy control subject and a lipodystrophy patient with seipin mutations ( p . T109Nfs*5 & p . P65Gfs*28 ) were transfected with cherry-LiveDrop and incubated with oleic acid immediately before imaging with spinning disk confocal microscopy . Images were taken at 10 s intervals for 30 min . Red , cherry-LiveDrop; green , BODIPY . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 015 Across species , seipin contains highly conserved residues in the two transmembrane domains and the ER luminal loop , but the short N- and C-terminal cytosolic regions are not conserved . To test whether the conserved regions are important for its function in iLD formation , we overexpressed fluorescently tagged regions of seipin in a setting of seipin depletion in the S2 cell model . We knocked down seipin with dsRNA targeting the 3′UTR to avoid degradation of expressed seipin proteins . These knockdowns were somewhat less efficient than targeting the open reading frame , with ~65% of cells exhibiting accumulation of BODIPY-negative LiveDrop puncta ( Figure 4A , B ) . When full-length Drosophila seipin was expressed in the knockdown cells , the numbers of LiveDrop puncta were dramatically reduced in most cells . Tagging with a fluorescent protein localized to the N- or C-terminus resulted in similar rescue efficiency 30 min after oleate loading ( Figure 4A , B ) . Deletion of the N- or C-terminal cytosolic domain or both did not affect the rescue efficiency of seipin overexpression . In contrast , expression of the N- or C-terminus alone did not rescue the phenotype ( Figure 4A , B ) . These results suggest that the evolutionarily conserved regions of seipin ( i . e . , the ER luminal domain and transmembrane domains ) are required for seipin’s function in organizing iLD formation . In agreement with evolutionary conservation of seipin’s function , expression of human seipin in seipin-depleted Drosophila S2 cells completely abolished abnormal LiveDrop puncta accumulation , and mature iLDs formed normally in all transfected cells ( Figure 4C , D ) . 10 . 7554/eLife . 16582 . 016Figure 4 . The conserved transmembrane domains and ER luminal loop of seipin are important for its function in initial LD formation in Drosophila S2 cells . ( A ) Prevention of the seipin-depletion phenotype by expression of the transmembrane domains and ER loop , but not by the N- or C-terminus of Drosophila seipin . Gene knockdowns were performed with control or seipin dsRNAs targeting 3′UTR region of seipin . Cells were then co-transfected with GFP-LiveDrop ( green ) and cherry-tagged Drosophila seipin truncation mutants ( red ) . Seipin constructs used are illustrated on the right . Cells were treated with oleic acid for 30 min before imaging . Arrows , BODIPY-positive LDs; arrowheads , BODIPY-negative LiveDrop puncta . Bars , 5 μm . Quantification of cells with abnormal accumulations of BODIPY-negative LiveDrop puncta are shown in ( B ) . Representative results from two independent experiments are shown . 40 cells from each condition were quantified . ( C ) Expression of human seipin ( hSeipin ) prevents the seipin-depletion phenotype . Gene knockdowns were performed with control or dsRNAs targeting coding region of seipin . Cells were then co-transfected with cherry-LiveDrop ( red ) and human seipin ( green ) . Cells were treated with oleic acid for 30 min before imaging . Arrows , BODIPY-positive LDs; arrowheads , BODIPY-negative LiveDrop puncta . Bars , 5 μm . Quantification of cells with abnormal accumulations of BODIPY negative LiveDrop puncta are shown in ( D ) . Representative results from two independent experiments are shown . 40 cells from each condition were quantified . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 016 To visualize the molecular architecture of the LiveDrop puncta at nanometer resolution , we analyzed control and seipin-knockout SUM 159 cells by transmission electron microscopy . Consistent with images obtained by live cell imaging , wildtype cells contained numerous mature iLDs ranging in diameter between 200–500 nm , with an average of ~300 nm , after 60 min of oleate treatment ( Figure 5A , B ) . In contrast , seipin-deficient cells showed a large accumulation of nascent LDs of uniform size of less than 200 nm diameter , with an average of ~170 nm ( Figure 5A , B ) . These nascent LDs were most often located in close proximity to the ER . To determine whether the monolayer covering them was continuous with or separate from the ER membrane , we used electron tomography to examine thick sections of cells . We found that the nascent LDs in seipin-deficient cells were always in close proximity to the ER and that their surface monolayer was separated from the ER membrane ( Figure 5C , D and Videos 7 , 8 , 9 and 10 ) . Modeling of the tomograms indicated that ribosomes appeared to be absent from the space between the nascent LDs and the ER and that this space sometimes had electron density Figure 5D and Video 11 . Taken together , these data suggest that the LiveDrop puncta represent nascent LDs that are separate from , but in close contact with the ER , and that fail to grow to the size of normal , mature iLDs . 10 . 7554/eLife . 16582 . 017Figure 5 . Seipin-deficient cells accumulate nascent LDs that are close to the ER . ( A ) Control and seipin knockout SUM159 cells were fixed , embedded , and stained with uranyl acetate and lead citrate . Thin sections were imaged with transmission electron microscopy . Arrows , mature iLDs; arrowheads , nascent LDs . Close-up images ( bottom row ) are from different cell areas , representing different levels of proximity to the ER . Bars , 0 . 5 μm ( top ) and 0 . 2 μm ( bottom ) . ( B ) Quantification of number and size of LDs in control and seipin-knockout SUM159 cells . Total of 15 areas from each phenotype were quantified . ***p<0 . 0001 . ( C ) Electron tomography of a nascent LD in the seipin-knockout SUM159 cell . Thick sections of cells were imaged and reconstituted with IMOD . Serial sections from one example are shown . Numbers on top represent the relative position . Arrows , membrane contact zone; arrowheads , possible filamentous structures between nascent LDs and the ER . Bar , 0 . 2 μm . ( D ) Modeling of electron tomograms . Green , ER; blue , nascent LDs; orange , ribosomes; yellow , possible filamentous structure between nascent LDs and the ER . Rendering was performed with IMOD software . Movies of the tomograms are shown in Video 11 . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01710 . 7554/eLife . 16582 . 018Figure 5—figure supplement 1 . Seipin-deficient cells accumulate small , nascent LDs that are close to the ER . ( A ) Thin-section TEM images of human fibroblasts from control and lipodystrophy patients with seipin loss-of-function mutations . Arrows , mature iLDs; arrowheads , nascent LDs . Close-up images ( bottom row ) are from different cell areas , representing different levels of proximity to the ER . Bars , 0 . 5 μm ( top ) and 0 . 2 μm ( bottom ) . ( B ) Thin-section TEM images of control and seipin knockdown Drosophila S2 cells . Note the presence of pre-existing larger LDs ( white arrows ) along with the smaller nascent LDs ( arrowheads ) in seipin knockdown . Bars , 0 . 5 μm ( top ) and 0 . 2 μm ( bottom ) . ( C ) A gallery of small , nascent LDs in the proximity of the ER , generated from additional dual-axis ( a-c ) or single-axis ( d-f ) electron tomograms of SUM159 seipin knockout cells . The original tomograms are shown in Videos 8 , 9 and 10 . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01810 . 7554/eLife . 16582 . 019Video 7 . Electron tomography of LDs in control and seipin knockout SUM159 cells with dual axis ( related to Figure 5C ) . Bar , 100nm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 01910 . 7554/eLife . 16582 . 020Video 8 . Electron tomography of LDs in control and seipin knockout SUM159 cells with dual axis ( related to Figure 5C ) . Bar , 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 02010 . 7554/eLife . 16582 . 021Video 9 . Electron tomography of LDs in control and seipin knockout SUM159 cells with single axis ( related to Figure 5C ) . Bar , 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 02110 . 7554/eLife . 16582 . 022Video 10 . Electron tomography of LDs in control and seipin knockout SUM159 cells with single axis ( related to Figure 5C ) . Bar , 200nm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 02210 . 7554/eLife . 16582 . 023Video 11 . Modeling of electron tomograms shown in Video 7 ( related to Figure 5D ) . Green , ER; blue , nascent LDs; orange , ribosomes; yellow , possible filamentous structure between nascent LDs and the ER . Rendering was performed with IMOD software . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 023 Because our data suggested a role for seipin in the growth of nascent LDs to mature iLDs , we used live imaging to visualize the protein during LD formation . We fluorescently tagged the N-terminus of seipin with GFP at its endogenous locus in S2 cells by CRISPR/Cas9-mediated genome editing ( Housden et al . , 2015 ) . Without oleic acid , GFP-seipin formed distinct foci in cells that overlapped with an ER marker ( ss-BFP-KDEL; ( Dayel et al . , 1999 ) ( Figure 6A ) . These foci did not localize to any specific domain of the ER ( e . g . , tubules , sheets , tips ) ( Figure 6A and Figure 6—figure supplement 1A ) or three-way junctions marked by lunapark ( Figure 6—figure supplement 1B ) , and they moved rapidly along the ER ( 1 . 06 ± 0 . 22 μm/s; Figure 6B , Figure 6—figure supplement 1A and Video 12 ) . Upon adding oleic acid to form LDs , LiveDrop puncta initially were distinct from GFP-seipin foci . However , by 5 min , many LiveDrop puncta and seipin foci became co-localized . Once this occurred , the individual LiveDrop punctum was stabilized and began to grow in volume ( Figure 3C and Video 13 ) . 10 . 7554/eLife . 16582 . 024Figure 6 . Endogenous seipin forms discrete , mobile foci in the ER that co-localize with initial LDs during formation in Drosophila S2 cells . ( A ) Endogenous seipin was tagged with GFP at the N-terminus ( crGFP-seipin , green ) by CRISPR/Cas9 genome editing . Cells were transfected with BFP-KDEL ( red ) , and movies were captured with live-cell microscopy at 0 . 31 s frame intervals in the absence of oleic acid . The first frame is shown . Bar , 5 μm . ( B ) Seipin forms discrete foci that are highly mobile along the ER . Movement of GFP-seipin foci from movies taken in ( A ) was tracked with FIJI software . Total distance and average velocity were calculated . n= 3 cells , 8 foci per cell . ( C ) Seipin foci become co-localized with LiveDrop puncta to form nascent growing LDs . crGFP-seipin cells expressing cherry-LiveDrop were incubated with oleic acid for 5 min before live-cell images were taken ( ~2 . 05 s frame interval ) for 2 min . Images were de-convolved as described in Methods . Top: image of a whole cell at indicated time point; inlay: magnification of boxed area; arrowhead: an event where a LiveDrop punctum becomes associated with seipin and grows . A time series of this event is magnified under the image . Bottom: two more examples of similar events in other cells . Green , GFP-seipin; red , cherry-LiveDrop . Bar , 5 μm . ( D ) Seipin foci associated with LDs become less mobile . crGFP-seipin cells expressing cherry-LiveDrop were treated with oleic acid for 30 min . Live-cell images were taken at max speed ( ~0 . 27 s frame interval ) for 30 s , and the movement of seipin foci that are associated ( arrowheads ) or not associated ( arrows ) with LDs were traced and analyzed with FIJI . Representative tracks overlaid on image are shown ( middle ) . Total distance and average velocity were calculated ( bottom ) . n= 3 cells , 6 puncta per cell . Bar , 5 μm . **p<0 . 005 . ( E ) The majority of formed LDs are associated with seipin foci . crGFP-seipin cells were treated with oleic acid for 3 hr , and LDs were stained with LipidTox ( red ) . Images were taken as Z stacks , and middle slices are shown . Number of LDs with or without a seipin focus within a cell was quantified . Bar , 5 μm . n=10 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 02410 . 7554/eLife . 16582 . 025Figure 6—figure supplement 1 . Seipin foci are localized in the ER but not at specific ER domains . ( A ) crGFP-seipin ( green ) cells expressing BFP-KDEL ( red ) were imaged as described in Figure 3A . A montage of frames is show . Seipin can be found at tubules ( e . g . , 9 . 98 and 10 . 56 s ) , junctions ( e . g . , 3 . 69 and 7 . 56 s ) , and tips ( e . g . , 0 . 00 and 2 . 50 s ) . Arrowhead follows the movement of a single seipin focus overtime . ( B ) crGFP-seipin does not specifically localize at three-way junctions of the ER . crGFP-seipin cell line expressing cherry-lunapark , a marker for ER three-way junction , and BFP-KDEL , an ER marker , were imaged without fixation . crGFP-seipin does not colocalize with lunapark . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 02510 . 7554/eLife . 16582 . 026Video 12 . Endogenously tagged GFP-seipin forms foci that move dynamically along the ER in S2 cells ( related to Figure 6B ) . crGFP-seipin cells were prepared and imaged as described in Figure 6A . Time is presented as sec: msec . Seipin foci move very dynamically and do not localize to specific ER domain . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 02610 . 7554/eLife . 16582 . 027Video 13 . Endogenously tagged GFP-seipin encounters and stabilizes LiveDrop puncta ( related to Figure 6C ) . Cells were treated , imaged and deconvolved as described in Figure 6C . Frame rate: 2 s . Note that a LiveDrop punctum becomes stabilized and grows in volume after association with a seipin focus . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 027 These data suggest a model in which , rather than fixed structures , seipin foci are highly mobile and survey the ER for nascent LDs . When they encounter one , as detected by LiveDrop , they interact with the nascent LD to enable its growth to a mature iLD . Consistent with this model , seipin foci associated with nascent LDs became relatively immobile ( 0 . 52 ± 0 . 09 μm/s ) , whereas those not associated with nascent LDs were highly dynamic ( 0 . 89 ± 0 . 19 μm/s; Figure 6D ) . At later times during LD formation , seipin remained in close proximity to LDs . At 3 hr or later after adding oleic acid , almost every LD was associated with at least one seipin focus ( Figure 6E ) . However , there were many more seipin foci than LDs , and many seipin foci did not localize to LDs . We showed that seipin deficiency leads to abnormal LDs ( Figure 1A , B ) . It was unclear , however , how arrested growth early in LD formation leads to the phenotype of giant LDs in late stages of formation . Accumulation of nascent iLD intermediates in the early stages might indirectly result in fewer but larger LDs . We hypothesized that the LD expansion pathway , catalyzed by TG synthesis on eLDs , is aberrantly activated during seipin deficiency . In the LD expansion pathway , GPAT4 targets to selected LDs to initiate localized TG synthesis to expand LDs , and GPAT4 is thus a marker for eLDs ( Wilfling et al . , 2013 ) . Immunofluorescence of endogenous GPAT4 in seipin-depleted cells showed that GPAT4 targets to LDs as early as 10 min after LD formation is induced by oleic acid . In control cells , GPAT4 localizes to LDs much later ( Figure 7A ) ( not until 1 hr after induction of LDs , not shown ) . At 20 min after adding oleic acid , 18% of LDs in seipin knockdown cells were marked by GPAT4 , but only 1–2% in control cells were GPAT4-positive ( Figure 7A , C ) . After prolonged oleic acid treatment ( 8 hr ) , 8% of LDs in wild-type cells were GPAT4-positive , compared with 72% in seipin-depleted cells ( Figure 7B , C ) . In fact , most of the GPAT4 pool was localized to LDs in seipin-depleted cells , and little remained in the ER ( Figure 7B ) . This observation suggests that accumulation of nascent iLDs in seipin deficiency caused aberrant GPAT4 targeting to LDs , thereby initiating localized TG synthesis and LD expansion prematurely . 10 . 7554/eLife . 16582 . 028Figure 7 . Aberrant LD formation in seipin deficiency increases the expanding LD population in Drosophila S2 cells . ( A , B ) Earlier and increased GPAT4 targeting to initial LDs with seipin depletion . Control or seipin knockdown cells were incubated with oleic acid for indicated times . Targeting of endogenous GPAT4 to LDs during early ( A ) or late ( B ) stage of LD formation were determined by immunofluorescence . Images are presented as projected Z stacks ( A ) , or single optical slices ( B ) . Red , GPAT4; green , LDs , stained with BODIPY . Bars , 5 μm for ( A ) and 10 μm for ( B ) . Note that GPAT4 targeting to LDs takes place as early as 10 min after adding oleic acid in seipin knockdown . ( C ) Quantification of GPAT4-positive LDs at 20 min and 8 hr after adding oleic acid . n=30 cells . ***p<0 . 001 . ( D ) The formation of giant LDs in seipin deficiency depends on LD-localized TG enzymes . Expression of TG synthesis enzymes that are localized to LDs ( LD-isoforms ) or elsewhere ( Non-LD isoforms ) were inhibited with specific dsRNAs in combination with control dsRNA or dsRNA against seipin . LD phenotypes were determined 16 hr after LD formation was induced with oleic acid . Bar , 10 μm . Quantification of LD size from 10 cells in each treatment is shown . Lines show mean values . Yellow box indicates giant LDs of diameter > 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 028 If the model of aberrant eLD formation in seipin deficiency is correct , generation of giant LDs in seipin deficiency should depend on GPAT4 and the LD expansion pathway . To test this prediction , we examined the effect of seipin knockdown , combined with knocking down TG synthesis enzymes in the LD expansion pathway ( i . e . , GPAT4 , AGPAT3 , DGAT2 ) , and compared this with knockdowns of TG synthesis enzymes that do not mediate LD expansion ( i . e . , GPAT1 , AGPAT2 , DGAT1; Wilfling et al . , 2013 ) . Knockdown of enzymes of the LD expansion pathway abolished the giant LDs ( diameter > 2 . 5 μm ) of seipin deficiency , but knockdown of enzymes of the non-LD pathway did not affect LD size in seipin knockdown cells ( Figure 7D , yellow box ) . Rapid LD expansion and increased localized TG synthesis in seipin-deficient cells could lead to insufficient phospholipid surfactants on the surface monolayers of LDs to cover the hydrophobic cores . This , in turn , would lead to increased surface tension on these eLDs and likely to LD coalescence ( Krahmer et al . , 2011 ) . To test this prediction , we directly measured [14C]-oleic acid incorporation into TG and phospholipids in LDs from control or seipin-depleted cells . Consistent with our earlier measurements , incorporation of radioactivity into total cellular PC and TG was similar in wild-type and seipin knockdown cells ( not shown ) . However , incorporation of label into phospholipids ( PC and PE ) specifically on LDs was decreased in seipin-depleted cells at 5–8 hr after adding oleic acid to cells , indicating a relative deficiency of phospholipids in LDs of seipin-depleted cells at these late times . As a consequence , the ratio of surface phospholipids to TG ( i . e . , PC/TG or PE/TG ) decreased markedly at this time ( Figure 8A ) . This observation is consistent with the time course of LD formation ( Figure 1A ) , with giant LDs forming 5–8 hr after adding oleic acid , and suggests phospholipid deficiency on LDs results in their coalescence to form giant LDs . 10 . 7554/eLife . 16582 . 029Figure 8 . Seipin deficiency affects phospholipid composition on LDs at later stages of formation in Drosophila S2 cells . ( A ) Phospholipids are deficient on LDs in seipin-knockdown cells 8 hr after initiating formation . Control or seipin knockdown cells were pulse-labeled with [14C]-oleic acid ( 100 μCi/μmol ) for indicated times . LD fractions were purified by gradient centrifugation , and phospholipids and neutral lipids were extracted and separated by TLC . The TLC plate was exposed on an imaging screen , and the intensity of bands was quantified with FIJI . Values are integrated density normalized to protein concentration or ratio of integrated density of indicated lipid classes . n=3 . **p<0 . 005; ***p<0 . 001 ( B ) Targeting of CCT1 to LDs is increased and earlier with seipin depletion . Control or seipin-knockdown cells were incubated with oleic acid for 1 , 3 , or 8 hr , and localization of endogenous CCT1 was determined by immunofluorescence . Image shows a typical result at 8 hr . Red , CCT1; green , BODIPY . Bar , 5 μm . Targeting of CCT1 to LDs was quantified and is expressed as percentage of CCT1 signal on LDs over total cellular CCT1 signal . Data are presented as box plot with Tukey’s test . n=20 . *p<0 . 05; **p<0 . 01; ***p<0 . 001 . ( C ) Giant LD phenotype in seipin deficiency is rescued by adding PC or choline . Control , seipin , or CCT1 knockdown cells were treated with liposomes containing PC , PE or choline for 24 hr and then incubated with oleic acid for 16 hr . LD phenotype from representative cells are shown . Bar , 10 μm . Quantification of LD size from 10 cells in each treatment is shown . Lines show mean values . Yellow box indicates giant LDs of diameter > 2 . 5 μm . ( D ) Giant LD phenotype in seipin deficiency is worsened by blocking PC synthesis and ameliorated by blocking PE synthesis . Expression of CCT1 or ECT was inhibited by respective dsRNAs , in addition to control or seipin dsRNAs . Representative LD phenotypes after 16 hr of oleic acid treatment are shown . Bar , 5 μm . Quantification of LD size from 10 cells in each treatment is shown . Lines show mean values . Yellow box indicates giant LDs of diameter > 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 029 Our previous studies in Drosophila S2 cells established that PC-deficient LDs recruit CCT1 , the rate-limiting enzyme of PC synthesis , to the surface monolayer , where the enzyme becomes activated for maintaining PC homeostasis ( Krahmer et al . , 2011 ) . If phospholipids , specifically PC , are deficient on eLDs in seipin-deficient cells , CCT1 should have increased targeting to LDs during formation . Indeed , seipin-depleted cells exhibited earlier and markedly increased CCT1 targeting to LDs , compared with control cells ( Figure 8B ) . Further , when seipin-deficient cells were treated with PC liposomes or choline supplementation , no giant LDs ( diameter > 2 . 5 μm ) in seipin deficiency were formed ( Figure 8C , yellow box ) . In contrast , PC liposomes corrected the phenotype of CCT1 deficiency , and as expected , choline supplementation did not . We also altered phospholipid synthesis by knock down of CCT1 or CTP:phosphoethanolamine cytidylyltransferase ( ECT ) and tested the effect on seipin knockdown . ECT knockdown is predicted to shift the balance of phospholipid synthesis to PC from PE . As expected , CCT1 depletion aggravated , and ECT depletion abolished , the seipin knockdown phenotype ( Figure 8D ) . These data suggest that faster LD expansion in seipin-depleted cells leads to more eLDs with transient PC deficiency , and this likely leads to LD coalescence to form giant LDs .
In the current study , we identify a discreet step in LD formation—the conversion of small , nascent LDs ( <200 nm diameter ) to larger , mature initial LDs ( typically 300–500 nm diameter ) —and show that seipin acts at this step of LD biogenesis . Oligomers of seipin form highly mobile foci in the ER that interact with small , nascent LDs at ER-LD contact sites and enable their growth to mature iLDs , likely by promoting neutral lipid transport . Without seipin , the precursors of this step , small nascent LDs , accumulate to large numbers adjacent to the ER but fail to grow . With ongoing TG synthesis , some of these intermediates , or possibly other LDs arising from random coalescence of ER TG collections , appear to prematurely engage the eLD pathway , and undergo expansion mediated by LD-localized lipid synthesis enzymes ( e . g . , GPAT4 and CCT1 ) . As a late phenotype of seipin deficiency , giant LDs ( >1–2 µm diameter ) form eventually from these fewer abnormal eLDs , probably by coalescence due to a relative lack of PC on their surfaces . This model explains both aspects of the seipin-deficient phenotype—many small LDs and occasional giant LDs—that we and others ( Fei et al . , 2011b , 2008; Szymanski et al . , 2007; Tian et al . , 2011 ) have observed . Our data uncover a specific step in LD formation that was previously unrecognized , namely the growth and maturation of nascent LDs to mature iLDs . In wildtype cells , LD formation intermediates at this step are difficult to observe , likely due to their rapid transition through the process . However , with seipin depletion , progression through this step is blocked , resulting in the apparent accumulation of large numbers of intermediates ( small , nascent LDs ) , found in our studies by electron tomography , that fail to mature . These findings indicate that seipin functions downstream of initial lens formation and TG budding , at a more distal step in the maturation of nascent LDs . How does seipin function in nascent LD maturation ? Our studies of the tagged endogenous protein show that seipin localizes to multimeric foci that are highly mobile within the ER , as if scanning the ER for nascent LDs . We also found that some seipin foci encountered and associated with LiveDrop puncta ( nascent LDs ) , became relatively less mobile , and at this point iLD growth ensued . These observations are consistent with the model of seipin engaging nascent LDs and facilitating their growth , likely via transfer of additional neutral lipids , to mature iLDs . How seipin molecularly interacts with nascent LDs to facilitate LD growth is unknown . Our data are consistent with findings in yeast that implicated a role for seipin oligomers in early stages of LD formation ( Binns et al . , 2010; Cartwright et al . , 2015 ) . These studies indicated that purified seipin forms oligomers in a toroid-structure comprising approximately 8–12 seipin proteins ( Binns et al . , 2010; Cartwright and Goodman , 2012; Sim et al . , 2012 ) . However , from live-cell imaging , we estimate that the seipin foci include considerably more molecular units in cells ( not shown ) . Possibly seipin oligomers anchored in the ER mediate specific contacts with the surface monolayer of nascent LDs and enable the transfer of lipids ( such as TG ) to the nascent LDs , allowing them to grow ( see model , Figure 9 ) . Precisely how seipin mediates this transfer is currently unknown . 10 . 7554/eLife . 16582 . 030Figure 9 . Model for the role of seipin in LD formation . Seipin oligomers localize at the contact sites between nascent iLDs and the ER , enabling transfer of lipids ( such as TG ) to the nascent LDs to convert nascent iLD of <200 nm diameter to mature iLD of 300–500 nm diameter . Without seipin , lipid transfer is inhibited , and LD growth is arrested , leading to the accumulation of nascent iLDs . The maintenance of the contact site between nascent iLDs and the ER does not require seipin; an unknown protein is likely to be involved . DOI: http://dx . doi . org/10 . 7554/eLife . 16582 . 030 Our findings in Drosophila and mammalian cells indicate that seipin localizes to ER-LD contact sites and suggest that seipin is part of a protein machinery acting at ER-LD contact sites to promote LD maturation . At later time points in LD biogenesis , we observed that each LD was associated with at least one seipin punctum . Our findings are consistent with several studies in yeast indicating that seipin localizes to ER-LD contact sites ( Cartwright et al . , 2015; Grippa et al . , 2015; Wang et al . , 2014 ) . We also found that small nascent LDs almost always associate with the ER , even when seipin was absent , suggesting these LDs are still connected to the ER via contact sites . Consistent with the notion of ER-LD contact sites , the electron-tomograms revealed an absence of ribosomes between the nascent iLDs and the ER membrane . The close connections of the organelles and possible contact sites suggest that other proteins might be involved in establishing ER-LD contact sites and are still present in seipin deficiency . Our findings suggest that seipin’s function in converting nascent LDs to mature iLDs is likely its ancient , primary function . In support of this , seipin deficiency resulted in similar phenotypes of larger numbers of LiveDrop-positive foci in cells from flies , human mammary carcinoma cells , and human fibroblasts from seipin-deficient subjects . Additionally , the evolutionarily conserved transmembrane domains and luminal ER loop were sufficient for seipin’s function in LD formation . The role of the variable N-terminal domain is less clear . We found that the overexpressed N-terminus of Drosophila seipin localized to LDs ( Figure 4A ) , suggesting that this part of the protein , though not required for formation , may aid in interaction with LDs . In yeast , the N-terminus was found to be functionally important with respect to the timing of LD formation ( Cartwright et al . , 2015 ) . Although our findings indicate that seipin functions in an early step in LD formation , they also potentially explain the phenotype of giant LDs found in nearly all cells with seipin deficiency . In the absence of seipin , we found that large numbers of nascent iLDs accumulate , and some of these LDs aberrantly entered the LD expansion pathway . In support of this , lipid synthesis enzymes , such as GPAT4 and CCT1 , that are normally found only on late-forming eLDs were aberrantly targeted to iLDs at early time points , and this mislocalization of eLD proteins to iLDs was crucial for the development of the giant LD phenotype at later time points . Similar abnormal protein targeting to LDs was recently found in yeast lacking seipin orthologues ( Grippa et al . , 2015; Wang et al . , 2014 ) . Several other models have proposed a primary role for seipin in maintaining ER homeostasis , with associated indirect effects on LD formation . Our data do not support these models . For example , seipin was hypothesized to primarily regulate lipid metabolism in the ER , including phospholipid synthesis ( Fei et al . , 2008 , 2011c; Tian et al . , 2011 ) , and this in turn affects LD formation . We found no evidence to support this model , either in studies of glycerolipid synthesis rates or in the activities of specific enzymes ( such as GPAT; not shown ) . We also found no evidence of accumulation of PA in membranes of seipin-deficient S2 cells , a finding that has been reported by several groups for seipin deficiency in yeast ( Fei et al . , 2011c; Sim et al . , 2012; Tian et al . , 2011; Wolinski et al . , 2015 ) . We also found no evidence of seipin primarily affecting ER morphology or ER stress activation , which would normally be associated with defects in calcium homeostasis , as has been suggested ( Bi et al . , 2014 ) . Since seipin appears to act downstream of the budding of nascent LDs , it might be expected that ER functions are largely conserved and unaffected by seipin deficiency . In summary , we provide evidence that seipin functions at a previously unrecognized discrete step in LD biogenesis , enabling nascent LDs to grow to mature iLDs . Cells lacking seipin can still form LDs , but these LDs have irregular size and abnormal lipid and protein composition , which likely leads to cellular dysfunction with respect to storing and reclaiming lipids for cellular needs . In support of this notion , seipin deficiency in humans leads to severe generalized lipodystrophy , with a near absence of adipose mass ( Magré et al . , 2001 ) . Seipin therefore is a crucial part of the cellular protein machinery that serves to organize oil emulsification for fat storage . Our findings also suggest that seipin likely works in concert with other proteins that mediate ER-LD contact to enable the growth of nascent LDs .
Drosophila S2 cells were kindly provided from the laboratory of Ron Vale ( UCSF ) , SUM159 cells ( RRID:CVCL_5423 ) were from the laboratory of Tomas Kirchausen ( Harvard Medical School ) , and human fibroblasts were from the laboratories of David Savage ( Univ . Cambridge ) and Abhimanyu Garg ( UT Southwestern ) . Drosophila S2 cells were cultured in Schneider's Drosophila Medium ( Life Technologies , Carlsbad , CA ) containing 10% FBS and 50 unit/ml penicillin and 50 μg/ml streptomycin . SUM159 cell were maintained in DMEM/F-12 GlutaMAX ( Life Technologies ) containing 5% FBS , 100 unit/ml penicillin , and 100 μg/ml streptomycin , 1 μg/ml hydrocortisone ( Sigma , St . Louis , MO ) , 5 μg/ml insulin ( Cell Applications , San Diego , CA ) and 10 mM HEPES , pH 7 . 0 . Human fibroblasts were cultured in high-glucose DMEM GlutaMAX ( Life Technologies ) containing 10% FBS , 1 mM sodium pyruvate and 1x NEAA ( Life Technologies ) . Where applicable , cells were incubated with media containing 1 mM oleic acid complexed with 0 . 5% BSA for various times to induce LD formation . Knockdown of target genes by RNAi in S2 cells was performed as described ( Krahmer et al . , 2011 ) . Primers used to synthesize dsRNAs for RNAi are listed in Supplementary file 1 . Cells were analyzed after 4–8 days of RNAi treatment . Transfection of plasmids in S2 cells and mammalian cells was performed with Effectene Transfection Reagent ( Qiagen , Valencia , CA ) and lipofectamine 3000 ( Life Technologies ) , respectively , according to manufacturer’s instructions . Unless otherwise mentioned , transfections were done 18–24 hr before experiment . Plasmids used for transfection are listed in Supplementary file 1C . Polyclonal antibodies against dGPAT4 ( CG3209 ) and dCCT1 ( CG1049 ) were custom-made ( GenScript ) and affinity purified ( Krahmer et al . , 2011; Wilfling et al . , 2013 ) . Mouse polyclonal antibody against hSeipin ( A02 ) was from Abnova ( RRID:AB_627320; Taipei City , Taiwan ) . Antibodies against BiP , Ire1 , and CHOP were from Cell Signaling Technology ( Danvers , MA ) . Antibodies against Tubulin ( monoclonal ) and Calnexin ( polyclonal ) were from Sigma-Aldrich . A Cas9 and sgRNA co-expression plasmid was generated as described ( Housden et al . , 2015 ) , targeting close to the start codon of the Seipin gene ( sgRNA target sequence: GCGCAGCAGGATGTTCATGG ) . A donor construct was also generated by PCR amplifying homology arms flanking the intended insertion site from genomic DNA extracted from S2R+ cells ( genomic coordinates of homology arms ( genome release July 2014 ) : X: 2 , 619 , 765–2 , 620 , 308 and X:2 , 620 , 312–2 , 620 , 964 ) . These homology arms and GFP coding sequence were assembled into a custom backbone vector ( Housden et al . , 2014 ) using Golden Gate assembly . Expression and donor constructs were mixed in a 1:1 ratio and transfected into S2R+ cells with Effectene Transfection Reagent ( QIAGEN ) , according to manufacturer’s instructions . Four days after transfections , GFP-expressing cells were isolated using fluorescence-activated cell sorting ( FACS ) and cultured under standard conditions . Cas9 and gRNA expression plasmid ( pX459 ) was obtained from Addgene . The sequence 5′-GACTAAGGGTGGACGTGATC-3′ was used as a gRNA to direct Cas9 into the exon 3 of the seipin locus . pX459 plasmid ( 500 ng ) was transfected into 80 , 000 cells , following a published protocol ( Ran et al . , 2013 ) . Briefly , 1 day after transfection , cells were treated with puromycin for 3 days to select for transfected cells . Cells were re-plated into 150 mm dishes at clonal density . Individual colonies were then isolated in 24-well dishes . Screening of positive clones was performed by qPCR ( sense primer: 5′- GCATGTTCTTGGTCACCATTTC-3′ and antisense primer 5′-CAAATAGCAGGAGGCTAGA GAAG-3′ ) using power SYBR green ( Life Technologies ) . Genomic DNA of clones showing mRNA expression defect were extracted ( Quick Extract DNA extraction solution; Epicentre ) , and the genomic sequence surrounding the target exon of seipin were amplified by PCR ( sense 5′-GCAAAGAAGGTGTATGGATGGAC-3′ and antisense 5′-CCGGCCAGTCTCTTAT TACTC-3′ ) . PCR products were subcloned into a plasmid ( Zero Blunt TOPO PCR cloning kit; Life Technologies ) to validate the edited region of positive knockout clones by sequencing . Cells were lysed with lysis buffer ( 150 mM NaCl , 50 mM Tris-HCl , pH 7 . 4 , 1% Triton X-100 ) and denatured in Laemmli buffer at 37°C for 10 min . Proteins were separated on 10% SDS-PAGE gel and transferred to a PVDF membrane ( BioRad , Hercules , CA ) . The membrane was blocked with blocking buffer ( TBST containing 5% milk ) for 24 hr and incubated with BSCL2 antibody ( A02 , Abnova ) at 1:1000 dilution in blocking buffer for 24 hr . The membrane was then washed with TBST for 10 min x 3 times , and incubated in mouse secondary antibody ( Santa Cruz Biotechnology , Dallas , TX ) at 1:5000 dilution in blocking buffer . Membrane was washed again with TBST for 10 min x 3 times and revealed using the SuperSignal West Femto kit ( ThermoFisher , Waltham , MA ) . Total RNA was isolated using the RNeasy Kit ( Qiagen ) , according to the manufacturer’s instructions . Complementary DNA was synthesized using iScript cDNA Synthesis Kit ( Bio-Rad ) , and quantitative real-time PCR ( qPCR ) was performed in triplicates using SYBR Green PCR Master Mix Kit ( Applied Biosystems , Waltham , MA ) . Sequences of the qPCR primers used are listed in Supplementary file 1B . Microscopy was performed on spinning disk confocal ( Yokogawa CSU-X1 ) set up on a Nikon Eclipse Ti inverted microscope with a 100× ApoTIRF 1 . 4 NA objective ( Nikon , Melville , NY ) in line with 2x amplification . Fluorophores were excited with 405- , 488- , or 561 nm laser lines , and fluorescence was detected by an iXon Ultra 897 EMCCD camera ( Andor , Belfast , UK ) . Unless otherwise mentioned , bandpass filters ( Chroma Technology , ) were applied to all acquisitions . Where applicable , Z stacks of 0 . 13 μm slices were obtained with piezo Z-stage . For live-cell imaging , temperature , humidity and CO2 were controlled during imaging using a stage-top chamber ( Oko Lab ) . S2 cells cultured in 24-well plates were transferred to glass-bottom dishes ( MatTek Corporation ) or glass-bottom plates ( In Vitro Scientific ) coated with Concanavalin A ( Sigma ) 1 hr before live-cell imaging or immunofluorescence staining . SUM159 and human fibroblasts were cultured directly in 24-well glass-bottom plates . For live-cell imaging of initial LD formation , equal volumes of media containing 2x oleic acid were added to wells immediately before image acquisition . Where applicable , 0 . 5 μg/ml BODIPY 493/503 ( Life Technologies ) was added before and with oleic acid supplementation to stain LDs . To image late LD phenotypes , BODIPY was added 5 min before imaging . For immunofluorescence against dGPAT4 or dCCT1 , cells were fixed , permeabilized and immunostained as described ( Wilfling et al . , 2013 ) . Alexa Fluor 555 goat anti-rabbit ( Life Technologies ) was used as secondary antibody . Acquired images were processed and quantified manually with FIJI software ( http://fiji . sc/Fiji ) . All quantifications were done on raw images . Where necessary , deconvolution of images was performed using Huygens Professional 15 . 05 ( Scientific Volume Imaging ) with CMLE algorithm using a measured PSF for each wavelength . Lattice light sheet microscopy was done as previously described ( Chen et al . , 2014 ) . Control or seipin knockdown Drosophila S2 cells expressing Cherry-LiveDrop and GFP-sec61 were imaged from 5 mm coverslips using a lattice light-sheet microscope with a square lattice light-sheet . Images were acquired on a Hamamatsu ORCA-Flash 4 . 0 sCMOS camera , where each plane of the cell was excited sequentially using 488 nm and 560 nm laser lines exposed for 18 ms each , and with an excitation inner/outer numerical aperture of 0 . 44/0 . 55 respectively and a corresponding light-sheet length of 10 µm . At each time point , the cells were imagined by scanning the objective and the dithered light-sheet at 200 nm step sizes , thereby capturing a volume of ~50 µm x 50 µm x 20 µm ( corresponding to 512 x 512 x 101 pixels ) every 4 . 12 s ( which includes a 80 ms pause between time points ) . The image z-stacks for both channels were obtained between 2–10 min post exposure to 1 mM oleic acid by continuous imaging periods of ~20 min in duration . The deconvolution was performed using GPU compiled Lucy-Richardson deconvolution algorithm using a measured PSF for each wavelength for 15 iterations . Cells in petri dishes were fixed in 2 . 5% gluteraldehyde in 0 . 1 M sodium cacodylate buffer , pH 7 . 4 , for 1 hr . Buffer rinsed cells were scraped in 1% gelatin and spun down in 2% agar . Chilled blocks were trimmed and postfixed in 1% osmium tetroxide for 1 hr . The samples were rinsed three times in sodium cacodylate rinse buffer and postfixed in 1% osmium tetroxide for 1 hr . Samples were then rinsed and en bloc stained in aqueous 2% uranyl acetate for 1 hr followed by rinsing , dehydrating in an ethanol series and infiltrated with Embed 812 ( Electron Microscopy Sciences ) resin and baked over night at 60 C . Hardened blocked were cut using a Leica UltraCut UC7 . Sections ( 60 nm ) were collected on formvar/carbon-coated nickel grids and contrast stained with 2% uranyl acetate and lead citrate . They were viewed using an FEI Tencai Biotwin TEM at 80Kv . Images were taken on a Morada CCD using iTEM ( Olympus ) software . For tomography , 250 nm sections were collected on formvar/carbon-coated copper grids and contrast stained with 2% uranyl acetate and lead citrate , and 15 nm gold particles were used as fiducial markers . These were viewed using FEI Tecnai TF20 at 200 Kv , rotating angle from 60° to −60° . Data were collected using FEI Eagle 1X1 and reconstructed using IMOD ( Mastronarde , 2008 ) . Hairpin-cherry-bound lipid droplets were recovered as described ( Kory et al . , 2015 ) . Purified LD solution ( 20 µl ) was added to 4 µl of purified Arf1-alexa488 at 100 µM , and 1 µl of GTP at 10 mM and EDTA ( 2 mM final concentration ) , as described ( Thiam et al . , 2013a ) to generate the 'aqueous solution' . To prepare the 'triolein solution' , 15 µl of DOPC ( Avanti ) solubilized in chloroform ( 25 mg/ml ) was added to 5 µl of DOPE ( Avanti ) also in chloroform ( 25 mg/ml ) . The chloroform was evaporated under vacuum , and the phospholipid film was re-solubilized in 400 µl of triolein ( Avanti ) by vortexing . 100 µl of the 'triolein solution' containing phospholipids and 5 µl of the prepared 'aqueous solution' were used to generate water-in-oil droplets ( Kory et al . , 2015 ) bound by both the hairpin ( from purified LDs ) and Arf1 . Two aqueous drops close enough spontaneously adhere to form a phospholipid bilayer between them . Images of adhering drops were acquired by a Lecia SP5 confocal scanning microscope . Cells were harvested , washed with ice-cold PBS , resuspended in 1 ml of homogenization buffer ( 250 mM sucrose , 20 mM Tris-HCl , 1 mM EDTA , pH 7 . 4 ) in the presence of complete protease inhibitor tablet ( Roche ) , and lysed with a motor-driven Potter-Elvehjem homogenizer ( Wheaton ) . Lysates were sequentially centrifuged at 600 × g for 5 min and 8000 × g for 15 min to remove unbroken cells , nucleus and mitochondria . To isolate microsomes , the 8000 × g supernatant was subject to ultracentrifugation at 100 , 000 × g for 1 hr . The resulted pellet was then rinsed with TBS ( 20 mM Tris-HCl , 150 mM NaCl , pH 7 . 4 ) and resuspended for analysis . To purify LDs , the 8 , 000 × g supernatant was mixed with 2 M sucrose ( 1:1 ) , sequentially overlaid with 1 ml of homogenization buffer and 1 ml of TBS . The gradient was ultracentrifuged at 100 , 000 × g for 1 hr in a swinging bucket rotor ( TLS55 , Beckman-Coulter ) . The fat cake layer ( LDs ) floating on top the gradient was collected for analysis . Protein concentration was determined by BCA ( Thermo Scientific ) or Bradford assay ( Bio-Rad ) . To determine the enrichment of LiveDrop in the LD fraction , seipin KO SUM159 cells were transfected with Cherry-LiveDrop and incubated with oleate for 1 hr . Cells were then collected and homogenized , and the supernatant from 600 g , 5 min centrifugation was adjusted to 30% Optiprep and overlaid with series of Optiprep gradient . After centrifugation , 9 fractions were collected and analyzed by western blotting . To label intracellular lipids with radioactive isotopes , cells were incubated in triplicate with media containing 1 mM [14C] oleic acid ( 25 μCi/μmol ) for 0 . 5 , 1 , 3 , 5 , and 8 hr . Labeled cells were lysed by sonication , or fractionated by subcellular fractionation as described above . Lipids from same amount of protein were extracted with chloroform/methanol ( 2:1 ) ( Folch et al . , 1957 ) , separated on silica gel TLC plates ( Merck ) in a two-solvent system , with CHCl3/methanol/acetic acid/formic acid/H2O ( 70:30:12:4:1 ) for phospholipids and n-heptane/isopropyl ether/acetic acid ( 60:40:4 ) for neutral lipids . TLC plates were then exposed to an Imaging Screen-K ( Bio-Rad ) and revealed with a MPI gel-imaging system ( Bio-Rad ) . The revealed bands were quantified with densitometry . To quantify lipids in control and seipin knockout SUM159 cells with LC-MS/MS , lipids were extracted with chloroform/methanol ( 2:1 ) as described ( Folch et al . , 1957 ) . Extracted lipids were first separated on an Accucore C18 column ( 2 . 1 x 150 mm , 2 . 6 µm; Thermo ) connected to an Ultimate 3000 HPLC ( Thermo ) , using a binary solvent system ( mobile phase A: 50:50 ACN/H2O , 10 mM NH4HCO2 , 0 . 1% formic acid; mobile phase B: 10:88:2 ACN/IPA /H2O , 2 mM NH4HCO2 , 0 . 02% formic acid ) . The HPLC was connected on-line to a Q-exactive mass spectrometer equipped with an electrospray ionization source ( ESI; Thermo ) . Mass spectra were acquired in data-dependent mode to automatically switch between full scan MS and up to 15 data-dependent MS/MS scans . The 15 most intense ions from the survey scan were selected and fragmented with higher energy collision dissociation ( HCD ) with stepped normalized collision energies of 20 , 30 and 40 in negative mode and 25 and 35 in positive ion mode . Peaks were analyzed using the Lipid Search algorithm ( MKI , Tokyo , Japan ) , defined through raw files , product ion and precursor ion accurate masses . Candidate molecular species were identified by database ( >1 , 000 , 000 entries ) search of positive and negative ion adducts . Mass tolerance was set to 5 ppm for the precursor mass . Samples were aligned within a time window and results combined in a single report . Internal standards spiked in prior to extraction were used for normalization and calculation of the molar amounts of lipids . To measure metabolic flux in S2 cells with shotgun mass spectrometry , cells were pulse labeled with [13C5]-oleic acid ( Sigma ) for 3 hr . Cells were then homogenized and lipids were extracted . Quantification of lipids by high-resolution shotgun mass spectrometry was performed as described ( PMID: 25391725 ) ( Ejsing et al . , 2009 ) . Data are presented as means ± SD unless otherwise stated . Statistical significance was evaluated by two-tailed Student’s t-test or paired t-test and plotted with GraphPad Prism 6 software . | Living organisms often store energy in the form of fat molecules called triglycerides . Enzymes in a compartment of the cell called the endoplasmic reticulum catalyze the chemical reactions needed to make these triglycerides . The cell then stores the triglycerides in a different structure called the lipid droplet . Lipid droplets form from the endoplasmic reticulum in an organized manner , but little is known about the cellular machinery that gives rise to lipid droplets . A protein called seipin is thought to be involved in lipid droplet formation . Seipin resides in the endoplasmic reticulum and a shortage of this protein in cells leads to abnormal lipid droplets – that is , cells often have lots of tiny lipid droplets or a few giant ones . People who lack seipin lose much of their fat tissue and instead store fat in the wrong places , such as the liver . Now , Wang et al . have studied the seipin protein in insect and human cells grown in the laboratory . The experiments confirmed that cells that lack the seipin protein form lots of tiny dot-like structures containing triglycerides that fail to grow into normal-sized lipid droplets . These lipid droplets have different proteins on their surface , which may impair their ability to store fat . Wang et al . also discovered that in normal cells , the seipin protein is found at distinct spots in the endoplasmic reticulum . This distribution appears to allow seipin to come into contact with the small , newly formed lipid droplets and enable them to grow . Together these findings suggest that the seipin protein could form part of a molecular machine that allows more triglycerides to be added into newly formed lipid droplets causing the droplets to grow as normal . When seipin is not present the newly formed lipid droplets initially become stuck in a smaller form . As a consequence , a few of these tiny droplets later enter a different cellular pathway of lipid droplet expansion , which turns them into abnormally large lipid droplets . Future challenges will be to determine precisely how seipin enables newly formed lipid droplets to grow . It will also be important to confirm whether seipin works with other proteins as part of a molecular machine and , if so , to investigate how these proteins affect the formation and growth of lipid droplets . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2016 | Seipin is required for converting nascent to mature lipid droplets |
Clathrin-mediated endocytosis is an evolutionarily ancient membrane transport system regulating cellular receptivity and responsiveness . Plasmalemma clathrin-coated structures range from unitary domed assemblies to expansive planar constructions with internal or flanking invaginated buds . Precisely how these morphologically-distinct coats are formed , and whether all are functionally equivalent for selective cargo internalization is still disputed . We have disrupted the genes encoding a set of early arriving clathrin-coat constituents , FCHO1 and FCHO2 , in HeLa cells . Endocytic coats do not disappear in this genetic background; rather clustered planar lattices predominate and endocytosis slows , but does not cease . The central linker of FCHO proteins acts as an allosteric regulator of the prime endocytic adaptor , AP-2 . By loading AP-2 onto the plasma membrane , FCHO proteins provide a parallel pathway for AP-2 activation and clathrin-coat fabrication . Further , the steady-state morphology of clathrin-coated structures appears to be a manifestation of the availability of the muniscin linker during lattice polymerization .
The precise purpose of clathrin-coated vesicles was correctly deduced 50 years ago from thin-section electron micrographs ( Roth and Porter , 1964 ) . Noticing the very close apposition of a compacted lumenal protein load and a cytosol-oriented ‘bristle’ coat on opposite faces of striking membrane invaginations , it was astutely reasoned that the bulbous plasmalemma structures ferry macromolecules into the cell interior in a highly selective manner . While it was soon clear clathrin-mediated endocytosis is a rapid activity ( Brown and Goldstein , 1976 ) , and is importantly involved in many fundamental cellular processes ( McMahon and Boucrot , 2011; Brodsky , 2012; Kirchhausen et al . , 2014 ) , the regularity of the static coated structures in micrograph views masks the true underlying molecular dynamics and complexity . It is also challenging to identify the very earliest stages of clathrin coat assembly in thin sections because of the restricted size , ill-defined morphological features , and a lack of prominent membrane curvature . With the recent live-cell clarification of the linked temporal steps that define clathrin-coated structure initiation , growth , invagination and scission ( Kaksonen et al . , 2003; Taylor et al . , 2011 ) , it is clear that a large number of gene products are involved . The choreographed arrival and departure of these numerous proteins matches well their defined biochemical activities and the morphological progression of the bud ( Taylor et al . , 2011 ) . Yet , what exactly defines a restricted plasma membrane zone that is destined to become a clathrin-coated bud remains elusive . In this regard , the clathrin coats at the cell surface are distinct from many other intracellular cargo sorting coats in that the selected assembly site is not demarcated by deposition of a GTP-bound small GTPase ( Ren et al . , 2013 ) . The lipid phosphatidylinositol 4 , 5-bisphosphate ( PtdIns ( 4 , 5 ) P2 ) is required for clathrin assembly ( Boucrot et al . , 2006; Zoncu et al . , 2007; Nunez et al . , 2011 ) , but this phosphoinositide is broadly distributed over the inner leaflet of the plasma membrane . Ordered polymerization of a clathrin patch in fact appears to be a variable and probabilistic process ( Ehrlich et al . , 2004; Taylor et al . , 2011; Cocucci et al . , 2012; Brach et al . , 2014 ) . The nucleation stage displays the widest temporal variation and likely involves the largest number of protein cofactors ( Taylor et al . , 2011 ) . While the principal heterotetameric cargo-selective adaptor , designated AP-2 , is certainly important ( Mitsunari et al . , 2005; Cocucci et al . , 2012; Aguet et al . , 2013 ) , its initial occurrence at inchoate bud sites is paralleled by a sizable suite of early-arriving pioneers , including eps15 , intersectin , epsin , Fcho1 and Fcho2 , CALM ( Taylor et al . , 2011 ) and Necap 1 ( Ritter et al . , 2013 ) . The Fcho family ( Fer/Cip4 homology only proteins ) , collectively termed the muniscins ( Reider et al . , 2009 ) , is posited to act as internalization site founders on the basis of the biochemical properties of the proteins and the endocytic consequences of RNAi-mediated transcript silencing ( Henne et al . , 2010 ) . Here we show that munsicins are indeed consequential endocytic pioneers , but that clathrin-mediated endocytosis still persists in their absence ( Uezu et al . , 2011; Mulkearns and Cooper , 2012; Mayers et al . , 2013; Brach et al . , 2014 ) . We delineate a key biochemical aspect of the early operation of these endocytic proteins that regulates the rate and extent of clathrin-lattice fabrication at the cell surface .
An iterative TALEN-based approach to disrupt the FCHO1 gene in clone #64 cells was therefore used . A tract within exon 5 , encoding residues Lys24–Ser29 on the predicted EFC domain α2 helix ( Henne et al . , 2007 ) and housing an internal ApaI site , was selected for disruption ( Figure 4A ) . Similar to the FCHO2 TALEN lesions , the selected clones display an ApaI-resistant band ( s ) in addition to the two 387- and 150-bp cleavage products of the 537-bp PCR amplicon ( Figure 4B ) . Also , parallel AseI digests confirm retention of the disrupted FCHO2 alleles ( Figure 4C ) . Sequencing of a representative clone ( #64/45F1 ) , uncovers three differently disrupted and a single wild-type FCHO1 allele . Yet this does not represent a mixed cell population , as three independent subclones derived from this line all display the identical genotype ( Figure 4D ) . Because HeLa cells are aneuploid ( Bottomley et al . , 1969 ) , and chromosome 19 is duplicated in some HeLa lines ( Adey et al . , 2013; Landry et al . , 2013 ) , we interpret this to be TALEN-mediated disruption of three of the four FCHO1 alleles in HeLa SS6 cells . We used clone #64/1 . E ( designated clone 1 . E cells ) for all further analysis . More generally , our results highlight potential complications when using gene editing on cultured cell lines with non-diploid genomes . 10 . 7554/eLife . 04137 . 006Figure 4 . TALEN targeting of the FCHO1 locus in HeLa clone #64 cells . ( A ) Chromosomal location and genomic organization of the FCHO1 gene with pertinent details of TALEN design . The repeat variable diresidues ( RVD ) selective for the different deoxynucleotides are color-coded ( single letter amino acid notation ) . The position of the internal ApaI restriction site within the targeted exon is highlighted ( yellow ) . ( B ) ApaI restriction digest analysis of exon-specific FCHO1 PCR products from parental wild-type ( WT ) HeLa SS6 and selected FCHO1-targeting TALEN pair transfected cell lines . The PCR product marked with an asterisk represents the 71-bp deletion allele in the #64/45F1 line and derived subclones . The pool PCR was prepared from genomic DNA isolated from the TALEN-transfected population of clone #64 HeLa cells before clone selection . ( C ) AseI digests of FCHO2 amplicons to verify the FCHO2−/− genotype of all the second- and third-round clones . ( D ) Sequencing results from PCR product amplified from genomic DNA extracted from clone #64/1 . E cells . As indicated by the ApaI digest ( B ) , all the clones derived from clone #64/45F1 have the identical genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 00610 . 7554/eLife . 04137 . 007Figure 4—figure supplement 1 . FCHO1 transcript silencing does not alter the arrangement of surface clathrin-coated structures . ( A–C ) HeLa SS6 cells were either mock transfected ( A ) or transfected with siRNA oligonucleotides directed against the FCHO1 ( B ) or FCHO2 ( C ) mRNA ( Umasankar et al . , 2012 ) . Silenced cells were fixed and stained with an antibody ( mAb AP . 6 ) against the α subunit of the AP-2 adaptor complex . Note that the randomly-scattered similarly-sized clathrin-coated structures typical of control HeLa SS6 cells are also present in FCHO1-silenced cells . The FCHO2-depleted cells exhibit the characteristic enlarged and more sparsely arrayed clathrin patches , analogous to FCHO2 gene-edited HeLa cone #64 cells . The efficacy of FCHO1 and FCHO2 transcript silencing using this siRNA protocol has been documented ( Umasankar et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 007 Genotype–phenotype analysis corroborates diminished steady-state mRNA levels for both FCHO1 and FCHO2 ( Figure 3A ) . Biochemical pull-down assays further reveal exceptionally low levels of FCHO1 protein in clone 1 . E cells compared with K562 , which normally express this protein . Despite typical enrichment of cytosolic munsicins by affinity isolation on immobilized GST-EPS15 ( 595–889 ) ( Umasankar et al . , 2012 ) , there is no immunoreactivity for either protein in the TALEN-edited cells ( Figure 5A ) . Still , the AP-2 in each of the cell extracts binds avidly to the GST-EPS15 , while EPS15R remains within the supernatant fractions . These results thus resemble ( neuronal ) dynamin 1 expression in SK-MEL-2 cells; despite mRNA recognition there is no immunodetectable protein and this GTPase is ∼70-fold less abundant than the ubiquitious dynamin 2 ( Grassart et al . , 2014 ) . Yet , if ectopically expressed , either GFP-tagged full-length FCHO1 or Sgip1 are readily detectable in whole HeLa clone 1 . E cell lysates ( Figure 5B ) . By contrast , AP-2 and numerous other endocytic and trafficking proteins are present in equivalent amounts in HeLa and clone 1 . E cells ( Figure 5B ) . 10 . 7554/eLife . 04137 . 008Figure 5 . HeLa clone #64/1 . E cells have undetectable levels of muniscin proteins . ( A ) Samples of 100 μg of GST or GST-EPS15 ( 595–896 ) prebound to glutathione-Sepharose beads were incubated with cell lysates from HeLa SS6 , clone #64 , clone #64/1 . E and K562 cells . After washing , aliquots of the supernatant ( S ) and pellet ( P ) fractions were resolved by SDS-PAGE and stained with Coomassie blue or replicates transferred to nitrocellulose . Positions of the molecular mass standards ( in kDa ) are shown . Results from both an anti-FCHO1 mAb directed against the μHD and an affinity-purified polyclonal raised against the EFC domain and weakly cross-reactive with FCHO2 are shown . ( B ) Parental HeLa SS6 or clone #64/1 . E cells , mock transfected or transfected with plasmid DNA encoding GFP-FCHO1 or GFP-Sgip1 as indicated , were collected after 16 hr and whole cell lysates subject to SDS-PAGE analysis . Gels were either stained with Coommassie blue or transferred to nitrocellulose . Duplicate blots were probed with designated antibodies . Positions of the molecular mass standards ( in kDa ) are shown . ( C–J ) Representative images of HeLa SS6 ( C , E , G , I ) or HeLa clone #64/1 . E ( D , F , H , J ) cells after immunolabeling with anti-AP-2 mAb AP . 6 , affinity-purified anti-CALM antibodies , anti-clathrin heavy chain ( HC ) mAb X22 , affinity purified anti-DAB2 antibodies , anti-intersectin 1 mAb , affinity-purified anti-EPS15 antibodies , and anti-HRB serum . Arrows indicate the Golgi/endosomal population of clathrin present in both control and clone #64/1 . E HeLa cells . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 008 Not surprisingly , the overall disposition of AP-2-positive endocytic structures in the clone 1 . E cells ( Figure 5D ) differs from HeLa SS6 cells ( Figure 5C ) and resembles closely the clone #64 cells , from which they were derived ( Figure 2H ) . And the clathrin-coated regions in the clone 1 . E cells are populated with a full cohort of early-arriving clathrin coat pioneers despite the abnormal surface arrangement . At steady state , surface deposited AP-2 and clathrin are accompanied by CALM , DAB2 , intersectin 1 , EPS15 and HRB ( Figure 5C–J ) , all members of the pioneer module of endocytic proteins ( Taylor et al . , 2011 ) and found in HeLa cell clathrin-coated vesicles ( Borner et al . , 2012 ) . Thus , in obvious contrast to other observations ( Henne et al . , 2010 ) , neither clathrin nor AP-2 is completely soluble in the clone 1 . E cells in spite of the extremely low level of muniscins compared with unedited cells . Further , these results demonstrate plainly that congregation of EPS15/EPS15R and intersectin 1 at bud zones does not depend upon stoichiometric FCHO1/2 ( Henne et al . , 2010 ) . Our findings are congruent with results with yeast ( Boettner et al . , 2009; Reider et al . , 2009; Stimpson et al . , 2009 ) and nematode ( Mayers et al . , 2013 ) model systems that show deletion of the single muniscin expressed in these organisms is not lethal . Comparative quantitative analysis of the distribution of AP-2 in HeLa SS6 and the clone 1 . E cells indicates a sharp increase in the very smallest ( <60 nm2 ) coated structures on the ventral surface . In the parental cells , these account for <25% of the AP-2 signal while in the FCHO1/2-depleted cells these small puncta represent almost 45% of the total AP-2 . There is also an increase in the largest ( >600 nm2 ) clathrin assemblages , as previously seen in FCHO2 silenced HeLa cells ( Mulkearns and Cooper , 2012; Umasankar et al . , 2012 ) . After disruption of the FCHO1 and FCHO2 genes , 46% of membrane-associated AP-2 is present in structures >600 nm2 while only 29% of AP-2 occurs within this size category in the HeLa SS6 cells . Analogous results are obtained upon measuring the distribution of EPS15 in the two cell types . Engineered loss of muniscins therefore shifts the equilibrium of clathrin lattices at the cell surface to the very small and the largest forms . The ultrastructural basis for the altered surface distribution of clathrin assemblies in clone 1 . E cells can be appreciated in deep-etch EM replicas ( Figure 6 ) . At steady state , the ventral surface of the parental HeLa SS6 cell line typically exhibits a range of clathrin-coated assemblies and invaginated buds ( Figure 6A , B ) . Both solitary coated buds and spherical coated invaginations adjacent to regions of planar clathrin lattice are regularly dispersed over the membrane . By contrast , conspicuous expanses of plasma membrane devoid of polyhedral clathrin typify the clone 1 . E cells , and overall the number of deeply invaginated buds is reduced ( Figure 6C–E ) . Whereas abutting planar assemblages are abundant , individual , spatially isolated de novo forming coated buds are present in clone 1 . E cells , as well as small patches of flat polygonal assemblies ( Figure 6C , D ) . Beside the spatial effects of FCHO1/2 gene disruption , structural alterations within the assembled clathrin layer are also apparent . Higher magnification views reveal that , in contrast to the well ordered , principally hexagonal flat sheets in control HeLa cells ( Figure 6B , Figure 6—figure supplement 1 ) , the clone 1 . E cells display geometrically poorly ordered lattices ( Figure 6E , Figure 6—figure supplement 1 ) . So , despite having a complement of co-assembled AP-2 , EPS15 , DAB2 , CALM , intersectin 1 and HRB , the lattices do not appear to assemble as regularly as in control HeLa cells . 10 . 7554/eLife . 04137 . 009Figure 6 . Ultrastructural analysis of gene-edited clone #64/1 . E cell clathrin lattices . ( A–E ) Selected but representative deep etch-EM replicas revealing the glass-attached ventral surface of control HeLa SS6 ( A and B ) or clone #64/1 . E ( C–E ) cells . Polyhedral clathrin assemblies are pseudocolored purple . Higher magnification views of control ( B ) and 1 . E ( E ) cells highlight irregular assembly of clathrin trimers in the absence of FCHO1 and FCHO2 . The numerous surface-attached spherical buds in ( D ) are caveolae not clathrin-coated structures; they completely lack a characteristic polyhedral coat . Scale bars: 500 nm in A , C and D; 100 nm in B and E . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 00910 . 7554/eLife . 04137 . 010Figure 6—figure supplement 1 . Lattice assembly defects in the FCHO1/2-depleted HeLa clone #64/1 . E cell line . ( A–D ) Similarly sized planar clathrin lattice regions from freeze-etch replicas of HeLa SS6 control ( A and B ) or HeLa clone #64/1 . E cells ( C and D ) representative of the population of clathrin-coated structures in each cell line . Clathrin assemblages are pseudocolored purple while caveolae , with the characteristic striated surface , are colored yellow . The regular arrangement of parental flat lattices ( A and B ) is illustrated with superimposed broken lines ( orange ) , highlighting the parallel arrangement of adjacent rows of clathrin polyhedra . By contrast , in the FCHO1/2-deficient clone #64/1 . E cells ( C and D ) there are geometric abnormalities within the planar assemblies ( orange spheres and ellipses ) and linear columns of adjacent polyhedra are frequently twisted within the lattice . Areas of polygonal discontinuity do occur in the HeLa SS6 cell lattices , but these appear typically where the lattice begins to invaginate from the adjoining planar region or two planar sections adjoin ( arrows ) . Scale bar: 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 010 A two-minute pulse of fluorescent transferrin at 37°C labels surface AP-2 puncta and peripheral early endosomes in control HeLa cells ( Figure 7A ) , as the unsynchronized ligand fluxes through the cell surface into the endosomal compartment . The clone 1 . E cells have the larger , less regular clathrin-coated structures also labeled with transferrin at 2 min , although a diffuse pool of unclustered receptors is clearly evident on the cell surface ( Figure 7B ) . This confirms endocytic entry is rate limiting in FCHO1/2-depleted cells ( Uezu et al . , 2011; Mulkearns and Cooper , 2012 ) . A buildup of transferrin receptors at the cell surface may be explained , in part , by lattice defects and a reduced number of invaginated buds . After 10 min , the bulk of transferrin is within endosomes in both the control ( Figure 7C ) and clone 1 . E cells , but residual transferrin at the enlarged clathrin-coated structures and on the plasma membrane in the edited cells is apparent . ( Figure 7D ) . 10 . 7554/eLife . 04137 . 011Figure 7 . Clathrin-dependent cargo internalization in HeLa clone #64/1 . E cells . ( A and B ) Representative confocal optical sections of HeLa SS6 ( A ) or clone #64/1 . E ( B ) cells incubated with 25 μg/ml Alexa Fluor488-conjugated transferrin ( green ) for 2 min at 37°C before washing on ice . Fixed cells were stained with the anti-AP-2 α subunit mAb AP . 6 ( red ) . The borders of some cells in the field are outlined ( orange ) and color-separated views of the boxed regions are shown on the right . Scale bar ( for A–D ) : 10 μm . ( C and D ) Analogous images from control HeLa SS6 ( C ) or clone #64/1 . E cells ( D ) after a 10-min pulse of transferrin . ( E–H ) Confocal image of adherent ventral region of HeLa clone #64/1 . E cells transiently transfected with GFP-FCHO1 ( 1-889 ) ( green ) before addition of a 2-min pulse of 25 μg/ml Alexa Fluor568-labeled transferrin ( red ) as in A and B . Fixed cells were immunolabeled for AP-2 with mAb AP . 6 ( blue ) . A cell in each field is outlined ( orange ) and a low GFP-FCHO1 expressing cell ( arrow ) and adjacent untransfected FCHO1/2-depleted cells ( asterisks ) are indicated . Scale bar ( for E–L ) : 10 μm . ( I–L ) Medial plane optical section of GFP-FCHO1 ( 1–889 ) transfected HeLa clone #64/1 . E cells following a 10-min pulse of transferrin . The perimeter of a non-complemented clone #64/1 . E cell is indicated ( orange ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 01110 . 7554/eLife . 04137 . 012Figure 7—figure supplement 1 . Reconstitution of transferrin capture within clathrin-coated structures in HeLa clone #64/1 . E cells . ( A–C ) HeLa clone 1 . E cells transiently transfected with full-length GFP-FCHO1 ( 1–889 ) ( green ) were observed by total internal fluorescence microscopy after addition of 25 μg/ml Alexa Fluor568 transferrin ( red ) . A single representative frame from the time-lapse sequence is shown . Compared with the gene-edited cells , FCHO1 expressing cells rapidly concentrate labeled transferrin at the more numerous reconstituted ventral clathrin assemblies . Scale Bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 012 Given the coat protein densities within the aberrant clathrin-coated structures typical of clone 1 . E cells , and the accumulation of transferrin receptors at the cell surface , the extent of transferrin clustering is reduced relative to the HeLa cell control line . Because the transferrin receptor has a YXXØ-type YTRF internalization signal that depends on AP-2 for uptake ( Hinrichsen et al . , 2003; Motley et al . , 2003; Huang et al . , 2004 ) , our results suggest that AP-2 cargo capture may be compromised in the clathrin assemblies in clone 1 . E cells . The disparity in transferrin sequestration between steady-state clone 1 . E and ordinary clathrin-coated structures is better appreciated by transfecting GFP-tagged full-length FCHO1 into the gene-edited cells ( Figure 7E–L ) . Forced reexpression of FCHO1 normalizes the AP-2 arrangement on the ventral surface ( Figure 7F ) . A 2-min pulse of transferrin concentrates the ligand near these AP-2/FCHO1-positive spots ( Figure 7E–H ) . Adjacent untransfected clone 1 . E cells have the characteristic dispersed transferrin signal over the plasma membrane , and even strongly AP-2-positive structures do not concentrate transferrin similar to relatively dimmer AP-2 puncta in the transfected cells ( Figure 7E ) . Further analysis of transferrin binding to GFP-FCHO1-transfected HeLa clone 1 . E cells by total internal reflection microscopy ( Figure 7—figure supplement 1 ) corroborates the defect in transferrin clustering in the clone 1 . E cell line . After a 10-min pulse of transferrin , medial plane confocal optical sections illustrate that both the clone 1 . E cells and neighboring GFP-FCHO1-transfected cells accumulate the ligand in perinuclear endosomes ( Figure 7I–L ) , albeit internalization is more efficient in the presence of expressed FCHO1 . Stalled ligand-bound receptor still persists at the surface of the clone 1 . E cells . Overall , the data obtained from the TALEN-edited cells are fully consistent with the siRNA results in HeLa cells ( Mulkearns and Cooper , 2012; Uezu et al . , 2012; Umasankar et al . , 2012 ) . We conclude that clathrin-coat assembly and endocytosis does not cease entirely when cellular levels of muniscin proteins are extremely low . These proteins also do not appear obligatory for sustained clathrin lattice growth ( Cocucci et al . , 2012 ) , but rather impact the fidelity of the polyhedral construction . Transiently expressed GFP-FCHO1 clusters at surface puncta and changes the general AP-2 configuration compared with the surrounding non-transfected clone 1 . E cells ( Figure 7F , G , 8B , B′ and Figure 8—figure supplement 1 ) . By contrast , transfected GFP alone is diffusely cytosolic , and has no effect on the abnormal clathrin patches in the gene-edited cells ( Figure 8A , A′ ) . This effect of FCHO1 on clathrin coat arrangement is not restricted to gene-edited lines . The native topology of surface clathrin lattices in cultured MCF-7 cells , a breast adenocarcinoma , is even more exaggerated and extensively abutting than the gene-edited clone 1 . E cells ( Figure 8—figure supplement 2 ) . A considerable size distribution of clathrin assemblies can thus occur naturally in different cell types ( Maupin and Pollard , 1983; Gaidarov et al . , 1999; Akisaka et al . , 2003; Ehrlich et al . , 2004; Grove et al . , 2014; Sochacki et al . , 2014 ) . Strikingly , temporary introduction of GFP-FCHO1 ( or GFP-FCHO2 ) into MCF-7 cells switches the AP-2 assemblies to considerably smaller puncta ( Figure 8—figure supplement 2 ) . Because the altered clathrin distribution in the clone 1 . E ( and MCF-7 ) cells is extremely penetrant and robust , we used complementation transfection experiments to delineate the domain ( s ) of FCHO1 required to revert the morphology to the wild-type pattern . 10 . 7554/eLife . 04137 . 013Figure 8 . The muniscin unstructured linker sector regulates clathrin-coated structure topology . ( A–H′ ) Representative single confocal optical sections of HeLa clone #64/1 . E cells transiently transfected with GFP ( A and A′ ) , GFP-FCHO1 ( 1–889 ) ( B and B′ ) , GFP-FCHO1 EFC domain ( 1–275 ) ( C and C′ ) , GFP-FCHO1 μHD ( 609–889 ) ( D and D′ ) , GFP-FCHO1 ( 1–609 ) ( E and E′ ) , GFP-FCHO1 ( 1–416 ) ( F and F′ ) GFP-FCHO1 ( 265–609 ) ( G and G′ ) or GFP-FCHO1 ( 1–889; Δ316–467 ) ( H and H′ ) . Fixed cells were stained for AP-2 using the anti-AP-2 α-subunit mAb AP . 6 . Merged channel images ( A–H ) and the corresponding AP-2 α subunit channel alone ( A′–H′ ) are shown . Transgene-dependent refashioning of the irregular 1 . E cell clathrin-coated structures to a more uniformly dispersed pattern is indicated ( arrows ) . Scale bar for all panels: 10 μm . ( I ) Cartoon diagram of the overall domain organization of FCHO1 with the relative locations of the various N-terminally-tagged truncation and deletion constructs tested shown schematically . The minimal region necessary to correct the surface clathrin morphology is boxed . ( J–J′ ) Selected confocal section of HeLa clone #64/1 . E cells transfected with GFP-Sgip1 ( 1–514 ) analyzed as in ( A–H′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 01310 . 7554/eLife . 04137 . 014Figure 8—figure supplement 1 . Comparative expression of GFP-tagged FCHO1 protein fragments in HeLa cells . Total cell lysates from HeLa SS6 cells transiently transfected with the indicated GFP-tagged FCHO1 protein fusions were resolved by SDS-PAGE and transferred in duplicate to nitrocellulose . Blots were probed with either affinity-purified anti-GFP or affinity-purified anti-FCHO1 EFC domain antibodies . Positions of the molecular mass standards ( in kDa ) are shown on the left . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 01410 . 7554/eLife . 04137 . 015Figure 8—figure supplement 2 . Oversized and clumped clathrin-coated structures in MCF-7 cells are normalized by ectopic FCHO1 or FCHO2 expression . ( A–D ) . MCF-7 cells transiently transfected with either fill-length GFP-tagged FCHO1 ( residues 1–889 ) ( A ) , GFP-FCHO2 ( residues 1–810 ) ( B ) , GFP-FCHO1 EFC + linker ( residues 1–609 ) ( C ) or GFP-FCHO1 EFC domain ( residues 1–275 ) ( D ) were fixed and stained with an antibody ( mAb AP-6 ) against the α subunit of the AP-2 adaptor complex ( red ) . Forced expression of FCHO1 or FCHO2 remodels the endogenous clathrin-coated structures into more uniform arrays , as visualized by AP-2 ( A , B and C ) . Although full-length GFP-FCHO1 or FCHO2 ( green ) concentrate at AP-2 positive surface spots , GFP-FCHO1 EFC + linker fully corrects AP-2 patterning while being diffusely associated with the plasma membrane in a featureless manner ( C ) similar to the EFC domain alone ( D ) . Enlarged views of the boxed regions are shown on the lower right . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 015 On expressing alone the lipid-binding EFC domain ( Henne et al . , 2007; Edeling et al . , 2009 ) in clone 1 . E cells , while membrane associated , it neither clusters at clathrin-coated structures nor modifies their appearance ( Figure 8C , C′ ) . By contrast , the GFP-μHD does concentrate at plasma membrane clathrin assemblages , because the domain functions as an interaction hub ( Reider et al . , 2009; Dergai et al . , 2010; Henne et al . , 2010; Mulkearns and Cooper , 2012; Umasankar et al . , 2012 ) but , again , does not refashion these atypical coated zones ( Figure 8D , D′ ) . Yet expression of proteins containing the polypeptide stretch linking the EFC domain to the μHD reestablishes a more regular arrangement of AP-2-positive sites in clone 1 . E cells ( Figure 8E–G′ , Figure 8—figure supplement 1 ) . Systematic truncation and deletion experiments better pinpoints the active region of the FCHO1 linker ( Figure 8I ) . In various proteins , inclusion of FCHO1 residues 339–416 promotes the redistribution of AP-2 into more a uniform and evenly dispersed form over the ventral plasma membrane . Deletion of residues 316–467 from full-length GFP-tagged FCHO1 prevents complementation of the clathrin distribution despite deposition of the transfected protein at clathrin pincta ( Figure 8H , H′ ) . The linker section of FCHO1 , but neither the EFC domain nor μHD , also repatterns the bulky clathrin in MCF-7 cells to more regularly dispersed arrays ( Figure 8—figure supplement 2 ) . Forced expression of the N-terminal linker-containing segment of Sgip1 ( residues 1–514 ) has a similar consequence in clone 1 . E cells , in spite of the absence of an EFC domain ( Figure 8J , J′ ) . Together , these experiments show that the central linker in FCHO1 , FCHO2 and Sgip1 , despite likely being intrinsically disordered , is required to revert the organization of clathrin-coated structures to the more uniform format typical of control cells . A remarkable aspect of this activity is that at very low expression levels , the linker can alter the steady-state organization of coated regions without being enriched at assembly zones , and can even operate when expressed as a soluble protein . The linker tract is the most divergent portion of the muniscin family members , with generally <25% identity overall . Yet alignments of this region of Fcho1 , Fcho2 , Sgip1 ( Katoh , 2004 ) reveal they share small blocks of phylogenetically conserved residues , particularly rich in acidic side chains and phospho-modifiable residues ( Figure 9A ) . As the linker regions from FCHO1 , FCHO2 and Sgip1 can each correct the HeLa clone 1 . E cell coat morphology , it seems likely that these blocks of common residues are functionally important . In fact , post-translational phosphorylation events are enriched at protein–protein interaction surfaces ( Nishi et al . , 2011 ) , and high-throughput mass spectrometry identifies numerous phosphosites within the shared region of the linker domain ( Figure 9A ) . 10 . 7554/eLife . 04137 . 016Figure 9 . Functionally significant phylogenetic conservation within the muniscin central linker domain . ( A ) Muniscin–AP-2 interactions . T-Coffee ( Notredame et al . , 2000 ) generated multiple sequence alignment of the phylogenetically conserved linker region within muniscin members . Amino acid regions of Fcho1 from selected species: Homo sapiens ( Hs; NP_001154829 ) , Rattus norvegicus ( Rn; XP_006252925 ) , Python bivittatus ( Pb; XP_007436694 ) , Fcho2: Hs ( NP_620137 ) , Danio rerio ( Dr; NP_001018617 ) , Fcho2-like ( Fcho2l ) : Aplysia californica ( Ac; XP_005111676 ) , the single FCHO member in Drosophila melanogaster ( Dm; NP_001097723 ) , and Daphnia pulex ( Dp; EFX87825 ) , as well as Sgip1: Mus musculus ( Mm; NP_659155 ) and Dr ( XP_005165952 ) are shown with appropriate residues numbers indicated . Identical residues are highlighted in magenta , highly similar residues in pale pink and conservatively substituted amino acids in yellow . The location of mass-spectrometry authenticated phosphosites ( Hornbeck et al . , 2012 ) are shown ( red font ) . ( B ) Samples of 100 μg of GST , GST-FCHO1 ( 316–467 ) or GST-ARH ( 180–308 ) immobilized on glutathione-Sepharose were incubated with rat brain cytosol , washed and samples of each supernatant ( S ) and pellet ( P ) fraction separated by SDS-PAGE . Replicate gels were either stained with Coomassie blue ( left ) or immunoblotted with the designated antibodies ( right ) . The positions of the molecular mass standards ( in kDa ) are indicated on the left . ( C ) Schematic illustration of the overall chain and domain composition of the AP-2 adaptor complex . ( D ) Samples of 100 μg of GST , GST-FCHO1 ( 316–467 ) , GST-FCHO2 ( 314–444 ) or GST-Sgip1 ( 77–214 ) immobilized on glutathione-Sepharose were used in pull-down assays with the purified AP-2 heterameric core complex as in ( B ) . The identity of the large subunit trunk polypeptides ( α and β2 subunits ) and the myc-tagged μ2 subunit is confirmed on immunoblots with mAb clone 8 , mAb 100/1 and mAb clone 31 , respectively . ( E ) Aliquots of 100 μg of GST or either 25 μg or 100 μg of GST-EPS15 ( 595–896 ) or GST-EPS15 ( 595–896/Δ617–636 ) immobilized on glutathione-Sepharose were incubated with soluble lysate from K562 cells . After washing , portions of the supernatant and pellet fractions were analyzed as in ( B ) . FCHO1 antibody 1 is a mAb while antibody 2 is an affinity-purified antibody that also recognizes FCHO2 weakly . Non-specific cross-reactive bands are indicated with asterisks; reactivity of the GST-EPS15 fusion proteins with the anti-FCHO1 and anti-FCHO2 antibody preparations is also indicated with asterisks . Notice the decreased FCHO1 and FCHO2 binding upon deletion of the minimal μHD sequence within the EPS15 C-terminal region that correlates with reduced clathrin association . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 01610 . 7554/eLife . 04137 . 017Figure 9—figure supplement 1 . Cytosolic AP-2 and the Necap 1 PHear domain ( residues 1–133 ) bind to FCHO1 and FCHO2 . ( A ) Samples of ∼200 μg of GST , GST-FCHO1 ( 316–467 ) , GST-FCHO2 ( 314–740 ) or GST-Sgip1 ( 77–214 ) immobilized on glutathione-Sepharose were used in pull-down assays as in with rat brain cytosol as in Figure 9B . HC; heavy chain . The anti-Necap 1 antibody has extensive non-specific reactivity with the GST-fusion proteins ( yellow asterisks ) but , compared with the GST control , there is little loss of the Necap 1 from the supernatants after incubation with GST-muniscin linker fusions , as is seen for AP-2 subunits . Thus Necap 1 interacts only weakly with the FCHO1 linker segment . ( B ) Sample of ∼200 μg of GST , GST-FCHO1 ( 316–467 ) or GST-EPS15 ( 595–740 ) immobilized on glutathione-Sepharose were used in pull-down assays with rat brain cytosol alone or supplemented with 5 μM or 25 μM purified AP-2 α appendage . The marked recovery of the α appendage together with the GST-EPS15 fusion is not paralleled in the GST-FCHO1 linker pellets , and competition with soluble intact AP-2 is not evident . The reduced apparent abundance of the μ2 subunit in the pellet fractions from the GST-FCHO1 linker pull-down assays is due to comigration of the μ2 subunit with the GST fusion protein . ( C ) Samples of 250 μg of GST or 50 μg or 250 μg of GST-Necap 1 PHear domain ( residues 1–133 ) immobilized on glutathione-Sepharose were incubated with K562 cell Triton X-100 lysate . After washing , aliquots of the supernatant ( S ) and pellet ( P ) fractions were resolved by SDS-PAGE and either stained with Coomassie blue ( top ) or transferred to nitrocellulose in replicate . Blots were probed with affinity-purified antibodies directed against FCHO1 or FCHO2 or with a mAb ( C-8 ) the α subunit of AP-2 . FCHO1 , FCHO2 and AP-2 all show dose-dependent interactions with the Necap 1 PHear domain ( Ritter et al . , 2004 , 2007 , 2013 ) , although FCHO1 clearly displays the highest apparent affinity . In general , FCHO2 shows a weaker capacity to correct the clathrin distribution phenotype in HeLa clone #64/1 . E cells , which is correlated with poorer binding to AP-2 and Necap 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 017 Several distinct lines of evidence indicate that the central linker acts with AP-2: First , the anomalous transferrin clustering at coated patches noted in clone 1 . E cells indicates muniscins could impact AP-2 cargo engagement . Second , in GST pull-down assays , the minimal FCHO1 linker ( residues 316–467 ) binds physically to cytosolic AP-2 ( Figure 9B ) . All four subunits of the heterotetramer sediment with the GST-FCHO1 linker , but engagement of AP-2 is not as efficient as the C-terminal portion of the cargo-selective monomeric adaptor ARH that binds directly to the β2 subunit appendage ( He et al . , 2002; Mishra et al . , 2005 ) , and clathrin binding is limited . We find no evidence for the linker polypeptide engaging lipid modifying enzymes ( PIP5KIγ [Figure 9B] , PLD , DGKδ1 [not shown] ) that could , in principal , affect clathrin coat assembly without physically being enriched at coat assembly zones . Fused to GST , the conserved central segments of FCHO2 ( residues 314–444 ) and Sgip1 ( residues 77–214 ) also bind to cytosolic AP-2 , although the interaction with FCHO2 is comparatively weaker ( Figure 9—figure supplement 1A ) . The independently folded α- and β2-subunit appendages of AP-2 ( Figure 9C ) can engage FCHO1 ( Umasankar et al . , 2012 ) . Yet titrating an excess of α appendage into pull-down assays does not alter the extent of soluble AP-2 heterotetramer engagement , so the major contact surface for the AP-2 on the FCHO1 linker tract does not appear to interact with the α appendage ( Figure 9—figure supplement 1B ) . In the same experiment , the added α appendage does bind to the C-terminal region of EPS15 and concomitantly diminish cytosolic AP-2 association , in a dose-dependent manner ( Figure 9—figure supplement 1B ) . Binary interaction assays reveal that the purified tetrameric core of AP-2 binds physically to the linker regions of FCHO1 , FCHO2 and Sgip1 ( Figure 9D ) , with relative affinities similar to interactions with intact cytosolic AP-2 ( Figure 9—figure supplement 1A ) . The sequence-related linkers of the muniscin members thus contact the AP-2 core directly . The other currently known interaction partner for the FCHO1/2 linker segment is Necap 1 ( Ritter et al . , 2013 ) . Soluble Necap 1 in brain cytosol binds only very weakly to the GST-FCHO1 linker and not to the analogous portions of FCHO2 and Sgip1 ( Figure 9—figure supplement 1A ) . However , when this association is assayed using immobilized GST-Necap 1 PHear domain ( residues 1–133 ) and K562 cell lysates , a strong interaction with FCHO1 is apparent ( Figure 9—figure supplement 1C ) . This suggests either that Necap 1 in brain cytosol is in an inhibited conformation or that the optimal binding site upon the FCHO1 linker lies outside of the minimal region ( residues 316–467 ) necessary for correction of the clone 1 . E cell clathrin phenotype . A clue to the mechanistic connection between FCHO1/2 and AP-2 comes from additional biochemical studies . The muniscin μHD is a densely wired endocytic interaction hub , physically contacting the pioneer proteins Eps15/R , intersectin , Dab2 , Hrb and CALM ( Reider et al . , 2009; Henne et al . , 2010; Mulkearns and Cooper , 2012; Umasankar et al . , 2012 ) . The C-terminal portion of EPS15 is the highest affinity μHD binding partner ( Figure 5A ) . A limited sequence tract between residues 595 and 636 represents the principal interaction region in EPS15 ( Umasankar et al . , 2012 ) . Deletion of part of this region ( residues 617–636 ) within the GST-EPS15 ( 595–896 ) fusion protein has a strong inhibitory effect on both FCHO1 and FCHO2 interactions from K562 cell lysates ( Figure 9E ) . AP-2 also binds to the C-terminal third of EPS15 ( Figure 5 , Figure 9—figure supplement 1B ) , but with lower apparent affinity and utilizing the tandemly arrayed Asp-Pro-Phe ( DPF ) tripeptide repeats positioned distal to the μHD-binding region ( Benmerah et al . , 1996; Iannolo et al . , 1997 ) . Consequently , the association of cytosolic AP-2 with the GST-EPS15 ( Δ617–636 ) is not different from the binding to the whole GST-EPS15 ( Figure 9E ) . Yet there is a strong decrease in the amount of clathrin associated with the immobilized Δ617–636 fusion , which parallels the diminished engagement of FCHO1 and FCHO2 . The majority of the soluble clathrin remains in the Δ617–636 assay supernatant fractions . Because AP-2 cannot engage clathrin trimers productively without undergoing allosteric activation ( Kelly et al . , 2014 ) , it is as if FCHO1/2 promote conformational rearrangement of AP-2 to facilitate clathrin binding . The role of the FCHO1 linker polypeptide was further explored using a tailored fusion to the cytosolic aspect of Tac ( CD25 ) , a type I transmembrane protein ( Uchiyama et al . , 1981 ) . Transient expression of Tac in HeLa SS6 cells leads to surface accumulation because this α chain of the IL-2 receptor is poorly internalized ( Humphrey et al . , 1993 ) . At steady state , individual low-level-expressing cells have a prominent surface pool of Tac and an internal population that overlaps partly with the Golgi-region marker GPP130 ( Linstedt et al . , 1997 ) and with the trans-Golgi network ( TGOLN2/TGN46 ) . The intracellular population in these cells thus represents primarily the biosynthetic trafficking of new Tac molecules en route to the surface ( Figure 10A , A′ ) . Cells expressing high levels of the Tac transgene have , in addition , a second internal accumulation of Tac , which colocalizes well with a pulse of internalized transferrin ( Figure 10A ) . Switching a Tac-linker ( FCHO1 residues 265–609 ) hybrid-encoding plasmid for Tac leads to similar but not identical expression and spread in HeLa SS6 cells ( Figure 10B–B″ ) . Increased localization of Tac-linker with the transferrin endocytic marker indicates that this customized fusion is more rapidly internalized from the cell surface than the native Tac protein . Analogous results are obtained upon expressing a Tac-FCHO1 μHD ( residues 609–889 ) hybrid ( Figure 10C , C′ ) . Consequently , we find that ectopic Tac and Tac-fusion proteins display a dichotomous distribution of surface and internal pools , the composition of the latter differing depending on the level of Tac protein driven by the transfected plasmid DNA and on the amino acid sequence of the cytosolic portion . 10 . 7554/eLife . 04137 . 018Figure 10 . Trafficking of exogenous Tac and Tac-FCHO1 fusion proteins in HeLa cells . ( A–A′ ) Maximal projection ( A ) and a selected medial plane ( A′ ) of deconvolved confocal image z-stacks of HeLa SS6 cells transiently transfected with a Tac-encoding plasmid . Prior to fixation , the transfected cell population was pulsed with 25 μg/ml Alexa Fluor568 transferrin ( red ) for 10 min . Fixed cells were then stained with a mAb ( 7 G7B6 ) directed against the lumenal domain of Tac ( green ) , anti-GPP130 antibodies ( blue ) . Relative accumulation of the Tac protein in the Golgi ( cyan color ) and transferrin-positive endosomes ( yellow ) varies in different individual transfected cells . The cell perimeter of two selected , non-transfected cells is outlined ( orange ) . Scale bar ( for all panels ) : 10 μm . ( B–C ) Maximal projection ( B and C ) or selected medial planes ( B′ and C′ ) from deconvolved z-stacks acquired from HeLa SS6 cells transfected with either a Tac-FCHO1 linker ( residues 265–610; B , B′ ) or a Tac-FCHO1 μHD ( residues 609–889; C , C′ ) . Fixed cells were processed identically to ( A ) . Accumulation of he Tac-FCHO1 chimeras in the ER , as marked by nuclear envelope staining , is also apparent in some cells . The cell perimeter of two selected , non-transfected cells in each panel is outlined ( orange ) . ( B″ ) Enlargements of the color-separated channels from the boxed region in B′ showing the colocalization of the transfected Tac-FCHO1 linker fusion with both the Golgi marker GPP130 and internalized transferrin . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 018 Two conspicuous changes in clathrin coat organization occur when just the full FCHO1 linker is expressed in HeLa SS6 cells in a membrane-tethered form . First , assembled AP-2 , normally restricted entirely to puncta on the inner leaflet of the plasma membrane , becomes mistargeted to juxtanuclear regions ( Figure 11A–A″ ) , located similarly to the Golgi and endosome compartment markers that overlap with intracellular Tac . The ectopic AP-2 is accompanied by colocalized EPS15 ( Figure 11A ) . We infer that the high local density of the Tac-FCHO1 linker , as it traverses the biosynthetic pathway following ER export or concentrates on maturing endosomes , promotes anomalous recruitment of AP-2 . Yet , when a corresponding Tac fusion with the FCHO1 μHD ( residues 609–889 ) is transfected , very strong EPS15 and intersectin 1 ( not shown ) deposition on the Golgi/endosome region occurs , but strikingly this is not paralleled by AP-2 binding ( Figure 11B–B″ ) . The effect of overexpression of a Tac-linker + μHD ( FCHO1 residues 316–889 ) hybrid appears similar to the linker alone , albeit with more robust EPS15 binding accompanying juxtanuclear AP-2 ( Figure 11C–C″ ) . Visualizing the Tac-FCHO1 fusion-expressing cells with an anti-Tac mAb ( Figure 11D–D″ ) clarifies that the juxtanuclear pool of endocytic proteins is indeed recruited to Tac-positive internal structures . Thus , EPS15 ( and intersectin 1 ) recruited by the Tac-anchored FCHO1 μHD concentrated on Golgi/endosome membranes does not attract cytosolic AP-2 , but AP-2 associated with these compartments through the membrane-linked linker polypeptide results in deposition of EPS15 . Impressively , the massing of AP-2 and EPS15 upon perinuclear biosynthetic and endocytic organelles in Tac-FCHO1 linker ( Figure 12A–C ) or Tac-FCHO1 linker + μHD ( Figure 12D–F ) expressing cells occurs in the absence of any conspicuous PtdIns ( 4 , 5 ) P2 enrichment in these compartments , as judged by a GFP-tagged PLCδ1 PH domain sensor selective for PtdIns ( 4 , 5 ) P2 . 10 . 7554/eLife . 04137 . 019Figure 11 . An artificial transmembrane FCHO1 linker protein misrecruits AP-2 onto internal membrane structures . ( A–A″ ) Selected medial ( A and A′ ) or basal ( A″ ) optical sections of deconvolved confocal z-stacks collected from HeLa SS6 cells transiently transfected with Tac-FCHO1 linker ( residues 265–609 ) . Fixed cells were stained with anti-AP-2 α subunit mAb AP . 6 ( green ) and affinity purified anti-EPS15 antibodies ( red ) and mounted in Hoechst 33342 to label DNA ( blue ) prior to imaging . Aberrant accumulation of AP-2 and EPS15 adjacent to the nucleus ( arrows ) in Tac-transfected cells ( asterisks ) correlates with smaller clathrin-coated puncta in the basal optical section of the same cell population ( A″ ) . Scale bar ( for A–C″ ) : 10 μm . ( B–C″ ) Chosen middle ( B , B′ , C and C′ ) or maximal projection ( B″ and C″ ) optical sections of HeLa SS6 cells transfected with Tac-FCHO1 μHD ( residues 609–889 ) or Tac-FCHO1 linker + μHD ( residues 265–889 ) and prepared analogously to ( A–A″ ) . The Tac-fused μHD produces dramatic relocalization of EPS15 ( B ) , diminishing this protein in surface AP-2-positive puncta ( B″ ) yet AP-2 does not relocate similarly . With ectopic expression of the FCHO linker + μHD , prominent accumulation of irregular intracellular AP-2 and EPS15 in a juxtanuclear locations again correlates with diminished surface clathrin spots , which are again relatively deficient in EPS15 because of the massive deposition of this endocytic pioneer component upon intracellular membranes . ( D–D″ ) Single ventral optical section of HeLa SS6 cells transfected with Tac-FCHO1 linker + μHD ( residues 265–889 ) and stained with anti-Tac mAb ( 7 G7B6; green ) and affinity-purified anti-EPS15 antibodies ( red ) . Relocalization of EPS15 occurs only in Tac overexpressing cells ( arrows ) , and the dose-dependent massing of EPS15 in the perinuclear region limits the amount of EPS15 present in surface clathrin-coated structures . The limiting membrane of the Tac-expressing cells is delineated ( orange ) . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 01910 . 7554/eLife . 04137 . 020Figure 12 . PtdIns ( 4 , 5 ) P2 is not enriched at sites of intracellular AP-2 accumulation . ( A–F ) HeLa SS6 cells cotransfected with a mixture of either Tac-FCHO1 linker ( residues 265–609 ) ( A–C ) or Tac-FCHO1 linker + μHD ( residues 265–889 ) ( D–F ) and GFP-PLCδ1 PH domain encoding plasmids were fixed and stained with anti-AP-2 α subunit mAb AP . 6 ( red ) and affinity purified anti-EPS15 antibodies ( blue ) . The GFP fluorescence in the medial region of the Tac-linker expressing cells ( C ) represents the soluble pool of this fluorescent lipid probe adjacent to the nucleus . Note also the decrease in size and more regular arrangement of the AP-2- and EPS15-positive clathrin coated surface structures in the Tac-linker ( A ) and Tac-linker + μHD ( D ) expressing cells , identified by mislocalized deposition in a juxtanuclear position ( arrows ) . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 020 Second , the Tac-linker fusion , when located on the plasma membrane , reconfigures the apparent dimensions of the clathrin-coated structures on the ventral surface relative to adjacent untransfected HeLa SS6 cells ( Figures 11C and 12A ) . This overall shift to small , more uniformly grouped puncta also happens with transfection of the Tac-linker + μHD ( Figures 11D″ and 12D ) , but not with the Tac-μHD fusion ( Figure 11B″ ) , even though the relocation of EPS15 is so dramatic in these cells that the protein becomes depleted from the surface AP-2-positive clathrin assemblies . The phenotypic consequence of forced expression of the tailored Tac-FCHO1 chimeras is even more notable in the muniscin-depleted clone 1 . E cells . While Tac alone has no noticeable effect on the characteristic distribution of AP-2 and EPS15 in these cells ( Figure 13A–D ) , the Tac-FCHO1 linker promotes deposition of a fraction of these proteins onto juxtanuclear membranes and remodels the abnormal surface clathrin patches into more homogenous and regularly scattered assemblies ( Figure 13E–H ) . By contrast , the Tac-FCHO1 μHD gives rise to noteworthy sequestration of EPS15 on juxtanuclear structures , as occurs in the parental HeLa SS6 cells , but again without accompanying AP-2 recruitment or refashioning of the surface clathrin structures in the clone 1 . E cells ( Figure 13I–L ) . Transient expression of the Tac-FCHO1 linker + μHD amalgamates the separate effects of the Tac-linker and Tac-μHD proteins , with marked AP-2 and EPS15 intracellular congregation and a drop in the dimensions of clathrin-coated puncta on the plasma membrane ( Figure 13M–P ) . Parallel experiments utilizing the variously transfected HeLa clone 1 . E cells and stained for Tac expression ( Figure 13—figure supplement 1 ) recapitulate all these findings . That the ectopic transmembrane anchored FCHO1 linker alters the morphologic characteristics of surface clathrin patches toward uniform small puncta strengthens the proposition that enlarged coats in the clone 1 . E cells originate directly from limiting cellular muniscin concentrations . 10 . 7554/eLife . 04137 . 021Figure 13 . Forced expression of the Tac-FCHO1 linker fusion restores clathrin coat distribution in gene-edited HeLa cells . ( A–P ) Basal ( A , E , I , M ) and medial ( B , F , J , N ) confocal sections from deconvolved z-image stacks from HeLa clone #64/1 . E cells immunolabeled with antibodies against AP-2 ( mAb AP . 6; green ) and EPS15 ( red ) and Hoechst 33342 for DNA ( blue ) . The clone #64/1 . E cells were short-term transfected with Tac ( A–D ) , Tac-FCHO1 linker ( residues 265–609 ) ( E–H ) , Tac-FCHO1 μHD ( residues 609–889 ) ( I–L ) or Tac-FCHO1 linker + μHD ( residues 265–889 ) ( M–P ) before fixation . Tac-expressing cells , identified by recruitment of AP-2/EPS15 onto internal membranes ( asterisks ) are indicated , and enlarged , color-separated views of the rectangular regions in the medial sections are presented on the right . Scale Bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 02110 . 7554/eLife . 04137 . 022Figure 13—figure supplement 1 . Mislocalized EPS15 in Tac-linker expressing HeLa clone #64/1 . E cells overlaps with intracellular Tac . ( A–D ) HeLa clone #64/1 . E cells were short-term transfected with Tac ( A ) , Tac-FCHO1 linker ( residues 265–609 ) ( B ) , Tac-FCHO1 μHD ( residues 609–889 ) ( C ) or Tac-FCHO1 linker + μHD ( residues 265–889 ) ( D ) . Fixed cells were stained with an anti-Tac mAb ( 7 G7B6 ) ( green ) and an affinity-purified antibody against EPS15 ( red ) . Single representative optical sections at the basal surface of the cells are shown with merged color overlay ( A–D ) and the EPS15 signal ( A′–D′ ) . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 022 Crucially , the discrete clathrin/AP-2 surface assemblies that arise in Tac-linker-expressing cells are operational . Live-cell imaging of a HeLa SS6 cells stably transfected with a YFP-tagged AP-2 β2 subunit ( β2-YFP ) ( Keyel et al . , 2008 ) shows that forced expression of the Tac-FCHO1 linker ( residues 265–609 ) in these cells again reconfigures the overall arrangement of clathrin-coated structures ( Figure 14A ) . Using total internal reflection fluorescence microscopy , the more uniform and evenly spread β2-YFP puncta at the ventral surface of Tac-linker-containing cells differ from the adjacent untransfected cells ( Figure 14A ) . After 2 min at 37°C , added fluorescent transferrin rapidly clusters at the β2-YFP-containing structures ( Figure 14B ) but , again , the relative ratio of transferrin to YFP differs between the Tac-linker producing and untransfected cells ( Figure 14A vs Figure 14B ) , suggesting that the AP-2 spots in the Tac-FCHO1 linker-expressing cells are better able to concentrate transferrin receptors . 10 . 7554/eLife . 04137 . 023Figure 14 . Direct regulation of coat morphology and cargo packaging by the FCHO1 linker domain . ( A and B ) Color channel separated total internal reflection images of HeLa SS6 cells stably expressing β2-YFP ( A ) and pulsed for 2 min with 25 μg/ml Alexa Fluor647 transferrin ( B ) . The cells were previously transfected with Tac-FCHO1 linker ( 265–609 ) and Tac-expressing cells are indicated ( asterisks ) . ( C ) Comparison of individual frames from the transferrin channel at 2 min and 12 min by pseudocoloring the initial fame green and overlaying the last frame colored blue . The Tac-expressing cells with the rearranged β2-YFP spots are indicated ( asterisks ) . ( D and E ) Enlarged , color-channel separated views of the labeled boxed regions in ( C ) . ( F ) After completion of time-resolved image acquisition , Alexa Fluor546-conjugated anti-Tac mAb 7 G7B6 was added to identify unambiguously the Tac-transfected cells . Following incubation with the anti-Tac , a final image was collected . Internalized transferrin within endosomes visible in the total internal reflection field is indicated ( arrows ) . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04137 . 023 In the time-lapse data set , comparing the initial ( 2 min ) transferrin signal ( pseudocolored green ) with a frame acquired 10 min later ( colored blue ) illustrates that the Tac-transfected cells , have more dynamic clathrin-coated vesicle uptake ( Figure 14C ) . Numerous green or blue puncta are present in the transfected cells ( Figure 14C–E ) , indicating that these transferrin-laden structures disappeared/appeared between the two acquisition time points . A fluorescently conjugated anti-Tac antibody added after the time-lapse recording confirms that the cells exhibiting the smaller scattered β2-YFP structures are indeed expressing the Tac-linker chimera ( Figure 14F ) . Also , at this time point transferrin concentrated within endosomes is evident in the evanescent field . We conclude that the linker polypeptide in munsicin proteins affects the operation of AP-2 on the cell surface; the linker facilitates the production of more numerous cargo-filled clathrin-coated structures .
The polyhedrally interdigitated outer lattice is emblematic of the extensive protein–protein interactions that underpin the formation of a clathrin-coated vesicle . Yet perhaps the most impressive array of protein contacts occurs at the outset of clathrin polymerization on the limiting cell membrane . The early module of proteins implicated in the nucleation of surface clathrin coats is very highly interconnected ( McMahon and Boucrot , 2011; Reider and Wendland , 2011 ) ; microscopic temporal connections ( Taylor et al . , 2011 ) are enabled by nanoscopic macromolecular associations . The principal interaction hub is undoubtedly AP-2 , which can , through the independently folded α and β2 appendages , bind to >20 other endocytic factors utilizing four separate contact interfaces ( Traub , 2009; Kelly and Owen , 2011; McMahon and Boucrot , 2011 ) . Numerous other pioneer proteins bind physically to PtdIns ( 4 , 5 ) P2 , similar to AP-2 ( Höning et al . , 2005; Jackson et al . , 2010 ) , and some contain EH and SH3 domains , producing a densely populated interactome to coordinate membrane remodeling , coat biogenesis and cargo selection . The inherent redundancy and functional overlap between several of these early-arriving proteins is responsible for the robust nature of clathrin-mediated endocytosis . For example , CALM , a PtdIns ( 4 , 5 ) P2 binding clathrin-associated sorting protein for the R-SNAREs VAMP2 , -3 and -8 ( Koo et al . , 2011; Miller et al . , 2011; Sahlender et al . , 2013 ) , interacts with the μHD of FCHO1 ( Henne et al . , 2010; Umasankar et al . , 2012 ) . CALM also contains an FXDXF-type interaction motif that engages directly the PHear domain of Necap 1 ( Ritter et al . , 2007 ) , has several DXF- and FXSXF-based AP-2-associating peptide sequences ( Meyerholz et al . , 2005 ) , and can interact physically with clathrin ( Tebar et al . , 1999; Meyerholz et al . , 2005 ) . The protein also has two NPF tripeptide motifs that bind to EH domains ( Salcini et al . , 1997 ) , as found in Eps15/R and intersectins . Yet CALM knock-out mice are born , and mouse embryonic fibroblasts can be cultured and analyzed ( Scotland et al . , 2012; Suzuki et al . , 2012 ) . In fact , a considerable number of clathrin coat components and ancillary proteins have been experimentally deleted or incapacitated without complete disruption of clathrin-dependent internalization ( Garcia et al . , 2001; Kang-Decker et al . , 2001; Morris et al . , 2002; Kamikura and Cooper , 2003; Holmes et al . , 2007; Koh et al . , 2007; Wang et al . , 2008; Chen et al . , 2009; Mullen et al . , 2012; Pozzi et al . , 2012; Scotland et al . , 2012; Suzuki et al . , 2012; Umasankar et al . , 2012; Kononenko et al . , 2013; Tsushima et al . , 2013; Alazami et al . , 2014 ) . Although Fcho1 and Fcho2 were proposed to be all-important ( Henne et al . , 2010 ) , we have carefully evaluated the overall veracity of this provocative idea . HeLa cells with undetectable levels of FCHO1 and FCHO2 still fabricate clathrin lattices and internalize clathrin-dependent transmembrane cargo . We have maintained these muniscin-depleted cells in culture for many months . That is not to say that these proteins are insignificant . The cellular concentration of muniscins regulates clathrin-coated structure organization and dynamics and , thereby , the efficiency of cargo internalization . Clearly , this underscores the importance of these particular pioneers in clathrin lattice biogenesis and function . The enlarged clustered clathrin assemblies characteristic of essentially-munsicin-null HeLa cells appear to be linked to physiological compensation . Operationally , they resemble hot spots , areas of repeated vesicle budding from non-terminal clathrin puncta within a spatially restricted area ( Nunez et al . , 2011 ) . Hot spots were actually reported the first time GFP-tagged clathrin-coated structures were visualized live ( Gaidarov et al . , 1999 ) . Endocytic activity at hot spots is more resistant to perturbations in the levels AP-2 and PtdIns ( 4 , 5 ) P2 , so-called nucleation factors ( Nunez et al . , 2011 ) . Thus , the extended long-lived lattices are likely better adapted to coupled budding events to conserve and reuse limiting factors and maintain steady , but slowed , clathrin-mediated endocytosis . The reduced capacity to sequester transferrin ( receptors ) within the clathrin-coated structures that we note in clone 1 . E cells could indicate that the unstructured linker domain of FCHO1/2 stabilizes the activated , open form of AP-2 to bias cargo capture . This idea is in line with the increased abundance of the smallest population of coated structures in the clone 1 . E ( and clone #64 ) cells , presumably abortive structures because of the overall reduced rate of transferrin uptake in these cells . However , live-cell imaging discloses that FCHO1/2 arrive at nascent clathrin bud sites on the cell surface with or before AP-2 ( Henne et al . , 2010; Taylor et al . , 2011 ) , and current copy number estimates suggest that AP-2 outnumbers FCHO2 in HeLa cells by an order of magnitude and FCHO1 by at least a factor of 1000 ( Kulak et al . , 2014 ) . So the interaction between AP-2 and FCHO1/2 is unlikely to be stoichiometric within the lattice , as borne out by proteomic analysis ( Borner et al . , 2012 ) . An obvious question , though , is why AP-2 is not able to properly engage the transferrin receptor YTFR sorting signal despite being positioned in the surface lattices of clone 1 . E cells . As indicated above , it is well established that the heterotetramer displays multiple modes of attachment to membrane-apposed clathrin . In addition to four spatially discrete PtdIns ( 4 , 5 ) P2 binding sites ( Gaidarov and Keen , 1999; Jackson et al . , 2010 ) , AP-2 has at least two dedicated cargo-specific interactions surfaces ( Jackson et al . , 2010 ) , a clathrin box ( Shih et al . , 1995 ) to contact the clathrin heavy chain terminal domain β-propeller ( ter Haar et al . , 2000 ) , as well as a second site in the β2 subunit appendage ( Edeling et al . , 2006 ) that binds to the extended α-solenoid region of the clathrin heavy chain designated the ankle ( Knuehl et al . , 2006 ) . Both the α- and β2-subunit appendages interact with a large overlapping set of early-arriving clathrin-associated sorting and accessory proteins . The appendages do not depend on allosteric changes within the core to promote binding because AP-2 can be quantitatively affinity purified with immobilized appendage binding motifs or immunoprecipitated with mAb AP . 6 ( Chin et al . , 1989 ) , which binds to the platform interaction site of the α appendage . We suspect that the wide range of interaction possibilities allows deposition of some incompletely rearranged AP-2 at clathrin lattices formed in the absence of muniscins . AP-2 can transition between metastable states; a partially open form bound to the dileucine sorting signal has been characterized ( Kelly et al . , 2008 ) . An incomplete conformational transition in all membrane-associated AP-2 in the clone 1 . E cells may also be linked to the alterations in lattice arrangement we see . Using synthetic liposome templates , AP-2 driven clathrin-coated buds are spherical ( Kelly et al . , 2014 ) while clathrin coats nucleated by a monomeric clathrin binding domain are flat but stabilize membrane remodeling induced by thermal fluctuations ( Dannhauser and Ungewickell , 2012 ) . The crescent-shaped EFC domain in the FCHO proteins prompted the conjecture that initial dimpling of a spherical membrane patch defined a bud site ( Henne et al . , 2010 ) . Further , if the munsicins , and their pioneer binding partners collectively assembled into a circumferential rim defining the outer edge of the indented membrane , they could stabilize nascent buds possibly by preventing lateral diffusional escape of phosphoinositides ( Zhao et al . , 2013 ) necessary for successful placement of incoming inner layer coat constituents . Yet two lines of evidence argue against this notion: First , the linker region alone is able to correct or remodel the abnormal distribution of clathrin structures , even when not directly membrane associated . Second , the Tac-linker + μHD chimera , while membrane spanning , is unlikely to similarly remodel membrane topology adjacent to the linker , but it is still able to restore the organization of surface clathrin in clone #64/1 . E cells . Instead , the unstructured linker tract between the folded EFC and μHD domains in FCHO1 , and the sequence-related portion of Sgip1 , are able to drive membrane deposition of AP-2 and partner engagement . In this sense , muniscins could operate in a conceptually analogous manner to the small GTPase Arf1 activating the structurally and functionally related heterotetrameric AP-1 clathrin adaptor ( Ren et al . , 2013 ) . Our findings allow us to rationalize a number of recent studies with an integrated model for the biogenesis of clathrin lattices at the cell surface . We concur that both AP-2 and Fcho1/2 are central regulators during clathrin coat initiation; both are clearly core members of the pioneer module of endocytic components ( Henne et al . , 2010; Taylor et al . , 2011; Cocucci et al . , 2012 ) . A pivotal assembly step involves the conformational rearrangement of AP-2 ( Collins et al . , 2002; Höning et al . , 2005; Kelly et al . , 2008; Jackson et al . , 2010; Cocucci et al . , 2012; Kelly et al . , 2014 ) . This shift in the equilibrium from the closed cytosolic basal conformation to the open , assembly-competent form of AP-2 can be regulated synergistically by PtdIns ( 4 , 5 ) P2 and tyrosine ( YXXØ ) - or dileucine ( [DE]XXXL[LIM]-based internalization signals ) . Reorganization of AP-2 by these allosteric regulators could account for the reported minimal stoichiometry of two AP-2 heterotetramers and one or two clathrin trimers to nucleate a clathrin-coated structure ( Cocucci et al . , 2012 ) . Indeed targeted gene disruption of the AP-2 μ2 subunit is pre-implantation lethal in mice ( Mitsunari et al . , 2005 ) . But in C . elegans and S . cerevisae , AP-2 is dispensable ( Huang et al . , 1999; Yeung et al . , 1999; Mayers et al . , 2013 ) . In yeast , seven early module ( pioneer ) proteins can be deleted without fully disrupting clathrin-dependent internalization ( Brach et al . , 2014 ) . As noted above , loss of numerous mammalian pioneer factors ( Eps15 , CALM , Hrb , epsin , Dab2 , intersectin , Necap 1 ) similarly does not terminate clathrin-mediated endocytosis . One interpretation of these results is that there are several parallel pathways to arrive at an internalization competent clathrin assemblage at the plasma membrane . However , extinguishing the muniscins has a more severe effect than loss of a set of pioneers that physically engage the μHD . This is plainly apparent in cells expressing the tailored Tac-μHD protein—despite massive misrouting of EPS15 ( and EPS15R and intersectin ) to the juxtanuclear region , AP-2-positive puncta at the cell surface persist . Our findings indicate that FCHO1 , and specifically the linker segment , interacts with AP-2 and can promote membrane deposition of the heterotetramer in the absence of elevated levels of PtdIns ( 4 , 5 ) P2 . Further , because AP-2 cannot bind productively to clathrin in the closed basal conformation ( Kelly et al . , 2014 ) , our biochemical experiments showing that FCHO1 can modulate the clathrin binding properties of EPS15-bound AP-2 suggest that FCHO1 can provide an alternative pathway to rearrange AP-2 to promote cargo capture , clathrin lattice assembly and budding . The remarkable effect that a membrane-attached FCHO1 linker has on the disposition of surface clathrin structures supports this idea . The appearance of expanded planar clathrin sheets , which often contain assembly defects in cells depleted of muniscins ( FCHO2 principally ) , argues additionally for an important role of these proteins in overseeing swift and regular polymerization of the clathrin coat . Necap 1 also exerts a regulatory input on AP-2 by competitively engaging several binding surfaces on the heterotetramer ( Ritter et al . , 2013 ) . As FCHO1 and Necap 1 associate with each other , whether the effect of the FCHO1 linker directly involves Necap 1 will be important to establish . Finally , our results defining regulation of overall size of clathrin-coated buds raise the possibility that kinetic uncoupling between the outer and inner layer components of other vesicular coats could also alter the morphology of the forming tubulovesiclular carrier to accommodate variably sized cargo .
All TALENs were designed using open-access software ( Mojo Hand; http://www . talendesign . org ) as described ( Bedell et al . , 2012 ) . The repeat variable di-peptide ( RVD ) -containing units for FCHO2 TAL-1 ( NG NN NN NI NG HD NG NG NN NG NG NI NN NI NI NI ) and FCHO2 TAL-2 ( NI HD NG NG HD NG NN NI NI HD NG NG HD HD NG NG ) or FCHO1 TAL-1 ( HD HD NG NN NG NI HD HD NI HD NI NN HD NN NG NN NI ) and FCHO1 TAL-2 ( HD HD NN HD HD NI NN HD NG HD HD NG NG NN NN NG NN ) were assembled using the Golden Gate approach ( Cermak et al . , 2011 ) . After assembly , the RVDs were cloned into the pC-Goldy TALEN destination vector ( Bedell et al . , 2012; Carlson et al . , 2012 ) suitable for expression in mammalian cells ( Addgene , Boston , MA ) . For TALEN transfections , equal amounts of each pair of TALENs were introduced into a HeLa cells grown in a 35 × 10 mm tissue culture dish using Lipofectamine 2000 ( Invitrogen/Life Technologies , Grand Island , NY ) . Transfected cells were cultured 3 days at 30°C , before splitting for indel analysis and plating for clone selection . Individual clones were collected by a standard limiting dilution approach in 96-well plates . Clonal populations were expanded and screened by SDS-PAGE and immunoblotting with antibodies specific for human FCHO2 . To confirm gene disruption in FCHO1 mutants , genomic DNA was isolated from individual clonal populations or from the transfected-cell pool using DNeasy Blood and Tissue kit ( Qiagen , Valencia , CA ) . The genomic region surrounding the target site was PCR amplified with human FCHO1 locus specific primers using the genomic DNA template obtained from each clone . PCR amplicons were initially digested with appropriate restriction enzymes and the mutations were assessed by loss of restriction enzyme digestion . Then the poly-allelic mixture of the individual clones was ligated to TOPO-TA vector ( Invitrogen ) and analyzed by sequencing . GFP-FCHO1 was designed as described previously ( Reider et al . , 2009; Umasankar et al . , 2012 ) . To generate GFP-FCHO2 , full length human FCHO2 ( 1–810 ) was PCR amplified from cDNA clone #9021067 ( Open Biosystems , Lafayette , CO ) and inserted into pEGFP-C1 using SalI restriction sites and Cold Fusion cloning technology ( System Biosciences , Mountain View , CA ) . Similarly , GFP-Sgip1 was constructed by Cold Fusion cloning of full-length mouse Sgip1 ( 1–806 ) from cDNA clone #4482298 ( Open Biosystems ) into pEGFP-C1 using BglII restriction sites . GFP-FCHO1 ( 265–889 ) , GFP-FCHO1 ( 265–609 ) and GFP-FCHO1 ( 609–889 ) were described previously ( Umasankar et al . , 2012 ) . The various N-terminally-tagged truncation constructs used for the functional complementation studies were obtained by introducing stop codons at appropriate sites using QuikChange site-directed mutagenesis ( Stratagene/Agilent Technologies , Santa Clara , CA ) . The deletion constructs GFP-FCHO1 ( 1–889; Δ316–467 or Δ316–339 ) or GST-eps15 ( 595–896; Δ617–636 ) were produced using Phusion site-directed mutagenesis ( Thermo Scientific , Pittsburgh , PA ) . The muniscin linker regions fused to an N-terminal GST ( GST-FCHO1 ( 316–467 ) , GST-FCHO2 ( 314–444 ) and GST-Sgip1 ( 77–214 ) plasmids ) were generated by the insertion of appropriate PCR amplicons into EcoRI site in pGEX-4T-1 by Cold Fusion cloning . The GST-β2 appendage ( rat residues 701–937 ) , GST-αC appendage ( mouse residues 701–938 ) , GST-EPS15 ( 595–896 ) and GST-ARH constructs have been described previously . The plasmid encoding the GST-PHear domain of mouse Necap 1 was kindly provided by Dr Brigitte Ritter . Tac in the pcDNA3 . 1 plasmid is explained elsewhere ( Jha et al . , 2012 ) . FCHO1 µHD ( residues 609–889 ) or linker + µHD ( residues 265–889 ) were PCR amplified from GFP-FCHO1 and cloned into the Tac plasmid between EcoRV and NotI restriction sites to replace the endogenous Tac cytosolic segment . A stop codon at residue 610 in Tac-FCHO1 linker + µHD yielded Tac-FCHO1 linker ( 265–609 ) . All constructs were verified by automated dideoxynucleotide sequencing ( Genewiz , South Plainfield , NJ ) , and the primer and restriction site details , sequences ( Supplementary file 1 ) , and maps are all available upon request . HeLa SS6 ( Elbashir et al . , 2001 ) , HeLa clone #64 , clone #64/1 . E , the neuronal SH-SY5Y ( Biedler et al . , 1978 ) and MCF-7 ( Soule et al . , 1973 ) cells were cultured in DMEM supplemented with 10% fetal calf serum and 2 mM L-glutamine at 37°C in an atmosphere of 5% CO2 . HeLa SS6 β2-YFP stably expressing cells ( Keyel et al . , 2008 ) were grown in the same medium containing 0 . 5 mg/ml G418 . K562 cells ( Lozzio and Lozzio , 1975 ) were grown in suspension in RPMI media supplemented with 5% fetal calf serum and 2 mM L-glutamine at 37°C in 5% CO2 . Cells were transfected with plasmids using Lipofectamine 2000 ( Invitrogen ) or with siRNA oligonucleotides using Oligofectamine ( Invitrogen ) according to the manufacturer's recommendations ( Umasankar et al . , 2012 ) . After 18–24 hr , cells were fixed with 4% paraformaldehyde in PBS ( pH 8 . 0 ) and quenched and permeabilized with a mixture of 75 mM NH4Cl , 20 mM glycine and 0 . 1% Triton X-100 . The cells were blocked in 5% normal goat serum diluted in PBS/fish skin gelatin/saponin mixture and stained with various antibodies . For fluorescent transferrin binding and uptake assays , cells were preincubated in DMEM , 25 mM Hepes , 0 . 5% BSA at 37°C for 1 hr to remove bound transferrin . RNA was isolated from HeLa SS6 and SH-SY5Y cells by dissolving pelleted cells directly in TRIzol ( Invitrogen ) followed by chloroform extraction . The aqueous layer was precipitated with isopropanol and resuspended in DEPC water . To generate cDNA , the SuperScript III kit ( Invitrogen ) was used with 4 μg DNase-treated RNA according to the manufacturer's instructions . PCR of endocytic protein transcripts was performed with Taq polymerase ( Genscript , Piscataway , NJ ) and the following primers: FCHO1 sense 5′-CTG GCG CTG TGC CAC CTG GAA CT-3′ , FCHO1 antisense 5′-GTA CTC TCC CTC CGC AGC CGC TC-3′ , FCHO2 sense 5′-CCC CAG CAA TAT CTA GAC ACA GTC C-3′ , FCHO2 antisense 5′-TAC AGA AAG AGG AGT TGT GGG CC-3′ , SGIP1 sense 5′-GAA GTG GCA AGA CCC AGG CGT TCC-3′ , SGIP1 antisense 5′-GGA GGT GTT CCA GTG GGA AAA GGC C-3′ , β-actin sense 5′-GAG AGG CAT CCT CAC CCT GAA GTA C-3′ , β-actin antisense 5′-GCA CAG CCT GGA TAG CAA CGT ACA T-3′ , clathrin heavy chain sense 5′-GGA AGG AGA GTC TCA GCC AGT GAA A-3′ , clathrin heavy chain antisense 5′-TAT GTA ACT TCC CTC CAG CTT GGC C-3′ , EPS15 sense 5′-TTG TTG CAG CAA GCG ATT CAG CCA-3′ , EPS15 antisense 5′-AGG GCA GGG TCT TGT TGG AGT TCC-3′ , EPS15R sense 5′-AGC CTC AAC AGC ACA GGG AGC CTG-3′ , EPS15R antisense 5′-AAG GTT CTG GGT GAG GCC CGA GTG-3′ . The PCR cycling conditions used were: 94°C for 3 min , then 94°C for 30 s , 53°C ( FCHO2 and clathrin ) or 55°C ( remainder ) for 30 s and 72°C for 1 min , repeated for 30 cycles , and then final extension at 72°C for 7 min . The affinity-purified rabbit anti-FCHO1 antibody 1 ( 1:2500 ) , anti-FCHO2 ( 1:2500 ) , anti-Dab2 ( 1:1000 ) , anti-epsin 1 ( 1:10000 ) and anti-AP-1/2 β1/β2-subunit GD/1 ( 1:2500 ) antibodies were produced for our laboratory . A second affinity-purified anti-FCHO2 ( 1:2500 ) antibody was kindly provided by Dr Harvey McMahon . Affinity-purified rabbit anti-eps15 polyclonal antibody and anti-AP-1/2 β1/β2-subunit mAb 100/1 were gifts from Dr Ernst Ungewickell , the anti-clathrin HC mAbs TD . 1 ( 1:5000 ) and X22 and the anti-AP-2 α-subunit mAb AP . 6 were generously provided by Dr Frances Brodsky , the rabbit R11-29 anti-AP-2 μ2-subunit ( 1:3000 ) antiserum kindly provided by Dr Juan Bonifacino , and the rabbit anti-intersectin 1 ( 1:2000 ) and rabbit anti-NECAP 1 ( 1:5000 ) antibodies were a gift from Dr Peter McPherson . The goat anti-Hrb C-19 ( 1:1000; sc-1424 ) , rabbit anti-eps15 C-20 ( 1:500; sc-534 ) and goat anti-CALM C-18 ( 1:500; sc-6433 ) polyclonal antibodies and rabbit anti-SHIP2 mAb E−2 ( 1:500; sc-166641 ) from Santa Cruz Biotechnology ( Dallas , TX ) , rabbit anti-EPS15R ( 1:5 . 000; EP-1145Y ) antibody and rabbit anti-FCHO1 antibody 2 ( 1:1000; ab84740 ) from AbCAM ( Cambridge , MA ) and the mouse anti-myc mAb 9E10 ( 1: 1000; MMS-150P ) from Biolegend ( Dedham , MA ) were used . The anti-β tubulin mAb E7 ( 1:2500 ) was purchased from the Developmental Studies Hybridoma Bank . The mAbs directed against the AP-2 α subunit clone 8/Adaptin α ( 1:1000; 610502 ) and μ2 subunit clone 31/AP50 ( 1:500; 611350 ) were from BD Transduction Laboratories ( San Jose , CA ) . Another AP-2 α subunit mAb C-8 ( 1:500; sc-17771 ) came from Santa Cruz and the anti-Tac mAb 7 G7B6 was from Ancell ( Bayport , MN ) . The secondary antibodies used were donkey anti-rabbit ( 1:5000; NA934V ) - or anti-mouse ( 1:5000; NA931V ) - horseradish peroxidase conjugates from GE Healthcare Life Sciences ( Pittsburgh , PA ) or rabbit anti-goat ( 1:5000; A4174 ) peroxidase conjugate from Sigma ( St . Louis , MO ) . The purification of GST , various GST-fusion proteins , and the preparation of rat brain cytosol for GST-based pull-down assays have been thoroughly described previously ( Umasankar et al . , 2012 ) . The lysates from parental HeLa SS6 , HeLa clone #64 or HeLa clone #64/1 . E cells were prepared from cells detached with Cellstripper ( Cellgro/Mediatech , Manassas , VA ) . K562 cells grown in suspension were pelleted directly . After washing , cell pellets were solubilized on ice for 30 min in 25 mM Hepes-KOH pH 7 . 2 , 125 mM potassium acetate , 5 mM magnesium acetate , 2 mM EDTA , 2 mM EGTA , and 2 mM dithiothreitol ( assay buffer ) supplemented with 1% Triton X-100 , 1 mM PMSF and complete protease inhibitor cocktail ( Roche , Indianapolis , IN ) . Lysates were centrifuged at 20 , 000×g before use in binding assays . Assays were in assay buffer , usually in a final volume of 300 μl . After incubation at 4°C for 1–2 hr , glutathione-Sepharose beads were sedimented and washed four times with ice-cold PBS . Aliquots of the supernatant and washed pellet factions were resolved by SDS-PAGE and either stained with Coomassie blue or transferred to nitrocellulose for immunoblotting . Detection was with enhanced chemiluminescence and X-ray film . The heterotetrameric core of AP-2 was produced in Escherichia coli BL21 ( DE3 ) pLysS after transfection with two bicistronic plasmids encoding the β2-subunit trunk and myc-tagged μ2 subunits ( Ampr ) and the GST-α-subunit trunk and σ2 subunit ( Kanr ) generously provided by David Owen . Purification was exactly as described ( Collins et al . , 2002; Kelly et al . , 2008; Jackson et al . , 2010 ) . For the binary interaction assays , a modified assay buffer composed of 25 mM Hepes-KOH pH 7 . 2 , 25 mM Tris–HCl , 125 mM potassium acetate , 5 mM magnesium acetate , 2 mM EDTA , 2 mM EGTA , 10 mM dithiothreitol , 0 . 2% Igepal CA-630 ( Nonidet P40 substitute ) and 0 . 1 mg/ml BSA was used . After incubation of immobilized GST-fusion proteins with 20 μg/ml purified AP-2 core in modified assay buffer for 60 min at 4°C , the glutathione-Sepharose beads were recovered at 1000 × g for 1 min . Following a first wash in ice-cold modified assay buffer and centrifugation at 1000×g for 1 min , the beads were washed two additional times with ice-cold PBS and centrifugation at 10 , 000×g for 1 min . Confocal fluorescence images were collected on an Olympus Fluoview FV1000 microscope as described previously ( Keyel et al . , 2006; Umasankar et al . , 2012 ) . The z-stacks were collected with a 0 . 25-µm step size between optical sections and the stacks were deconvolved using the blind deconvolution algorithm within Autoquant X3 ( Media Cybernetics , Rockville , MD ) . Quantitation of objects in fluorescent images was with the Nikon Elements software ( version 4 . 30 , Nikon , Melville , NY ) . For three-color total internal reflection fluorescence microscopy , adherent cells were imaged on a Nikon Eclipse Ti inverted microscope with a 60 × 1 . 49 NA oil-immersion objective . Cells were maintained in DMEM supplemented with 10% fetal calf serum and 25 mM HEPES , pH 7 . 2 at 37°C on MatTek dishes ( MatTek Corporation , Ashland , MA ) and imaged continuously at 5 s/frame . GFP/YFP was excited with a 488 nm laser , Alexa Fluor546 conjugated anti-Tac mAb with a 561 nm laser , and transferrin-Alexa 647 ( Molecular Probes/Life Technologies , Grand Island , NY ) with a 647 nm laser line . Images were collected using an Andor ( Belfast , Ireland ) Xyla 5 . 5 camera; at full resolution under these conditions the pixel size with a 1 × coupler matches Nyquist sampling ( 120 nm xy exactly ) . Data sets were acquired acquired using Nikon Elements . Adherent plasma membrane from parental HeLa SS6 and clone #64/1 . E cells were prepared for rapid-freeze , deep-etch electron microscopy as described previously ( Heuser , 2000 ) . Briefly , cells grown on oriented pieces of glass cover slip were disrupted by sonication and fixed in a glutaraldehyde and paraformaldehyde mixture before flash freezing in liquid helium ( Keyel et al . , 2006 ) . | Cells can take proteins and other molecules that are either embedded in , or attached to , their surface membrane and move them inside via a process called endocytosis . This process often involves a protein called clathrin working together with numerous other proteins . Early on , a complex of four proteins , called the adaptor protein-2 complex , interacts with both the ‘cargo’ molecules that are to be taken into the cell , and the cell membrane . Clathrin molecules then assemble into an ordered lattice-like coat , on top of the adaptor protein complex layer . This deforms a small patch of the cell membrane and curves it inwards . The clathrin molecules coat this pocket as it grows in size , until it engulfs the cargo . The pocket quickly pinches off from the membrane to form a bubble-like structure called a vesicle , which is brought into the cell . A family of proteins termed Muniscins were thought to be involved in the early stages of endocytosis and have to arrive at the membrane before the adaptor protein-2 complex and clathrin . But experiments to test this idea—that reduced , or ‘knocked-down’ , the production of Muniscins—had given conflicting results . As such , it remained unclear how the small patches of membrane carrying cargo molecules are marked as being destined to become clathrin-coated vesicles . Now Umasankar et al . have studied the role that these proteins play in the early stages of endocytosis in human cells grown in a laboratory . A gene-editing approach was used to precisely disrupt a gene that codes for a Muniscin protein called FCHO2 . Umasankar et al . observed that these ‘edited’ cells formed clathrin coats that were more irregular compared with those that form in normal cells . Nevertheless , clathrin-mediated vesicles still formed when this protein was absent , though the process of endocytosis was slower . Similar results were seen when Umasankar et al . used the same approach to disrupt the gene for a related protein called FCHO1 in the same cells . A short fragment of the Muniscin proteins , called the linker , was shown to bind to , and activate , the adaptor protein-2 complex . The linker then recruits this complex to the specific regions of the cell membrane where clathrin-coated vesicles will form . Several dozen other proteins also accumulate where clathrin pockets form; as such , one of the next challenges will be to investigate if this mechanism of locally activating the cargo-gathering machinery is common in living cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2014 | A clathrin coat assembly role for the muniscin protein central linker revealed by TALEN-mediated gene editing |
Vertebrate neck musculature spans the transition zone between head and trunk . The extent to which the cucullaris muscle is a cranial muscle allied with the gill levators of anamniotes or is instead a trunk muscle is an ongoing debate . Novel computed tomography datasets reveal broad conservation of the cucullaris in gnathostomes , including coelacanth and caecilian , two sarcopterygians previously thought to lack it . In chicken , lateral plate mesoderm ( LPM ) adjacent to occipital somites is a recently identified embryonic source of cervical musculature . We fate-map this mesoderm in the axolotl ( Ambystoma mexicanum ) , which retains external gills , and demonstrate its contribution to posterior gill-levator muscles and the cucullaris . Accordingly , LPM adjacent to the occipital somites should be regarded as posterior cranial mesoderm . The axial position of the head-trunk border in axolotl is congruent between LPM and somitic mesoderm , unlike in chicken and possibly other amniotes .
The evolution of a mobile neck was a key innovation at the origin of tetrapods ( Daeschler et al . , 2006 ) . It involved expansion of muscles , some derived from the head ( cranial muscles ) and some from the trunk , to support the skull apart from the pectoral girdle and permit a greater range of movement of the head relative to the rest of the body . Cranial muscles support a variety of functions , including feeding , respiration , vision , facial expression and vocalization . They are distinct from trunk muscles in genetic regulation and susceptibility to disease ( Noden , 1983; Noden et al . , 1999; Sambasivan et al . , 2009; reviewed by Bismuth and Relaix , 2010; Diogo et al . , 2015; Noden and Francis-West , 2006; Tzahor , 2009 ) . Developmentally , they are non-somitic , arising instead from cranial paraxial and splanchnic mesoderm ( Couly et al . , 1992; Noden , 1983; Evans and Noden , 2006; reviewed by Noden and Trainor , 2005 ) . Cranial muscle regulatory factors include Isl1 , Tbx1 , MyoR , Capsulin and Pitx2 , which operate in specific muscle groups ( Hacker and Guthrie , 1998; Sambasivan et al . , 2009; Lu et al . , 2002; Mootoosamy and Dietrich , 2002; Harel et al . , 2009 ) . Pitx2 , for example , specifies mandibular arch mesoderm but not hyoid arch mesoderm in the mouse ( Shih et al . , 2007a ) . In contrast to cranial muscle , formation of trunk muscle is Pax3-dependent ( Tajbakhsh et al . , 1997 ) . The domain of the vertebrate neck contains two muscle groups: the hypobranchial muscles ventrally and the cucullaris dorsally . Hypobranchial muscles are derived from occipital somites , which form the hypoglossal cord and migrate towards the tongue ( Noden , 1983; O'Rahilly and Müller , 1984 ) . The number of occipital somites contributing to cranial structures varies among species , however . For example , somites 2 and 3 form both hypobranchial musculature and the occipital arch in the axolotl ( Piekarski and Olsson , 2007; 2014 ) , whereas in chicken somites 2–5 form both the occipital region of the skull and tongue musculature ( Couly et al . , 1993; Huang et al . , 1999; 2000 ) . The cucullaris muscle , a feature of gnathostomes , connects the head to the pectoral girdle , thus spanning the transition zone between cranial and trunk myogenic signaling regimes ( Kuratani , 1997 ) . It is the putative homologue of the trapezius and sternocleidomastoid in amniotes ( Lubosch et al . , 1938 ) . In sharks and the Queensland lungfish , the cucullaris elevates the gill arches and protracts the pectoral girdle . It originates near the skull and continues caudally and ventrally to insert on the scapular region of the pectoral girdle; a ventral fascicle extends to the posteriormost branchial bar ( Edgeworth , 1926; 1935; Allis , 1917; Vetter , 1874; Greenwood and Lauder , 1981 ) . The cucullaris is a thin muscle , and it can be difficult to visualize its three-dimensional position vis-à-vis adjacent skeleton and musculature . Hence , it is poorly described in many taxa with regard to both its shape and its relation to other cranial and trunk musculature . It is innervated by the accessory ramus of the vagus ( X ) nerve in anamniotes , but primarily by the accessory ( XI ) nerve in amniotes ( Edgeworth , 1935 ) . While in chicken the connective-tissue component of hypobranchial muscles and the ventrolateral neck region is derived from neural crest ( Le Lièvre and Le Douarin , 1975 ) , the cucullaris is reported to have somite-derived connective tissue ( Noden , 1983 ) . The derivation of connective-tissue components of the mouse trapezius is not fully resolved; both lateral plate mesoderm ( Durland et al . , 2008 ) and neural crest ( Matsuoka et al . , 2005 ) are reported sources . While a somitic derivation of the hypobranchial muscles is widely accepted , the embryonic origin of the cucullaris is controversial ( reviewed by Tada and Kuratani , 2015; Ericsson et al . , 2013 ) . Historically , the cucullaris was considered a branchiomeric cranial muscle based in part on its anatomical relation to the gill levators ( Vetter , 1874; Edgeworth , 1935; Piatt , 1938 ) . Subsequent fate mapping of anterior somites in chicken and axolotl , though , demonstrated a somitic ( trunk ) contribution ( Noden , 1983; Couly et al . , 1993; Huang et al . , 1997; 2000; Piekarski and Olsson , 2007 ) . More recent fate mapping in chicken and genetic analysis in mouse reveal that the trapezius is primarily a lateral plate mesoderm-derived structure that employs a cranial , rather than trunk , myogenic program ( Theis et al . , 2010; Lescroart et al . , 2015 ) . These data leave unresolved whether the lateral plate origin of the cucullaris is the result of a posterior shift of the head myogenic program or if instead head mesoderm extends caudally into the region adjacent to the anterior somites . To distinguish between these hypotheses , it is important to define the posterior limit of myogenic cranial mesoderm in an organism with a relatively conservative cervical and branchial region . Amniote branchial-arch musculature is reduced in comparison to that of piscine sarcopterygians and aquatic salamanders such as the axolotl , which has a relatively plesiomorphic arrangement of cranial muscle . Moreover , axolotls possess bushy external gills and their associated musculature , which likely were present in the larvae of Paleozoic tetrapods , as well as a robust gill skeleton , which was present in the earliest limbed stem tetrapods ( Schoch and Witzmann , 2011 ) . Here , we address this problem from a combined morphological , genetic and developmental perspective . In the axolotl , we locate the head-trunk boundary within unsegmented cranial mesoderm . In addition , we use micro-computed tomography ( CT ) to describe the morphology of the cucullaris and gill levators in a phylogenetically diverse series of gnathostome taxa , including limbless caecilians and the coelacanth , the sister taxon to all other extant lobe-finned fishes . In previous studies , the cucullaris was investigated largely by gross dissection . In many species , however , the cucullaris is a thin , superficial muscle embedded in several layers of fat and connective tissue . It can be difficult to expose without damaging its in situ context with respect to the trunk and the pectoral girdle . Our CT-based reconstructions reveal such three-dimensional relationships without tissue disruption . In those taxa that have both branchial levators and a cucullaris , the cucullaris consistently appears to be in series with the levators . This suggests that the cucullaris is a serial homolog of the levators , thus supporting a cranial muscle identity of the cucullaris . Likewise , in the axolotl , although the cucullaris in adults assumes a large , triangular 'trapezius-like' morphology , the larval cucullaris is clearly in series with the levators . The ubiquity of the cucullaris further supports the hypothesis that it is a critical component of the head-trunk connection in gnathostomes . To study the development of tissues in the transitional region spanned by the levators and cucullaris , we extend modern fate-mapping techniques and gene-expression analysis of cranial mesoderm to the axolotl . We show that unsegmented mesoderm adjacent to the anterior three somites contributes to the cucullaris as well as to the gill-levator muscles in a manner consistent with their apparent serial homology , which supports categorization of the cucullaris as a branchiomeric muscle . Cranial mesoderm markers , including isl1 and tbx1 , also are expressed in the developing cucullaris region . We find molecular regionalization of the cranial muscles at stage 40 , with distinct expression patterns in the mandibular and hyoid arch musculature . Adductor muscles within the mandibular arch have distinct gene expression patterns as well . We argue that the posterior limit of cranial mesoderm in the axolotl extends caudally to the axial level of somite 3 and that the head-trunk boundary is consistent between the somites and lateral plate mesoderm . We discuss the importance of posterior cranial mesoderm in the evolution of the vertebrate neck .
New soft-tissue-contrast staining methods for high-resolution CT afforded us the opportunity to examine the volumetric anatomy of muscles in a sample of vertebrates spanning Gnathostomata ( Figure 1; Video 1; Figure 1—figure supplements 1–8 ) . In Chondrichthyes , such as the chimaera , the cucullaris is a massive muscle that may incorporate anterior gill levators; in this respect , it may not be strictly homologous , in its entirety , with the cucullaris of osteichthyans . In piscine osteichthyans , such as bichir and lungfish , the cucullaris is a thin , strap-like muscle , sometimes called the 'protractor pectoralis' ( e . g . , Greenwood and Lauder , 1981 ) . It diverges from the anterior gill levators , but otherwise it is in series with them , is located in the same connective tissue sheath , and shares muscle fibers with the immediately anterior levators . It originates from the posterior region of the head and inserts on the pectoral girdle and , when present , the fifth gill arch . In some amphibians and in amniotes , the cucullaris is a large wedge-shaped muscle , sometimes termed the trapezius ( Owen , 1866; Edgeworth , 1935; Gegenbaur et al . , 1878 ) . This morphology is seen clearly in an Anolis lizard and in an opossum , which exhibits the primitive mammalian condition ( Figure 1F; Figure 1—figure supplements 7 , 8 ) . 10 . 7554/eLife . 09972 . 003Figure 1 . Cranial muscle evolution based on contrast-stained CT scans and an MRI scan ( coelacanth adult ) . ( A−F ) Left lateral views of gill-levator musculature and the cucullaris ( or its homologue ) in representative gnathostomes , showing its insertion on the pectoral girdle ( except in caecilians , where it inserts on ventral fascia ) . ( G , H ) Left lateral views of gill levators and the cucullaris in relation to the branchial skeleton in a coelacanth . The cucullaris attaches to the posteriormost gill arch . Box in G is enlarged in H . ( I ) Left dorsolateral view of the cucullaris in a caecilian . The gill-levator musculature is shaded green , the cucullaris blue , and the pectoral girdle white . In the lower panels , the fifth ceratobranchial is in pink and the anterior branchial skeleton in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00310 . 7554/eLife . 09972 . 004Figure 1—figure supplement 1 . Stereo image of chimaera from Figure 1 with skeletal elements and muscles segmented . See Figure 1 for color guide . Stereos created in VGStudio Max v2 . 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00410 . 7554/eLife . 09972 . 005Figure 1—figure supplement 2 . Stereo image of bichir in dorsolateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00510 . 7554/eLife . 09972 . 006Figure 1—figure supplement 3 . Stereo image of lungfish in dorsolateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00610 . 7554/eLife . 09972 . 007Figure 1—figure supplement 4 . Stereo image of coelacanth in dorsolateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00710 . 7554/eLife . 09972 . 008Figure 1—figure supplement 5 . Stereo image of axolotl in dorsolateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00810 . 7554/eLife . 09972 . 009Figure 1—figure supplement 6 . Stereo image of caecilian in dorsolateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 00910 . 7554/eLife . 09972 . 010Figure 1—figure supplement 7 . Stereo image of anole in dorsolateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01010 . 7554/eLife . 09972 . 011Figure 1—figure supplement 8 . Stereo image of opossum in lateral view . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01110 . 7554/eLife . 09972 . 012Video 1 . Video of representative gnathostome cranial and pectoral regions spinning around their long axes with skeletal elements and muscles segmented . See Figure 1 for color guide . Videos rendered in VGStudio Max v2 . 2 at 2048 X 1536 resolution , 25 frames per second ( 20 s ) , Windows AVI format , 85% quality . Sides segmented represent original specimens , whereas some images in figure panels were reversed so that all specimens are oriented in the same direction . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 012 The cucullaris has not been described in the musculoskeletally conservative coelacanth , nor in the limbless caecilian amphibians . Based on novel dissections , Greenwood and Lauder ( 1981 ) reported the cucullaris absent from the coelacanth . Millot and Anthony ( 1958 ) , however , had earlier briefly described in the coelacanth a fifth gill levator that originates on the anocleithrum of the pectoral girdle , unlike the first four gill levators , which originate in the otic region . We examined this muscle in both a CT scan of a contrast-stained coelacanth pup and an MRI ( magnetic resonance imaging ) scan of an adult . The muscle in these specimens is larger than previously described , with several heads originating on the pectoral girdle ( Figure 1C ) . It is angled differently from the other levators , but its fibers remain in close association with them and extend from the anocleithrum to insert on the fifth ceratobranchial . In the adult , an anterior portion of this muscle extends dorsally to attach on the fascia of the epaxial musculature ( Figure 1G , H ) . Based on its morphology and location , we regard the fifth gill levator as the homolog of the cucullaris . Accordingly , the coelacanth cucullaris retains the ancestral connection between the posteriormost branchial bar and the pectoral girdle , which is seen in at least some sharks and lungfish ( Edgeworth , 1926; 1935; Allis , 1917; Vetter , 1874; Greenwood and Lauder , 1981 ) . Even though actinopterygians lack an ossified fifth gill arch , the cucullaris in these taxa sometimes joins the fibers of the posteriormost gill levator ( Greenwood and Lauder , 1981 ) . Although the cucullaris in the coelacanth does not attach to the head , it retains a division that extends rostrally , terminating upon the dorsal fascia posterior to the cranium; it also retains the ancestral connection between the pectoral girdle and the cranial skeleton by its attachment to the branchial skeleton . In the caecilian Typhlonectes natans we examined the m . levator arcus branchiales complex , previously described in Dermophis mexicanus ( Bemis et al . , 1983 ) . The muscle is also termed the m . cephalodorsosubpharyngeus ( Wilkinson and Nussbaum , 1997; Lawson , 1965 ) . Based on our examination , the m . levator arcus branchiales complex is a triangular structure that originates from the otic capsule and dorsal trunk muscle fascia and inserts ventrally on the the posteriormost ceratobranchial ( Figure 1E , I ) . A posterior division of the muscle , the pars posterosuperficialis ( Wilkinson and Nussbaum , 1997 ) , inserts on the fascia separating the rectus abdominus from the interhyoideus . Based on these topographic relationships , we homologize the posterior division of the m . levator arcus branchialis complex with the cucullaris . It is unclear if a portion of the anterior division should also be considered part of the cucullaris , connecting to the adult hyobranchial skeleton . The cucullaris and its homologs comprise a highly conserved connection between head and trunk . In general , the cucullaris is intimately associated , and sometimes partially continuous , with the gill levators . In numerous taxa , it attaches to the skull and gill skeleton , both cranial elements . In amphibians , the cucullaris has also been termed the protractor pectoralis ( Ziermann and Diogo , 2013 ) and the trapezius ( Piatt , 1938 ) . In juvenile axolotls , the cucullaris resembles the condition in bichir; it is morphologically similar to and in series with the anterior gill levators , whereas in adults it expands into a broad , thin sheet ( Figure 2A , B ) . Given the conservative morphology of branchiomeric musculature in the axolotl ( Ericsson et al . , 2004; Ericsson and Olsson , 2004; Ziermann and Diogo , 2013 ) , we began fate-mapping head mesoderm that contributes to the pharyngeal arches . In chicken and axolotl , mesoderm from somites 1 and 2 contributes to the cucullaris ( Piekarski and Olsson , 2007 ) . In chicken , however , the majority of the cucullaris is derived from lateral plate mesoderm adjacent to the occipital somites ( Theis et al . , 2010 ) . Consequently , we suspected that the axolotl cucullaris might also have a dual origin from both somitic and unsegmented mesoderm . 10 . 7554/eLife . 09972 . 013Figure 2 . Development of the cucullaris muscle in the axolotl . ( A , B ) Morphology of the developing cucullaris , with the four gill-levator muscles ( lab I–IV ) shaded light green and the cucullaris ( cuc ) blue . More-posterior muscles are shaded dark green . Anterior is to the left . ( A ) Dorsolateral view of an OPT scan of a juvenile axolotl stained with the 12/101 muscle antibody . ( B ) Contrast-stained CT scan of an adult axolotl in lateral view . The cucullaris is expanded into a broad sheet that inserts on the scapula . ( C ) Schematic depiction of an orthotopic transplantation of unsegmented mesoderm lateral to somites 1−3 at stage 21 . Lateral views; anterior is to the left . nc , neural crest; nt , neural tube; s3 , somite 3 . ( D−I ) GFP labeling following stage-21 transplantation of unsegmented mesoderm lateral to somites 1−3 . ( D−F ) Labeling of the levator arcuum branchiarum anlagen ( laba ) dorsal to the developing gills is visible in lateral ( D , E ) and dorsal ( F ) views . Anterior is to the left . ( G ) Gill-levator muscles ( levator arcuum branchiarum , lab ) of arches 3 and 4 , the cucullaris ( cuc ) and the dilatator laryngis ( dil ) are labeled in a juvenile axolotl . Dorsal view; anterior is to the left . ( H , I ) Transverse sections through the posterior occipital region ( H ) and anterior trunk ( I ) of a juvenile axolotl . GFP labeling is visible in the gill levators and anterior cucullaris ( H ) and in the posterior cucullaris near its attachment with the scapula ( I; sc ) . Lateral is to the left; dorsal is to the top . ( J−M’ ) isl1 expression in albino embryos . ( J ) At stage 34 , isl1 is expressed in ventral mesoderm , in the developing heart region ( arrowheads ) and around the dorsal cranial placodes ( cp ) . Arrow indicates several stripes of expression dorsal to the developing gills . ( K ) At stage 36 , isl1 marks the profundal ( gPr ) /trigeminal ( gV ) placode region and earlier expression is maintained dorsal to the gills ( arrow ) . ( L ) At stage 40 , isl1 is expressed in neurons within the dorsal spinal cord ( arrow ) and in the gill-levator region ( arrowheads ) . Dorsal view . ( M , M’ ) Transverse section of a stage-40 embryo with isl1 expression in the dorsal gill levator region ( arrows ) and ganglia ( arrowheads ) . Box in M is enlarged in M’ . Scale bars , 100 μm , except G , 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01310 . 7554/eLife . 09972 . 014Figure 2—figure supplement 1 . Additional stages of embryonic isl1 expression in A . mexicanum . ( A ) At stage 16 , isl1 is expressed in the region of the developing heart field ( arrow ) . ( B ) By stage 21 , expression has expanded dorsally ( arrowheads ) . ( C ) Ventral region of isl1 at stage 21 . nt , neural tube . ( D ) isl1 expression at stage 28 , including the branchial arches . ( E ) Frontal section dorsal to the developing gill arches at stage 36 . Inset panel indicates plane of section ( dashed red line ) . Lateral is to the top . A , B and D , lateral views; C , ventral view . Anterior is to the left in all panels . Scale bars , 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01410 . 7554/eLife . 09972 . 015Figure 2—figure supplement 2 . Embryonic expression of tbx1 and msc in A . mexicanum . All embryos are depicted in lateral view except E , J and O , which are ventral views; anterior is to the left . ( A , B ) Bilateral stripes of tbx1 expression ( arrows ) are present in mid-neurula stages . ( C ) Two bilateral stripes of tbx1 expression are visible ( arrows ) . ( D ) tbx1 is expressed in the region of the developing branchial arches ( asterisk ) . ( E ) Patches of tbx1 expression in the mandibular arch ( arrows ) . ( F , G ) tbx1 is expressed in mandibular ( m ) , hyoid ( h ) and branchial ( b ) regions . ( H ) tbx1 is expressed in the otic vesicle ( ov ) . ( I , J ) At stage 38 , tbx1 is expressed in developing muscle groups . mm mandibular arch muscle; hm , hyoid arch muscle; gm , gill musculature . ( K , L ) msc is expressed anteriorly at neurula stages . ( M ) Patch of msc expression just posterior to the eye ( arrow ) . ( N ) msc is expressed in the gill arch muscles ( gm ) . ( O ) msc expression at stage 38 . Scale bars , 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01510 . 7554/eLife . 09972 . 016Figure 2—figure supplement 3 . Mesoderm fate mapping in A . mexicanum embryos . Boundaries between regions 1 and 2 and regions 2 and 3 are approximate . LPM s , lateral plate mesoderm adjacent to somite; CM , cranial mesoderm; lme , levator mandibulae externus; lma , levator mandibulae anterior; im , intermandibularis posterior; bhe , branchiohyoideus externus; dm , depressor mandibulae; ldb , levatores et depressores branchiarum; lab , levatores arcus branchiarum; cuc , cucullaris; dil , dilatator laryngis s3 , somite 3; nc , neural crest , nt , neural tube . In region key anterior is to the left; dorsal is to the top . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 016 We transplanted GFP+ mesoderm adjacent to the first three somites into a white ( d/d ) host ( Figure 2C ) . By stage 35 , GFP+ cells were visible dorsal to the developing gill buds in the region of the presumptive gill muscles , the cucullaris and the dilatator laryngis ( Figure 2D−G ) . In cross section , labeled cells were present throughout the length of the cucullaris but absent from the somitic hypobranchial or epaxial muscles , thus indicating little if any somitic contamination ( Figure 2H , I ) . Additional transplantations were then performed for three regions of cranial mesoderm anterior to the first somite . In most of these transplants , the first gill-levator muscle originated from mesoderm just anterior to the first somite , whereas the posterior three gill levators arose from unsegmented mesoderm at the level of somites 1−3 ( Figure 2—figure supplement 3 ) . Transcription factors involved in cranial muscle development are expressed in gill levator/cucullaris muscle territory . At neurula and tailbud stages , isl1 is expressed in anterior cranial mesoderm associated with the second heart field ( Figure 2—figure supplement 1; Sefton et al . , 2015 ) . In subsequent stages , isl1 expression expands dorsally to encompass the entire dorsoventral length of the gills ( Figure 2—figure supplement 1D ) , but later ( stage 34 ) it is reduced near the heart and appears in the developing cranial placodes ( Figure 2J ) . From stage 36 through at least stage 40 , stripes of expression are present dorsal to the developing gills , including the levator anlage ( Figure 2K−M’ ) . The cranial-mesoderm marker tbx1 ( sequences of tbx1 , msc and pitx2 by pers . comm . from J . Whited , B . Haas and L . Peshkin ) is expressed in the developing gill muscle region at stages 35 and 38 ( Figure 2—figure supplement 2I ) . By stage 38 , msc is also expressed in the gill region . Unlike isl1 , expression of tbx1 and msc extends distally into the external gills ( Figure 2—figure supplement 2O ) . In axolotl , jaw adductor muscles include the levator mandibulae externus ( lme ) and the levator mandibulae anterior ( lma ) ; the latter muscle is also called the pseudotemporalis ( Ziermann and Diogo , 2013 ) . Both of these muscles develop within the mandibular arch ( Figure 3B , I ) . We examined expression of lhx2 , a LIM-domain transcription factor involved in pharyngeal muscle specification in the mouse ( Harel et al . , 2012 ) . As seen in mouse and Xenopus , lhx2 is expressed in axolotl in the brain and eye ( Figure 3; Figure 4—figure supplement 1A−E; Viczian et al . , 2006; Atkinson-Leadbeater et al . , 2009 ) . At stage 34 , lhx2 is expressed in the mesodermal core of the pharyngeal arches ( Figure 3E , F ) , but by stage 40 it becomes more restricted to specific muscle groups , including the lme and ventral hyoid arch musculature , but expression was not visible in the lma at stage 40 ( Figure 3G , H ) . At stage 38 , the anlage of the lma is located posterior to the eye ( Figure 3I ) . By stage 40 , this region expresses isl1 , including the superficial lma and developing ganglia , while isl1 expression was not visible in the lme ( Figure 3J−K’’ ) . 10 . 7554/eLife . 09972 . 017Figure 3 . Fate-mapping and gene expression in mandibular adductor muscles . ( A ) Schematic depiction of orthotopic transplantations of anterior cranial mesoderm . nc , neural crest; nt , neural tube; s3 , somite 3 . ( B ) Labeling of mandibular arch mesoderm at stage 35 following transplantation at stage 19 . Arrow points to the anlage of the levator mandibulae externus ( lmea ) . ( C ) Specimen in ( B ) at stage 43 , with labeling of the levator mandibulae externus ( lme ) . ( D ) Transverse section through the eye region of a stage-55 axolotl following transplantation at stage 20 . The levator mandibulae externus is labeled ventral to the eye . Mc , Meckel’s cartilage . Lateral is to the left; dorsal is to the top . ( E ) At stage 34 , lhx2 is expressed in the pharyngeal mesoderm of all six arches ( 1–6 ) as well as the forebrain ( fb ) and eye . ( F ) Frontal section through the head at stage 36 showing lhx2 expression in the mesodermal core of the third pharyngeal arch ( arrow ) . Anterior is to the top . Inset panel depicts plane of section ( dashed red line ) . ( G ) At stage 40 , lhx2 is expressed in the levator mandibulae externus ( lme ) and in the interhyoideus ( ih ) , a ventral cranial muscle . ( H ) Transverse section at the level of the eye at stage 40 , showing lhx2 expression in the levator mandibulae externus . Dorsal is to the top . ( I ) Labeling of mandibular and hyoid arch mesoderm in a stage-38 embryo , including the anlage of the levator mandibulae anterior ( lmaa ) , an anterior jaw adductor . Lateral view; anterior is to the left . ( J−K” ) isl1 expression in albino embryos . ( J ) At stage 40 , isl1 is expressed dorsal to the gills ( arrowhead ) and in the pineal gland ( pg ) . Expression posterior to the eye ( arrow ) overlaps with the region forming the levator mandibulae anterior . ( K−K’’ ) Transverse sections of a stage-40 embryo . Box in K is enlarged in K’ and K” . ( K ) isl1 is expressed in the lateral portion of the developing levator mandibulae anterior ( lma ) and in the trigeminal nerve ( arrowhead ) . ( K’’ ) Desmin staining of muscle cells , including those expressing isl1 . Scale bars , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01710 . 7554/eLife . 09972 . 018Figure 4 . Origin of the mandibular depressor muscle and expression of pitx2 in the hyoid arch . ( A ) Schematic depiction of orthotopic transplantation of cranial mesoderm . nc , neural crest; nt , neural tube; s2 , somite 2 . Somite 1 is small and triangular in shape . ( B ) GFP labeling of dorsal hyoid-arch mesoderm at stage 35 following transplantation at stage 20 includes the anlage of the depressor mandibulae ( dma ) . ( C ) Specimen in ( B ) at stage 57 , with labeling of the depressor mandibulae ( dm ) and otic capsule ( oc ) . ( D ) Labeling of the depressor mandibulae in a transverse section through the jaw region of a stage-57 juvenile axolotl . sq , squamosal bone . Dorsal is to the top; lateral is to the left . ( E ) At stage 34 , pitx2 is expressed in the eye , the ventral mandibular arch ( m ) , the hyoid arch ( arrow ) and more faintly in migrating somitic cells ( arrowhead ) . ( F ) At stage 38 , pitx2 expression is maintained in the hyoid arch ( arrow ) and is also present in the otic vesicle ( arrowhead ) . Scale bars , 100 μm , except C , 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 01810 . 7554/eLife . 09972 . 019Figure 4—figure supplement 1 . Embryonic expression of lhx2 and pitx2 in A . mexicanum . ( A , B ) In neurula stages , lhx2 is expressed in anterior neuroectoderm . A: anterior view , dorsal is to the top . B: lateral view , anterior is to the left . ( C ) lhx2 is expressed in cranial mesoderm at stage 22 . ( D ) During middle-tailbud stage , lhx2 is expressed in the mandibular-arch ( m ) and branchial-arch mesoderm as well as the eye field ( e ) and forebrain ( fb ) . ( E ) lhx2 expression is maintained in the eye and brain , as well as in mandibular and hyoid ( h ) arches . ( F ) pitx2 is expressed in anterior cranial mesoderm . ( G ) Transverse section through the anterior neural folds ( nf ) reveals expression through the ectoderm and mesoderm at neurula stages . ( H ) At stage 31 , pitx2 is expressed in oral ectoderm . ( I ) At stage 35 , pitx2 is expressed asymmetrically in the left lateral-plate mesoderm ( arrow ) and oral region ( o ) . Ventral view . ( J ) pitx2 is also expressed in hyoid arch mesoderm ( arrow ) at the same stage . ( K ) At stage 40 , pitx2 is strongly expressed in the oral epithelium . It is also present in two hyoid arch muscles , the branchiohyoideus externus ( bhe; arrowheads ) and the depressor mandibulae ( dm ) . ( L ) pitx2 is expressed in the tongue muscles ( arrowhead ) . Ventral view; anterior is to the left . ( M ) Transverse section at stage 40 shows pitx2 expression in the tongue muscles ( arrowheads ) and hyoid arch musculature ( arrow ) . Dorsal is to the top; lateral is to the left . Inset panel indicates plane of section ( red dashed line ) . C−F , H , J , K , lateral views; anterior is to the left . Scale bars A−J , 500 μm; K−M , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 019 A third gene provides an additional example of genetic heterogeneity in cranial muscle development . In mouse , Pitx2 is broadly expressed in developing muscle; it is required to specify mesoderm of the mandibular arch but not of the hyoid arch ( Shih et al . , 2007a; 2007b ) . In axolotl , as in chick , pitx2 is expressed in anterior ectoderm and oral region from neurula stages through at least tailbud stages ( Figure 4E , F; Figure 4—figure supplement 1F−M; Bothe and Dietrich , 2006 ) . A stripe of expression in hyoid arch mesoderm and in the migrating hypobranchial muscle precursors also appears by stage 34 ( Figure 4E ) . It is maintained in hyoid arch derivatives and by stage 40 is concentrated in the hyoid musculature , including the depressor mandibulae and branchiohyoideus externus ( Figure 4F; Figure 4—figure supplement 1K , M ) . The depressor mandibulae anlage ( dma ) is in the dorsal/proximal pharyngeal arch at stage 35 ( Figure 4B , C ) and then extends to insert on Meckel’s cartilage at stage 42 ( Ericsson and Olsson , 2004 ) . At stage 40 , pitx2 is not expressed in gill musculature , although it is strongly expressed in tongue musculature ( Figure 4—figure supplement 1L , M ) . We investigated the myogenic properties of lateral plate mesoderm adjacent to the somites to determine if local signals in the cranial mesoderm of the mandibular and hyoid arch regions could instruct lateral plate mesoderm at various axial levels to adopt cranial muscle fate . Myogenic lateral plate mesoderm in the cucullaris region adjacent to somite 2 was transplanted into anterior cranial mesoderm following extirpation of a region of host mesoderm ( Figure 5A ) . Transplanted cells were incorporated into both dorsal and ventral mandibular arch or hyoid arch muscle ( Figure 5B , C’’ ) . These muscles displayed normal innervation from the mandibular branch of the trigeminal nerve ( Figure 5C−C’’ ) and the facial nerve , respectively . Local cues appear sufficient to pattern myogenic lateral plate mesoderm from the cucullaris region and to promote mandibular or hyoid arch muscle development . 10 . 7554/eLife . 09972 . 020Figure 5 . Heterotopic transplantation of lateral plate mesoderm . ( A ) Schematic depiction of a caudal-to-cranial heterotopic transplantation of lateral plate mesoderm from somite level 2 ( donor ) to mandibular arch mesoderm ( host ) . Lateral views; anterior is to the left . nc , neural crest; nt , neural tube; s3 , somite 3 . ( B−C’’ ) Stage-45 larva following the heterotopic transplantation shown in ( A ) . ( B ) GFP+ cells contribute to mandibular arch muscles ( arrow ) . Lateral view; anterior is to the left . ( C ) Lateral plate mesoderm contributes to the levator mandibulae externus ( lme ) . ( C’ ) Innervation of the levator mandibulae externus by the mandibular branch of the trigeminal nerve ( V ) is normal . VII , facial nerve . ( C’’ ) The intermandibularis ( im ) , a ventral mandibular muscle , is also labeled . Ventral view; anterior is to the top . ( D ) Schematic depiction of a caudal-to-cranial heterotopic transplantation of lateral plate mesoderm from somite level 5 to mandibular arch mesoderm . ( E ) Stage-45 larva following the heterotopic transplantation shown in ( D ) . Ventral view; anterior is to the left . No muscle fibers are formed , but labeled cells contribute to cranial vasculature ( arrowheads ) . Scale bars , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 020 Next , more posterior lateral plate mesoderm , adjacent to somite 5 , was transplanted heterotopically to mandibular arch mesoderm at stage 21 ( Figure 5D ) . While transplanted cells were present among mandibular arch structures , in 9 of 10 larvae they did not incorporate into muscle ( Figure 5E ) . Neither mandibular nor hyoid arch mesoderm appears sufficient to induce posterior non-myogenic lateral plate mesoderm to form muscle .
We provide evidence from comparative morphology , embryonic fate mapping and gene expression that the cucullaris is a branchiomeric muscle in series with the gill levators and that it is stably conserved across gnathostomes as a link between head and trunk . Accordingly , we propose the fifth gill levator of the coelacanth to be homologous to the cucullaris , which , as in some sharks , rays and lungfish , attaches the pectoral girdle to the posteriormost gill bar ( Edgeworth , 1935; Greenwood and Lauder , 1981 ) . We regard this interpretation of data from coelacanth , viz . , the cucullaris has reduced its dorsal attachment to the head/epaxial muscle fascia , more parsimonious than a previous interpretation that the cucullaris is absent and that a gill levator has entirely shifted its origin from the head to the pectoral girdle ( Greenwood and Lauder , 1981; Millot and Anthony , 1958 ) . In larval Ichthyophis kohtaoensis , a caecilian , the fourth gill levator is substantially larger than the anterior three levators ( Kleinteich and Haas , 2007 ) . The cucullaris could potentially develop from the caudalmost gill levator , as has been suggested in urodeles ( Edgeworth , 1935; Ziermann and Diogo , 2013 ) . The cucullaris of caecilians , lacking an insertion to the absent shoulder girdle , instead has evolved a patent connection between the otic capsule ( as well as the dorsal trunk fascia ) and fascia associated with ventral trunk musculature . It is uncertain if the anterior portion of the levator arcus branchiales complex , which inserts on the posteriormost gillbar , is also part of the cucullaris or instead represents only gill levators that did not degenerate following metamorphosis . In the former case , this unusual configuration might express an ancestral potential of the cucullaris to attach to the gill skeleton , and it evokes reports that the paired fin/limb apparatus has surprising developmental resemblance to the gill arches ( Gillis et al . , 2009 ) . The cucullaris has evolved to perform distinct functions in different lineages . In placoderms , for example , it may have depressed the head ( Trinajstic et al . , 2013 ) . The morphology of the cucullaris in sharks and rays suggests the muscle in gnathostomes originates ancestrally from the pectoral girdle and inserts on two parts of the cranial skeleton: the posterior gill bar and the caudal region of the head . The connection to the gill arches was likely lost in early tetrapods ( but possibly later reappeared in caecilians ) , while an alternate attachment to the clavicle evolved in some lineages . The cucullaris is purportedly absent in turtles and snakes , but recent work suggests that it may be present in both groups . In turtles , it has been proposed that the muscle originates on the shell ( carapace ) , which incorporates parts of the pectoral girdle ( Lyson et al . , 2013 ) . In snakes , the pectoral girdle is absent and the origin of the cucullaris has concomitantly shifted to the body wall ( Tsuihiji et al . , 2006 ) . The cucullaris is located in a complex transition zone between head and trunk; in the axolotl , this complexity is reflected in the muscle’s dual embryonic derivation from both somitic and cranial mesoderm . An origin from both the caudal branchial levator and somites was suggested in the spotted salamander , Ambystoma maculatum , based on serial sections and dissection ( Piatt , 1938 ) . Our finding that unsegmented mesoderm adjacent to the anterior somites forms the posterior gill-levator muscles , a laryngeal muscle , the levatores et depressores branchiarum and the cucullaris indicates that the posterior limit of cranial mesoderm is at somite 3 . The presence of labeled cranial mesoderm cells in a laryngeal muscle in axolotl betrays the deep phylogenetic conservation of a relationship between the cucullaris and laryngeal muscles , which was revealed in a recent analysis demonstrating that mouse laryngeal muscles are clonally related to the trapezius and absent following mutation of the gene Tbx1 ( Lescroart et al . , 2015 ) . Moreover , expression of the genes isl1 and tbx1 in the gill-levator region suggests these muscles develop through the cranial muscle regulatory network , consistent with their classical anatomical classification as cranial muscles . Our analysis of cranial mesoderm markers in axolotl provides additional evidence for genetic heterogeneity in cranial muscle development in anamniotes , which has been demonstrated in mouse and chicken ( Nathan et al . , 2008; Dong et al . , 2006; Kelly et al . , 2004; Marcucio and Noden , 1999 ) . Surprisingly , our data reveal that differentiation of mandibular adductor muscles is present in amphibians at the level of gene expression . At stage 40 , isl1 is expressed in the superficial anterior adductor , while lhx2 is expressed in the external adductor . In mouse , the LIM homeodomain gene Isl1 is required for normal second heart field ( SHF ) development and its expression in SHF progenitors is downregulated following differentiation ( Cai et al . , 2003 ) . Genetic fate mapping in mouse demonstrates a large contribution of Isl1-positive cells to the ventral intermandibular muscle and the cucullaris ( Nathan et al . , 2008; Theis et al . , 2010 ) . In the axolotl at stage 40 , pitx2 is expressed in hyoid arch and tongue musculature but is not in the gill musculature . Taken together , these findings underscore the regionalization of developmental programs that underlies cranial muscle formation , both among pharyngeal arches and even within the mandibular adductor complex . Moreover , the broad phylogenetic diversity of the model species involved suggests that such regionalization may be an ancestral feature of tetrapod vertebrates that is retained in living taxa and may also exist in their piscine outgroups . In the axolotl embryo , somites and pharyngeal arches are located at the same post-otic axial level , which is a basic feature of morphologically conservative vertebrates ( Kuratani , 1997 ) . Lateral plate mesoderm adjacent to somites 1 and 2 is located in the intermediate region between head and trunk and is important for morphogenetic movements associated with the migration of hypobranchial muscle progenitors ( Lours-Calet et al . , 2014 ) . Moreover , lengthening of the amniote neck is associated with the caudal shift of the heart into the thorax ( Hirasawa et al . , 2016 ) . The head-trunk interface at the paraxial level is marked by the path of circumpharyngeal neural crest cells as they migrate ventral to the occipital somites to form the circumpharyngeal ridge caudal to the pharynx ( Kuratani , 1997 ) . Specialized muscles occur at this paraxial level , including the trapezius/cucullaris and , in axolotls , the gill levators . Our finding that the posterior gill-levator muscles and the cucullaris originate from cranial mesoderm adjacent to the first three somites supports categorization of the cucullaris as a branchiomeric muscle . Moreover , it may help explain why lateral plate mesoderm in the embryonic ‘trunk’ in chicken has myogenic capacity . Our fate-mapping data suggest that this mesoderm , which gives rise to the cucullaris in amniotes , is not a novel source of musculature , but instead is cranial mesoderm associated with the most posterior pharyngeal arch ( 5th , 6th or 7th , depending on species ) . We propose that , in the axolotl , somite 3 is the posterior limit of mesodermal contribution to cranial structures in both paraxial and lateral mesoderm ( Figure 6A ) . In our heterotopic transplantations , cranial mesoderm that forms the cucullaris is able to follow the myogenic program of cranial muscles in the mandibular and hyoid arches . Although the chicken lacks many of the cartilages and muscles associated with the posterior pharyngeal arches in other tetrapods , it retains cucullaris progenitors in the same anatomical position as in the axolotl ( Figure 6B ) . 10 . 7554/eLife . 09972 . 021Figure 6 . The cucullaris and the transition zone between the head and trunk . ( A ) In the axolotl embryo , the head-trunk boundary in unsegmented mesoderm is closely congruent with that in the somites . Paraxial and lateral mesoderm anterior to somite 3 form cranial structures ( including the heart ) . The illustration depicts a stage-21 embryo with epidermis removed; anterior is to the left . Somite fate-mapping data are from Piekarski and Olsson ( 2007 ) ; Piekarski and Olsson ( 2014 ) . ( B ) In the chicken , the axial level of the head-trunk boundary in somitic mesoderm is posterior to the border in unsegmented mesoderm . Somite fate-mapping data are from Couly et al . ( 1993 ) and Huang et al . ( 1999 ) ; cucullaris data are from Theis et al . ( 2010 ) ; mandibular depressors and branchiomandibular data are from Noden ( 1983 ) and Evans and Noden ( 2006 ) . ( C−E ) Contrast-stained CT images of the lungfish branchial skeleton , pectoral girdle , posterior skull , gill levators and cucullaris . All structures except the branchial skeleton are segmented on the left side only . The anterolateral view depicts only the fifth gill arch , with its attachment to the cucullaris; the body is rendered transparent . The lungfish cucullaris retains the ancestral tripartite attachment: origin from the pectoral girdle ( o ) and insertions on the posterior skull ( i1 ) and fifth ceratobranchial ( i2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 021 The head-trunk boundary in the axolotl is congruent between cranial mesoderm and somitic mesoderm , but in the chicken ( and probably other amniotes ) the head-trunk boundary in somites is posterior to that in unsegmented cranial mesoderm ( Figure 6A–B; Couly et al . , 1993; Piekarski and Olsson , 2014; Huang et al . , 2000 ) . It remains to be determined whether this congruence , as seen in the axolotl , is the plesiomorphic condition for tetrapods . Heterotopic transplantations in chicken suggest that somitic mesoderm has greater regional plasticity than lateral plate mesoderm . Somites that contribute to the posterior skull are able to generate vertebrae when transplanted to a more posterior position , independent of Hox gene expression ( Kant and Goldstein , 1999 ) , whereas caudal cranial mesoderm that gives rise to the cucullaris is unable to generate muscle when transplanted to a more posterior location ( Theis et al . , 2010 ) . It will be of interest to identify the mechanisms responsible for the incorporation of somites into the posterior skull during tetrapod evolution and to determine if the posterior limit of cranial mesoderm is less evolutionarily labile than somitic contribution to cranial structures .
CT scans were prepared from anatomical specimens of Hydrolagus sp . ( MCZ 164893 ) , Polypterus bichir ( MCZ 168418 ) and Protopterus sp . ( MCZ 54055 ) from the Museum of Comparative Zoology at Harvard University , as well as Typhlonectes natans ( YPM HERA 012618 ) and Monodelphis domestica ( YPM MAM 10713 ) from the Yale Peabody Museum of Natural History at Yale University . Latimeria chalumnae ( AMNH 32949h ) was obtained from the American Museum of Natural History . For contrast staining , specimens were immersed in 5% Lugol solution in 70% ethanol for 7−10 d at room temperature . Specimens were washed in 70% ethanol for 2 d , changing solution daily . Three-dimensional images were taken using an XRA-002 microCT scanner ( X-Tek; Tring , United Kingdom ) at the Center for Nanoscale Systems at Harvard University . Reconstructions were performed with VGStudio Max 2 . 0 ( Volume Graphics ) . MRI data for Latimeria chalumnae ( SIO 75–347 ) from the Scripps Institution of Oceanography were obtained from the Digital Fish Library hosted by the University of California , San Diego , through the generosity of Lawrence Frank and Rachel Berquist . White mutant ( dd ) , GFP+ white mutant and albino ( aa ) embryos of the Mexican axolotl ( Ambystoma mexicanum ) were obtained from the Ambystoma Genetic Stock Center at the University of Kentucky and from the Hanken laboratory breeding colony at Harvard University . Before grafting , embryos were decapsulated manually by using watchmaker forceps and then staged ( Bordzilovskaya et al . , 1989; Nye et al . , 2003 ) . Explants of unsegmented cranial mesoderm ( stages 18–22 ) from donor embryos were grafted unilaterally into stage-matched hosts in place of comparable regions that had been extirpated . Stage-matched donors were of similar size and form . In heterotopic transplantations , anterior cranial mesoderm from regions that contribute to mandibular or hyoid arch musculature was partially extirpated in hosts . In one set of heterotopic experiments , GFP+ mesoderm adjacent to somite 2 was moved into either region 1 or region 2 of host anterior cranial mesoderm ( integrating into either the mandibular or hyoid arch ) . In a second set of experiments , GFP+ mesoderm adjacent to somite 5 was transplanted into host anterior cranial mesoderm . Fixation , embedding and sectioning were performed as previously described for A . mexicanum ( Sefton et al . , 2015 ) . For GFP labeling , sections were incubated with rabbit polyclonal anti-GFP ab290 ( 1:2000; Abcam , Cambridge , MA ) , followed by AlexaFluor-488 goat anti-rabbit ( 1:500; Life Technologies , Carlsbad , CA ) . DAPI ( 0 . 1–1 μg/ml in PBS ) was used to label cell nuclei . Some sections were stained with the skeletal muscle marker 12/101 monoclonal antibody ( 1:100; Developmental Studies Hybridoma Bank , Iowa City , IA ) . Additionally , desmin ( 1:100; Monosan , PS031; Uden , Netherlands ) was used to label muscle in stage-40 embryos . Acetylated alpha-tubulin ( 1:100; Sigma , T6793; St . Louis , MO ) was used to detect developing axons , followed by AlexaFluor-568 goat anti-mouse ( 1:500; Life Technologies , Carlsbad , CA ) . A specimen for OPT ( Sharpe et al . , 2002 ) was stained with 12/101 followed by AlexaFluor-568 goat anti-mouse as described above . Clearing and embedding were performed at the University of Washington , where the larva was dehydrated in ethanol , cleared in 1:2 benzyl alcohol/benzylbenzoate , and imaged with a Bioptonics 2100M scanner . In situ hybridization was performed on albino ( aa ) embryos . Antisense riboprobes were generated from the cloned fragment ( DIG RNA labeling kit; Roche Diagnostics , Indianapolis , IN ) . In situ hybridization was carried out as previously described ( Henrique et al . , 1995 ) , with an additional MAB-T wash overnight at 4°C ( 100 mM maleic acid , 150 mM NaCl , pH 7 . 5 , 0 . 1% Tween 20 ) . Hybridization was performed at 65°C . Primers are included in Table 1 . Amplified PCR fragments were subcloned into the pCR II vector ( Life Technologies ) . 10 . 7554/eLife . 09972 . 022Table 1 . Primer sequences . DOI: http://dx . doi . org/10 . 7554/eLife . 09972 . 022TranscriptForward ( 5’ to 3’ ) Reverse ( 5’ to 3’ ) Product Sizelhx2AACAGTGACGCAAACAGTGGTTGAAGCAGTTAGCGCAGAA755 bpmscACCAGCAGACACCAAGCTCTTGTGTCCTCCTCTGATGTGAA708 bppitx2AGATCGCCGTGTGGACTAACGGTGGTAGCGAGTTTTGGAA809 bptbx1GGAGTACGACCGAGATGGAAATGAAGCGCTGATGACAGTG688 bp | Muscles in the head and trunk ( main body ) form from different parts of the embryo , and their development uses different genes . Trunk muscles are derived from somites – paired blocks of cells arranged in segments on either side of the midline ( which divides the body into left and right halves ) . By contrast , cells that give rise to head muscles are arranged in a continuous mass . But what about neck muscles ? Some studies claim they develop like head muscles; others suggest they are trunk muscles . These studies commonly examine mice or chickens . By examining species that have a more primitive complement of head and neck muscles , Sefton et al . now show that a neck muscle should be considered a kind of head muscle . Gill muscles are definitive head muscles . Sefton et al . found that the cucullaris , a prominent neck muscle in fishes and amphibians , forms from the same mass of cells that gives rise to gill muscles . Moreover , studying muscle development in Mexican axolotls showed that cells that contribute to gill muscles extend into the trunk , which is further back in the embryo than was previously known . Previous studies reported the cucullaris muscle is absent in a “lobe-finned” fish called the coelacanth , which is closely related to four-limbed animals . However , by using a technique called micro-computed tomography to visualize the neck muscles of this fish , Sefton et al . show that the cucullaris muscle is present and connects the rear-most gill to the shoulder . The finding that neck muscles form like head muscles in the axolotl confirms a previous claim that was based on studies of bird embryos . A future challenge is to understand the molecular and genetic mechanisms that establish the boundary between head and trunk muscles , and work out how those mechanisms might have influenced how the neck evolved . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"evolutionary",
"biology"
] | 2016 | Evolution of the head-trunk interface in tetrapod vertebrates |
Colistin is an antibiotic of last resort , but has poor efficacy and resistance is a growing problem . Whilst it is well established that colistin disrupts the bacterial outer membrane ( OM ) by selectively targeting lipopolysaccharide ( LPS ) , it was unclear how this led to bacterial killing . We discovered that MCR-1 mediated colistin resistance in Escherichia coli is due to modified LPS at the cytoplasmic rather than OM . In doing so , we also demonstrated that colistin exerts bactericidal activity by targeting LPS in the cytoplasmic membrane ( CM ) . We then exploited this information to devise a new therapeutic approach . Using the LPS transport inhibitor murepavadin , we were able to cause LPS accumulation in the CM of Pseudomonas aeruginosa , which resulted in increased susceptibility to colistin in vitro and improved treatment efficacy in vivo . These findings reveal new insight into the mechanism by which colistin kills bacteria , providing the foundations for novel approaches to enhance therapeutic outcomes .
The emergence of multi-drug-resistant Gram-negative pathogens such as Escherichia coli , Klebsiella pneumoniae and Pseudomonas aeruginosa has led to the increased use of polymyxin antibiotics , which are often the only viable last-resort therapeutic option ( Velkov et al . , 2010; Garg et al . , 2017; Biswas et al . , 2012; Thomas et al . , 2019 ) . Two closely related polymyxin antibiotics are used clinically , colistin ( polymyxin E ) and polymyxin B , which share a high degree of structural similarity , consisting of a cationic peptide ring of 7 amino acids connected to a hydrophobic acyl tail by a linear chain of three amino acids ( Velkov et al . , 2010; Biswas et al . , 2012 ) . Polymyxins are rapidly bactericidal towards Gram-negative bacteria in vitro but are considerably less efficacious in vivo , with up to 70% of patients failing to respond to colistin treatment ( Falagas et al . , 2010; Paul et al . , 2018; Linden et al . , 2003 ) . Restrictions on dosage due to the nephrotoxicity of polymyxins mean that only 50% of people with normal renal function achieve a steady state serum concentration sufficient to kill bacteria ( Tran et al . , 2016; Satlin et al . , 2020 ) . As such , there is a desperate need to develop new approaches to enhance the efficacy of polymyxin antibiotics . Barriers to increasing polymyxin efficacy include the significant gaps in our understanding of their mode of action . Whilst it is well established that the binding of polymyxins to lipopolysaccharide ( LPS ) on the surface of Gram-negative bacteria leads to disruption of the outer membrane ( OM ) , it is unclear how this results in cell lysis and bacterial death ( Figure 1—figure supplement 1; Biswas et al . , 2012; MacNair et al . , 2018 ) . It is hypothesised that damage to the LPS monolayer enables polymyxins to traverse the OM via a process of ‘self-directed uptake’ , although this has not been demonstrated experimentally ( MacNair et al . , 2018; Powers and Hancock , 2003 ) . Once across the OM , polymyxins permeabilise the cytoplasmic membrane ( CM ) , which is required for bacterial lysis and killing ( Velkov et al . , 2010; Garg et al . , 2017; Biswas et al . , 2012 ) . However , the mechanism by which colistin damages the CM is unclear ( Powers and Hancock , 2003; Trimble et al . , 2016 ) . It has been proposed that the surfactant activity of polymyxins , conferred by the positively charged peptide ring and hydrophobic tail , is sufficient to compromise the phospholipid bilayer of the CM via a detergent-like effect ( Velkov et al . , 2010; Biswas et al . , 2012 ) . In support of this , polymyxins can interact with mammalian cell membranes , leading to changes in epithelial monolayer permeability ( Berg et al . , 1996 ) . Polymyxin antibiotics also have some inhibitory activity against the Gram-positive bacterium Streptococcus pyogenes , where the CM is formed of a phospholipid bilayer ( Betts et al . , 2016 ) . However , several lines of evidence call into doubt the ability of physiologically relevant concentrations of polymyxins to disrupt phospholipid bilayers . Firstly , the concentrations of polymyxin B required to disrupt mammalian epithelial cells or inhibit the growth of S . pyogenes ( 8–16 µg ml−1 ) are above typical serum concentrations of the antibiotic , and colistin at clinically relevant concentrations displays no activity against other Gram-positive organisms such as Staphylococcus aureus or Enterococcus faecalis ( Berg et al . , 1996; Betts et al . , 2016; Si et al . , 2018; Kouidhi et al . , 2011 ) . Furthermore , colistin has very little activity against synthetic phospholipid bilayer membranes unless LPS is present , a finding that explains why polymyxins are 30–100-fold less active against colistin-resistant Acinetobacter baumanii isolates that are LPS-deficient , with an OM composed of a phospholipid bilayer ( Khadka et al . , 2018; Moffatt et al . , 2010; Zhang et al . , 2018 ) . Finally , molecular dynamics simulations show that the interaction of colistin with phospholipid bilayers is unlike what has been reported for other antimicrobial peptides that target phospholipid bilayers ( Fu et al . , 2020 ) . Together , these observations call into question whether , at physiologically relevant concentrations , colistin disrupts the CM of Gram-negative bacteria via the engagement of the polymyxin antibiotic with membrane phospholipids . In addition to the mode of action of colistin , there are also gaps in our understanding of the mechanisms by which colistin resistance protects bacteria from polymyxin antibiotics . In Gram-negative bacteria , LPS is synthesised in the cytoplasm via the Raetz pathway , during which it is introduced into the inner leaflet of the CM ( Raetz et al . , 2007; Simpson and Trent , 2019 ) . It is then flipped to the outer leaflet of the CM by MsbA before transportation to the OM via the LptABCDEFG machinery ( Okuda et al . , 2016; Zhou et al . , 1998; Li et al . , 2019 ) . To date , 10 mobile colistin resistance ( mcr ) gene variants have been described , all of which encode phosphoethanolamine ( pEtN ) transferases that modify the lipid A component of LPS with pEtN as it is trafficked through the CM on the way to the OM ( Liu et al . , 2016; Carroll et al . , 2019; Nang et al . , 2019; Skov and Monnet , 2016; Wang et al . , 2020 ) . Colistin resistance can also arise via mutations in genes encoding two-component regulatory systems such as PhoPQ , PmrAB , or BasRS ( Poirel et al . , 2017; Janssen et al . , 2020 ) . This typically leads to the addition of 4-amino-4-deoxy-l-arabinose ( L-ara4N ) and/or pEtN groups to LPS , with this modification also occurring at the CM ( Simpson and Trent , 2019; Poirel et al . , 2017 ) . Despite the association between MCR-mediated LPS modification and colistin resistance , there is evidence that it does not prevent polymyxin-mediated damage of the OM . For example , colistin has been shown to permit ingress of the N-phenyl-1-napthylamine ( NPN ) fluorophore into the OM of E . coli expressing mcr-1 ( MacNair et al . , 2018 ) . Furthermore , colistin greatly enhances the activity of hydrophobic antibiotics such as rifampicin against polymyxin-resistant bacteria via disruption of the OM ( Brennan-Krohn et al . , 2018 ) . However , despite colistin damaging the OM of resistant bacteria , it is unable to kill or lyse them ( MacNair et al . , 2018 ) . This suggests that the modification of LPS with pEtN and/or L-ara4N protects the CM from colistin , but it is not clear how ( MacNair et al . , 2018; Brennan-Krohn et al . , 2018 ) . Improving our knowledge of how colistin kills bacteria is essential to help devise new approaches to enhance the efficacy of last resort polymyxin antibiotics ( Liu et al . , 2016; Carroll et al . , 2019; Nang et al . , 2019; Skov and Monnet , 2016; Wang et al . , 2020 ) . To do this , we set out to better understand how mcr-1 protects bacteria from colistin and to then use this information to elucidate the mode of action of colistin , with the ultimate aim of exploiting this information to improve colistin efficacy .
The first issue we wanted to resolve was whether MCR-1 protected the CM and/or OM of bacteria from colistin . To do this , we used an isogenic E . coli MC1000 strain pair , one of which expresses mcr-1 from the IPTG-inducible vector pDM1 ( mcr-1 ) to ensure consistent expression under our experimental conditions , and the other transformed with the pDM1 vector alone as a control ( pEmpty ) ( Dortet et al . , 2018; Key resources table ) . As expected , we found that E . coli MC1000 expressing mcr-1 had a significantly greater colistin minimum inhibitory concentration ( MIC , 2 µg ml−1 ) compared to the MC1000 pEmpty control strain ( 0 . 25 µg ml−1 ) , which was similar to that seen for clinical isolates ( Dortet et al . , 2018; Figure 1—figure supplement 2 ) . This confirmed that the E . coli cells were producing functional MCR-1 . To fully characterise the LPS-modifying activity of MCR-1 , we undertook MALDI-TOF-based lipidomic analysis of both whole E . coli cells and E . coli spheroplasts that lacked an OM ( Weiss and Fraser , 1973 ) . We confirmed spheroplast formation by microscopy and used FITC labelling of OM surface proteins to demonstrate removal of the OM ( Figure 1—figure supplement 3 , Figure 1—figure supplement 4 ) . Our lipidomic analysis revealed the presence of LPS modified with pEtN in both the CM and OM of mcr-1 expressing bacteria , consistent with the location of MCR-1 in the CM ( Dortet et al . , 2018; Furniss et al . , 2019; Figure 1—figure supplement 5 ) . Of note , whilst 42 ± 19% of total cellular LPS from MCR-1-producing E . coli was unmodified , the proportion of unmodified LPS in the CM was just 21 ± 2% ( Figure 1A , Figure 1—figure supplement 5 ) . We next assessed the effect of colistin on the integrity of the E . coli OM using the hydrophobic NPN dye , which fluoresces upon contact with phospholipids exposed by damage to the LPS monolayer ( MacNair et al . , 2018; Helander and Mattila-Sandholm , 2000 ) . As expected , colistin caused permeabilisation of the OM of the E . coli pEmpty strain in a dose-dependent manner ( Figure 1B ) . In agreement with previous findings , we found that colistin also disrupted the OM of E . coli expressing mcr-1 to a similar degree to E . coli pEmpty ( Figure 1B; MacNair et al . , 2018 ) . To further investigate permeabilisation of the OM by colistin , we assessed the susceptibility of bacteria to rifampicin in the presence of the polymyxin antibiotic . Rifampicin cannot normally cross the OM , which makes E . coli intrinsically resistant to the antibiotic . However , in keeping with previous work , we found that colistin sensitised E . coli expressing mcr-1 to rifampicin , with a fractional inhibitory concentration index ( FICI ) value of 0 . 14 indicating synergy between the two antibiotics in a checkerboard assay ( Figure 1—figure supplement 6; MacNair et al . , 2018 ) . This confirmed that colistin disrupted the OM of resistant bacteria producing MCR-1 . Therefore , MCR-1-mediated changes to LPS did not prevent permeabilisation of the OM by colistin , which reflects the presence of the relatively large quantity of unmodified LPS in the OM as determined in our lipidomic analysis ( Figure 1A , Figure 1—figure supplement 5 ) . Next , we assessed damage to the CM structure in the E . coli strain pair during colistin exposure , using the membrane impermeant dye propidium iodide ( PI ) . PI fluoresces upon contact with DNA in the bacterial cytoplasm , and thus is indicative of permeabilisation of the both the OM and CM of whole bacterial cells ( Allison and Lambert , 2015; Pietschmann et al . , 2009 ) . As expected , colistin exposure resulted in a strong PI signal from E . coli pEmpty cells , indicative of CM permeabilisation ( Figure 1C ) , which gradually declined , most likely due to nucleases released from lysed bacteria ( Lee et al . , 2017 ) . However , despite colistin permeabilising the OM of E . coli expressing mcr-1 , the CM of these bacteria remained intact , as demonstrated by the lack of PI-mediated fluorescence ( Figure 1C ) . In keeping with these findings , colistin caused lysis of E . coli pEmpty cells , as seen by a reduction in OD595nm readings over time ( Figure 1D ) . By contrast , E . coli cells producing MCR-1 grew in the presence of colistin despite the damage the polymyxin caused to the OM , as demonstrated by an increase in OD595nm measurements over time ( Figure 1D ) . As such , the damage caused to the OM of E . coli MCR-1 strain by colistin is likely to be minor . Taken together , these data demonstrate that MCR-1 protects the CM but not the OM from colistin-mediated permeabilisation . Although colistin was able to permeabilise the OM of E . coli expressing mcr-1 , it was possible that the pEtN modifications might reduce the ability of the antibiotic to access the periplasm and thus the CM . To negate this possibility and focus on whether MCR-1 mediated LPS modification directly protected the CM from colistin , we performed experiments using spheroplasts of our E . coli strains that lacked both OM and cell wall . To test whether LPS modifications altered the biophysical properties of the CM , we measured both membrane fluidity and surface charge of the E . coli spheroplasts using established methods . There were no differences in fluidity of the CM between E . coli pEmpty or mcr-1-expressing cells ( Figure 2—figure supplement 1 ) . As might be expected , there was a slight increase in the positive charge of the CM of the mcr-1-expressing E . coli relative to the pEmpty control , indicative of the presence of cationic pEtN modifications to LPS ( Figure 2—figure supplement 1 ) . To investigate whether this slight increase in membrane positivity was likely to be sufficient to repel colistin from the membrane , we determined the susceptibility of spheroplasts from E . coli MC1000 mcr-1 or pEmpty to colistin and compared it with the cationic antimicrobial peptides ( CAMPs ) daptomycin or nisin . Both CAMPs are well characterised for their ability to permeabilise phospholipid bilayers and , like colistin , they are positively charged , enabling us to detect whether the change to membrane charge conferred by MCR-1-modified LPS in the CM contributed specifically to polymyxin resistance ( Karas et al . , 2020; Zendo et al . , 2010 ) . Importantly , increased membrane positive charge is a common mechanism of resistance to daptomycin ( Karas et al . , 2020 ) . In the absence of treatment , there was a small but progressive loss of CM integrity over time due to the fragile nature of spheroplasts . However , allowing for this , the CM of spheroplasts of E . coli MC1000 mcr-1 was resistant to damage by colistin , but susceptible to daptomycin and nisin ( Figure 2A–C ) . By contrast , colistin , daptomycin , and nisin all permeabilised the CM of spheroplasts of E . coli pEmpty ( Figure 2A–C ) . In keeping with the data from assays measuring CM damage , colistin , daptomycin , and nisin all caused lysis of spheroplasts of E . coli pEmpty , whilst the spheroplasts from E . coli expressing mcr-1 were undamaged by colistin , but were lysed by both daptomycin and nisin ( Figure 2D–F ) . Combined , these data demonstrated that the protection afforded to the CM by MCR-1 is specific for colistin , and that the polymyxin antibiotic does not share the same target as the phospholipid-targeting CAMPs . To further explore the specificity of MCR-1-mediated LPS modifications in the CM for protection against colistin , we produced spheroplasts of E . coli with different levels of LPS modification . This revealed a clear dose-dependent relationship between the abundance of unmodified LPS in the CM and the susceptibility of spheroplasts to colistin-mediated CM damage and lysis ( Figure 2—figure supplement 2 ) . Therefore , since MCR-1 specifically modifies LPS and this modification selectively protects the CM from colistin in a dose-dependent manner , we concluded that LPS is the CM target of colistin , just as it is in the OM . To understand how colistin targeting of LPS in the CM leads to membrane disruption , we studied the role of cation bridges which are crucial for stabilising interactions between LPS molecules , by exposing spheroplasts from E . coli pEmpty cells to colistin in the absence or presence of excess magnesium . In keeping with a role for cation bridges , we found that magnesium chloride conferred dose-dependent protection from colistin-mediated CM disruption ( Figure 3A ) . To rule out a general protective osmotic effect from the higher salt concentration , we demonstrated that identical concentrations of sodium chloride did not protect spheroplasts from colistin ( Figure 3B ) . Furthermore , the presence of exogenous magnesium had no significant effect on reducing spheroplast CM damage caused by daptomycin or nisin ( Figure 3C , D ) , confirming that these CAMPs do not have the same CM target as colistin . In conclusion , our findings demonstrate that colistin targets LPS in the CM of polymyxin-susceptible E . coli , leading to the displacement of cationic inter-LPS bridges , membrane disruption , and ultimately bacterial lysis . This is similar to the mechanism by which colistin disrupts the OM of bacteria and synthetic phospholipid bilayer membranes containing low levels of LPS ( Khadka et al . , 2018; Moore and Hancock , 1986; D'amato et al . , 1975 ) . However , the high levels of LPS modified by pEtN in the CM of MCR-1-producing E . coli prevent colistin from targeting LPS in the CM , protecting the membrane and conferring resistance to the polymyxin antibiotic . Having determined that colistin kills bacteria by targeting LPS in the CM , we wanted to use this information to develop a new therapeutic approach to enhance colistin efficacy . Murepavadin is a first-in-class peptide-based inhibitor of the LptD component of the LptABCDEFG complex of P . aeruginosa that transports LPS from the CM to the OM ( Andolina et al . , 2018 ) . Thus , inhibition of the Lpt system in P . aeruginosa leads to LPS accumulation in the CM , which we hypothesised would increase the susceptibility of the bacterium to colistin ( Andolina et al . , 2018; Sperandeo et al . , 2008 ) . To test our hypothesis , we first used a checkerboard MIC assay and found that colistin synergised with murepavadin against P . aeruginosa PA14 cells ( FICI value of 0 . 375 ) , revealing that sub-lethal concentrations of the LptD inhibitor sensitised the bacterium to colistin ( Figure 4A; Odds , 2003 ) . To confirm that sub-lethal concentrations of murepavadin altered LPS abundance in the CM , P . aeruginosa was incubated with murepavadin , before the amount of LPS in whole cells and spheroplasts was measured using the well-established Limulus Amoebocyte Lysate ( LAL ) assay , since previous work has shown this approach to be a highly accurate way of quantifying LPS in whole cell lysates ( Figure 4—figure supplement 1 , Figure 4—figure supplement 2; Hoppe Parr et al . , 2017 ) . The suitability of the LAL assay was further confirmed using a MALDI-TOF-based lipidomic analysis of spheroplast lysates , which confirmed that the LPS in both murepavadin-exposed and untreated bacteria was unmodified and thus able to be accurately detected and quantified ( Figure 4—figure supplement 3; Takayama et al . , 1984 ) . Sub-lethal concentrations of murepavadin caused a slight reduction in LPS levels in the OM of P . aeruginosa cells , but a significant increase in the amount of LPS in the CM compared to untreated spheroplasts ( Figure 4—figure supplement 3 , Figure 4—figure supplement 4 ) . Moreover , our lipidomic analysis revealed that the ratio of lipid A:phospholipid increased in P . aeruginosa spheroplasts pre-exposed to murepavadin , confirming that the LptD inhibitor caused LPS to accumulate in the CM ( Figure 4—figure supplement 3 ) . Next , we proceeded to test whether LPS accumulation in the CM increased the susceptibility of P . aeruginosa to colistin . We started by examining the effect of colistin on the OM and CM of P . aeruginosa exposed , or not , to murepavadin . Despite the slight reduction of LPS at the OM caused by murepavadin , colistin permeabilised the OM to the same extent as bacteria that had not been exposed to murepavadin , with similar levels of NPN uptake ( Figure 4B ) . By contrast , however , murepavadin significantly enhanced permeabilisation of the CM by colistin in whole cells of P . aeruginosa , as determined via uptake of PI ( Figure 4C ) . Thus , an increase in LPS levels in the CM promoted colistin-mediated damage , in keeping with our conclusion that LPS in the CM is the target of the polymyxin antibiotic . Furthermore , P . aeruginosa cells exposed to murepavadin were more rapidly lysed by colistin than untreated cells ( Figure 4D ) . Taken together , these findings indicated that LPS accumulation in the CM increased susceptibility to the polymyxin antibiotic . However , an alternative explanation for the synergy between colistin and murepavadin was that polymyxin-mediated damage to the OM enabled murepavadin greater access to LptD in the periplasm . To test this , we first examined whether the OM permeabilising agent polymyxin B nonapeptide ( PMBN ) also synergised with murepavadin . However , we did not see synergy in MIC checkerboard or bactericidal assays between PMBN and murepavadin ( Figure 4—figure supplement 5 , Figure 4—figure supplement 6 ) . Next , we pre-treated P . aeruginosa with murepavadin alone to cause LPS accumulation in the CM and then removed the murepavadin by washing before converting the whole cells to spheroplasts and exposing these to colistin alone . The murepavadin pre-treated spheroplasts were much more susceptible to colistin-induced CM damage and lysis than untreated spheroplasts ( Figure 4—figure supplement 7 ) . Therefore , colistin does not sensitise P . aeruginosa to murepavadin by compromising the OM; rather , murepavadin-induced accumulation of LPS in the CM potentiates the activity of colistin . Taken together , these experiments provided further evidence that colistin targets LPS in the CM and suggest that murepavadin and colistin might form a useful combination therapy . P . aeruginosa is a major cause of chronic lung infection in people with cystic fibrosis ( CF ) and bronchiectasis . In both conditions , disease severity can rapidly increase during episodes of ‘exacerbation’ which must be treated aggressively to quickly restore lung function and reduce long-term damage ( Polverino et al . , 2017; Karampitsakos et al . , 2020; Stanford et al . , 2021 ) . Therefore , having shown that murepavadin sensitised the CM to colistin-mediated damage , we wanted to determine whether this translated into enhanced antibacterial activity against relevant clinical isolates and increased treatment efficacy in vivo . We found that a sub-lethal concentration of murepavadin sensitised P . aeruginosa PA14 to a normally sub-lethal concentration of colistin ( 2 µg ml−1 ) , resulting in >10 , 000-fold reduction in c . f . u . counts relative to bacteria incubated with murepavadin or colistin alone after 8 hr ( Figure 5A ) . We also found that murepavadin potentiated the activity of even lower concentrations of colistin ( 1 µg ml−1 ) , with exposure to the LPS transport inhibitor increasing the ability of the polymyxin antibiotic to damage the CM , triggering bacterial lysis and cell death ( Figure 5—figure supplement 1 ) . We next examined a panel of 15 multi-drug resistant P . aeruginosa clinical strains , isolated from the sputum of CF patients , to investigate whether murepavadin increased colistin-mediated bacterial killing ( Figure 5—figure supplement 2 ) . Of these 15 clinical isolates , 14 were susceptible to murepavadin alone , whilst one strain was resistant to the LptD inhibitor ( Supplementary file 1 ) . In 11 out of 14 murepavadin-susceptible CF isolates tested ( 79% ) , sub-lethal concentrations of murepavadin caused a significant increase in the bactericidal activity of colistin against P . aeruginosa ( Figure 5B ) . Importantly , murepavadin did not affect the bactericidal activity of colistin against the strain that was resistant to the LptD inhibitor ( Figure 5B ) . This confirmed that the potentiating effects of the LptD inhibitor on polymyxin-mediated killing were not due to off-target effects . Next , we employed a high inoculum P . aeruginosa murine lung infection model and used a short treatment duration to assess how quickly combined colistin and murepavadin therapy could reduce bacterial burden , relative to mono-therapy . Mice were inoculated via the intranasal route with P . aeruginosa PA14 to cause a lung infection , and then treated intranasally with PBS alone , or PBS containing colistin only ( 5 mg kg−1 ) , murepavadin only ( 0 . 25 mg kg−1 ) , or colistin and murepavadin combined at the concentrations used for mono-treatment . These concentrations were based on those used previously to mimic treatment of human lung infections , and the route of delivery is similar to that used clinically ( Bernardini et al . , 2019; Yapa et al . , 2014; Melchers et al . , 2019; Aoki et al . , 2009 ) . Mono-therapy with colistin alone or murepavadin alone had very little effect on the bacterial load assessed after 3 hr treatment compared with the no-treatment control ( Figure 5C ) . By contrast , combination therapy with colistin and murepavadin caused a ~500-fold reduction in c . f . u . counts relative to the no-treatment control ( Figure 5C ) . Therefore , murepavadin synergises with colistin both in vitro and in vivo , suggesting it may be useful as a combination therapeutic approach for lung infections caused by P . aeruginosa .
Colistin is an increasingly important last-resort antibiotic used to treat infections caused by multi-drug-resistant Gram-negative pathogens , including P . aeruginosa , K . pneumoniae , and E . coli ( Garg et al . , 2017; Biswas et al . , 2012; Thomas et al . , 2019 ) . However , treatment failure occurs frequently , and resistance is a growing concern ( Falagas et al . , 2010; Paul et al . , 2018; Linden et al . , 2003; Tran et al . , 2016; Satlin et al . , 2020; MacNair et al . , 2018 ) . Efforts to address these issues are compromised by a poor understanding of colistin’s bactericidal mode of action . Whilst the initial interactions of colistin with LPS in the OM of Gram-negative bacteria were well established , it was unclear how the antibiotic traversed the OM and damaged the CM to cause cell lysis ( Figure 1—figure supplement 1 ) . In this work , we demonstrate that colistin targets LPS in the CM , resulting in membrane permeabilisation , bacterial lysis , and killing ( Figure 6 ) . Our conclusion that colistin targets LPS in the CM was based initially on experiments with E . coli expressing the mcr-1 colistin resistance determinant . MCR-1 modifies lipid A with a pEtN moiety as it passes through the CM on its way to the OM ( Liu et al . , 2016; Nang et al . , 2019 ) . Since MCR-1 specifically protected spheroplasts from colistin but not nisin or daptomycin , which both target phospholipid bilayers , it was clear that colistin did not share the same target as the other two CAMPs . Given the only difference between E . coli spheroplasts expressing mcr-1 and pEmpty spheroplasts was modified LPS , our data reveal that colistin targets LPS in the CM , leading to disruption of the CM , which is a pre-requisite for subsequent cell lysis and bacterial killing ( Allison and Lambert , 2015 ) . These findings were then supported by experiments showing that LPS accumulation in the CM of P . aeruginosa sensitised this bacterium to colistin . Similar to Salmonella and E . coli , the abundance of LPS in the CM of P . aeruginosa was found to be ~100 times lower than the OM , indicating that only about 1% of total LPS is present in the CM ( Zendo et al . , 2010; Osborn et al . , 1972 ) . However , studies with model membranes have shown that the presence of low concentrations of LPS ( 1% total composition ) in phospholipid bilayer membranes was both necessary and sufficient for colistin-mediated permeabilisation ( Khadka et al . , 2018 ) . Therefore , our conclusion explains how an antibiotic with a high degree of specificity for LPS could damage both the OM and CM ( Velkov et al . , 2010 ) . The reason why MCR-1 protected the CM but not OM from colistin-mediated damage is most likely due to the lower proportion of unmodified LPS at the CM ( 21 ± 2% ) relative to the OM ( 42 ± 19% ) ( Figure 1A ) . Furthermore , the overall abundance of LPS in the CM is very low , resulting in very few targets ( i . e . unmodified LPS molecules ) for colistin in the CM of mcr-1-expressing E . coli . By contrast , the OM of MCR-1-producing E . coli contains many more unmodified LPS molecules that can be bound by colistin , explaining why colistin is able to damage this structure , but cannot permeabilise the CM of E . coli expressing mcr-1 at a physiologically relevant concentration ( Khadka et al . , 2018 ) . Whether colistin resistance conferred by chromosomal mutations in two-component systems is also mediated by modified LPS in the CM remains to be tested ( Carroll et al . , 2019; Nang et al . , 2019; Skov and Monnet , 2016; Wang et al . , 2020; Raetz et al . , 2007; Simpson and Trent , 2019 ) . Our data showing that colistin requires unmodified LPS to be present in the CM to kill bacteria explains how an antibiotic with high affinity and specificity for LPS causes disruption to both the OM and CM ( Velkov et al . , 2010 ) . Furthermore , these findings provide support for the observations that colistin does not damage the OM of colistin-resistant A . baumannii isolates where LPS has been replaced by a phospholipid bilayer , and that polymyxins cause only minimal disruption to model phospholipid membranes unless LPS is present ( Khadka et al . , 2018; Moffatt et al . , 2010; Zhang et al . , 2018; Fu et al . , 2020 ) . Whilst the interaction of colistin with LPS in the CM is likely to share similarities with the same process at the OM , there are also likely to be differences owing to the differing concentrations of LPS between the two membranes ( Khadka et al . , 2018; Zendo et al . , 2010; Osborn et al . , 1972 ) . In the OM , LPS is a highly abundant component with molecules tightly packed and stabilised with cation bridges . By contrast , LPS is a minority component in the CM , which may affect the rate and degree to which the CM is disrupted by polymyxins . In support of this , whilst colistin induced OM damage within minutes of bacterial exposure to the antibiotic , disruption of the CM took much longer . Even when spheroplasts lacking an OM were exposed to colistin , it still took more than 2 hr for CM permeabilisation to occur . Therefore , it appears that colistin-mediated disruption of the CM is considerably less efficient than that of the OM , likely due to the much lower levels of LPS present in the CM . In addition to disruption of both the OM and IM , it has been proposed that the lethal activity of polymyxin antibiotics may be due , at least in part , to: disruption of NADH-quinone reductase; the generation of reactive oxygen species ( ROS ) ; the binding of the antibiotic to ribosomes; and the fusion of the OM and CM , leading to phospholipid exchange ( Cajal et al . , 1996; Clausell et al . , 2006; Clausell et al . , 2007; Deris et al . , 2014; Ajiboye et al . , 2018; Li and Velkov , 2019; Ayoub Moubareck , 2020; El-Sayed Ahmed et al . , 2020 ) . However , whilst these have been considered as discrete events or alternative mechanisms of action , it is possible that all these phenomena occur as downstream consequences of colistin-mediated CM disruption . For example , NADH-quinone reductase is a component of the electronic transport chain ( ETC ) , which is located within the CM and may therefore be disrupted by membrane damage , whilst the generation of ROS may arise via disruption of the ETC as has been proposed for the CAMP LL-37 ( Deris et al . , 2014; Ayoub Moubareck , 2020; Choi et al . , 2017 ) . The fusion of the OM and CM and subsequent exchange of lipids appears to depend upon the interaction of the polymyxin with , and presumably disruption of , both membranes ( Cajal et al . , 1996; Clausell et al . , 2006; Clausell et al . , 2007; Ayoub Moubareck , 2020; El-Sayed Ahmed et al . , 2020 ) . Finally , the interaction of polymyxins with ribosomes requires the antibiotic to pass through the CM to access the cytoplasm ( McCoy et al . , 2013 ) . Therefore , whilst the disruption of the CM by polymyxin antibiotics is the key step required for bacterial killing , this may be due to multiple deleterious effects on cellular processes . Our findings provide strong evidence that colistin targets LPS in the CM , in addition to the OM , and that this is required for the bactericidal and lytic activity of the antibiotic at clinically relevant concentrations . This insight into the mode of action of colistin enabled us to devise a new therapeutic approach to enhance colistin efficacy . Using the LptD inhibitor murepavadin , which is in development as an inhaled treatment for P . aeruginosa infections , we triggered LPS accumulation in the CM , and thereby increased the susceptibility of bacteria to colistin ( Lehman and Grabowicz , 2019 ) . The potential clinical utility of this approach was demonstrated by showing enhanced activity of colistin–murepavadin combination therapy against a panel of clinical CF isolates , as well as potent efficacy in a murine model of P . aeruginosa lung infection . It is anticipated that a combination of colistin and murepavadin could enhance the low treatment efficacy of polymyxin antibiotics and may also limit the toxic side effects associated with both compounds by enabling the use of lower doses of the drugs ( Lehman and Grabowicz , 2019 ) . Interestingly , whilst we found that blocking LPS transport to the OM sensitised bacteria to colistin , previous work has shown that novobiocin increases the susceptibility of bacteria to colistin by increasing LPS transport to the OM ( Mandler et al . , 2018 ) . This might suggest that novobiocin reduces LPS levels in the CM and thus contradicts our findings . However , transport of LPS to the OM is regulated such that this process does not deplete LPS in the CM ( Xie et al . , 2018 ) . Furthermore , LPS biosynthesis is tightly regulated at the CM in response to the abundance of LPS by PbgA , LapB , and FtsH ( Clairfeuille et al . , 2020; Guest et al . , 2020; Fivenson and Bernhardt , 2020; O'Rourke et al . , 2020; Lee et al . , 2006 ) . Therefore , it is expected that LPS synthesis would be increased to address the elevated rate of transport to the OM , which might lead to elevated levels of LPS in the CM as it is produced to meet demand . In support of this , novobiocin exposure increases the expression of lpxC in E . coli , which encodes the enzyme that is the first committed step in LPS biosynthesis and provides a key checkpoint in LPS production ( Raetz et al . , 2007; Simpson and Trent , 2019; O'Rourke et al . , 2020 ) . However , the effect of novobiocin on LPS abundance in the CM remains to be tested . It should be noted that whilst bacteria can modulate LPS biosynthesis to maintain LPS abundance in the CM , there is no mechanism to remove LPS from the CM , which is why LPS accumulation occurs with murepavadin . In summary , this work contributes to our understanding of the mechanism of action of colistin by demonstrating that polymyxin antibiotics target LPS in both the OM and the CM , and that this leads to the disruption of both membranes , resulting in the bactericidal and lytic activities of the antibiotic . Modulation of LPS levels in the CM can enhance colistin activity , providing the foundations for new approaches to enhance the efficacy of this antibiotic of last resort .
The bacterial strains used in this study are listed in Key resources table . For each experiment , all strains were grown in Luria broth ( LB; Thermo Fisher Scientific , USA ) for 18 hr to stationary phase at 37°C with shaking ( 180 r . p . m . ) . For routine culture of bacteria on solid media , strains were grown on LB supplemented with 1 . 5% technical agar ( BD Biosciences , USA ) . Liquid and solid growth media were supplemented with tetracycline ( 12 . 5 μg ml−1; Sigma-Aldrich , USA ) and isopropyl-β-D-thiogalactoside ( IPTG , Melford Laboratories , UK; used at 0 . 5 mM unless stated otherwise ) where required . Enumeration of bacterial c . f . u . was done by plating 10-fold serial dilutions of bacterial cultures on to Mueller-Hinton agar ( MHA; Thermo Fisher Scientific ) plates . Inoculated agar plates were incubated statically for 18 hr in air at 37°C . The MIC of colistin and murepavadin against bacterial strains was determined by the broth microdilution protocol ( Wiegand et al . , 2008 ) . A microtitre plate was used to prepare a range of antibiotic concentrations in 200 μl cation-adjusted Mueller-Hinton broth ( CA-MHB; Thermo Fisher Scientific ) by two-fold serial dilutions . For certain experiments , checkerboard analyses were performed by preparing two-fold serial dilutions of two antibiotics in different directions , generating an 8 × 8 matrix to assess the MICs of the relevant antibiotics in combination , with FICI values calculated as previously described ( Odds , 2003 ) . Stationary-phase bacteria were diluted 1000-fold in fresh CA-MHB and seeded into each well of the microtitre plate to a final concentration of 5 × 105 c . f . u . ml−1 . The microtitre plates were then incubated statically at 37°C for 18 hr in air , after which point the MIC was defined as the lowest antibiotic concentration at which there was no visible growth of bacteria . In some cases , the extent of bacterial growth after 18 hr incubation was also determined by obtaining OD595nm measurements using a Bio-Rad iMark microplate absorbance reader ( Bio-Rad Laboratories , USA ) . Stationary-phase bacteria were diluted 1000-fold in fresh CA-MHB , and 4 μl was seeded into the wells of a microtitre plate containing 200 μl CA-MHB , and for some experiments the LptD inhibitor murepavadin , to give a final inoculum of 5 × 105 c . f . u . ml−1 . The microtitre plate was incubated with shaking ( 180 r . p . m . ) at 37°C for 16 hr in a Tecan Infinite 200 Pro multiwell plate reader ( Tecan Group Ltd . , Switzerland ) and optical density measurements were taken at 600 nm every 15 min . Spheroplasts of E . coli and P . aeruginosa strains lacking an OM and cell wall were generated as previously described ( Weiss and Fraser , 1973 ) . Briefly , stationary-phase bacteria grown overnight were washed twice by centrifugation ( 12 , 300 × g , 3 min ) followed by resuspension in CA-MHB , and added at a final inoculum of 108 c . f . u . ml−1 to 9 ml CA-MHB containing for some experiments varying concentration of IPTG , or the LPS transport inhibitor murepavadin . Cultures were then incubated at 37°C with shaking ( 180 r . p . m . ) for 2 hr . After the incubation , bacteria were washed twice by centrifuging ( 3273 × g , 20 min , 4°C ) and resuspending first in 10 ml Tris buffer ( 0 . 03 M , pH 8 . 0; Sigma-Aldrich ) , and subsequently in Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose . EDTA ( 250 µl , 10 mg ml−1; Sigma-Aldrich ) and lysozyme ( 1 ml , 10 mg ml−1; Roche , Switzerland ) were added to remove the OM and cell wall respectively , and the cell suspension was incubated for 1 hr in a water bath shaker at 30°C . Trypsin ( 500 µl , 10 mg ml−1; Sigma-Aldrich ) was then added , and the culture again incubated at 30°C in a water bath shaker for 15 min . The resulting spheroplasts produced were harvested by mild centrifugation ( 2000 × g , 20 min , 4°C ) , with the supernatant containing the removed OM extracted for further analysis . Successful conversion of bacterial whole cells into spheroplasts was confirmed using phase-contrast microscopy , as detailed below . Whole cells of E . coli and P . aeruginosa grown overnight were washed twice by centrifugation ( 12 , 300 × g , 3 min ) and resuspension in CA-MHB , added at a final inoculum of 108 c . f . u . ml−1 to 9 ml CA-MHB , and incubated for 2 hr at 37°C with shaking ( 180 r . p . m . ) . OM proteins of these bacteria were subsequently labelled with fluorescein isothiocyanate ( FITC , Sigma-Aldrich ) as previously described ( Loh and Ward , 2012 ) . Bacterial cells were washed twice by centrifugation ( 3273 × g , 20 min , 4°C ) and resuspension in 10 ml Labelling Buffer ( 50 mM Na2CO3 , 100 mM NaCl , pH 8 . 0 ) , to which FITC was added at a final concentration of 0 . 5 mg ml−1 . Bacteria were incubated for 30 min at room temperature , before labelled cells were harvested by centrifuging ( 3273 × g , 20 min , 4°C ) and washed thrice by resuspending in 10 ml Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose . 1 ml of FITC-labelled bacteria was extracted and centrifuged ( 12 , 300 × g , 3 min ) , and the cells were fixed in 4% paraformaldehyde ( PFA ) in phosphate-buffered saline ( PBS ) . The remaining 9 ml of FITC-labelled cells were converted into spheroplasts , as described above . The spheroplasts produced were recovered by mild centrifugation ( 2000 × g , 20 min , 4°C ) and resuspension in 9 ml Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose , before 1 ml of spheroplasts were fixed in the same way as with whole cells . The amount of FITC fluorescence in the OM of whole cells and CM of spheroplasts was observed using fluorescence microscopy , as described below . For quantification of FITC fluorescence , 200 µl samples of the fixed bacterial suspensions were seeded into the wells of a black-walled microtitre plate , and fluorescence measured with a Tecan Infinite 200 Pro multiwell plate reader , using an excitation wavelength of 490 nm and an emission wavelength of 525 nm . For phase-contrast and fluorescence microscopy , a 5 µl sample of fixed bacterial whole cells or spheroplasts was spotted onto a thin 1 . 2% agarose gel patch prepared in distilled water on a microscope slide . Bacteria were imaged using an Axio Imager . A2 Zeiss microscope ( Carl Zeiss Microscopy GmbH , Germany ) at 1000× magnification with an oil immersion objective lens . The ZEN 2012 software was used for image acquisition , whilst analysis of cell length:width ratios was done using the FIJI/ImageJ software by measuring two perpendicular lines drawn through the centre of bacteria . For each experiment , all microscopy images were acquired and processed using identical settings throughout . Spheroplasts from bacterial cells were prepared as described above and then resuspended in ddH2O ( 200 μl ) , before mild acid hydrolysis was performed via the addition of 2% ( vol/vol ) acetic acid in ddH2O ( 200 μl ) and incubation at 100°C for 30 min . For experiments with whole cells , bacteria grown overnight to stationary-phase were washed three times by centrifuging and resuspending in ddH2O , and a mild acid hydrolysis was performed on these whole cells as described for spheroplasts . Acid-treated whole cells or spheroplasts were recovered by centrifugation ( 17 , 000 × g , 2 min ) , and the resulting pellet was washed before being resuspended in 50 μl ultrapure water . The whole cell or spheroplast suspension ( 0 . 5 μl ) was then loaded immediately onto the target and overlaid with 1 . 2 μl of a matrix consisting of 9H-Pyrido[3 , 4-B]indole ( Norharmane ) ( Sigma-Aldrich ) dissolved in 90:10 ( vol/vol ) chloroform/methanol to a final concentration of 10 mg ml−1 . The bacterial suspension and matrix were then mixed on the target before gentle drying under air at room temperature . MALDI-TOF mass spectroscopy analysis was undertaken with a MALDI Biotyper Sirius system ( Bruker Daltonics , USA ) , using the linear negative-ion mode as described previously ( Furniss et al . , 2019 ) . Manual peak picking at masses relevant to phospholipids or lipid A was performed on the mass spectra obtained , and the corresponding signal intensities at the defined masses were determined . Peaks were considered only if their signal/noise ratio was at least 5 . To determine the ratio of modified lipid A to unmodified lipid A , the area under the pETN-modified lipid A peak ( m/z 1 , 919 . 2 ) was divided by the area under the peak corresponding to native lipid A ( m/z 1 , 796 . 2 ) . To determine the relative abundance of LPS , the sum of the area under the lipid A peaks ( m/z 1447–1700 ) was divided by the sum of the area under representative phospholipid peaks ( phospholipid 34:1 , 2 , m/z 717–747 ) . All mass spectra were generated and analysed with three biological replicates and two technical replicates . To detect damage to the OM of bacteria , the well-established NPN uptake assay was used ( Helander and Mattila-Sandholm , 2000 ) . Stationary-phase bacterial cells were washed in fresh CA-MHB and diluted to an optical density ( OD600nm ) of 0 . 5 in 5 mM pH 7 . 2 HEPES buffer ( Sigma-Aldrich ) . This bacterial suspension was added to wells containing the relevant antibiotics in HEPES buffer , as well as the fluorescent probe N-phenyl-1-naphthylamine ( NPN; Acros Organics , USA ) at a final concentration of 10 μM . Samples were placed in a black microtitre plate with clear-bottomed wells and fluorescence measured immediately in a Tecan Infinite 200 Pro multiwell plate using an excitation wavelength of 355 nm and an emission wavelength of 405 nm . Fluorescence measurements were obtained every 30 s for 10 min , and the degree of OM permeabilisation , referred to as the NPN Uptake Factor , was calculated using the following formula:Fluorescence of sample with NPN−Fluorescence of sample without NPNFluorescence of HEPES buffer with NPN−Fluorescence of HEPES buffer without NPN To measure CM disruption of whole cells , bacteria grown to stationary-phase overnight were washed and inoculated into 3 ml MHB containing the relevant antibiotics . Cultures were incubated at 37°C with shaking ( 180 r . p . m . ) for up to 8 hr , and every 30 min , aliquots ( 200 μl ) were taken and bacteria isolated by centrifugation ( 12 , 300 × g , 3 min ) . Cells were then washed in sterile PBS before being added to the wells of a black-walled microtitre plate , and PI ( Sigma-Aldrich ) was added to each well at a final concentration of 2 . 5 μM . Fluorescence was measured immediately in a Tecan Infinite 200 Pro multiwell plate reader ( excitation at 535 nm , emission at 617 nm ) . To measure disruption of the CM in spheroplasts , spheroplasts of E . coli and P . aeruginosa generated as detailed above were washed by centrifugation ( 4000 × g , 5 min ) and resuspension in Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose . Spheroplast samples ( 20 µl ) were then added in the wells of a black-walled microtitre plate to 180 µl of Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose , the relevant antibiotics , and PI at a final concentration of 0 . 25 μM . The microtitre plate was incubated with shaking ( 180 r . p . m . ) at 37°C for up to 8 hr in a Tecan Infinite 200 Pro multiwell plate reader and fluorescence ( excitation at 535 nm , emission at 617 nm ) measured every 15 min using a gain of 80 or 100 . For both whole bacterial cells and spheroplasts , to account for differences in fluorescence values arising from variations in cell number , relative fluorescence unit ( r . f . u . ) measurements were corrected for OD at 600 nm . In the case of whole bacterial cells , washed stationary-phase bacteria were inoculated into 3 ml CA-MHB containing the relevant antibiotics , as described above . Cultures were then placed in a shaking incubator ( 37°C , 180 r . p . m . ) for 8 hr , and every 30 min , samples ( 200 μl ) were transferred to a microtitre plate , where OD595nm measurements were obtained using a Bio-Rad iMark microplate absorbance reader . For spheroplasts , washed spheroplasts ( 20 µl ) were added to 180 µl of Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose and the relevant antibiotics in a microtitre plate as detailed above . The microtitre plate was incubated for up to 8 hr at 37°C with shaking ( 180 r . p . m . ) in a Tecan Infinite 200 Pro multiwell plate reader , and readings of OD600nm were made every 15 min . The fluidity of the CM of spheroplasts was assessed using the fluorescent dye Laurdan , as previously described ( Müller et al . , 2016 ) . Washed spheroplasts of E . coli ( 500 µl ) prepared as described above were incubated at room temperature for 5 min in Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose and Laurdan at a final concentration of 100 µM . Spheroplast samples were washed by three rounds of centrifugation ( 4000 × g , 5 min ) and resuspension in Tris buffer containing 20% sucrose , then 200 µl was transferred to the wells of a black-walled microtitre plate . Membrane fluidity was measured using a Tecan Infinite 200 Pro multiwell plate reader , with fluorescence determined using an excitation wavelength of 330 nm , and emission wavelengths of 460 nm and 500 nm . Generalised Polarisation ( GP ) values were calculated using the following formula:GP=Emission intensity at 460 nm−Emission intensity at 500 nm Emission intensity at 460 nm+Emission intensity at 500 nm A higher GP value indicated a membrane with reduced fluidity , with altered water penetration into the membrane affecting the fluorescence of the Laurdan dye . The charge of the CM of spheroplasts was measured using FITC-labelled Poly-L-Lysine ( PLL ) ( Jones et al . , 2008 ) . Washed E . coli spheroplasts ( 300 µl ) generated as described above were incubated in the dark for 10 min at room temperature in Tris buffer ( 0 . 03 M , pH 8 . 0 ) containing 20% sucrose and FITC-PLL at a final concentration of 20 µg ml−1 . To remove any unbound PLL , spheroplasts were subsequently washed thoroughly by three rounds of centrifugation ( 4000 × g , 5 min ) and resuspension in Tris buffer containing 20% sucrose . Spheroplast samples ( 200 µl ) were seeded into the wells of a black-walled microtitre plate , and FITC fluorescence was quantified in a Tecan Infinite 200 Pro multiwell plate reader ( excitation at 490 nm , emission at 525 nm ) . Reduced PLL binding to the surface of spheroplasts indicated a more positively charged membrane , with the cationic FITC fluorophore having less affinity for the CM . Stationary-phase bacteria grown overnight were washed and grown for 2 hr , before conversion to spheroplasts as described above . During formation of spheroplasts , the OM extracted from the bacterial cells was recovered , and the concentration of LPS in the OM , as well as the concentration of LPS in the CM of spheroplasts , was determined using the chromogenic Limulus Amebocyte Lysate ( LAL ) assay ( all reagents from Thermo Fisher Scientific ) as described previously ( Lam et al . , 2014 ) . OM samples and spheroplasts lysed by freeze-thaw to release LPS were diluted in 10-fold steps , and 50 μl of each sample was equilibrated to 37°C and loaded into the wells of a microtitre plate at the same temperature . LAL reagent ( 50 μl ) was added to each well , and the mixture incubated at 37°C for 10 min . Chromogenic substrate solution ( 100 μl , 2 mM ) was subsequently added to each well and the microtitre plate was incubated for a further 6 min at 37°C . The enzymatic reaction was stopped by adding 50 μl of 25% acetic acid to each well , and the presence of LPS was determined by measuring absorbance at 405 nm in a Tecan Infinite 200 Pro multiwell plate reader . A standard curve was generated using an E . coli endotoxin standard stock solution , which enabled the conversion of A405nm values into concentrations of LPS . As described above , stationary-phase bacteria grown overnight were washed twice and added at a final inoculum of 108 c . f . u . ml−1 to 3 ml CA-MHB containing colistin and/or murepavadin . Cultures were incubated with shaking ( 37°C , 180 r . p . m . ) for up to 8 hr . Bacterial survival was determined after 2 , 4 , 6 , and 8 hr by serially diluting cultures in 10-fold steps in 200 μl sterile PBS ( VWR International , USA ) , before enumeration of c . f . u . counts on MHA plates . The use of mice was performed under the authority of the UK Home Office outlined in the Animals ( Scientific Procedures ) Act 1986 after ethical review by Imperial College London Animal Welfare and Ethical Review Body ( PPL 70/7969 ) . Wild-type C57BL/6 mice were purchased from Charles River ( UK ) . All mice were female and aged between 6 and 8 weeks . Mice were housed with five per cage with Aspen chip 2 bedding and 12 h light/dark cycles at 20–22°C . Mice were randomly assigned to experimental groups . Water was provided ad libitum and mice were fed RM1 ( Special Diet Services ) . To establish colonisation of the lungs , mice were anesthetised and intranasally inoculated with 107 c . f . u . of P . aeruginosa PA14 in 50 μl of PBS , as described previously ( Clarke , 2014; Brown et al . , 2017 ) . Infection was allowed to establish for 5 hr , before mice were again anaesthetised and treated via the intranasal route with 50 μl of PBS alone , or PBS containing colistin ( 5 mg kg−1 ) , murepavadin ( 0 . 25 mg kg−1 ) , or a combination of colistin and murepavadin for 3 hr . To enumerate bacterial load in the lungs , mice were humanely sacrificed , their lungs removed and homogenised in PBS , and then plated onto Pseudomonas isolation agar ( Thermo Fisher Scientific ) . Experiments were performed on at least three independent occasions , and the resulting data are presented as the arithmetic mean of these biological repeats , unless stated otherwise . Error bars , where shown , represent the standard deviation of the mean . For single comparisons , a two-tailed Student’s t-test was used to analyse the data . For multiple comparisons at a single time point or concentration , data were analysed using a one-way analysis of variance ( ANOVA ) or a Kruskal–Wallis test . Where data were obtained at several different time points or concentrations , a two-way ANOVA was used for statistical analyses . Appropriate post hoc tests ( Dunnett’s , Tukey’s , Sidak’s , Dunn’s ) were carried out to correct for multiple comparisons , with details provided in the figure legends . Asterisks on graphs indicate significant differences between data , and the corresponding p-values are reported in the figure legend . All statistical analyses were performed using GraphPad Prism 7 software ( GraphPad Software Inc , USA ) . | Antibiotics are life-saving medicines , but many bacteria now have the ability to resist their effects . For some infections , all frontline antibiotics are now ineffective . To treat infections caused by these highly resistant bacteria , clinicians must use so-called ‘antibiotics of last resort’ . These antibiotics include a drug called colistin , which is moderately effective , but often fails to eradicate the infection . One of the challenges to making colistin more effective is that its mechanism is poorly understood . Bacteria have two layers of protection against the outside world: an outer cell membrane and an inner cell membrane . To kill them , colistin must punch holes in both . First , it disrupts the outer membrane by interacting with molecules called lipopolysaccharides . But how it disrupts the inner membrane was unclear . Bacteria have evolved several different mechanisms that make them resistant to the effects of colistin . Sabnis et al . reasoned that understanding how these mechanisms protected bacteria could reveal how the antibiotic works to damage the inner cell membrane . Sabnis et al . examined the effects of colistin on Escherichia coli bacteria with and without resistance to the antibiotic . Exposing these bacteria to colistin revealed that the antibiotic damages both layers of the cell surface in the same way , targeting lipopolysaccharide in the inner membrane as well as the outer membrane . Next , Sabnis et al . used this new information to make colistin work better . They found that the effects of colistin were magnified when it was combined with the experimental antibiotic murepavadin , which caused lipopolysaccharide to build up at the inner membrane . This allowed colistin to punch more holes through the inner membrane , making colistin more effective at killing bacteria . To find out whether this combination of colistin and murepavadin could work as a clinical treatment , Sabnis et al . tested it on mice with Pseudomonas aeruginosa infections in their lungs . Colistin was much better at killing Pseudomonas aeruginosa and treating infections when combined with murepavadin than it was on its own . Pseudomonas aeruginosa bacteria can cause infections in the lungs of people with cystic fibrosis . At the moment , patients receive colistin in an inhaled form to treat these infections , but it is not always successful . The second drug used in this study , murepavadin , is about to enter clinical trials as an inhaled treatment for lung infections too . If the trial is successful , it may be possible to use both drugs in combination to treat lung infections in people with cystic fibrosis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2021 | Colistin kills bacteria by targeting lipopolysaccharide in the cytoplasmic membrane |
We study the effects of inbreeding in a dioecious plant on its interaction with pollinating insects and test whether the magnitude of such effects is shaped by plant individual sex and the evolutionary histories of plant populations . We recorded spatial , scent , colour , and rewarding flower traits as well as pollinator visitation rates in experimentally inbred and outbred , male and female Silene latifolia plants from European and North American populations differing in their evolutionary histories . We found that inbreeding specifically impairs spatial flower traits and floral scent . Our results support that sex-specific selection and gene expression may have partially magnified these inbreeding costs for females , and that divergent evolutionary histories altered the genetic architecture underlying inbreeding effects across population origins . Moreover , the results indicate that inbreeding effects on floral scent may have a huge potential to disrupt interactions among plants and nocturnal moth pollinators , which are mediated by elaborate chemical communication .
Plant-pollinator interactions are of central importance for the emergence as well as the maintenance of global biodiversity ( Crepet and Niklas , 2009; Ollerton , 2017 ) and provide ecosystem services with tangible sociocultural and economic value ( Gill et al . , 2016; Porto et al . , 2020 ) . Global change continues to disrupt these interactions by altering the physiology , phenology , and particularly the spatial distribution of component species ( Burkle et al . , 2013; Vanbergen , 2013; Glenny et al . , 2018 ) . Habitat degradation and fragmentation reduce the size and connectivity of plant populations , which results in lowered pollinator visitation rates ( Aguilar et al . , 2006; Dauber et al . , 2010 ) . Plant population retraction and isolation may also affect interactions with pollinators at the plant individual level by increasing inbreeding rates ( Carr et al . , 2014 ) . The mating among closely related plant individuals may compromise floral traits attracting pollinators and hence cause negative feedback on pollinator visitation . Mechanistic insight into the effects of inbreeding on plant-pollinator interactions and intrinsic factors shaping the magnitude of such effects is limited but urgently required for the conservation of component species . Inbreeding increases homozygosity in the offspring generation . This may enhance the phenotypic expression of deleterious recessive mutations ( i . e . , dominance ) and reduce heterozygote advantage ( i . e . , over-dominance ) , which can result in severe declines of Darwinian fitness in inbred relative to outcrossed offspring ( i . e . , inbreeding depression ) ( Charlesworth and Willis , 2009 ) . Inbreeding may in addition disrupt plant-insect interactions . While it is well established that inbreeding can increase a plant’s susceptibility to herbivores by diminishing morphological and chemical defences ( Campbell et al . , 2013; Kariyat et al . , 2012; Kalske et al . , 2014 ) , its effects on plant-pollinator interactions are less well understood . Inbreeding may reduce a plant’s attractiveness to pollinating insects by compromising the complex set of floral traits involved in interspecific communication . These traits comprise ( i ) the spatial arrangement of individual flowers ( e . g . , size , shape ) and multiple flowers within an inflorescence ( e . g . , number , height above ground , degree of aggregation ) , hereinafter referred to as spatial flower traits ( Dafni et al . , 1997 ) ; ( ii ) the scent bouquet as determined by the composition of floral volatile organic compounds ( VOC ) such as terpenoids , benzenoids , and phenylpropanoids ( Muhlemann et al . , 2014; Borghi et al . , 2017 ) ; ( iii ) flower colour as defined by the composition of pigments with wavelength-selective light absorption and the backscattering of light by petal surface structures ( van der Kooi et al . , 2016; Borghi et al . , 2017 ) ; and ( iv ) the quality and quantity of rewards such as nectar , pollen , oviposition sites , or shelter ( Simpson and Neff , 1981 ) . These cues are particularly efficient in attracting pollinators across either long , medium , or short distances and act synergistically in determining visitation rates ( Dafni et al . , 1997; Muhlemann et al . , 2014 ) . Although in a few cases inbreeding has been shown to alter single floral traits ( Ivey and Carr , 2005; Ferrari et al . , 2006; Haber et al . , 2019 ) , insight into more syndrome-wide effects is restricted to a single study . Kariyat et al . , 2021 demonstrated that inbred Solanum carolinense L . display reduced flower size , pollen and scent production , and receive fewer visits from diurnal generalists . It is necessary to broaden such integrated methodological approaches to other plant-pollinator systems ( e . g . , nocturnal specialist pollinators ) and further floral traits ( e . g . , flower colour ) . The magnitude and slope of inbreeding effects in plants can vary across environments , since local conditions partly determine the selective value of recessive alleles unmasked by inbreeding ( Fox and Reed , 2011 ) . While the influence of environmental stress on the expression of inbreeding depression is well studied , the effects of plant sex , which considerably shapes an individual’s interaction with its environment , remain largely unexplored . Individuals of dioecious plant species invest into either male or female reproductive function . This partitioning goes along with different life histories , resource demands , stress susceptibilities , and consequently sex-specific selection regimes in identical habitats ( Moore and Pannell , 2011; Barrett and Hough , 2013 ) . Sex-specific selection may modify the magnitude of inbreeding depression in dioecious plants . Studies on animals reported higher inbreeding depression in females than males resulting from higher reproductive investment and prolonged life cycles in the former ( Ebel and Phillips , 2016 ) . In plants , such relations have rarely been investigated ( Teixeira et al . , 2009 ) , and if so , not with a focus on floral traits . If inbreeding effects on floral traits are more pronounced in female than male plants , the relative frequency of pollinator visits may be biased towards the latter sex , with devastating consequences for the effective size and persistence of populations . Studies on sex-specific inbreeding effects on floral traits are thus needed to improve the risk assessment for the conservation of dioecious plant species . Plant populations may escape progressive retractions under increased inbreeding rates by purging . Inbreeding unmasks deleterious recessive mutations , which facilitates their selective removal from the population gene pool and may result in a rebound of fitness when the demographic bottleneck is intermediate ( Crnokrak and Barrett , 2002 ) . Plant species that have successfully colonised distant geographic regions provide perfect models for studying the relevance of purging in natural plant populations . As colonisation events are associated with successive demographic bottlenecks , purging is expected to be one determinant for the successful establishment and proliferation of plant populations in novel habitats ( Facon et al . , 2011; Schrieber and Lachmuth , 2017 ) . However , only few empirical studies verified the role of purging in plant colonisation success by revealing significantly lower inbreeding depression following experimental crossings in invasive than native plant populations ( Rosche et al . , 2016; Schrieber et al . , 2019b ) . Again , the focus of these studies was on fitness components rather than floral traits . Yet , attractiveness to pollinators is a key for successful colonisation in species introduced into novel communities ( Morales and Traveset , 2009 ) , especially if plants are not capable of selfing ( e . g . , dioecious ) . Floral syndromes are likely under strong selection in such species , which should rapidly purge deleterious recessive mutations affecting spatial , scent , colour , and rewarding flower traits . In the present study , we investigated the effects of inbreeding on plant-pollinator interactions and tested whether the magnitude of such effects depends on plant sex and population origin using the Silene latifolia Pior . ( Caryophyllaceae ) and its crepuscular moth pollinators . Natural S . latifolia populations partly suffer from biparental inbreeding due to limited seed and pollen dispersal ( McCauley , 1997 ) . As inbreeding reduces not only fitness ( Teixeira et al . , 2009 ) but also impairs interactions with herbivorous insects in S . latifolia ( Schrieber et al . , 2019a; Schrieber et al . , 2019b ) , a disruption of plant-pollinator interactions can be expected for this species . Moreover , the dioecious reproductive system of S . latifolia provides the opportunity to quantify variation in the magnitude of inbreeding effects in these traits among females and males . Finally , the species expanded successfully from parts of its native distribution range in Europe to North America in the early 19th century ( Keller et al . , 2009; Keller et al . , 2012 ) , which may have given rise to purging events . We assessed spatial flower traits , headspace floral scent composition , flower colour and rewards , and quantified pollinator visitation in experimentally inbred and outbred male and female S . latifolia individuals from European and North American populations . We hypothesised that ( i ) inbreeding compromises floral traits , ( ii ) these inbreeding effects are more pronounced in female than male plants , ( iii ) inbreeding effects are more pronounced in European than North American populations , and ( iv ) the combined effects of inbreeding , sex , and population origin cause feedback on pollinator visitation rates .
S . latifolia shows a distinct moth pollination syndrome with large , white , and funnel-shaped flowers ( Dafni et al . , 1997 ) . The flowers open from dusk till mid-morning to release a scent bouquet composed of more than 60 VOC , whereby emission peaks around dusk ( Dötterl and Jürgens , 2005; Dötterl et al . , 2009; Mamadalieva et al . , 2014 ) . During the daytime , no measurable floral scent is emitted ( Dötterl et al . , 2005 ) . Nectar production peaks 3–4 days after flower opening and is just as floral scent emission reduced after pollination ( Gehring et al . , 2004; Dötterl and Jürgens , 2005; Muhlemann et al . , 2006 ) . S . latifolia exhibits various sexual dimorphisms with male plants producing more and smaller flowers that excrete lower volumes of nectar with higher sugar concentrations as compared to females ( Gehring et al . , 2004; Delph et al . , 2010 ) . The quality of floral scent exhibits no clear sex-specific patterns , while male plants have been shown to emit higher or equal total amounts of VOC as compared to females in different studies ( Dötterl and Jürgens , 2005; Waelti et al . , 2009 ) . Various diurnal generalist pollinators as well as crepuscular moths visit S . latifolia flowers . The latter , including the specialist Hadena bicruris Hufn . ( Lepidoptera: Noctuidae ) , were shown to be the most efficient pollinators for S . latifolia ( Young , 2002 ) , which is the reason why we exclusively focus on nocturnal pollination in our study . All nocturnal pollinators are rewarded with nectar , while the specialist H . bicruris is additionally rewarded with oviposition sites . S . latifolia and H . bicruris form a well-studied nursery pollination system , in which female moths pollinate female plants while ovipositing on the flower ovaries to provide their larvae with developing seeds . Pollination services provided by male H . bicruris likely over-compensate the costs of seed predation by their offspring ( Labouche and Bernasconi , 2010 ) . A substantial fraction of floral VOC produced by S . latifolia triggers antennal and behavioural responses in male and female H . bicruris moths ( Dötterl et al . , 2006 ) . The activity of H . bicruris peaks at dusk between May and July ( Bopp and Gottsberger , 2004 ) . H . bicruris is abundant in 90% of European S . latifolia populations but has not yet been introduced to North America . Other nocturnal moths including the specialist Hadena ectypa Morrison ( Lepidoptera: Noctuidae ) provide main pollination services to S . latifolia in the invaded range without imposing costs by seed predation ( Young , 2002; Castillo et al . , 2014 ) . We collected seed capsules from five female individuals ( maternal families ) in each of eight European and eight invasive North American S . latifolia populations ( Figure 1; Figure 1—figure supplement 1 ) . Seeds from all maternal families ( consisting of full-sibs and/or half-sibs , hereinafter referred to as sibs ) were germinated and plants were grown under controlled greenhouse conditions for experimental crossings within populations . Each female individual from the P-generation received pollen from a male derived from the same maternal family ( inbreeding ) and pollen from a male derived from a different maternal family within the same population ( outcrossing ) at separate flowers ( Figure 1—figure supplement 2 ) . During the crossings , plants were kept at randomised positions in the greenhouse . Female flower buds were covered with mesh bags prior to opening until fruit maturation and opened flowers were released from bags only for directed pollen transfer . The field sampling , rearing conditions , and experimental crossing are described in detail in Schrieber et al . , 2019a , Schrieber et al . , 2019b . Seeds were dried and stored at room temperature until further use . For the experiment , we grew plants from the F1 generation under greenhouse conditions ( 16/8 hr light/dark at 20/10°C±6°C ) . After the onset of flowering , we randomly chose one female and one male individual per breeding treatment ( inbred , outbred ) × maternal family ( 1−5 ) ×population ( 1−8 ) × origin ( Europe , North America ) combination , resulting in 320 plant individuals for the experiment ( Figure 1—figure supplement 3 ) . Using these individuals , we assessed the combined effects of breeding treatment , plant sex , and population origin on different flower traits and pollinator visitation rates over the summers 2019 and 2020 . Plants were grown in 3 L ( 2019 ) and 6 L ( 2020 ) pots filled with a 3:1 mixture of potting soil ( TKS2 Instant Plus , Floragrad , Oldenburg , Germany ) and pine bark ( Pine Bark 1–7 mm , Neede , Oosterbeek Humus Producten , The Netherlands ) . They were kept in pots with randomised positions either in the greenhouse , a common garden in Kiel , Germany ( Europe ) with sealed ground ( 54 . 346794°N , 10 . 107990°E , 19 m elevation ) or a field site in Kiel , Germany ( Europe ) , covered by an extensively used meadow ( 54 . 347742°N , 10 . 107661°E , 19 m elevation ) for different parts of data acquisition . For an overview of the time schedule , locations , and exact sample sizes for data acquisition , see Table 1 . Plants received water and fertilisation ( UniversolGelb 12-30-12 , Everris-Headquarters , Geldermalsen , The Netherlands ) when necessary for the entire experimental period and were prophylactically treated with biological pest control agents under greenhouse conditions to prevent thrips ( agents Amblyseius barkeri and Amblyseius cucumeris ) and aphid ( agent Chrysoperla carnea ) infestation ( Katz Biotech GmbH , Baruth , Germany ) . For characterisation of flower scent , we trapped the headspace VOC of S . latifolia flowers on absorbent polydimethylsiloxane ( PDMS ) tubing following the method of Kallenbach et al . , 2014 , Kallenbach et al . , 2015 . We placed the plants in a spatial distance of 50 cm to one another in the greenhouse and maintained high air ventilation 1 week prior to and during VOC collection . We selected one well-developed flower per individual and enclosed it in a VOC collection unit ( Figure 1—figure supplement 5 ) . The collection units consisted of polypropylene cups with lids ( 50 mL , Premium Line , Offenburg , Tedeco-Gizeh , Germany ) , both having holes ( diameter 15 mm ) to prevent heat and waterlogging . They were fixed via wooden sticks at the exterior of the plant pot . In addition , 14 control collection units were fixed on empty plant pots and positioned throughout the greenhouse . Prior use , the absorbent PDMS tubes ( length 5 mm , external diameter 1 . 8 mm , internal diameter 1 mm; Carl Roth , Karlsruhe , Germany ) were cleaned with solvents and heat as described in Kallenbach et al . , 2014 . Two PDMS tubes were added to each collection unit and remained in the floral headspace between 9 p . m . and 5 a . m . , which is the time of peak scent emission in S . latifolia ( Dötterl and Jürgens , 2005 ) . Afterwards , the PDMS tubes were removed and stored at −20°C in sealed glass vials until analysis via thermal desorption–gas chromatography–mass spectrometry ( TD-GC-MS , TD 30 – GC 2010plus – MS QP2020 , Shimadzu , Kyoto , Japan ) . All samples were measured in a single trial in a fully randomised order . Trapped VOC were desorbed from PDMS tubes for 8 min at 230°C under a helium flow of 60 mL min−1 and adsorbed on a Tenax cryo-trap with a temperature of –20°C . From the trap , compounds were desorbed at 250°C for 3 min , injected to the GC in a 3:10 split mode , and migrated with a helium flow of 1 . 6 mL min−1 on a VF5-MS column ( 30 m × 0 . 25 mm + 10 m guard column , Agilent Technologies , Santa Clara , CA ) . The GC temperature program started at 40°C for 5 min and increased to 125°C at a rate of 10°C min−1 with a hold time of 5 min and to 280°C at a rate of 30 °C min−1 with a hold time of 1 min . Line spectra ( 30–400 m/z ) of separated compounds were acquired in quadrupole MS mode . An alkane standard mix ( C8-C20 , Sigma-Aldrich , Darmstadt , Germany ) was analysed under the same conditions in order to calculate Kovats retention indices ( KI ) for targeted compounds ( Kováts , 1958 ) . Compounds were identified by comparing the KI and mass spectra with those of synthetic reference compoundsor with library entries of the National Institute of Standards and Technology ( NIST ) ( Smith et al . , 2004 ) , Pherobase ( El-Sayed , 2011 ) , the PubChem database ( Kim et al . , 2016 ) , and Adams , 2007 . Control samples ( collection units without flowers ) , and blanks ( cleaned PDMS tubes ) were used to identify and exclude contaminations , leaving a total number of 70 VOC ( Supplementary file 1 ) . Compounds were not quantified but the intensity of the total ion chromatogram of peaks was compared among treatment groups ( hereinafter referred to as intensity ) . A linear relationship among peak areas and compound concentrations has been validated for the passive sorption method in Kallenbach et al . , 2014 . The intensities of VOC were not corrected for flower size because we wanted to capture all variation in scent emission that is relevant for the receiver , that is , the pollinator . For targeted statistical analyses , we focused on those VOC that evidently mediate communication with H . bicruris according to Dötterl et al . , 2006 . We analysed the Shannon diversity per plant ( calculated with R-package: vegan v . 2 . 5–5 , Oksanen et al . , 2019 ) for 20 floral VOC in our data set that were shown to elicit electrophysiological responses in the antennae of H . bicruris ( Supplementary file 1 ) . Moreover , we analysed the intensities of three lilac aldehyde isomers , which trigger oriented flight and landing behaviour in both male and female H . bicruris most efficiently when compared to other VOC in the floral scent of S . latifolia . Furthermore , H . bicruris is able to detect the slightest differences in the concentration of these three compounds at very low dosages ( Dötterl et al . , 2006 ) . Flower colour was quantified using a digital image transformation approach that accounts for the visual system of the pollinator as well as natural light conditions ( Troscianko and Stevens , 2015 ) . Images were acquired in the common garden after plants had acclimated to ambient light conditions for 3 weeks . All images were taken during 1 hr of dusk time on rain-free days in order to fit the natural light conditions perceived by H . bicruris ( Bopp and Gottsberger , 2004 ) . We picked one well-developed , fully opened flower per plant and inserted it into a black ethylene vinyl acetate platform equipped with two reflectance standards ( PTFB 10%; Spectralon 99% Labsphere , Congleton , UK ) and a size standard ( Figure 1—figure supplement 6 ) . The platform had a fixed location in the field and was oriented towards the setting sun . Raw images were taken with a digital camera ( Samsung NX1000 , Suwon , South Korea , ) converted to full spectrum sensitivity ( 300–1000 nm ) via removal of the sensor’s filter and fitted with an ultraviolet ( UV ) sensitive lens ( Nikon EL 80‐mm , Japan ) . We took images in the visible and in the UV part of the light spectrum by fitting an UV and infrared ( IR ) blocking filter ( UV/IR Cut , transmittance 400–700 nm , Baader Planetarium , Reutlingen , Germany ) and an UV pass plus IR block filter ( U-filter , transmittance 300–400 nm , Baader Planetraium , Reutlingen , Germany ) to the lens , respectively . All images were taken as RAWs with an aperture of 5 . 6 , an iso of 800 , and a shutter speed varying according to light conditions . Images were processed using the Multispectral Image Calibration and Analysis ( MICA ) -Toolbox plugin ( Troscianko and Stevens , 2015 ) in ImageJ 1 . 47 t ( Rueden et al . , 2017 ) . They were linearised to correct for the non-linear response of the camera to light intensity and equalised with respect to the two light standards in order to account for variation in natural light perceived among images ( Stevens et al . , 2007 ) . All petals were selected for analysis , and the reproductive organs and para-corolla were omitted . Linearised images were then mapped to the visual system of a nocturnal moth . As the visual system of H . bicruris is unexplored , we used the tri-chromatic visual system of Deilephila elpenor L . ( Lepidoptera: Sphingidae ) , which includes three rhodopsins with absorption maxima of 350 nm ( UV ) , 440 nm ( blue ) , and 525 nm ( green ) ( Johnsen et al . , 2006 ) . We considered this system to be comparable to that of H . bicruris , given the similar activity behaviour of adults , morphological similarity of the preferred plant species ( Lonicera periclymenum L . [Caprifoliaceae] with white-creamy funnel-shaped flowers ) and overlapping distribution ranges . We fitted the images to a cone catch model incorporating ( i ) the spectral sensitivity of our Samsung NX1000-Nikkor EL 80 mm 300–700 nm camera ( data derived from Troscianko and Stevens , 2015 ) ; ( ii ) the spectral sensitivities of the three photoreceptors in the D . elpenor compound eye ( data derived from Johnsen et al . , 2006 ) ; and ( iii ) the spectral composition of sun light during dusk ( data derived from Johnsen et al . , 2006 ) . As moths forage on liquids only , we measured nectar as floral reward ( Figure 1—figure supplement 7 ) . We selected one well-developed , closed flower bud per plant in the common garden and enclosed it in a transparent mesh bag ( Organza mesh bags , Saketos , Sieniawka , Poland ) until harvest to avoid pollination and nectar removal . All flowers were harvested at noon of the fourth day after opening and were stored immediately at 4°C until processing to prevent further nectar secretion . Nectar was extracted into 1–2 µL microcapillary tubes ( Minicaps NA-HEP , Hirschmann Laborgeräte , Eberstadt , Germany ) . The length of the nectar column was measured with a calliper to determine the exact volume . Nectar sugar content was analysed with a refractometer adjusted for small sample sizes ( Eclipse Low Volume 0–50°brix , Bellingham and Stanley , UK ) . Since nectar volume trades off against nectar quality in pollinator attraction ( Cnaani et al . , 2006 ) , we addressed floral rewards in S . latifolia via the total amount of sugar excreted per flower as calculated based on the following equation: gsugar=volume[L]∗ ( ∘brix∗ ( 1+4 . 25∗∘brix1000 ) ∗10 ) . We quantified visits by crepuscular pollinators belonging to the order of Lepidoptera at the field site . For this purpose , plants were arranged in plots ( 1 . 5 m × 1 . 5 m , distance among plants = 0 . 5 m ) that consisted of eight individuals representing all populations from one breeding treatment × sex × origin combination . Each of the possible combinations ( N = 8 ) was replicated five times at the level of maternal families , resulting in a total number of 40 plots ( N = 320 plants in total ) . Plots were spaced from each other at a distance of 6 m in order to provide pollinators with the choice of visiting specific breeding treatment × sex × origin combinations ( Glenny et al . , 2018 ) . The position of plots and plants within plots was fully randomised ( Figure 1—figure supplement 8 ) . We performed 14 observation trials between May and July to cover the annual peak activity of H . bicruris ( Bopp and Gottsberger , 2004 ) . Each trial comprised 5 min observation time for each of the plots ( total observation time: 2800 min , observation time per plot: 70 min ) and was completed within 1 hr in the dawn time by four observers . The exact daytime of observation was acquired at the plot level for each of the trials . Plant and flower visits were determined at the plant individual level . If a moth had first contact with a flower , this was counted as a plant visit . The number of approached flowers per plant during a visit was counted until a moth either left or switched to another plant . The number of plant and flower visits per trial was averaged at the plot level for further analyses . The number of visiting moth individuals and moth species was not determined . The vast majority of visits were performed by H . bicruris ( personal observation ) . Please note that North American S . latifolia populations were tested in their ‘away’ habitat only and that the observed plant performance and pollinator visitation rates can thus provide no direct implications for their ‘home’ habitat . However , we neither aimed at elaborating on the invasion success of S . latifolia nor on adaptive differentiation among European and North American populations , but at investigating inbreeding effects on plant-pollinator interactions in multiple plant populations in a common environment . Given the close taxonomic relationship of H . bicruris ( main pollinator in Europe ) and H . ectypa ( main pollinator in North America ) ( Young , 2002; Castillo et al . , 2014 ) , the behavioural responses of the former species to variation in the quality of its host plant were considered to overlap sufficiently with responses of the latter species . All statistical analyses were performed in R v4 . 0 . 3 ( R Development Core Team , 2020 ) with ( generalised ) linear mixed effects models ( LMMs: R-package lme4 v1 . 1–23 , Bates et al . , 2014 , GLMMs: R-package glmmTMB v1 . 0 . 2 . 1 , Brooks et al . , 2017 ) . Models for responses reflecting spatial flower traits , floral scent , colour , and rewards included the predictors breeding treatment , sex , and origin , as well as all possible interactions among these factors . The latitudinal coordinate of the population origin was included as covariate in all models , whereas the exact age of the plant individuals ( accounts for difference of 12 days in planting date ) was included only in models for flower scent , which was acquired in early phases of the experiment . Both covariates were centred and scaled ( i . e . , subtraction of mean and division by standard deviation ) . The random effects for floral trait models were population , affiliation of paternal plant in P-generation to field collected family nested within population , and affiliation of maternal plant in P-generation to field collected family nested within population . Models for pollinator visitation rates included the predictors breeding treatment , sex , and origin , as well as all possible interactions among them , the covariate daytime ( centred and scaled ) , and the random effects of plot and trail ( latitude of population origin , population , maternal and paternal affiliation not included , since data were averaged on plot level , see Pollinator visitation rates section ) . Several of the described models included count data responses with an access of zeroes ( intensities of lilac aldehydes and pollinator visitation rates ) . These models were additionally fitted with zero inflation formulas . The fit of lilac aldehydes models was best when including only an intercept model for zero inflation , whereas the fit of pollinator visitation rate models was best when including the same predictors and random effects in the conditional and zero inflation part of the model . All of the described models ( Table 1 ) were validated based on checking plots ( quantile-quantile , residual versus fitted ) and tests provided in the R-package DHARMa v0 . 3 . 3 . 0 ( Hartig , 2020 ) . Sum-to-zero contrasts were set on all factors for the calculation of type III ANOVA tables based on Wald χ² tests ( R-package: car v3 . 0–10 , Fox and Weisberg , 2018 ) . If origin , breeding treatment and/or sex were involved in significant interactions , we calculated post hoc contrasts on the estimated marginal means of their levels within levels of other factors involved in the respective interaction ( R-package: emmeans v1 . 5 . 1 , Lenth , 2020 ) . Variance components were extracted from all models using the R-package insight ( Lüdecke et al . , 2019 ) and are summarised in Figure 2—figure supplement 1 . Multivariate statistical analyses of the full VOC dataset are summarised in Figure 2—figure supplement 2 .
Spatial flower traits of S . latifolia varied pronouncedly between plants of different breeding treatments , sexes , and population origins ( Table 2 ) . Synflorescences of inbreds had lower maximal height above ground than those of outbreds ( p<0 . 001 , χ² ( 1DF ) =37 . 31 , Figure 2a ) . Flower number ( Figure 2b ) was higher in plants from North America than Europe ( p=0 . 005 , χ² ( 1DF ) =8 . 01 ) and additionally depended on the interaction breeding treatment × sex ( p=0 . 003 , χ² ( 1DF ) =8 . 99 ) . Inbred plants generally produced fewer flowers than outbreds , and this effect was more severe in females ( 35% reduced by inbreeding , ppost <0 . 001 ) than males ( 12% reduced by inbreeding , ppost = 0 . 011 ) . The number of flowers produced was lower in male than female plants in both inbreds ( 78% reduced in females , ppost <0 . 001 ) and outbreds ( 71% reduced in females , ppost <0 . 001 ) . The area of petal limbs ( Figure 2c ) was smaller in female than male plants ( p<0 . 001 , χ² ( 1DF ) =51 . 35 ) and reduced by inbreeding ( p=0 . 002 , χ² ( 1DF ) =9 . 25 ) . The expansion of the corolla depended on the interaction breeding treatment × sex ( p=0 . 004 , χ² ( 1DF ) =8 . 17 ) . Inbreeding reduced corolla expansion in females by 17% ( ppost <0 . 001 ) but had no effect in male plants , and differences between sexes in corolla expansion were consequently apparent in inbreds ( 23% lower in females than males , ppost <0 . 001 ) but not in outbreds ( Figure 2d ) . Corolla expansion additionally depended on the interaction sex × origin ( p=0 . 009 , χ² ( 1DF ) =6 . 86 ) . It was lower in female than male plants in populations originating from North America only ( 23% lower in females , ppost <0 . 001 ) . Breeding treatment , sex , and population origin affected floral VOC in S . latifolia interactively ( Table 2 ) . The Shannon diversity of those VOC known to elicit antennal responses in H . bicruris depended on the interaction breeding treatment × origin ( p=0 . 016 , χ² ( 1DF ) =5 . 83 , Figure 2e ) . Inbreeding reduced the Shannon diversity of these VOC by 7% in European plants ( ppost = 0 . 013 ) but had no significant effect on the Shannon diversity of the VOC in plants from North America . The intensity of lilac aldehyde A depended on the interaction breeding treatment × sex × origin in the conditional model ( p=0 . 025 , χ² ( 1DF ) =5 . 03 , Figure 2f ) . Post hoc comparisons yielded a marginally significant lower intensity of this compound in inbred than outbred females in plants from North America ( 41% reduced by inbreeding , ppost = 0 . 056 ) but no further differences occurred among other groups . Similar non-significant trends were observed for the other lilac aldehyde isomers ( Supplementary file 1 ) . Multivariate statistical analyses of 20 H . bicruris active VOC and all 70 VOC detected in S . latifolia revealed no clear separation of floral headspace VOC patterns for any of the treatments ( Figure 2—figure supplement 2 ) . In summary , the combined effects of breeding treatment , sex , and range on floral scent were rather weak . The proportion of flower colour detectable for crepuscular moths and the sugar excreted as reward with nectar were independent of breeding treatment and population origin but exhibited differences between plants of different sex ( Table 2 ) . Male flowers reflected more light in the spectrum detectable by the UV receptor ( 350 nm ) ( p<0 . 001 , χ² ( 1DF ) =41 . 92 , Figure 2g ) and the blue receptor ( 440 nm ) ( p<0 . 001 , χ² ( 1DF ) =39 . 59 , Figure 2h ) of moths than flowers of females . Likewise , the amount of sugar excreted with nectar was higher in male than female plants ( p<0 . 001 , χ² ( 1DF ) =14 . 16 , Figure 2i ) . The number of pollinator visits per plant by moths was shaped by the interaction breeding treatment × sex × origin in the conditional model ( p=0 . 016 , χ² ( 1DF ) =5 . 84 , Table 2 , Figure 3a ) . Post hoc comparisons yielded that plant visits were reduced by 79% following inbreeding in female plants from North America ( ppost = 0 . 007 ) , but unaffected by inbreeding in European females and males from both origins . Moreover , plant visits were fewer in female than male plants in European outbreds ( 83% fewer in females , ppost < 0 . 001 ) , European inbreds ( 78% fewer in females , ppost = 0 . 001 ) , and North American inbreds ( 87% fewer in females , ppost < 0 . 001 ) as well as 77% lower in plants from Europe than North America in outbred females ( ppost = 0 . 014 ) . The number of flowers approached per plant visit was likewise shaped by the interaction breeding treatment × sex × origin in the conditional model ( p=0 . 001 , χ² ( 1DF ) =10 . 61 , Figure 3b , Table 2 ) . Post hoc comparisons yielded that flower visits were 88% lower for inbred than outbred females ( ppost = 0 . 001 ) but 64% higher for inbred than outbred males ( ppost = 0 . 031 ) in plant populations from North America , whereas flower visits were unaffected by inbreeding in European male and female plants . Moreover , flower visits were reduced in females relative to males for European outbred plants ( 83% reduction in females , ppost = 0 . 003 ) , European inbreds ( 73% reduction in females , ppost = 0 . 027 ) , and North American inbreds ( 90% reduced in females , ppost = 0 . 002 ) but 59% higher in females than males in outbreds from North America ( ppost < 0 . 001 ) . Finally , flower visits were higher in North American than European outbred female plants ( ppost = 0 . 001 ) but lower in European than North American outbred males ( ppost = 0 . 001 ) . Both the number of plant and flower visits depended on the interaction of breeding treatment × sex × origin in the zero inflation part of the model as well ( Table 2 , Figure 3—figure supplement 1 ) . The direction and magnitude of these effects did not contrast with the conditional models .
In partial accordance with our first hypothesis , inbreeding compromised several , but not all floral traits in S . latifolia . Spatial flower traits suffered most strongly from inbreeding in males and females from both origins ( Figure 2a–c ) . These results are in line with previous studies on hermaphroditic , self-compatible species ( Ivey and Carr , 2005; Glaettli and Goudet , 2006 ) and support that the complex genetic architecture underlying such traits ( Feng et al . , 2019 ) gives rise to dominance and over-dominance effects at multiple loci . The chemodiversity and abundance of floral VOC involved in communication with H . bicruris moths was reduced in a sex- and origin-specific manner in inbred relative to outbred S . latifolia ( Figure 2e–g ) , while the full floral scent profile exhibited no differences among inbreds and outbreds ( Figure 2—figure supplement 2 ) . So far , lower emissions of floral VOC in inbreds have been reported for only few plant species pollinated by diurnal generalists ( Ferrari et al . , 2006; Haber et al . , 2019; Kariyat et al . , 2021 ) . Our study revealed such effects for plants pollinated by specialist moths that use scent as a major cue for plant location ( Riffell and Alarcón , 2013 ) . Dominance and over-dominance may either have directly interfered with genes involved in VOC synthesis and their regulation in S . latifolia or unfolded their effects by disrupting physiological homoeostasis and thereby inducing intrinsic stress that came at the cost of scent production ( Kristensen et al . , 2010; Fox and Reed , 2011 ) . A recent study indicates that the effects of inbreeding on the diversity of floral VOC in our study may even have been underestimated . Kergunteuil et al . , 2021 demonstrated that porous polymers may differ in their affinity with specific VOC and hence in their sensitivity in recording variation in VOC diversity entailing blind spots . They recommend a shift in practice from the use of single to multiple porous polymers ( e . g . , a combination of PDMS and Poropak Q ) for VOC collection in future plant ecological studies , which may uncover the full impact of plant inbreeding on the composition of floral volatiles . In contrast to spatial flower traits and scent , flower colour and the total amount of sugar excreted with nectar exhibited no differences among inbreds and outbreds in S . latifolia ( Table 2 ) . Flower colour is a trait that has , to our knowledge , not yet been studied in the context of inbreeding , despite its crucial role in flower identification and localisation ( Garcia et al . , 2019 ) . Our data suggest that the flower colour perceived by moths is not altered by inbreeding and generally seems to be a conserved trait in S . latifolia . Other species with high intraspecific variation in flower colour may be ideal models to further examine the relationship with inbreeding in the future by combining visual modelling with choice experiments ( Kelber et al . , 2003 ) . The independence of sugar excretion from inbreeding in S . latifolia indicates that the strong reduction of flower number in inbreds allows compensating the quality of floral rewards via a resource allocation trade-off . Overall , the observed inbreeding effects on floral traits were partially small and variable in their magnitude as compared to previous investigations . However , our findings highlight that even weak degrees of biparental inbreeding ( i . e . , one generation sib-mating ) can result in an impairment of multiple flower traits that is detectable against the background of natural variation among multiple plant populations from a broad geographic region . This observation indirectly supports that the selfing syndrome ( i . e . , smaller , less scented flowers observed in selfing relative to outcrossing populations of hermaphroditic plant species ) may not merely be a result of natural selection against resource investment into floral traits , but also a direct negative consequence of inbreeding ( Andersson , 2012 ) . Most importantly , we observed that variation in inbreeding effects was consistent in its dependency on plant sex , which gives insight into the role of intrinsic biological differences between males and females in the expression of inbreeding depression . Males outperformed females in all floral traits , except scent production ( Figure 2 ) . As such , our study confirmed previously observed sexual dimorphisms in S . latifolia ( nectar: Gehring et al . , 2004; flower number: Delph et al . , 2010 ) but also yielded contradicting results . As opposed to Delph et al . , 2010 , we observed larger instead of smaller flowers in males . This may base on the use of a size estimate that accounts for variation in flower shape or the comparably large geographic range and higher number of populations covered by our study . Moreover , we discovered a novel sexually dimorphic trait in the colour appearance of S . latifolia to crepuscular moths in the UV and blue light spectrum ( Figure 2g–h ) . Given that moths use blue light as a major cue to start feeding on nectar ( Cutler et al . , 1995 ) , the lower light reflectance observed for female flowers is another trait rendering them less attractive than males . The evolution of lower female attractiveness to pollinators is driven by sex-specific resource allocation , that is , high costs for production of ovaries and seeds may restrict allocation to floral traits ( Moore and Pannell , 2011; Barrett and Hough , 2013 ) . This process may also explain the larger magnitude of inbreeding effects in female plants of S . latifolia , which we observed in accordance with our second hypothesis ( Figure 2b , d ) . High reproductive expenditure in females may increase the frequency and intensity of resource depletion stress ( e . g . , drought ) under field conditions ( Obeso , 2002; Li et al . , 2004; Zhang et al . , 2010 ) . Consequently , females may suffer disproportionally from inbreeding when dominance and over-dominance affect loci mediating resistance to such stress ( Fox and Reed , 2011 ) . The assumption that sex-specific viability selection plays a role in the expression of inbreeding depression seems likely for S . latifolia . Previous research elaborated that females of this species experience resource depletion stress more often than males in the late growing season during fruit maturation ( Gehring and Monson , 1994 ) and that inbreeding reduces resistance to environmental stress ( Schrieber et al . , 2019a; Schrieber et al . , 2019b ) . Another non-exclusive explanation for the different inbreeding effects on female versus male S . latifolia may be found in sexual selection . Compared to females , the reproductive success of males is more limited by the availability of mates than by the availability of resources , which results in selection for increased attractiveness to pollinators ( Moore and Pannell , 2011; Barrett and Hough , 2013 ) . Competition for increased siring success among male plant individuals may create strong selection pressures that could rapidly purge deleterious recessive mutations in genes directly linked to attractiveness of male flowers to pollinators . Finally , a proportion of sex-specific inbreeding effects may be attributed to differential gene expression . S . latifolia harbours numerous genes with alleles that affect male and female fitness in opposite directions . These sexually antagonistic genes are partly subsumed in non-recombining regions of gonosomes ( Scotti and Delph , 2006 ) . Those located at the X-chromosome are always effectively dominant in males ( XY ) but may be recessive and therefore contribute to inbreeding depression in females ( XX ) . The remaining fraction of sexually antagonistic genes is located at autosomal regions but exhibits sex-specific expression as controlled by the gonosomes ( Scotti and Delph , 2006 ) . These genes may exhibit systematic differences in the abundance and effect magnitudes of deleterious recessive alleles between males and females , thus contributing to sex-specific inbreeding effects . Not only floral traits but also plant viability may exhibit sex-specific inbreeding depression in dioecious species . This could result in deviations from optimal sex ratio and , consequently , reductions of effective population sizes that accelerate local extinctions under global change ( Hultine et al . , 2016; Rosche et al . , 2018 ) . Future studies should aim at disentangling the relative contribution of sex-specific selection and gene expression to differences in the magnitude of inbreeding depression between males and females and at assessing their feedback on sex ratios to predict and manage these specific threats . Plants exhibited a general difference among geographic origins in merely one floral trait ( Figure 2b ) . Indeed , we had not expected broad differences in floral traits among European and North American S . latifolia plants ( i . e . , significant main effects of origin ) . A sufficient overlap in the composition of pollinator communities ( H . ectypa replaces H . bicruris in the invaded range , Castillo et al . , 2014 ) and appropriate pre-adaptations in floral traits were probably essential for S . latifolia as an obligate outcrossing plant species to successfully colonise North America . As discussed in detail in previous studies , higher flower numbers in North American S . latifolia ( Figure 2b ) may result from changes in the selective regimes for numerous abiotic factors ( Keller et al . , 2009 ) or from the release of seed predation . As opposed to H . bicruris , H . ectypa pollinates North American S . latifolia without incurring costs for seed predation , which may result in the evolution of higher flower numbers , specifically in female plants ( Elzinga and Bernasconi , 2009 ) . While adaptive differentiation among S . latifolia populations from different origins was not in the focus of this study , we hypothesised that North American populations purged genetic load linked to floral traits during the colonisation process ( i . e . , interaction breeding treatment × origin ) . In contrast to hypothesis iii , the magnitude of inbreeding effects was not consistently higher in European than North American populations . Instead , it was independent of origin for most floral traits , except flower scent , and either higher or lower in European plants for different scent traits ( Figure 2f–g ) . These findings provide no support for recent purging events in North American populations . They rather add to evidence that the magnitude of inbreeding effects is highly specific for the traits as well as the populations or population groups under investigation ( e . g . , Escobar et al . , 2008; Angeloni et al . , 2011 ) . This specifity roots in the composition of gene loci affected by dominance and over-dominance and is determined by the complex interplay of demographic population histories ( i . e . , size retractions and expansions , genetic drift , isolation , gene flow ) and the selective environment ( Charlesworth and Willis , 2009 ) . As such , the precise mechanisms underlying variation in inbreeding effects on different scent traits across population origins of S . latifolia can only be explored based on comprehensive genomic resources , which are currently not available . Future studies should also incorporate field data on the abundance of specialist pollinators and extend the focus from variation in the magnitude of inbreeding effects among geographic origins to variation among populations within geographic origins and individuals within populations . This would allow a detailed quantification of geographic variation in inbreeding effects and elaborating on the causes and ecological consequences of such variation ( Thompson , 2005; Schrieber and Lachmuth , 2017; Thompson et al . , 2017 ) . Pollinator visitation rates partially mirrored the above-discussed variation in flower traits . They depended on the breeding treatment in a highly sex- and origin-specific manner: In North American populations , inbred females received significantly fewer plant and flower visits than outbreds , whereas flower visits were higher in inbred than outbred males ( Figure 3 ) . We conclude that the severe inbreeding effects on spatial flower traits alone do not necessarily reduce moth visitation rates because these effects were observed for both plant sexes and origins ( Figure 2a , b , c ) . A floral trait that was negatively affected by inbreeding only in North American female plants , just like pollinator visitation rates , was the abundance of lilac aldehyde A ( Figure 2f ) . The other lilac aldehyde isomers exhibited similar but non-significant trends ( Supplementary file 1 ) . Although these findings provide limited support for our fourth hypothesis , they yield interesting insight into the relative importance of floral traits for the behaviour of a lepidopteran specialist pollinator . The seemingly low importance of inbreeding effects on spatial flower traits for pollinator visitation rates may be explained with the limited visual system of nocturnal insects ( van der Kooi et al . , 2021; Sondhi et al . , 2021 ) . Moths can likely perceive differences in the spatial arrangement of flowers on inbred versus outbred plants only when they have already approached close to them ( Barnett et al . , 2018 ) . Our setup for pollinator observations had relatively large distances among replicate inbred and outbred plots ( Figure 1—figure supplement 8 ) and hence may have not enabled a choice based on spatial flower traits by the moths . In contrast , differences in scent cues should be perceived across large distances . The antennae of H . bicruris can detect slight differences in lilac aldehyde concentrations at very low dosages and the compounds elicit oriented flight and landing responses in the moth more than any other VOC in the scent of S . latifolia flowers ( Dötterl et al . , 2006 ) . Consequently , a low lilac aldehyde abundance may have resulted in a low attraction of moths to North American inbred female plants from the distance in our experiment . The non-significant trend for higher abundances of lilac aldehydes in inbred than outbred males from North America ( Supplementary file 1 ) could also explain the corresponding variation observed in flower visitation rates . However , bioassays under more controlled conditions are needed to further evaluate a mechanistic relationship among pollinator visitation and intensities of lilac aldehydes . In summary , our research on S . latifolia suggests that in addition to inbreeding disrupting interactions with herbivores by changing plant leaf chemistry ( Schrieber et al . , 2019a ) , it affects plant interactions with pollinators by altering flower chemistry . Our observations are in line with studies on other plant species ( Ivey and Carr , 2005; Kariyat et al . , 2012; Kariyat et al . , 2021 ) and highlight that inbreeding has the potential to reset the equilibrium of species interactions by altering functional traits that have developed in a long history of co-evolution . These threats to antagonistic and symbiotic plant-insect interactions may mutually magnify in reducing plant individual fitness and altering the dynamics of natural plant populations under global change . As such , our study adds to a growing body of literature supporting the need to maintain or restore sufficient genetic diversity in plant populations during conservation programs . | Destroying habitats can reduce the size of local populations of many plants and animals . For plants , a smaller population means a greater chance of inbreeding , where individual plants that are closely related to each other mate and produce offspring . Inbreeding often results in offspring that are weaker than their parents which can reduce the plant’s chance of survival . Many plants rely on animals to help them to breed . For example , bees carry pollen – containing the male sex cell – to other flowers which then fertilize the plant to produce seeds . Flowers use a wide range of attributes to attract animals such as their colour , scent and providing them with food . However , inbreeding may alter these characteristics which could make it harder for inbred plants to reproduce , meaning that populations would end up shrinking even faster . To test this theory , Schrieber et al . studied flowers from white campions which use moths to breed . Inbred plants had smaller and fewer flowers , and had a different smell . In particular , they produced less of a chemical scent that is known to attract moths at night . Schrieber et al . then tracked moths visiting a mixed population of inbred and control plants . Fewer moths visited the inbred flowers , particularly the ones that were female . This shows that inbreeding may accelerate population loss and extinction by making flowers less attractive to animals . This work highlights the impact habitat destruction has on plants and shows how species can decline rapidly as populations shrink . This could help to support conservation efforts and inform ecology models to better understand our effect on the environment . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"ecology",
"evolutionary",
"biology"
] | 2021 | Inbreeding in a dioecious plant has sex- and population origin-specific effects on its interactions with pollinators |
Many mammalian cancer cell lines depend on glutamine as a major tri-carboxylic acid ( TCA ) cycle anaplerotic substrate to support proliferation . However , some cell lines that depend on glutamine anaplerosis in culture rely less on glutamine catabolism to proliferate in vivo . We sought to understand the environmental differences that cause differential dependence on glutamine for anaplerosis . We find that cells cultured in adult bovine serum , which better reflects nutrients available to cells in vivo , exhibit decreased glutamine catabolism and reduced reliance on glutamine anaplerosis compared to cells cultured in standard tissue culture conditions . We find that levels of a single nutrient , cystine , accounts for the differential dependence on glutamine in these different environmental contexts . Further , we show that cystine levels dictate glutamine dependence via the cystine/glutamate antiporter xCT/SLC7A11 . Thus , xCT/SLC7A11 expression , in conjunction with environmental cystine , is necessary and sufficient to increase glutamine catabolism , defining important determinants of glutamine anaplerosis and glutaminase dependence in cancer .
Altered metabolism is a hallmark of cancer ( Hanahan and Weinberg , 2011 ) and reflects the increased energetic and biosynthetic demands of proliferating cancer cells ( DeBerardinis and Chandel , 2016; Vander Heiden and DeBerardinis , 2017 ) . Indeed , one of the earliest noted biochemical differences between cancerous and normal tissues is increased glucose uptake and glycolysis in tumors ( Cori and Cori , 1925; Warburg et al . , 1927 ) . Beyond increased glucose metabolism in tumors , another long standing observation is that many cancer cells consume substantial amounts of glutamine in culture , far in excess of other amino acids ( Eagle , 1955; Jain et al . , 2012 ) . Consistent with this , many cancer cell lines depend on extracellular glutamine for proliferation in vitro , even though glutamine is nominally non-essential and can be synthesized from other nutrients . This contributed to the concept that some cancer cells are glutamine addicted ( Altman et al . , 2016; Wise and Thompson , 2010 ) . Glutamine can be used to support proliferation in multiple ways . It is a proteinogenic amino acid , can act as a nitrogen donor for the synthesis of amino acids as well as nucleotides , and glutamine can contribute carbon to the TCA cycle to replace cycle intermediates that are removed for the production of biomass , a process termed anaplerosis ( Altman et al . , 2016; Daye and Wellen , 2012; DeBerardinis and Cheng , 2010; Hensley et al . , 2013 ) . Analysis of the fate of glutamine nitrogen in proliferating cancer cells in vitro suggests that most glutamine nitrogen is excreted as ammonia and alanine , suggesting that the high rate of glutamine consumption is not due to nitrogen demand ( DeBerardinis et al . , 2007 ) . Instead , glutamine metabolism is important for TCA cycle anaplerosis in multiple contexts ( Altman et al . , 2016; DeBerardinis et al . , 2007; Yuneva et al . , 2007 ) . Across cancer cell lines , the majority of aspartate , glutamate and TCA cycle metabolites are glutamine-derived ( Altman et al . , 2016; Wise and Thompson , 2010 ) , with production of amino acids being a major fate of anaplerotic glutamine carbon ( Hosios et al . , 2016 ) . In line with these results , proliferation of many cell types following glutamine starvation is rescued by providing an alternative source of the TCA cycle intermediates α-ketoglutarate ( αKG ) ( van den Heuvel et al . , 2012; Wise et al . , 2008; Wise et al . , 2011 ) or oxaloacetate ( Patel et al . , 2016 ) . Collectively , these results suggest that anaplerosis contributes to the large cellular consumption of glutamine in culture , and to dependence on this amino acid for cell proliferation and survival . Glutamine can enter the TCA cycle through multiple metabolic routes . First , several transporters are capable of transporting glutamine into cells ( Hediger et al . , 2013; Hyde et al . , 2003 ) . Relevant to cancer , the neutral amino acid transporters ASCT2/SLC1A5 and LAT1/SLC7A5 are known to have higher expression in certain tumors , and can mediate glutamine uptake in cell lines derived from these tumors ( Bhutia et al . , 2015; Pochini et al . , 2014 ) . Intracellularly , glutamine is converted to glutamate either by donating the amide nitrogen for the production of nucleotides or asparagine , or by glutaminase activity ( encoded by GLS1 or GLS2 ) , which produces glutamate and ammonia from glutamine ( Curthoys and Watford , 1995; Krebs , 1935 ) . In proliferating cells , glutaminase activity can be the primary driver of glutamate production from glutamine , as amide nitrogen incorporation into mass is low compared to release of glutamine nitrogen as ammonia ( Brand , 1985; Brand et al . , 1986; Hosios et al . , 2016; Wise et al . , 2008 ) . Genetic or pharmacological loss of GLS1 activity depletes TCA metabolites and slows proliferation of a variety of cancer cell lines in culture ( Cheng et al . , 2011; Gameiro et al . , 2013; Gao et al . , 2009; Gross et al . , 2014; Le et al . , 2012; Seltzer et al . , 2010; Son et al . , 2013; Timmerman et al . , 2013; van den Heuvel et al . , 2012; Wang et al . , 2010; Yuneva et al . , 2012 ) . This has led to interest in targeting glutaminase activity therapeutically , and the glutaminase inhibitor CB-839 is being evaluated in clinical trials to treat cancer ( Gross et al . , 2014 ) . In the last step of glutamine carbon entry into the TCA cycle , glutamate produced from glutamine is converted to αKG by either transamination reactions or by glutamate dehydrogenase to produce αKG as an anaplerotic TCA cycle intermediate ( Moreadith and Lehninger , 1984 ) . Rapidly proliferating cells have been shown to preferentially use transamination reactions for αKG production , consistent with their increased need for nitrogen for biosynthetic demand ( Coloff et al . , 2016 ) . Finally , consistent with these observations of increased glutamine catabolism and dependence in rapidly proliferating cultured cells , glutamine catabolic pathways are controlled by oncogene expression and upregulated in many cancer cell lines ( Altman et al . , 2016 ) . Tumor cell environment can also influence dependence on glutaminase for anaplerosis and proliferation . Tracing of glucose and glutamine fate in tumors derived from human non-small cell lung cancer ( NSCLC ) and mouse KRAS-driven NSCLC found that these tumors rely more on glucose than on glutamine for TCA cycle anaplerosis ( Davidson et al . , 2016; Hensley et al . , 2016; Sellers et al . , 2015 ) . This correlated with resistance of these tumors to glutaminase inhibition in vivo; but surprisingly , cell lines derived from these tumors with decreased glutamine catabolism showed dramatically increased glutamine anaplerosis and sensitivity to glutaminase inhibition when cultured in vitro ( Davidson et al . , 2016 ) . These results suggest that , even for cells capable of glutamine catabolism , environmental differences between tumors in vivo and tissue culture dictates the use of glutamine for anaplerosis . We describe the identification of an environmental factor that contributes to differential glutamine anaplerosis between human NSCLC tumors growing in vivo and cell lines cultured in vitro . Nutrient availability is different for cancer cells growing as tumors in vivo compared to culture conditions in vitro , and we explored the role of nutrient environment by culturing the human A549 NSCLC cell line in adult bovine serum , a medium that more closely models in vivo nutrient levels . In this condition , the contribution of glutamine carbon to the TCA cycle decreases to levels observed for A549 tumors growing in vivo . Culture in adult bovine serum also induces glutaminase inhibitor resistance , a phenotype seen for NSCLC tumors in vivo . Deconvolution of differences in media composition showed that levels of a single nutrient , cystine ( the oxidized dimer of the amino acid cysteine ) , largely dictates the observed differences in glutamine anaplerosis and dependence . Further , we find that cystine regulation of glutamine anaplerosis depends on expression of the cystine/glutamate antiporter xCT ( encoded by the gene SLC7A11 ) . Lastly , we find that administration of cystine to tumor bearing mice increases glutamine use by tumors in vivo . Collectively , these results suggest that environmental cystine availability and xCT/SLC7A11 expression are critical determinants of glutamine anaplerosis and glutaminase dependence . They also highlight how nutrient conditions can impact cell metabolism .
Mutant Kras-driven mouse lung cancer cells exhibit differences in glutamine metabolism dictated by their environment , such that glutamine extensively labels TCA cycle intermediates in cell culture but not in tumors ( Davidson et al . , 2016 ) . To confirm this finding , we examined glutamine metabolism in mutant KRAS-driven human A549 lung cancer cells cultured in multiple environments . For these studies , we used gas chromatography-mass spectrometry ( GC-MS ) to trace the fate of glutamine where all five carbons are 13C labeled ( [U-13C5]glutamine ) in each environment , focusing on whether glutamine carbon contributed to the TCA cycle ( Figure 1A ) . First , we verified that glutamine was not a major source of TCA cycle carbon in subcutaneous A549 xenograft tumors . For these studies [U-13C5]glutamine was infused into tumor bearing animals for 6 hr to achieve a final enrichment of ~35% labeled glutamine in plasma and ~25% in tumors ( Figure 1B ) . Consistent with previous results ( Davidson et al . , 2016 ) , there was little labeling of intratumoral glutamate , TCA cycle intermediates , and aspartate from [U-13C5]glutamine despite the presence of labeled glutamine in the tumors ( Figure 1B ) . These findings suggest that glutamine is not a major source of TCA cycle carbon in A549-derived tumors in vivo . Normalization of tumor m + 5 glutamate label to m + 5 glutamine labeling indicates that only ~25% of glutamate is derived from glutamine in these tumors . In contrast , when A549 cells are cultured in RPMI-based media with [U-13C5]glutamine added to a similar enrichment of ~33% , we find that ~27% of glutamate is glutamine derived , which when normalized to glutamine enrichment indicates that 80% of glutamate is derived from glutamine under these conditions ( Figure 1C ) . Additionally , when the cells are cultured in RPMI , more than 60% of the carbon in aspartate and other TCA cycle intermediates was derived from glutamine ( Figure 1C ) . This is consistent with glutamine being a major TCA cycle carbon source for A549 cells in culture , and demonstrates that compared to standard culture conditions tumors derived from these cells use less glutamine in vivo . To begin to examine the environmental factors that contribute to this difference in glutamine metabolism , we cultured cells in 100% adult bovine serum with no additional nutrients based on the notion that this may better reflect the nutrient levels available from the circulation to the cells in tumors . A549 cells grow exponentially with a doubling time of ~48 hr when cultured in bovine serum ( Figure 1D ) and can be maintained indefinitely in such medium ( see Materials and methods for details ) . We added [U-13C5]glutamine to adult bovine serum to achieve an enrichment of ~33% that is comparable to the labeled glutamine enrichment observed in plasma and tumors in vivo . We then traced the fate of 13C carbon into the TCA cycle and found substantially less labeling of glutamate , aspartate and TCA cycle metabolites compared to cells cultured in RPMI-based media ( Figure 1C ) . In fact , only ~35% of glutamate was estimated to be derived from glutamine when cells were cultured in adult bovine serum . We verified that glutamine labeling reached isotopic steady state after 8 hr of culture in adult bovine serum ( Figure 1—figure supplement 1 ) , suggesting that these labeling differences were not explained by differences in labeling kinetics . Importantly , the decrease in glutamine-derived label observed in A549 cells cultured in adult bovine serum was not specific to bovine serum or to serum products , as the contribution of glutamine carbon to glutamate and TCA cycle intermediates was similarly reduced in adult bovine heparinized plasma and adult human serum ( Figure 1—figure supplement 2 ) . Reduced TCA cycle labeling from glutamine when cells are cultured in adult bovine serum is not limited to A549 cells . [U-13C5]glutamine labeling of TCA cycle intermediates was higher in cell lines derived from cancers arising in diverse tissues when cultured in RPMI compared to culture in adult bovine serum ( Figure 1—figure supplement 3 ) . Thus , culturing cancer cells in serum or plasma results in reduced glutamine catabolism to support TCA cycle anaplerosis . Sensitivity of Kras-driven mouse lung cancer cells to the glutaminase inhibitor CB-839 correlates with glutamine carbon contribution to the TCA cycle , such that tumors derived from these cells in vivo were insensitive to CB-839 while proliferation of the same cells is inhibited in vitro ( Davidson et al . , 2016 ) . Consistent with these findings , we find that A549 cells cultured in RPMI are also sensitive to CB-839 , but the same cells become resistant to CB-839 when cultured in adult bovine serum ( Figure 1E ) . Thus , A549 cells cultured in bovine serum exhibit decreased glutamine metabolism and are resistant to glutaminase inhibitors , adopting a metabolic phenotype with regards to glutamine metabolism that is more similar to lung tumors in vivo than to lung cancer cells cultured in standard tissue culture media . We next sought to identify factors that are different between RPMI and adult bovine serum that account for the observed differences in glutamine anaplerosis and dependence . Dialysis experiments were used to determine whether differences in the small molecule ( <3 . 5 kDa ) or large molecule ( >3 . 5 kDa ) fractions accounted for differences in glutamine contribution to the TCA cycle . First , we dialyzed a small volume of RPMI against large volumes of adult bovine serum using 3 . 5 kDa cutoff dialysis cassettes ( Figure 2A ) . This medium termed ‘adult bovine serum → RPMI , ’ contained the small molecule fraction from adult bovine serum and the large molecule fraction from RPMI-based media . Similarly , we dialyzed adult bovine serum against RPMI to generate ‘RPMI → adult bovine serum , ’ containing RPMI small molecules and the adult bovine serum large molecule fraction ( Figure 2A ) . Cells could be cultured in both media conditions , allowing us to trace the fate of [U-13C5]glutamine in A549 cells cultured in each media . Similar to cells grown in adult bovine serum , cells grown in ‘adult bovine serum → RPMI’ exhibited low fractional labeling of glutamate , aspartate and TCA cycle metabolites ( Figure 2B ) . In contrast , cells cultured in ‘RPMI → adult bovine serum’ exhibited higher labeling of aspartate and TCA cycle intermediates from glutamine , similar to cells cultured in RPMI ( Figure 2B ) . These differences in glutamine contribution to the TCA cycle also correlated with the ability of CB-839 to inhibit proliferation , as A549 cell proliferation is inhibited by CB-839 when cells are cultured in ‘RPMI → adult bovine serum’ but is largely unaffected when cultured in ‘adult bovine serum → RPMI’ ( Figure 2C ) . Similar dialysis experiments were performed with DMEM and adult bovine serum . Consistent with the RPMI dialysis experiments , ‘adult bovine serum → DMEM’ cultured cells displayed levels of glutamine anaplerosis and CB-839 sensitivity that were similar to cells cultured in adult bovine serum , while ‘DMEM → adult bovine serum’ had increased CB-839 sensitivity , similar to cells cultured in DMEM ( Figure 2—figure supplement 1 ) . Taken together , these experiments suggest that differences in the small molecule fraction of standard tissue culture media and adult bovine serum largely explain the differences in glutamine anaplerosis and glutaminase inhibitor sensitivity . We next considered differences in the small molecule fraction of adult bovine serum and RPMI/DMEM that could give rise to the differences in glutamine anaplerosis and CB-839 sensitivity . RPMI and DMEM both contain excess glucose , amino acids and micronutrients compared to blood ( Mayers and Vander Heiden , 2015 ) . In fact , re-formulation of DMEM to have more physiological concentrations of glucose , amino acids and other nutrients were previously reported to alter central carbon metabolism and glutamine dependence ( Schug et al . , 2015; Tardito et al . , 2015 ) . In order to determine if excess nutrients in standard media formulations potentiate glutamine anaplerosis and dependence , we determined the concentration of amino acids , glucose , and pyruvate in the sera used to culture cells in this study ( Table 1 ) . We supplemented the adult bovine serum with amino acids , glucose , and pyruvate to match levels found in RPMI or DMEM . Levels of vitamins and micronutrients were added to adult bovine serum at the RPMI or DMEM concentration . Addition of RPMI nutrients ( glucose , amino acids , vitamins and micronutrients ) to adult bovine serum resulted in increased glutamine contribution to the TCA cycle in A549 cells ( Figure 3A ) . Addition of nutrients to adult bovine serum to levels found in either RPMI or DMEM also caused increased A549 sensitivity to CB-839 ( Figure 3B , Figure 3—figure supplement 1A ) . Thus , culturing cells in nutrient levels found in standard tissue culture media is sufficient to increase glutamine catabolism and dependence in these cells . We next added different subsets of RPMI or DMEM nutrients to adult bovine serum to identify which nutrient ( s ) promote increased glutamine utilization and dependence . Addition of amino acids alone at RPMI levels to adult bovine serum increased glutamine contribution to the TCA cycle to the same extent as addition of the whole pool of all RPMI nutrients ( Figure 3A ) . Addition of amino acids alone at RPMI levels also induced CB-839 sensitivity ( Figure 3B ) , as did addition of amino acids to DMEM levels ( Figure 3—figure supplement 1A ) . We next sought to identify the amino acid ( s ) causing this increased use of glutamine . Because DMEM contains fewer amino acids , and causes the same phenotype as RPMI amino acids , we reasoned that the amino acids responsible must be present in DMEM . Therefore , we systematically determined CB-839 sensitivity of cells cultured in adult bovine serum supplemented with these amino acids minus subsets that share common transport mechanisms ( Hediger et al . , 2013; Hyde et al . , 2003 ) . Adding the DMEM amino acid pool lacking serine , glycine , threonine and cystine to adult bovine serum failed to cause increased sensitivity of cells to CB-839 , while omitting other subsets of amino acids had no effect on CB-839 sensitivity ( Figure 3—figure supplement 1A ) . These data argue that one or more of these four amino acids is responsible for the phenotype . Serine , glycine , threonine or cystine were individually increased to DMEM levels in adult bovine serum containing DMEM levels of other amino acids . Only cystine addition triggered sensitivity to CB-839 ( Figure 3—figure supplement 1B ) . Addition of cystine alone to adult bovine serum is sufficient to sensitize cells to CB-839 ( Figure 3B ) . Similarly , addition of cystine at RPMI levels to adult bovine serum causes an increase in glutamine labeling of TCA cycle intermediates that is comparable to the labeling pattern observed in cells cultured in adult bovine serum containing levels of all nutrients found in RPMI ( Figure 3A ) . Cystine levels in adult plasma ( ~20–50 µM ) are 4–10 fold lower than in DMEM or RPMI ( ~200 µM ) ( Table 1 ) . These experiments demonstrate that high levels of cystine found in standard tissue culture formulations cause enhanced glutamine anaplerosis and dependence . Glutamine derived glutamate has multiple possible fates in cells ( Figure 3—figure supplement 2A ) . It can contribute to TCA cycle metabolites , but also contributes to other metabolites like glutathione . Previous studies have suggested that glutamine-derived glutathione is essential for cell growth and further that depletion of glutathione upon glutaminase inhibition and subsequent loss of reactive oxygen species ( ROS ) buffering forms the basis of glutaminase dependence ( Biancur et al . , 2017; Jin et al . , 2015; Li et al . , 2015; Timmerman et al . , 2013; Ulanet et al . , 2014 ) . In contrast , other studies have found that maintenance of TCA cycle derived metabolite levels , such as nucleotides , forms the basis of glutaminase dependence ( Cox et al . , 2016; Gaglio et al . , 2009; Lee et al . , 2016; Okazaki et al . , 2017; Patel et al . , 2016 ) . We examined whether cystine-induced glutaminase dependence in A549 cells could be specifically ascribed to TCA cycle or glutathione depletion . Both TCA cycle metabolites and reduced glutathione ( GSH ) levels were decreased in A549 cells cultured in RPMI and exposed to CB-839 ( Figure 3—figure supplement 2B ) , suggesting that high cystine-driven use of glutaminase is required for both TCA cycle anaplerosis and maintenance of glutathione . CB-839 treatment also caused increased levels of DCFDA staining , indicating ROS levels increase following glutaminase inhibition , which was reversible by co-treatment with N-acetylcysteine ( NAC ) ( Figure 3—figure supplement 2C ) . These data show that both glutathione depletion and TCA cycle depletion occur upon glutaminase inhibition in the presence of high extracellular cystine . We next performed rescue experiments with dimethyl α-ketoglutarate or glutathione monoethyl ester to determine if replenishment of the TCA cycle or glutathione could prevent the anti-proliferative effect of glutaminase inhibition in RPMI which contains high levels of cystine . We also assessed if glutathione depletion with buthionine sulfoximine , an inhibitor of GSH synthesis , had similar anti-proliferative effects as glutaminase inhibition with CB-839 , and if NAC treatment could reverse the effects of CB-839 treatment . Treatment of cells with dimethyl α-ketoglutarate or glutathione monoethyl ester was able to rescue the proliferation of CB-839 treated A549 cells , albeit to differing extents ( Figure 3—figure supplement 2D ) . However , assessment of TCA cycle metabolite and GSH levels in cells rescued with dimethyl α-ketoglutarate , showed that this rescued both TCA metabolite pools and GSH levels ( Figure 3—figure supplement 2E ) . Similarly , glutathione monoethyl ester treatment increased levels of both TCA cycle metabolites and GSH in CB-839 treated cells , although glutathione monoethyl ester was less capable of rescuing both the anti-proliferative effect of CB-839 and metabolite levels at the concentration used ( Figure 3—figure supplement 2E ) . These data are consistent with previous reports showing A549 cells express γ-glutamyl transferase and are able to degrade GSH to replenish glutamate pools ( Kang et al . , 1994 ) . Thus , neither α-ketoglutarate or GSH rescue provided definitive insight into whether TCA cycle intermediates or GSH levels are more limiting for proliferation in high cystine upon glutaminase inhibition . However , as described previously ( Kang et al . , 1991 ) , we found that BSO treatment had no effect on A549 proliferation ( Figure 3—figure supplement 2D ) despite GSH depletion to undetectable levels ( Figure 3—figure supplement 2E ) . NAC treatment , while able to decrease DCFDA detectable ROS ( Figure 3—figure supplement 2C ) , did not rescue cell proliferation following CB-839 treatment ( Figure 3—figure supplement 2D ) . Thus , while we have not identified precisely which glutamate dependent metabolic processes are limiting for proliferation upon glutaminase inhibition in high cystine conditions , these data argue that glutathione depletion and the subsequent increase in ROS alone is not sufficient to explain decreased proliferation of these cells . How might exogenous cystine regulate the contribution of glutamine carbon to the TCA cycle ? Glutamine metabolism is linked to cystine metabolism via the xc– transporter system . This transporter system is composed of the transporter xCT ( encoded by SLC7A11 ) and the chaperone 4F2hc/CD98 ( encoded by SLC3A2 ) , and together they mediate the exchange of glutamate for cystine across the plasma membrane ( Lewerenz et al . , 2013 ) . We hypothesized that in the presence of high exogenous cystine , xCT-mediated transport of glutamate might deplete the intracellular glutamate/αKG pool , thus promoting glutamate regeneration from glutamine via glutaminase ( Figure 4A ) . To begin to test this , we measured glutamate release and uptake of both glutamine and cystine by A549 cells cultured in RPMI or RPMI containing low levels ( 10 µM ) of cystine . Consistent with the hypothesis that cystine potentiates glutamine anaplerosis by triggering enhanced glutamate secretion via xCT and increased glutamine catabolism , we observed that higher extracellular cystine increased both the glutamine and cystine uptake rates as well as the glutamate release rate ( Figure 4—figure supplement 1 ) in A549 cells . To determine if xCT/SLC7A11 was required for cystine induced glutamine anaplerosis and dependence , we generated A549 cells with shRNA-mediated stable knockdown of SLC7A11 , where knockdown was or was not rescued by expression of an shRNA-resistant SLC7A11 cDNA ( Figure 4B ) . We first confirmed that knockdown of SLC7A11 with two hairpins reduced the glutamine uptake and glutamate release rate of cells cultured in RPMI , but did not have a detectable effect on glutamine uptake and glutamate release when cells were cultured in RPMI containing low cystine ( 10 µM ) ( Figure 4—figure supplement 2A ) . Knockdown of SLC7A11 had no consistent effect on steady state intracellular levels of glutamate or the glutamate derived metabolite glutathione in either RPMI or RPMI with low cystine ( Figure 4—figure supplement 2B ) . We next traced the fate of [U-13C5]glutamine in SLC7A11 knockdown or control cells when cultured in adult bovine serum or adult bovine serum supplemented with RPMI levels of cystine . To quantify the extent to which cystine enhanced glutamine anaplerosis , we assessed the difference between normalized m + 5 labeled αKG in adult bovine serum and in adult bovine serum with high cystine . We found that knockdown of SLC7A11 substantially decreased the ability of cystine to potentiate glutamine anaplerosis , and that this effect was blunted by expression of the shRNA-resistant SLC7A11 cDNA ( Figure 4C ) . Additionally , knockdown of SLC7A11 with multiple hairpins ( Figure 4D ) abrogated CB-839 sensitivity of A549 cells in adult bovine serum with RPMI levels of cystine ( Figure 4E ) . Thus , xCT/SLC7A11 is necessary for cystine induced glutamine anaplerosis and dependence in A549 cells . We next asked if xCT/SLC7A11 expression influences cystine-induced glutamine anaplerosis across other human cancer cell lines . We traced [U-13C5]glutamine fate in a panel of human cell lines from cancers arising in multiple tissues and with multiple genetic drivers ( Figure 4—source data 1 ) . These cells were cultured in RPMI or RPMI with low ( 10 µM ) cystine and the extent of glutamine anaplerosis determined as in Figure 4C . We found a strong correlation between cystine-induced glutamine anaplerosis and xCT/SLC7A11 mRNA expression reported by the Cancer Cell Line Encyclopedia ( CCLE ) ( Barretina et al . , 2012 ) , suggesting that cystine levels and xCT/SLC7A11 expression contribute to the prominent use of glutamine as a anaplerotic TCA cycle substrate in many cancer cells in culture ( Figure 4F ) . Given that xCT/SLC7A11 expression correlates with the ability of cystine to induce glutamine anaplerosis , we reasoned that xCT/SLC7A11 expression might also be a determinant of glutaminase dependence in cancer cells . We compared reported CB-839 IC50 values ( Gross et al . , 2014 ) and xCT/SLC7A11 mRNA expression levels from the CCLE for a panel of breast cancer cell lines ( Figure 4G and Figure 4—source data 1 ) . Sensitive lines tend to have higher xCT/SLC7A11 expression than resistant lines , suggesting that xCT/SLC7A11 expression may contribute to a requirement for glutaminase in cultured cell lines . We asked if increased xCT/SLC7A11 expression in CB-839 resistant cell lines would be sufficient to cause cystine-induced glutamine anaplerosis and increase glutaminase inhibitor sensitivity . xCT/SLC7A11 was expressed in three CB-839 resistant breast cancer cell lines , all of which had low baseline expression of this transport system: MCF7 , MDA-MB-468 and AU565 ( Figure 4H ) . We first confirmed that expression of SLC7A11 in MCF7 cells increased the glutamine uptake and glutamate release rate of cells cultured in RPMI , and that this effect was enhanced by high cystine levels in the media ( Figure 4—figure supplement 3A ) . SLC7A11 expression had no effect on intracellular glutamate or GSH levels in standard RPMI , but did prevent the decrease in glutathione levels observed when MCF7 cells are cultured in RPMI with low cystine ( Figure 4—figure supplement 3B ) . We next traced [U-13C5]glutamine fate in these cell lines with and without SLC7A11 overexpression . xCT/SLC7A11 expression increased the incorporation of glutamine carbon into TCA cycle intermediates in the presence of RPMI levels of cystine in all three cell lines ( Figure 4I ) . xCT/SLC7A11 expression also potentiated CB-839 sensitivity when these cells are cultured in RPMI , but not RPMI with 10 µM cystine ( Figure 4J and Figure 4—figure supplement 4 ) . These results demonstrate that xCT/SLC7A11 is necessary and sufficient for increased glutamine anaplerosis and glutaminase addiction in the presence of high levels of extracellular cystine . Lastly , we asked whether the relatively low level of cystine available in vivo could limit glutamine anaplerosis of xCT/SLC7A11-expressing tumors . For these studies , we sought to increase cystine availability in A549 tumor bearing mice and monitor glutamine carbon incorporation into metabolites in the tumors . We found that an oral dose of 2 . 4 g/kg cystine could raise plasma cystine levels in mice from ~15 µM to ~70 µM for at least 4 hr ( Figure 5A ) . We next administered cystine and [U-13C5]glutamine to A549 tumor bearing mice to determine if cystine administration would increase glutamine anaplerosis . Because oral cystine administration to mice surgically prepared for long-term glutamine infusion as performed in Figure 1B was not technically feasible , [U-13C5]glutamine was administered to tumor bearing mice via multiple bolus intravenous injections ( Elgogary et al . , 2016; Lane et al . , 2015; Tardito et al . , 2015; Yuneva et al . , 2012 ) 30 min after cystine dosing ( Figure 5B ) . This labeled glutamine delivery method does not allow us to assess steady state contribution of glutamine carbon to the TCA cycle in tumors; however , glutamine intravenous bolus injections do allow examination of glutamine incorporation into glutamate and downstream metabolites during a pre-steady state , kinetic labeling period and this method has previously been used to corroborate predicted changes in tumor glutamine utilization ( Elgogary et al . , 2016; Lane et al . , 2015; Tardito et al . , 2015; Yuneva et al . , 2012 ) ( Figure 5B ) . From measuring glutamine uptake and glutamate release rates under high and low cystine conditions ( Figure 4—figure supplement 1 ) , we surmise that the intracellular pools of glutamine and glutamate turnover more rapidly with increasing extracellular cystine . Thus , a prediction of our model that would be captured by kinetic [U-13C5]glutamine labeling is that intratumoral labeling of glutamine , as well as glutamate and TCA cycle metabolites , will be higher in the presence of exogenous cystine . Terminal [U-13C5]glutamine plasma enrichment was similar in cystine-treated and untreated animals , while cystine-treated animals had substantially higher plasma cystine concentrations ( Figure 5B ) . Consistent with increased glutamine catabolism in cystine treated tumors , labeling of glutamine , glutamate and αKG from [U-13C5]glutamine in tumors was higher in cystine treated mice ( Figure 5C ) . These results suggest that increasing cystine levels in tumors can cause increased glutamine consumption and glutamate release , a phenomenon we have shown using in vitro models leads to enhanced glutamine anaplerosis . These results are consistent with a model where low cystine levels in vivo limit tumor glutamine catabolism and anaplerosis .
Environmental differences between in vitro culture conditions and tumors in vivo can lead to differential reliance on glutamine anaplerosis , causing some cells to be addicted to glutamine catabolism in vitro . We found that non-physiological cystine levels in tissue culture can explain many of the differences in glutamine metabolism between lung cancer cells in vitro and in tumors . The glutamate/cystine antiporter xCT/SLC7A11 is necessary for the observed cystine-induced glutamine anaplerosis in A549 cells . Thus , high levels of extracellular cystine and the cellular capacity to secrete glutamate in exchange for cystine cooperate to increase glutamine anaplerosis in these cells . Increased glutamate secretion in the presence of cystine increases intracellular glutamate turnover , which in turn will increase the kinetic rate of glutamine label incorporation into both glutamate and TCA cycle intermediates . However , there are multiple carbon sources that can contribute to the pool of TCA cycle metabolites , raising the question of why increased glutamate secretion also results in increased anaplerotic contribution of glutamine carbon relative to other nutrients . The enzymology of glutaminase suggests one potential explanation . GLS1 activity is strongly inhibited by glutamate ( Cassago et al . , 2012; Curthoys and Watford , 1995 ) suggesting that the drop in intracellular glutamate driven by high cystine could raise glutaminase activity and increase glutamine contribution to the TCA cycle relative to other pathways . Thus , it could be the action of non-physiological cystine concentrations constantly depleting the intracellular glutamate pool that causes glutamine anaplerosis to become dominant . xCT/SLC7A11 expression correlates with the ability of cystine to increase glutamine anaplerosis in a panel of cell lines derived from tumors of different oncogenotype and tissue of origin ( Figure 4F ) . This argues that xCT/SLC7A11 expression may be necessary for environmental cystine to enhance glutamine anaplerosis regardless of tumor type , and is not a phenomenon limited to NSCLC and A549 cells . Indeed , examination of panels of human breast cancer cell lines have found that xCT/SLC7A11 promotes glutamate secretion , providing further evidence that xCT/SLC7A11 can promote glutamine catabolism in cell types beyond NSCLC ( Briggs et al . , 2016; Timmerman et al . , 2013 ) . Collectively , these experiments suggest that glutaminase dependence in culture for many cancer cells derived from various tumor types will be strongly influenced by both environmental cystine and xCT/SLC7A11 expression . These findings have clear implications for the clinical use of glutaminase inhibitors that are being evaluated in trials to treat a variety of tumor types ( https://clinicaltrials . gov/ct2/show/NCT02071862 ) . First , assessing xCT/SLC7A11 expression may help identify patients likely to benefit from these drugs . While many cancer cells are considered to be glutamine addicted , not every cancer cell line requires glutamine for proliferation in vitro ( Cetinbas et al . , 2016; Cheng et al . , 2011; Gross et al . , 2014; Timmerman et al . , 2013; van den Heuvel et al . , 2012 ) . Glutamine dependence has been linked to various oncogenes including MYC ( Gao et al . , 2009; Wise et al . , 2008 ) , RAS ( Son et al . , 2013 ) , IDH ( Matre et al . , 2014; Seltzer et al . , 2010 ) and hormone receptor status in breast cancers ( Gross et al . , 2014 ) . However , the genetic and biochemical basis for glutamine dependence remains poorly understood . Thus , understanding the underlying factors that cause increased reliance on glutamine is essential for identifying patients that would likely benefit from clinical glutaminase inhibitors . Identification of high xCT/SLC7A11 expression as a marker for glutaminase inhibitor sensitivity may be useful to identify patients that are more likely to benefit from glutaminase inhibitor therapy regardless of genotype . Functional examination of glutamate/cystine exchange by tumor cells using imaging techniques such as xCT-specific PET probes ( Baek et al . , 2012 ) may also be helpful in selecting tumors that are dependent on glutamine metabolism . xCT/SLC7A11 expression is known to be governed by the antioxidant response transcription factor NRF2 ( Sasaki et al . , 2002 ) . Interest in modulating the antioxidant response pathway has led to the development of a number of NRF2-activating compounds , some of which are already in clinical use ( Davies et al . , 2016; Gao et al . , 2014; Magesh et al . , 2012 ) . xCT/SLC7A11 expression triggered by NRF2-activating drugs can sensitize tumors to glutaminase inhibitors ( Sayin et al . , 2017 ) . In addition , beyond the use of xCT/SLC7A11 expression as a marker for glutaminase inhibitor responsiveness , raising tumor cystine levels might also be a way to induce or enhance glutamine addiction . We found that raising tissue cystine levels may increase glutamine anaplerosis in tumors . Therefore , co-administration of cystine with glutaminase inhibitors may enhance response to glutaminase inhibitors . Oral cystine administration can be effective to raise plasma cystine levels in humans ( Morin et al . , 1971 ) , arguing that this strategy could be applicable to treating patients . Eagle’s medium and RPMI were formulated to identify the minimal set of nutrients required for mammalian cancer cells and leukocytes to rapidly proliferate ( Eagle , 1955; Moore et al . , 1967 ) , not to match physiological levels of nutrients available in vivo . Thus , these media lack certain nutrients available to tumors in vivo and also contain other nutrients in excess to what is found in blood . Several studies examining tumor metabolism in mice and humans have suggested that the way nutrients are used in tumors can be different from what is observed in cell culture . While increased glucose consumption and lactate secretion observed in most cells in culture is also observed in tumors ( Davidson et al . , 2016; Hensley et al . , 2016; Marin-Valencia et al . , 2012; Mashimo et al . , 2014; Sellers et al . , 2015; Tardito et al . , 2015 ) , glutamine oxidation and anaplerosis can differ between both lung and glial tumors and cell lines derived from these tumors in culture ( Davidson et al . , 2016; Marin-Valencia et al . , 2012; Tardito et al . , 2015 ) . Thus , standard tissue culture can cause some aspects of central carbon metabolism to be misrepresented and the identification of media that better model the environmental conditions in tumors may facilitate better understanding of cancer metabolism . Previous studies have suggested that altering tissue culture media to better reflect levels of amino acids found in human plasma can allow more accurate modeling of tissue metabolism ( Schug et al . , 2015; Tardito et al . , 2015 ) . Culturing glioblastoma cells in serum-like medium ( SMEM ) results in decreased glutamine utilization and a requirement to net produce glutamine from other carbon sources , similar to what is observed in glioblastoma in vivo ( Tardito et al . , 2015 ) . The differences between SMEM and standard media formulations that elicited this difference is unknown , but the fact that culturing cells in media with more physiological cystine levels resulted in a change in glutamine metabolism is consistent with our findings . In culture systems , the bulk of TCA cycle carbon can be accounted for from glucose and glutamine . However , the sources of anaplerotic TCA cycle carbon in tumors are not fully understood despite the increased contribution of glucose anaplerosis in some settings ( Davidson et al . , 2016; Hensley et al . , 2016; Sellers et al . , 2015 ) . Adult bovine serum differs from standard media in levels of many nutrients . Thus , additional nutritional differences beyond amino acids may contribute to why glutamine metabolism of A549 cells in serum more closely mimics A549 tumors in vivo and suggests factors other than cystine levels might be involved . Supporting this hypothesis , addition of cystine to adult bovine serum increases A549 incorporation of glutamine carbon into the TCA cycle , but not to the same levels observed when the same cells are cultured in RPMI ( Figure 3A ) . Additionally , A549 cells grown in adult bovine serum are also more resistant to CB-839 than A549 cells grown in RPMI with low levels of cystine ( 10 µM ) ( Figure 3B ) . These results suggest that alternative anaplerotic carbon source ( s ) are available in adult bovine serum that contribute to further glutaminase independence . Identification of what contributes to the TCA cycle in serum may yield new insight into the anaplerotic carbon sources used by cells in vivo . Beyond the contribution of oncogenic mutations to reprogramming cellular metabolism , environment also plays a fundamental role in determining the metabolic program of cancer cells . We find that environmental cystine level is one variable that can alter how nutrients are used by cancer cells . Many studies aim to identify the set of genes required for cancer cell proliferation , with the goal of identifying novel therapeutic targets ( Hart et al . , 2015; Shalem et al . , 2014; Wang et al . , 2017 ) . Based on our results , selection of media conditions will be essential to maximize the in vivo relevance of targets identified by screens to find metabolic gene vulnerabilities . Our results motivate further studies to characterize the nutritional content of the tumor microenvironment , and development of tissue culture systems to propagate and study cells under such conditions , as this may help translate metabolic dependencies of cancer cell lines into better cancer therapies .
All cell lines used in this study were directly obtained from ATCC ( Manassas , VA ) ( A549: ATCC Cat# CRM-CCL-185 , RRID:CVCL_0023; HCT116: ATCC Cat# CCL-247 , RRID:CVCL_0291; 143B: ATCC Cat# CRL-8303 , RRID:CVCL_2270; MDAMB231: ATCC Cat# HTB-26 , RRID:CVCL_0062; PC3: ATCC Cat# CRL-1435 , RRID:CVCL_0035; NCIH1299: ATCC Cat# CRL-5803 , RRID:CVCL_0060; TT: ATCC Cat# CRL-1803 , RRID:CVCL_1774; MDAMB468: ATCC Cat# HTB-132 , RRID:CVCL_0419; ASPC1: ATCC Cat# CRL-1682 , RRID:CVCL_0152; HS578T: ATCC Cat# HTB-126 , RRID:CVCL_0332; NCIH226: ATCC Cat# CRL-5826 , RRID:CVCL_1544; NCIH1395: ATCC Cat# CRL-5868 , RRID:CVCL_1467; MIAPACA2: ATCC Cat# CRM-CRL-1420 , RRID:CVCL_0428; AU565: ATCC Cat# CRL-2351 , RRID:CVCL_1074; BT474: ATCC Cat# CRL-7913 , RRID:CVCL_0179; MCF7: ATCC Cat# HTB-22 , RRID:CVCL_0031 ) and DMSZ ( Braunschweig , Germany ) ( CAL120: DSMZ Cat# ACC-459 , RRID:CVCL_1104 ) or were gifts from other laboratories ( PANC1: RRID:CVCL_0480; EVSAT: RRID:CVCL_1207 ) . All cell lines not obtained directly from ATCC or DMSZ were STR tested to confirm their identity prior to use ( University of Arizona Genetics Core , Tucson , AZ ) . All cell lines were regularly tested for mycoplasma contamination using the Mycoprobe mycoplasma detection kit ( R and D Systems , Minneapolis , MN ) . All cells were cultured in a Heracell ( Thermofisher , Waltham , MA ) humidified incubators at 37°C and 5% CO2 . Cell lines were routinely maintained in RPMI-1640 or DMEM ( Corning Life Sciences , Tewksbury , MA ) supplemented with 10% heat inactivated fetal bovine serum ( VWR Seradigm , Radnor , PA , Lot 120B14 ) . For continuous maintenance of A549 cells in adult bovine serum , the following modifications to standard tissue culture practices were made to maintain A549 cell viability during passage . First , cells were cultured in large volumes of adult bovine serum ( 8 mL/35 mm diameter well ) to prevent depletion of essential nutrients during the culture . Additionally , the adult bovine serum was replaced with fresh adult bovine serum every 48 hr . Second , passaging serum maintained cells using standard trypsin ( 0 . 025% ) /EDTA solution to detach cells leads to significant loss of viability . Therefore , serum grown cells were detached with a 1:1 mixture of 0 . 5% trypsin-EDTA ( Thermofisher ) and serum free Eagle’s minimal essential media ( Thermofisher ) . This allowed routine passaging and plating of cells with minimal loss of viability . In unpublished observations made while culturing many cells lines in adult bovine serum , we have found that not all cell lines proliferate as standard monolayer cultures in adult bovine serum as they do in RPMI . Preliminarily , we have observed that some of these cells lines can proliferate in adult bovine serum when coated with collagen I according to previously published methods ( Olivares et al . , 2017 ) , or by the addition of cystine to slightly higher levels ( 20–50 µM ) than those found in adult bovine serum . All stable isotope tracing and proliferation rate experiments were performed in RPMI or DMEM containing 10% heat inactivated fetal bovine serum ( Seradigm , Lot 120B14 ) that was repeatedly dialyzed against saline ( 150 mM NaCl ) using 3 . 5 kDa cutoff membranes ( Thermofisher ) to remove all small molecule metabolites from the serum . Adult bovine serum ( Sigma Aldrich , St . Louis , MO , Lot 16A041 ) , adult bovine heparinized plasma ( Pel-freeze , Rogers , AR ) and pooled adult human serum ( Innovative Research , Novi , MI , Lot 20211 ) were thawed at 37°C for 2 hr . and heat inactivated for 30 min . at 56°C then filtered through a 0 . 22 µm filter prior to use or storage at −20°C . To analyze absolute concentrations of amino acids , pyruvate and lactate in adult bovine serum , adult heparinized plasma and adult human serum , 10 μL of the serum or plasma was added to 10 μL of isotopically labeled internal standards of amino acids ( Cambrige Isotope Laboratory , Andover , MA , MSK-A2-1 . 2 , CLM-1822-H-PK , and CLM-8699-H-PK ) , pyruvate ( Cambridge Isotope Laboratories , CLM-2440-PK ) and lactate ( Sigma Aldrich , 485926 ) . These mixtures were then extracted in 600 μL of ice cold HPLC grade methanol , vortexed for 10 min , and centrifuged at 21 kg for 10 min . 450 μL of each extract was removed and dried under nitrogen gas and stored −80°C until further analysis by GC-MS . Glucose concentration of serum or plasma was measured using a YSI-2900D Biochemistry Analyzer ( Yellow Springs Instruments , Yellow Springs , OH ) according to the manufacturer’s instructions . To generate adult bovine serum containing added amino acids , stock amino acid mixtures were generated by adding individual amino acid powders ( all obtained from Sigma Aldrich ) at appropriate ratios in an electric blade coffee grinder ( Hamilton Beach , Glen Allen , VA , 80365 ) . The amino acids were then mixed using 10 pulses on the espresso setting . An appropriate amount of the amino acid mix was added to adult bovine serum , which was then sterilized using a 0 . 22 µm filter prior to use . To prepare the dialyzed media ‘Adult bovine serum → RPMI or DMEM’ , 210 mL of RPMI or DMEM was dialyzed twice overnight at 4°C against 4 L of adult bovine serum using 70 mL 3 . 5 kDa cutoff dialysis cassettes ( Thermofisher ) . For ‘RPMI or DMEM → adult bovine serum’ , 210 mL of adult bovine serum was dialyzed against RPMI or DMEM as above . These media were sterilized using a 0 . 22 µm filter prior to use . For addition of metabolites to RPMI-1640 , glutathione monoethyl ester ( Santa Cruz , Dallas , TX ) , N-acetylcysteine ( Sigma ) , dimethyl α-ketoglutarate ( Sigma ) were added at the indicated concentrations to RPMI-1640 , which was then sterilized using a 0 . 22 µm filter prior to use . Stable cDNA or shRNA expressing cell lines were generated by lentiviral or retroviral infection for 24 hr followed by selection in RPMI-1640 containing 1 µg/mL puromycin or 500 µg/mL hygromycin B . Mock infected cells were similarly selected , and selection was considered complete when no viable cells were detected in these mock infection controls . All virally manipulated cells were maintained under antibiotic selection at indicated concentrations until used in experimental assays . For reduction of xCT/SLC7A11 expression , lentiviral pLKO human SLC7A11 shRNA vectors ( SLC7A11 shRNA #1: TRCN0000288926 and SLC7A11 shRNA #2: TRCN0000288927 ) containing the puromycin resistance gene were obtained from Sigma Aldrich . A pLKO vector targeting GFP was used as a control . To rescue xCT/SLC7A11 expression , cDNA encoding human SLC7A11 was obtained from Origene ( Rockville , MD ) . The cDNA was then PCR amplified and cloned into the HindIII and ClaI sites of retroviral vector pLHCX , which contains the hygromycin resistance gene ( Clontech , Mountain View , CA ) , allowing for CMV promoter driven expression of SLC7A11 . Site directed mutagenesis ( Hemsley et al . , 1989 ) was performed to synonymously mutate the sequence of SLC7A11 targeted by SLC7A11 shRNA #1: TRCN0000288926 ( CCCTGGAGTTATGCAGCTAAT ) such that it would no longer be targeted by this shRNA ( GCCCGGCGTGATGCAATTGAT ) . To overexpress SLC7A11 , SLC7A11 cDNA was PCR amplified and ligated into the SalI and NotI sites of the vector pENTR4 ( no ccDB ) ( Addgene , Cambridge , MA ) ( Campeau et al . , 2009 ) . LR Gateway recombination ( Thermofisher ) was used to clone the SLC7A11 cDNA from pENTR4 ( no ccDB ) into the expression vector pLenti CMV Puro DEST ( w118-1 ) ( Addgene ) to allow for puromycin selectable CMV driven expression of SLC7A11 ( Campeau et al . , 2009 ) . All vectors constructed for this study had the entire coding sequence confirmed by Sanger sequencing ( Quintara Biosciences , Berkeley , CA ) prior to use . Cellular proliferation rate in different media and drug conditions was determined as previously described ( Sullivan et al . , 2015 ) . Briefly , cell lines proliferating in log phase in RPMI medium were trypsinized , counted and plated into six well dishes ( Corning Life Sciences ) in 2 mL of RPMI medium and incubated overnight . Initial seeding density was 20 , 000 cells/well for A549 cells , or 50 , 000 cells for MCF7 , AU565 and MDA-MB-468 cells . The next day , a six well plate of cells was trypsinized and counted to provide a number of cells at the start of the experiment . Cells were then washed twice with 2 mL of phosphate buffered saline ( PBS ) , and 8 mL of the indicated media premixed with indicated compounds or vehicles was added . This large volume of media was chosen to prevent severe nutrient depletion , especially when adding adult bovine serum medium . Cells were then trypsinized and counted 4 days after adding the indicated medias . Proliferation rate was determined using the following formula: Proliferation rate in doublings/day = [Log2 ( Final Day 4 cell count/Initial Day 0 cell count ) ]/4 days . Cells were counted using a Cellometer Auto T4 Plus Cell Counter ( Nexcelom Bioscience , Lawrence , MA ) . Measurement of intracellular DCFDA fluorescence was performed using the Abcam DCFDA Cellular Reactive Oxygen Species Detection Assay Kit ( Abcam , Cambridge , MA , ab113851 ) according to the manufacturer’s instructions . Briefly , 25 , 000 A549 cells were plated in each well of a 96 well plate in 200 µL RPMI without phenol red ( Sigma Aldrich ) . Cells were allowed to attach for 8 hr . Subsequently , the media on the cells was changed to 200 µL fresh media containing DMSO , CB-839 and NAC as indicated . After 24 hr of treatment , the cells were washed twice with 200 µL of 1x Buffer included in the assay kit . Cells were then incubated with 100 µL of 1x Buffer containing 10 µM DCFDA for 45 min . at 37°C . An unstained control was included , and incubated with 100 µL of 1x Buffer alone . After staining , the DCFDA containing solution was removed and cells were resuspended 100 µL of PBS . DCFDA fluorescence was measured using an Infinite M200Pro plate reader ( Tecan ) in fluorescence mode with excitation at 485 and emission at 535 . Cells were then detached with trypsin-EDTA solution and counted using a Cellometer Auto T4 Plus Cell Counter ( Nexcelom Bioscience ) . The ratio of DCFDA signal intensity to cell number was subsequently computed . For immunoblotting analysis , cell lines growing in log phase were trypsinized , counted and plated at a density of 400 , 000 cells/well of a six well dish . The following day , cells were washed with 2 mL of PBS and then lysed in 100 μL RIPA buffer [25 mM Tris-Cl , 150 mM NaCl , 0 . 5% sodium deoxycholate , 1% Triton X-100 , 1x cOmplete protease inhibitor ( Roche ) ] . Cells were scraped and the resulting lysate was clarified by centrifugation at 21 kg for 20 min . Protein concentration of the lysate was determined by BCA assay ( Thermofisher ) . Lysates were resuspended at 2 μg/μL in Laemmli SDS PAGE sample loading buffer ( 10% glycerol , 2% SDS , 60 mM Tris-Cl pH 6 . 8 , 1% β-mercaptoethanol , 0 . 01% bromophenol blue ) and denatured at 100°C for 5 min . Extracts ( 30 μg of protein ) were resolved by SDS PAGE using 10% acrylamide gels running at 120 V until the dye front left the gel . After SDS-PAGE resolution , protein extracts were transferred to nitrocellulose using an iBlot semidry transfer system ( Thermofisher ) . Membranes were subsequently incubated with primary antibody in Odyssey buffer ( Licor Biosciences , Omaha , NE ) , washed , and incubated with IRDye-conjugated anti-mouse and anti-rabbit IgG secondary antibodies in Odyssey buffer with 0 . 1% Tween-20% and 0 . 02% SDS . Blots were imaged using an Odyssey infrared scanner ( Licor Biosciences ) . Antibodies and dilutions used in this study were: 1:1000 rabbit anti-xCT ( Cell Signaling Technology , Danvers , MA , 12691S ) , 1:10000 mouse anti-Vinculin ( Abcam , ab18058 , RRID:AB_444215 ) , 1:10000 IR680LT dye conjugated goat anti-rabbit IgG ( Licor Biosciences , 925–68021 ) , 1:10000 IR800 dye conjugated goat anti-mouse IgG ( Licor Biosciences , 925–32210 ) . For amino acid consumption analysis , cell lines growing in log phase were trypsinized , counted and plated at a density of 100 , 000 cells/well of a six well dish and allowed to attach for 24 hr . The following day , cells were washed with 2 mL of PBS and then fed 2 mL of either RPMI or RPMI ( 10 µM cystine ) media . 1 mL of media was immediately removed , spun for 3 min . at 845 g to remove any cells from the media , and then frozen at −80°C for later analysis . Cell number was also determined using a Cellometer Auto T4 Plus Cell Counter ( Nexcelom Bioscience ) . 2 days later , media was harvested and cell number of the culture again determined . Amino acid concentration in the fresh or spent media was determined by GC-MS as described above in Media preparation and analysis . To calculate amino acid consumption rates , cell numbers at the initial and day two time points were used to fit an exponential growth function , and integration of these curves yielded the number of ( cell · days ) by which the media was conditioned . Changes in amino acid concentration for each culture were normalized to the integrated growth curve of each culture to yield amino acid consumption/release per cell per unit time . To determine steady state labeling of polar metabolites by glutamine in cultured cells , cell lines were seeded at an initial density of 200 , 000 cells/well in a six well dish in 2 mL of RPMI medium . Cells were incubated for 24 hr , and then washed twice with 2 mL of PBS . Cells were then incubated for 8 or 24 hr in the indicated media , to which [U-13C5]glutamine ( Cambridge Isotope Laboratories , CLM-1822-H-PK ) was added , such that the fractional enrichment of glutamine in the given medium would be ~33% . Following the labeling period , media was aspirated from cells and the cells were rapidly washed in ~8 mL of ice cold saline . The saline was subsequently aspirated and 600 μL of ice cold methanol:water ( 4:1 ) was added . Cells were scraped on ice , and the resulting extracts were vortexed for 10 min , and centrifuged at 21 kg for 10 min . 450 μL of each extract was removed and dried under nitrogen gas and stored −80°C until further analysis . Cell cultures treated as indicated were washed in blood bank saline and extracted in methanol:water as described above in Cell culture isotopic labeling experiments and metabolite extraction . Cellular extracts were then resuspended in 100 µL of acetonitrile:water ( 1:1 ) . LC-MS analysis was then performed using a QExactive orbitrap mass spectrometer using an Ion Max source and heated electrospray ionization ( HESI ) probe coupled to a Dionex Ultimate 3000 UPLC system ( Thermofisher ) . External mass calibration was performed every 7 days . Samples were separated by chromatography by injecting 10 µL of sample on a SeQuant ZIC-pHILIC 2 . 1 mm x 150 mm ( 5 µm particle size ) column . Flow rate was set to 150 µL/min . and temperatures were set to 25°C for the column compartment and 4°C for the autosampler tray . Mobile phase A was 20 mM ammonium carbonat , 0 . 1% ammonium hydroxide . Mobile phase B was 100% acetonitrile . The chromatographic gradient was: 0–20 min . : linear gradient from 80% to 20% mobile phase B; 20–20 . 5 min . : linear gradient from 20% to 80% mobile phase B; 20 . 5 to 28 min . : hold at 80% mobile phase B . The mass spectrometer was operated in full scan , polarity-switching mode and the spray voltage was set to 3 . 0 kV , the heated capillary held at 275°C , and the HESI probe was held at 350°C . The sheath gas flow rate was 40 units , the auxiliary gas flow was 15 units and the sweep gas flow was one unit . The MS data acquisition was performed in a range of 70–1000 m/z , with the resolution set to 70 , 000 , the AGC target at 1e5 and the maximum injection time at 20 msec . Relative quantitation of metabolites was performed with XCalibur QuanBrowser 2 . 2 ( Thermofisher ) using a five ppm mass tolerance , referencing an in-house retention time and m/z library of metabolite standards . In order to detect GSH and GSSG , the MS was operated in targeted selected ion monitoring ( tSIM ) mode with the quadrapole centered on M + H m/z 308 . 0811 ( GSH ) and m/z 613 . 1592 , with an isolation window of 1 m/z . The resolution was set at 70 , 000 , the AGC target was 1e5 , and the maximum injection time was 250 ms . All experiments performed in this study were approved by the MIT Committee on Animal Care ( IACUC ) . Nu/nu mice were purchased from Charles River ( Wilmington , MA ) ( RRID:IMSR_CRL:088 ) and housed on a 12 hr light and 12 hr dark cycle , with ad lib access to food and water . For subcutaneous xenograft studies , mice were injected with 2 , 000 , 000 A549 cells ( suspended in a volume of 100 μL PBS ) per site into the right and left flank . Continuous infusions were performed as previously described ( Davidson et al . , 2016 ) . 3–4 days prior to tracer studies , tumor-bearing mice underwent a surgery to implant a catheter into the jugular vein and were allowed to recover . Mice were infused with [U-13C5]glutamine at 3 . 7 mg/kg/min for 300 min . At the end of the infusion , mice were terminally anesthetized with sodium pentobarbital and blood collected via heart puncture . Tissues were rapidly collected , freeze-clamped in liquid nitrogen , and stored −80°C . Repeated bolus intravenous injection tracer studies were performed as previously described ( Lane et al . , 2015; Yuneva et al . , 2012 ) . Tumor-bearing mice were orally administered a 2 . 4 g/kg dose of L-cystine ( Sigma Aldrich ) . Cystine was formulated as a 240 mg/mL suspension in 0 . 1% Tween 80% and 0 . 5% methylcellulose . Forty-five minutes after gavage , mice were lightly anesthetized using isofluorane and intravenously injected with 200 μL of [U-13C5]glutamine tracer ( 36 . 2 mg/ml dissolved in saline ) . These injections were performed a total of 3 times at 15 min intervals . Forty-five minutes after the first injection ( and 15 min after the last injection ) , animals were euthanized , blood collected via heart puncture , and relevant tissues rapidly collected and freeze-clamped in liquid nitrogen . Frozen tissues were weighed ( 10–20 mg ) and pulverized using a cryomill ( Retsch , Haan , Germany ) . Metabolites were extracted in 1 . 3 mL chloroform:methanol:water ( 4:6:3 ) , vortexed for 10 min , and centrifuged at 21 kg for 10 min . Polar metabolites were dried under nitrogen gas and stored −80°C until further analysis . Blood collected from animals was immediately placed in EDTA-tubes ( Sarstedt , North Rhine-Westphalia , Germany ) and centrifuged to separate plasma . To analyze absolute concentrations of amino acids in the plasma , including cystine , 10 μL of plasma was added to 10 μL of a mixture of isotopically labeled amino acids of known concentrations ( Cambridge Isotope Laboratories , MSK-A2-1 . 2 ) . To analyze fractional enrichment of metabolites , 10 μL of plasma was diluted with 10 μL of water . Diluted plasma samples were extracted in 600 μL of ice cold HPLC grade methanol , vortexed for 10 min , and centrifuged at maximum speed for 10 min . 450 μL of each extract was removed and dried under nitrogen gas and stored −80°C until further analysis . Polar metabolites were analyzed by GC-MS as described previously ( Lewis et al . , 2014 ) . Dried and frozen metabolite extracts were derivitized with 16 μL MOX reagent ( Thermofisher ) for 60 min . at 37°C . Samples were then derivitized with N-tertbutyldimethylsilyl-N-methyltrifluoroacetamide with 1% tert-butyldimethylchlorosilane ( Sigma Aldrich ) 30 min . at 60°C . Following derivitization , samples were analyzed by GC-MS , using a DB-35MS column ( Agilent Technologies , Santa Clara , CA ) installed in an Agilent 7890A gas chromatograph coupled to an Agilent 5997B mass spectrometer . Helium was used as the carrier gas at a flow rate of 1 . 2 mL/min . One microliter of sample was injected in split mode ( all samples were split 1:1 ) at 270°C . After injection , the GC oven was held at 100°C for 1 min . and increased to 300°C at 3 . 5 °C/min . The oven was then ramped to 320°C at 20 °C/min . and held for 5 min . at this 320°C . The MS system operated under electron impact ionization at 70 eV and the MS source and quadrupole were held at 230°C and 150°C respectively . The detector was used in scanning mode , and the scanned ion range was 100–650 m/z . Mass isotopomer distributions were determined by integrating appropriate ion fragments for each metabolite ( Lewis et al . , 2014 ) using in-house software ( Young et al . , 2008 ) that corrects for natural abundance using previously described methods ( Fernandez et al . , 1996 ) . | Cancer cells need to consume certain nutrients in order to grow , and some cancer drugs work by affecting the ability of the cells to use these nutrients . For decades researchers have grown cancer cells in petri dishes with standard nutrient formulations ( also known as tissue culture ) , but the nutrients available to cancer cells in tissue culture are not the same as those found in the body . Cancer cells growing in tissue culture consume large amounts of a nutrient called glutamine . These cells die when exposed to a class of drugs called glutaminase inhibitors that prevent them from processing glutamine . However , when these same cancer cells grow as tumors in animals , they process less glutamine and are not affected by glutaminase inhibitors . So what differences are there between growing cancer cells in tumors and tissue culture that explain this different reliance on glutamine ? Muir et al . reasoned that changing the levels of nutrients available to cancer cells might change what these cells consume , and so grew human cancer cells in cow serum ( which has a similar nutrient content to blood in humans and other mammals ) . Indeed , these cells consumed less glutamine and responded to glutaminase inhibitors in a way that is similar to how tumors in the body respond to these drugs . Comparing the nutrient content of cow serum and typical tissue culture formulations revealed that high levels of the nutrient cystine cause the cells to rely more on glutamine . The results presented by Muir et al . suggest that cancer cells in tumors could be made to consume more glutamine and that this would make them sensitive to glutaminase inhibitors – a possibility that will be studied in future work . Exposing cultured cancer cells to nutrient levels closer to those found in the body may also better predict how tumor cells use nutrients and respond to some treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2017 | Environmental cystine drives glutamine anaplerosis and sensitizes cancer cells to glutaminase inhibition |
In many organisms , it remains unclear how X chromosomes are specified for dosage compensation , since DNA sequence motifs shown to be important for dosage compensation complex ( DCC ) recruitment are themselves not X-specific . Here , we addressed this problem in C . elegans . We found that the DCC recruiter , SDC-2 , is required to maintain open chromatin at a small number of primary DCC recruitment sites , whose sequence and genomic context are X-specific . Along the X , primary recruitment sites are interspersed with secondary sites , whose function is X-dependent . A secondary site can ectopically recruit the DCC when additional recruitment sites are inserted either in tandem or at a distance ( >30 kb ) . Deletion of a recruitment site on the X results in reduced DCC binding across several megabases surrounded by topologically associating domain ( TAD ) boundaries . Our work elucidates that hierarchy and long-distance cooperativity between gene-regulatory elements target a single chromosome for regulation .
Eukaryotic genomes are large , encompassing thousands of genes and spanning across millions of base pairs . In order to fit within the limiting confines of the nucleus , genomic DNA is compacted in a hierarchical manner . Linear DNA wraps around histones to form nucleosomes . A string of nucleosomes makes up the chromatin fibers , which further organize into variably sized topologically associating domains ( TADs ) ( Dekker and Heard , 2015; Matharu and Ahanger , 2015 ) . Finally , each chromosome occupies an individual territory within the nucleus ( Dekker and Mirny , 2016; Denker and de Laat , 2016 ) . Importantly , this genome organization is a dynamic process with implications for transcriptional regulation . For example , nucleosome positioning regulates transcription factor ( TF ) binding at promoters and enhancers , which in turn interact via long-range chromatin looping . At the domain-scale , the coordinated regulation of many genes across hundreds of kilobases of DNA has been observed in several organisms ( Cohen et al . , 2000; Scott-Boyer and Deschepper , 2013; Spellman and Rubin , 2002 ) . For instance , the coordinated regulation of gene clusters , such as the Hox genes , can span hundreds of kilobases and requires the specific recruitment of regulatory proteins ( Bauer et al . , 2016 ) . How gene regulatory complexes are specifically targeted to large chromosomal domains remains poorly understood . Research on the X-specific targeting of dosage compensation machinery in various animals helps to elucidate the molecular mechanisms that specifically target chromosomal domains for transcriptional co-regulation . X chromosome dosage compensation refers to mechanisms that equalize X chromosome gene expression between males ( XY ) and females ( XX ) of diploid species wherein sex is determined by differences in X chromosome copy number ( Adler et al . , 1997; Albritton et al . , 2014; Chen and Zhang , 2015; Deng et al . , 2011; Lin et al . , 2012 , Lin et al . , 2011; Veitia et al . , 2015; Wheeler et al . , 2016 ) . Here , we focus on the transcriptional regulatory mechanisms that act to restore X expression balance between the sexes . In wild type mammals , X expression balance is achieved via X inactivation , wherein one of the two female X chromosomes is transcriptionally silenced during development ( Heard and Disteche , 2006 ) . In Drosophila melanogaster , the male-specific lethal ( MSL ) complex binds to the single X chromosome in males where it upregulates transcription two-fold ( Conrad and Akhtar , 2012 ) . In Caenorhabditis elegans , both hermaphrodite X chromosomes are targeted by the dosage compensation complex ( DCC ) , resulting in a two-fold downregulation of expression from each ( Ercan and Lieb , 2009; Ercan , 2015 ) . Although the mechanisms are diverse , in each case , a protein complex is specifically targeted to the X chromosome in only the sex where it regulates transcription ( Ercan , 2015 ) . The working model for X-specific binding of the different dosage compensation complexes involves a two-step strategy: recruitment and spreading ( Ercan , 2015 ) . While the mechanism of spreading has been observed to be sequence-independent ( Ercan et al . , 2009; Kelley et al . , 1999; Sun and Birchler , 2009 ) , in both C . elegans and D . melanogaster , recruitment is accomplished in part by short DNA sequence motifs ( Alekseyenko et al . , 2012 , Alekseyenko et al . , 2008; Ercan et al . , 2007; Jans et al . , 2009 ) . Although important for DCC and MSL recruitment , the DNA sequence motifs are not sufficient to explain X-specificity . In both species , motif sequence is found across the genome , and is only slightly enriched on the X chromosome ( Alekseyenko et al . , 2012 , Alekseyenko et al . , 2008; Ercan et al . , 2007; Jans et al . , 2009 ) . This epitomizes an important open question in general mechanisms of domain-scale gene regulation: how are gene regulatory complexes specifically targeted to large chromosomal domains when their corresponding recruitment motifs lack domain-specificity . In this study , we combined systematic ChIP-seq analysis with recruitment mutants and targeted genome editing to determine how the C . elegans DCC is specifically recruited to the X chromosomes . Our work offers an answer to the specificity question , indicating that hierarchy and long-distance cooperation between a set of motif-containing recruitment sites restricts binding of the DCC to the X chromosomes . At the core of the DCC is a condensin complex ( hereafter condensin DC ) ( Csankovszki et al . , 2009a , Csankovszki et al . , 2009b ) ( Figure 1A ) . Condensins are evolutionarily conserved protein complexes , most often cited for their role in chromosome condensation and segregation during cell division ( reviewed in [Hirano , 2016] ) . Recent work also suggests key roles for condensins in gene regulation during interphase ( Cobbe et al . , 2006; Kranz et al . , 2013; Longworth et al . , 2012; Rawlings et al . , 2011; Dej et al . , 2004; Lupo et al . , 2001 ) . Condensins are composed of a dimerizing pair of structural maintenance of chromosomes proteins ( SMC-2 and SMC-4 ) that interact with three chromosome-associated polypeptides ( CAPs ) . C . elegans condensin DC shares four out of five subunits ( MIX-1 [Lieb et al . , 1998] , DPY-26 [Plenefisch et al . , 1989] , DPY-28 [Plenefisch et al . , 1989] , and CAPG-1 [Csankovszki et al . , 2009b] ) with the canonical condensin I , distinguished only by the SMC-4 variant , DPY-27 ( Csankovszki et al . , 2009a , Csankovszki et al . , 2009b ) . Condensin DC interacts with at least five non-condensin proteins , SDC-1 , 2 , 3 , DPY-30 , and DPY-21 , which together form the genetically defined DCC ( Meyer , 2005 ) . With the exception of SDC-1 and DPY-21 , all DCC subunits are essential ( Plenefisch et al . , 1989; Villeneuve and Meyer , 1990 ) . DPY-30 , in addition to its role in dosage compensation , is a subunit of the highly conserved MLL/COMPASS complex , which methylates histone H3 at lysine 4 ( H3K4 ) ( Pferdehirt et al . , 2011; Li and Kelly , 2011; Shilatifard , 2008; Hsu and Meyer , 1994 ) . 10 . 7554/eLife . 23645 . 003Figure 1 . DCC recruitment sites are defined using high resolution ChIP-seq analysis . ( A ) The C . elegans dosage compensation complex ( DCC ) is composed of a modified condensin complex ( condensin DC ) which is distinguished from condensin I by the SMC-4 variant , DPY-27 . The non-condensin subunits SDC-2 , SDC-3 , and DPY-30 are required for DCC recruitment to the X chromosomes . ( B ) Peak distribution across each of the six chromosomes . ( C ) Representative ChIP-seq enrichment for an 80 kb window that includes recruitment element on the X , rex-8 . Recruitment sites on the X chromosome were defined as a 400 bp window centered on the SDC-2 ChIP-seq summits that overlap with SDC-3 , DPY-30 , and DPY-27 peaks and that do not show significant H3K4me3 enrichment . ( D ) Recruitment sites identified in this study ( n = 64 ) largely overlap with sequences previously shown to recruit DCC to multi-copy extrachromosomal array ( n = 47 ) ( Jans et al . , 2009; Pferdehirt et al . , 2011; McDonel et al . , 2006 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 00310 . 7554/eLife . 23645 . 004Figure 1—figure supplement 1 . ChIP-seq data suggests that some previously defined rex-sites fail to recruit the DCC in the context of the X chromosome . ( A ) SDC-2 , SDC-3 , DPY-30 , DPY-27 , and H3K4me3 ChIP-seq enrichment data is plotted for a 1 Mb window . Three previously identified rex-sites ( rex-8 ( rank 1 ) , rex-28 ( rank 35 ) , and rex-29 ( rank 50 ) ) are identified as recruitment sites in this study . One previously identified rex-site ( rex-27 ) shows little DCC enrichment and high H3K4me3 enrichment . ( B ) Plots showing SDC-3 and DPY-27 average ChIP enrichment across a 1 kb window centered on the peak summit . The top three recruitment sites ( rank 1–3 ) are shown in pink; recruitment sites ranked 31–33 are shown in green; recruitment sites ranked 62–64 are shown in blue . Rex-22 , rex-27 , and rex-30 , shown previously to recruit to array but not called recruitment sites in this study , are shown in gray . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 00410 . 7554/eLife . 23645 . 005Figure 1—figure supplement 2 . Peaks excluded by H3K4me3 overlap do not resemble strong recruitment sites . ( A ) Boxplots indicating SDC-2 ChIP-seq enrichment score across each of the defined classes: strong recruitment sites ( n = 17 ) , intermediate recruitment sites ( n = 16 ) , weak recruitment sites ( n = 31 ) , and peaks excluded by H3K4me3 enrichment ( n = 9 ) . B ) Representative ChIP-seq enrichment for a 100 kb window that includes a peak excluded by H3K4me3 enrichment . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 005 Fluorescence microscopy using DCC-specific antibodies indicated that the DCC binds to both hermaphrodite X chromosomes ( Chuang et al . , 1996; Dawes et al . , 1999 ) . Subsequent high-resolution ChIP-chip and ChIP-seq experiments revealed a pattern of DCC binding that supports the recruitment and spreading hypothesis ( Ercan et al . , 2009 , Ercan et al . , 2007; Jans et al . , 2009 ) . The recruitment sites show high levels of DCC binding , while sites of spreading show comparatively weaker DCC binding and frequently overlap with promoters and enhancers ( Ercan et al . , 2009 , Ercan et al . , 2007; Kranz et al . , 2013 ) . DCC spreading is independent of X chromosome sequence as the complex is able to spread into autosomal sequence fused to the end of the X ( Ercan et al . , 2009 ) . Interestingly , condensin DC spreads more effectively than the recruiter proteins SDC-2 and SDC-3 ( Ercan et al . , 2009 ) , highlighting the distinction between recruitment and spreading . Recruitment of the DCC to the X chromosomes is dependent on SDC-2 , SDC-3 , and DPY-30: SDC-3 binding requires both SDC-2 and DPY-30 ( Davis and Meyer , 1997 ) ; DPY-30 binding requires both SDC-2 and SDC-3 ( Pferdehirt et al . , 2011 ) . Only SDC-2 can localize to the X chromosomes in the absence of other complex members ( Dawes et al . , 1999 ) . Identification of DNA sequences capable of autonomously recruiting the DCC has previously been conducted using extrachromosomal arrays ( Csankovszki et al . , 2004; McDonel et al . , 2006 ) . Briefly , the DNA fragment to be tested is injected into the worm where it forms a multi-copy array . Fluorescence microscopy is then utilized to assess DCC recruitment to the extrachromosomal sequence . However , the array-based assay is limiting in both function ( in that it is not feasible to test every possible X chromosome sequence ) and in specificity ( in that assaying recruitment to artificial , multi-copy arrays can result in false positives ) . Here , to overcome these limitations , we used a genomics approach to identify and categorize DCC recruitment sites in the context of the native X chromosome . To better understand how a limited number of recruitment sites function to specifically target the DCC to the X , we investigated the role of DNA sequence and chromatin accessibility on recruitment site activity , and performed targeted genome editing to test recruitment site function . Here , we found that DCC recruitment to the X chromosomes is initiated by a small number of strong recruitment sites that contain clusters of a 12 bp DNA sequence motif . Strong recruitment sites display significant overlap with annotated high occupancy transcription factor target ( HOT ) sites and are marked by DNA-encoded nucleosome occupancy , two features that distinguish motif clusters on the X from those on the autosomes . Strong recruitment sites are distributed along the length of the X chromosome , interspersed by weaker sites whose recruitment function is X-chromosome dependent . We addressed the X-dependent activity of recruitment sites by ectopically inserting recruitment sequence into both the autosomes and the X chromosome . While the ectopic insertion of a recruitment site in single copy failed to fully recruit the DCC to an autosome , both increased copy number and the insertion of additional recruitment sequences at a distance significantly increased DCC recruitment to an ectopic site , suggesting that cooperation between multiple recruitment sites is necessary for robust DCC binding on the X . Furthermore , deletion of a recruitment site resulted in moderate but significant reduction of DCC binding centered on the deleted site and surrounded by strong TAD boundaries . This result suggests that recruitment by multiple sites and subsequent spreading of the DCC in cis within distinct chromosomal domains establishes the full DCC binding pattern . Overall , our work demonstrates that cooperation between a hierarchical set of recruitment sites specifically targets the C . elegans dosage compensation complex to the X chromosomes .
Previously , extrachromosomal array based assays were used to demonstrate the ability of a DNA fragment to autonomously recruit the DCC , independent of the X chromosome . This technique was used to identify 47 recruitment elements on the X , named rex-1 through rex-47 ( Jans et al . , 2009; Pferdehirt et al . , 2011; McDonel et al . , 2006 ) . However , extrachromosomal arrays carry multiple copies of the sequence being tested leaving these assays susceptible to false positives . Many sequences shown to recruit the DCC to a multi-copy extrachromosomal array show little to no DCC binding in ChIP-seq experiments ( Figure 1—figure supplement 1A ) . Here , to identify sequences that recruit the DCC in the context of the native X chromosome , we used C . elegans mixed developmental stage embryos ( ~100 cell stage and higher obtained by bleaching adults and using similar growth conditions between strains ) to perform ChIP-seq analysis of SDC-2 , SDC-3 , and DPY-30 , the proteins required for DCC recruitment ( Pferdehirt et al . , 2011; Dawes et al . , 1999; Davis and Meyer , 1997 ) . Consistent with previous work ( Ercan et al . , 2009 ) , we found that SDC-2 binding is enriched on the X chromosome: 336 out of 659 ( 51% ) binding sites are on the X . Similarly , 54% ( 389/719 ) of SDC-3 and 29% ( 254/868 ) of DPY-30 binding peaks are on the X chromosome . The condensin DC portion of the complex , as assayed by DPY-27 ChIP-seq ( Kranz et al . , 2013 ) , is largely restricted to the X chromosome: over 99% ( 2546/2564 ) of DPY-27 binding peaks are on the X . Peak distribution ( Figure 1B ) and a representative window of ChIP-seq enrichment for SDC-2 , SDC-3 , DPY-30 , DPY-27 , and H3K4me3 ( a marker of COMPASS activity [Shilatifard , 2008 , Shilatifard , 2012; Xiao et al . , 2011] ) ( Figure 1C ) are shown . As a side note , among the 336 autosomal SDC-2 binding peaks , 206 ( 61 . 3% ) overlap with both SDC-3 and DPY-30 but fail to recruit condensin DC , suggesting a DCC-independent role for the recruiter proteins at these sites . Previous work has demonstrated that high levels of DCC ChIP enrichment are predictive of recruitment activity ( Jans et al . , 2009 ) . Thus , in order to define a list of DCC recruitment sites , we began by identifying the 100 strongest SDC-2 ChIP-seq binding peaks in the genome . Of these , 92 overlap with both SDC-3 and DPY-30 binding peaks . DPY-30 , as a member of MLL/COMPASS , is associated with the histone modification H3K4me3 ( Shilatifard , 2008 , Shilatifard , 2012; Xiao et al . , 2011 ) . To remove sites of COMPASS activity , we eliminated peaks that overlap with H3K4me3 enrichment . Lastly , we looked for the presence of condensin DC binding by requiring overlap with a DPY-27 binding peak on the X chromosome . Our criteria for defining recruitment sites was strict; 9 DCC binding loci on the X chromosome were eliminated due to high enrichment of H3K4me3 ( Figure 1—figure supplement 2 ) . Our finalized list of 64 recruitments sites contains 32 out of 47 ( 68% ) rex-sites that were shown previously to recruit the DCC to extrachromosomal arrays ( Figure 1D , Supplementary file 1D ) . The remaining 15 previously identified rex sites either lack binding of one or more DCC subunits ( 14/15 ) and/or overlap with H3K4me3 enrichment ( 6/15 ) , indicating COMPASS activity . For clarity , when referring to loci identified by our ChIP-seq enrichment method we will use the term ‘recruitment site . ’ When referring to a specific sequence that has been previously shown to recruit the DCC to array we will use the term ‘rex-site . ’ ChIP-seq enrichment at recruitment sites , both endogenous and ectopic , will be referred to as ‘DCC recruitment . ’ ChIP-seq enrichment at all other genomic loci will be referred to as ‘DCC binding . ’ Closer examination revealed that ChIP enrichment scores at the recruitment sites vary greatly; some sites appear to recruit the DCC much more robustly than others ( Figure 1—figure supplement 1B ) . Recruitment sites were ranked by their SDC-2 , SDC-3 , DPY-30 , and DPY-27 ChIP enrichment scores and separated into three strength classes using k-means clustering ( Figure 2A ) . We identified 17 strong , 16 intermediate , and 31 weak recruitment sites . Strong recruitment sites are spaced roughly 1 Mb apart along the length of the X chromosome and are interspersed by both weak and intermediate strength sites ( Figure 2B ) . This observed pattern of recruitment site localization and strength suggested that X-specific recruitment of the DCC is hierarchical , initialized by DCC binding at a subset of strong recruitment sites . 10 . 7554/eLife . 23645 . 006Figure 2 . Recruitment sites on the X chromosome are hierarchical ( A ) Heatmaps showing ChIP-seq enrichment for the DCC recruiter proteins , SDC-2 , SDC-3 , and DPY-30 , the condensin DC subunit DPY-27 , and the histone modification H3K4me3 . Heatmaps are plotted across a 3 kb window centered on the 64 recruitment sites . Based on the DCC ChIP enrichment scores , the recruitment sites are categorized into three strength classes ( strong , intermediate , weak ) using k-means clustering ( n = 3 ) . ( B ) The strong DCC recruitment sites ( pink , n = 17 ) are scattered across the X chromosome . Intermediate ( n = 16 ) and weak ( n = 31 ) recruitment sites ( gray ) are distributed in between . The median distance between recruitment sites is ~90 kb . SDC-2 enrichment across the length of the X chromosome was plotted using IGV . ( C ) In an sdc-3 null mutant , SDC-2 binds the strongest recruitment sites , indicating hierarchy of recruitment . As in ( A ) , heatmaps indicating SDC-2 ChIP-seq enrichment are plotted across a 3 kb window centered on the 64 recruitment sites . Data from wild-type , sdc-3 null , and sdc-2 null are all plotted on the same scale . Density plots above each heatmap indicate the average ChIP enrichment score for each recruitment class plotted across the 3 kb window . Strong sites are plotted in pink , intermediate in blue , and weak in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 00610 . 7554/eLife . 23645 . 007Figure 2—figure supplement 1 . Specific reduction in SDC-2 enrichment at weak recruitment sites in an sdc-3 null mutant . ( A ) Scatterplot comparing SDC-2 ChIP-seq enrichment between wild-type ( N2 ) and the sdc-3 null mutant . Briefly , the X chromosome was divided into 200 bp windows with a step size of 100 bp . Average SDC-2 enrichment across the window was calculated and plotted . Windows that overlap with a strong recruitment site are plotted in red , intermediate site in teal , and weak site in green . For each recruitment strength class , the best fit linear model was determined and plotted . Slopes of the best-fit lines are indicated ( strong = 28 . 190 , intermediate = 0 . 420 , weak = 0 . 16 ) . Arrowhead indicates windows overlapping with a region of the X chromosome deleted in the sdc-3 null background . ( B ) Scatterplot comparing SDC-2 enrichment in wild-type ( N2 ) versus the log2 change in enrichment in the sdc-3 null for all 64 recruitment sites . Each point represents an average across a 40 bp window centered at the recruitment site summit . Dividing the entire X chromosome into 40 bp window revealed an average log2 ratio of −1 . 280 ( horizontal line ) . Strong recruitment sites are shown in red , intermediate site in teal , and weak site in green . Arrowhead indicates window overlapping with region of the X chromosome deleted in the sdc-3 null background . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 007 Among the DCC subunits , only SDC-2 can localize to the X chromosome in the absence of other complex members and only SDC-2 is required for proper localization of all other dosage compensation proteins ( Dawes et al . , 1999 ) . Under a model of recruitment site strength hierarchy , we predicted that the strongest recruitment sites might be acting as initial DCC entry sites . If correct , then SDC-2 ChIP-seq performed in an sdc-3 null mutant strain ( in which only SDC-2 can localize to the X ) should reveal binding only at the strongest recruitment sites . Indeed , in sdc-3 ( y126 ) null mutant embryos , SDC-2 binds to 16 out of 17 strong , 11 out of 16 intermediate , and only 2 out of 31 weak recruitment sites ( Figure 2C ) . The single strong site that fails to recruit SDC-2 is in a region that is deleted in the sdc-3 null mutant . To test the possibility of an equivalent reduction of SDC-2 binding across all sites on the X chromosome in the sdc-3 mutant , we plotted SDC-2 enrichment across the X chromosome in wild-type animals versus the sdc-3 mutant ( Figure 2—figure supplement 1A ) . We observe a striking difference between changes in SDC-2 binding at the strong versus the weaker recruitment sites . Across the length of the X chromosome , the background SDC-2 enrichment is reduced ( median log2 ratio of −1 . 280 , Figure 2—figure supplement 1B ) . Weak recruitment sites exhibit a markedly lower log2 ratio ( median −3 . 042 ) . In contrast , strong recruitment sites have a median log2 ratio of −1 . 057 . Retainment of SDC-2 enrichment specifically at the strong recruitment sites in animals lacking sdc-3 is consistent with hierarchical binding of SDC-2 at recruitment sites . Additionally , the pattern of SDC-2 ChIP-seq enrichment in the sdc-3 null suggests that SDC-2 binding is restricted to the initial recognition sites . Compared to wild-type , we see only a narrow window of SDC-2 binding in the sdc-3 null strain , indicating that SDC-2 fails to spread in the absence of other DCC members ( Figure 2C ) . As a control , ChIP-seq analysis in sdc-2 ( y74 ) null mutant embryos revealed a complete lack of SDC-2 binding at all 64 recruitment sites ( Figure 2C ) . In sum , these observations suggest that our categorization of recruitment sites into stronger and weaker classes using the pattern of DCC ChIP enrichment reflects the hierarchy of recruitment . Strong recruitment sites are the initial entry points for DCC binding , which allows for DCC binding at weaker recruitment sites . SDC-2 is a novel 344 . 4kD protein unique to Caenorhabditis ( Nusbaum and Meyer , 1989 ) and the only hermaphrodite-specific member of the DCC . All other complex subunits are maternally loaded into the embryo ( Dawes et al . , 1999; Stoeckius et al . , 2014 ) and remain diffuse in the nucleus prior to the onset of sdc-2 expression at around the 40 cell stage ( Dawes et al . , 1999 ) . While we cannot exclude the possibility that , prior to sdc-2 expression , some condensin DC is loosely associated with all chromosomes , previous work indicates that SDC-2 is required for X-specific enrichment of condensin DC binding ( Dawes et al . , 1999 ) . Despite its importance in recruiting the DCC to the X chromosomes , it remains unclear how SDC-2 is able to specifically recognize and bind to the X . Toward understanding the mechanism underlying X-recognition by SDC-2 , we examined histone occupancy at recruitment sites , reasoning that initial recruitment to stronger sites might be the result of those sites being inherently more open prior to the onset of sdc-2 expression . We performed histone H3 ChIP-seq in wild-type and sdc-2 null mutant embryos , using the sdc-2 null strain as a proxy to measure nucleosome occupancy before the onset of sdc-2 expression . While wild-type H3 data indicated a tendency for open chromatin at all recruitment sites ( Figure 3A , left ) , we observed a marked increase in H3 enrichment , specifically at the strong recruitment sites , in the sdc-2 null mutant ( Figure 3A , right ) . Interestingly , a nucleosome occupancy model trained on in vitro sequence preferences ( Ercan et al . , 2011 ) predicts high nucleosome occupancy at DCC recruitment sites ( Figure 3B ) . These results suggest that SDC-2 is required for open chromatin at the recruitment sites which themselves encode for high nucleosome occupancy . 10 . 7554/eLife . 23645 . 008Figure 3 . SDC-2 is required for open chromatin at strong DCC recruitment sites that encode for high intrinsic nucleosome occupancy . ( A ) Heatmaps showing H3 ChIP-seq enrichment across a 1 kb window centered on the 64 recruitment sites . Wild-type data ( left ) indicates open chromatin at all recruitment sites . Data from an sdc-2 null strain ( right ) indicates increase in H3 enrichment , especially at the strong recruitment sites . Density plots above each heatmap indicate the average ChIP enrichment score for each recruitment class . ( B ) Heatmap indicates the probability of nucleosome occupancy across a 1 kb window centered on the recruitment sites . Probability ranges from 0 . 0 ( yellow ) to 1 . 0 ( green ) . DNA-encoded nucleosome occupancy signal is highest at the center of the recruitment sites . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 008 Previous work using either array-characterized rex-sites ( Jans et al . , 2009; McDonel et al . , 2006 ) or DCC ChIP-chip data ( Ercan et al . , 2009 , Ercan et al . , 2007 ) independently identified a short DNA sequence motif enriched at DCC recruitment sites . Extrachromosomal arrays bearing wild-type motif sequence were shown to recruit the DCC while arrays with motif mutations failed to recruit ( McDonel et al . , 2006 ) , indicating that motif sequence is important for DCC binding . However , the motif is only slightly enriched ( ~3 fold ) on the X ( Ercan et al . , 2007; Jans et al . , 2009 ) , and copies of the motif on autosomes do not recruit the DCC . To investigate the role of the motif in X-specific recruitment of the DCC , we first redefined the 12 bp motif using higher resolution ChIP-seq data ( Figure 4A ) . Similar to the previously identified sequences , the 12 bp motif contains an 8 bp GCGCAGGG core with variable specificities for the surrounding nucleotides . We scanned the whole genome for motif occurrences using the Transcription Factor Affinity Prediction ( TRAP ) tool ( Thomas-Chollier et al . , 2011 ) , generating a genome-wide list of motif match locations and associated affinity scores ( Supplementary file 1E ) . 10 . 7554/eLife . 23645 . 009Figure 4 . The 12 bp recruitment motif is enriched on the X . Bound motifs are characterized by DNA-encoded nucleosome occupancy . ( A ) The 12 bp recruitment motif identified using high-resolution SDC-2 ChIP-seq data . ( B ) Motif distribution is plotted for chromosomes I through V ( shades of gray ) and chromosome X ( purple ) . Motif enrichment is dependent on score-cutoff . Weaker motifs ( score <7 ) are randomly distributed across the chromosomes . Strong motifs ( score ≥7 ) are enriched on the X chromosome . Perfect score motifs ( score 10 , sequence ATCGCGCAGGGA ) show almost complete X-specificity: 10/11 ( 91% ) are on the X . ( C ) Heatmaps showing H3 ChIP-seq enrichment from wild-type ( left ) and sdc-2 null ( right ) embryos , plotted across a 1 kb window centered on all strong motifs of score ≥7 . Motifs are sorted by H3 enrichment in wild-type , and divided into three categories: Bound by DCC on the X ( n = 50 ) , unbound on the X ( n = 45 ) , and unbound on the autosomes ( n = 184 ) . Here , we define ‘bound’ as the overlap of both SDC-3 and DPY-27 binding . There are no bound motifs on the autosomes . ( D ) Heatmap indicates the probability of nucleosome occupancy across a 1 kb window centered on the motifs . As in ( C ) , motifs are sorted by H3 enrichment in wild-type . ( E ) Boxplot indicates intrinsic nucleosome occupancy for a 150 bp window centered on each motif . Bound motifs ( dark purple , median 0 . 616 ) have significantly higher DNA-encoded nucleosome occupancy compared to unbound motifs on the X ( light purple , median 0 . 4865 ) and motifs on the autosomes ( grey , median 0 . 503 ) . Significance calculated using one-tailed students t-test ( ** p-value<0 . 01 , *** p-value<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 009 We found that X-enrichment of the 12 bp recruitment motif is sensitive to affinity score cut-off . Motifs with a score of less than seven are randomly distributed throughout the genome while those with a score of seven or greater are approximately three fold enriched on the X . The X chromosome contains 43% ( 47/110 ) of motifs with a score of 8 or better , 64% ( 20/31 ) with a score of 9 or better , and 91% ( 10/11 ) of perfect match motifs ( ATCGCGCAGGGA ) ( Figure 4B ) . Based on this enrichment pattern , we define strong motifs to be those with a score of 7 or better . Strong motifs are enriched on , but not specific to , the X chromosome . There are 184 strong motifs distributed across the five autosomes , none of which are bound by the DCC . And while all motifs with a perfect score of 10 recruit the DCC in the context of the X chromosome ( i . e . all of them overlap with a recruitment site ) , the single perfect autosomal motif fails to recruit the DCC ( Figure 5—figure supplement 1 ) . Additionally , we find that , of the 95 strong motifs on the X chromosome , only 50 ( 52% ) are bound by the DCC . This result parallels the general observation that many TFs bind only a subset of their potential binding motifs ( D'haeseleer , 2006 ) . Because chromatin accessibility often functions to regulate TF binding , we reasoned that unbound motifs on both the X and the autosomes might be precluded from DCC binding due to the presence of nucleosomes . To test this , we plotted histone H3 ChIP-seq enrichment across 1 kb windows centered at all of the strong motifs in the genome ( n = 279 ) ( Figure 4C ) . We found that bound motifs are , indeed , depleted of histones . However , several motifs on both the X chromosome and the autosomes exhibit similar histone depletion , yet are not bound by the DCC . Therefore , motif accessibility is insufficient to predict DCC binding . Because we observe an increase in H3 enrichment at strong recruitment sites in the absence of SDC-2 ( Figure 3A ) , we reasoned that SDC-2 might be similarly required for open chromatin at strong motifs . In the absence of SDC-2 , H3 occupancy increases specifically at motif sequences that are bound by the DCC in wild-type animals ( Figure 4C ) . Increased H3 occupancy reflects the nucleosome occupancy predicted by the in vitro model ( Figure 4D ) . Interestingly , DCC bound motifs have , on average , higher predicted nucleosome occupancy than unbound motifs on either the X or the autosomes ( Figure 4E ) . Combined with our previous observation that recruitment sites are characterized by high intrinsic nucleosome occupancy ( Figure 3A ) , our data suggests a role for nucleosome positioning in DCC binding and recruitment . While enriched on the X chromosome , the 12 bp motif is also found on all five of the autosomes . However , unlike on the autosomes where motifs are more randomly dispersed , motifs on the X tend to cluster together in linear space ( Ercan et al . , 2007 ) . Here , we refined the analysis of motif clustering by defining a motif cluster to be a strong motif ( score ≥7 ) flanked within 200 bp by at least one other motif with a score of 5 or better . Using this definition , we identified 17 homotypic motif clusters on the X chromosome , all of which are bound by the DCC and all of which are contained within a recruitment site ( Figure 5A ) . Significantly , of the 10 perfect motif matches on the X chromosome , nine are contained within a motif cluster and are centered at a strong recruitment site . Two of the strongest recruitment sites , rex-14 and rex-32 , each contain two perfect motifs ( score = 10 ) . Overall , motif clustering is a characteristic of strong recruitment sites . Among the 17 strong sites , six contain a cluster of two motifs and five contain a cluster of three motifs . These observations suggest that motif clustering on the X is important for DCC recruitment . However , clustering alone cannot fully distinguish the X from the autosomes . Using our definition , we find five motif clusters on the autosomes , none of which recruit the DCC ( Figure 5—figure supplement 1 ) . All five autosomal clusters have only one strong motif and one weaker motif . In addition , each autosome contains at most , two motif clusters ( Figure 5A ) . 10 . 7554/eLife . 23645 . 010Figure 5 . Clustered motifs at HOT sites distinguish the X chromosome from the autosomes . ( A ) A motif cluster is defined as a strong motif ( score ≥7 ) flanked within 200 bp by a second motif . Motif locations plotted for each of the 22 motif clusters: 17 are X- linked ( 11 at strong recruitment sites , six at weaker recruitment sites ) and five are on the autosomes . Plots are centered on the strongest motif in each cluster . Weak motifs ( score 5–7 ) are shown in gray; strong motifs ( score 7–9 ) are shown in black; perfect matches ( score = 10 ) are shown in pink . For simplicity , motif scores ( Supplementary file 1E ) are rounded down to the nearest whole integer . ( B ) Plot demonstrating overlap between recruitment sites and motif clusters , motif strength , HOT sites , and tRNAs . Black box indicates presence and white box indicates absence of a given overlap . For motifs , pink box indicates presence of a perfect score motif ( score = 10 ) ; black box indicates presence of a strong motif ( score 7–9 ) ; grey box indicates presence of a weak motif ( score 5–7 ) . Significance of overlap is calculated for the entire set of 64 recruitment sites using permutation test ( overlapPermTest function of the regioneR package ( Fruciano et al . , 2016 ) , 100 permutations , ** p-value<0 . 01 ) . The reported ‘n . overlap’ refers to the number of indicated genomic annotation that overlaps with at least one recruitment site . Overlap with HOT site is a characteristic of strong recruitment sites . Overlap with tRNAs is a characteristic of weaker recruitment sites . ( C ) Heatmaps showing H3 ChIP-seq enrichment from wild-type ( left ) and sdc-2 null ( right ) embryos , plotted across a 3 kb window centered on annotated HOT sites . HOT sites on the X are divided into two categories: those that overlap with a recruitment site and those that do not . HOT sites from chromosome V are representative of all autosomal HOT sites . SDC-2 is required for opening chromatin at the HOT DCC recruitment sites on the X . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01010 . 7554/eLife . 23645 . 011Figure 5—figure supplement 1 . ChIP-seq data reveals that neither perfect affinity motifs nor motif clustering by itself is sufficient to explain X-specific recruitment of the DCC . SDC-2 , SDC-3 , DPY-30 , DPY-27 , and H3K4me3 ChIP-seq enrichment is plotted across a 3 kb window centered on a representative recruitment site from each strength class , and for two autosomal loci ( the perfect autosomal motif and the autosomal motif cluster used in Figure 6E ) that do not recruit the DCC . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01110 . 7554/eLife . 23645 . 012Figure 5—figure supplement 2 . Extended analysis of overlap between recruitment sites with each tested annotation . Black box indicates presence and white box indicates absence of a given overlap . Significance of overlap is calculated using permutation test ( overlapPermTest function of the regioneR package ( Fruciano et al . , 2016 ) , 100 permutations , ** p-value<0 . 01 ) ) . Red box indicates enrichment of overlap; blue box indicates depletion . The reported ‘n . overlap’ refers to the number of indicated genomic annotation that overlaps with at least one recruitment site . References for genomic annotations can be found in Supplementary file 1F . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 012 Alone , neither motif sequence nor motif clustering nor nucleosome occupancy is able to explain X-specificity of DCC recruitment . Subsequently , in order to identify additional factors that distinguish DCC recruitment sites on the X , we looked for significant overlap with various genomic annotations ( Figure 5B ) . References for annotations can be found in Supplementary file 1F , and a more exhaustive overlap plot can be found in Figure 5—figure supplement 2 . Previous work has demonstrated significant enrichment of condensin I binding at tRNA genes in yeast , chicken and C . elegans ( Kranz et al . , 2013; D'Ambrosio et al . , 2008; Haeusler et al . , 2008; Kim et al . , 2013 ) . Of the 638 annotated C . elegans tRNA genes , 286 ( 44 . 8% ) are on the X chromosome . Of these , 178 ( 62 . 2% ) overlap with a condensin DC binding site on the X chromosome , as assayed by DPY-27 ChIP-seq . Interestingly , 16 out of 64 DCC recruitment sites overlap with a tRNA gene ( Figure 5B ) . This overlap occurs largely at weak recruitment sites , suggesting that canonical condensin recruitment mechanisms at tRNAs may help to reinforce condensin DC binding on the X chromosome downstream of DCC recruitment at strong sites . Interestingly , the set of strong recruitment sites is distinguished by an overlap with HOT sites , defined as genomic regions that are bound by a high number of TFs ( 38 or more of the 57 TFs assayed in [Van Nostrand and Kim , 2013] ) without their corresponding motifs ( modENCODE Consortium et al . , 2010 ) ( Figure 5B ) . Among the 17 strong recruitment sites , 16 fall within the boundaries of an annotated HOT site ( Van Nostrand and Kim , 2013 ) . In C . elegans , HOT sites tend to be CpG-rich and typically show characteristics of open chromatin , including nucleosome depletion ( Chen et al . , 2014 ) . Our H3 data confirms a lack of histone binding at all annotated HOT sites ( Figure 5C ) . In the sdc-2 null mutant , H3 becomes enriched only at those HOT sites that overlap with a recruitment site , indicating that SDC-2 binding is necessary to keep these HOT sites open . Strikingly , while none of the five autosomal motif clusters are HOT , we find that 11 out of 12 motif clusters on the X chromosome both overlap with a HOT site and are characterized as strong recruitment sites . Importantly , this overlap between homotypic motif clusters and HOT sites on the X chromosome fully distinguishes the X from the autosomes . Our analysis has thus far demonstrated that the 17 strong recruitment sites are distinguished from other motif containing sites in the genome by motif clustering , chromatin structure , and overlap with annotated HOT sites . In addition , these strong recruitment sites are bound by SDC-2 in the absence of other DCC subunits , suggesting a hierarchy of DCC recruitment among the recruitment sites . The presence of weaker recruitment sites that are not as easily distinguishable from autosomal sequence suggests that these sites might be active only in the context of the X chromosome . X chromosome dependence of a subset of recruitment sites is consistent with previous work that demonstrated the inability of some large X chromosome duplications to recruit the DCC when detached from the rest of the chromosome ( Blauwkamp and Csankovszki , 2009 ) . To test X-dependence of recruitment site activity , we ectopically inserted ~400 bp of recruitment site sequence into autosomal loci that do not bind the DCC in wild-type animals . Rex-1 , here categorized as an intermediate strength site , was the first recruitment sequence identified by multi-copy extrachromosomal array analysis ( McDonel et al . , 2006 ) . We reasoned that if rex-1 is sufficient to autonomously recruit the DCC , then inserting 400 bp of rex-1 sequence into the chromosome II MosSci site should result in ectopic DCC binding . We used ChIP-qPCR to assay for the presence of both SDC-3 ( recruiter protein ) and DPY-27 ( condensin DC ) at the autosomal locus . ChIP-qPCR is a powerful tool for quantitatively determining the amount of antibody-immunoprecipitated DNA . However , the assay is sensitive to the inherent variability of ChIP between biological replicates , necessitating proper controls for normalization . To account for background noise , we used the ∆∆Ct method to calculate fold enrichment over input compared to a negative control locus . To account for between-replicate differences in ChIP efficiency , we internally normalized fold enrichment to a positive control ( i . e . the endogenous locus ) . Experiments were done in biological triplicate and final ChIP-qPCR results were plotted as percent of endogenous recruitment , allowing comparison across experimental set ups . ChIP-qPCR results indicate that rex-1 sequence recruits SDC-3 and DPY-27 at a much lower level when inserted in single copy into chromosome II ( Figure 6A ) . Respectively , SDC-3 and DPY-27 qChIP enrichment reaches only 25 . 79% and 12 . 40% of endogenous site levels . In contrast , sequence from the strongest recruitment site , rex-8 , inserted in single copy at the same locus binds SDC-3 and DPY-27 at 46 . 56% and 21 . 89% of endogenous rex-8 levels , respectively . We conclude that rex-8 , which contains a cluster of three motifs and overlaps with a HOT site in the native X chromosome , is a strong DCC recruiter and is sufficient to ectopically recruit the complex to an autosome , albeit at reduced levels compared to the endogenous site . 10 . 7554/eLife . 23645 . 013Figure 6 . Cooperation between recruitment sites is necessary for robust DCC recruitment . ( A ) SDC-3 ( left , blue ) and DPY-27 ( right , brown ) ChIP-qPCR data is plotted for the indicated ectopic insertions . ChIP-qPCR is plotted as percent of endogenous recruitment , that is , recruitment at the ectopic site is compared to recruitment at the corresponding endogenous sequence . MosSci was used to insert rex-1 and rex-8 sequence into the same locus on chromosome II . Bombardment was used to insert rex-1 sequence in three-copy on chromosome III and eight-copy on chromosome I . CRISPR was used to insert rex-1 sequence in single copy on chromosome X . Information on strain generation is available in Supplementary file 1A . Significance is tested using one-tailed t-test assuming unequal variance ( *p-value<0 . 05; **p-value<0 . 01 ) . ( B ) DPY-27 ChIP-seq data comparing wild-type average and three biological replicates of the ChIP-seq data in the strain bearing an ectopic insertion of rex-1 in single copy in chromosome X . The insertion site is indicated as the dashed line . Inserted rex-1 sequence is not shown , as ChIP-seq enrichment at the ectopic sequence is confounded by binding at the endogenous site . Location of the novel spreading peak in the insertion strain is indicated by the red arrowhead . ( C ) As in ( A ) , SDC-3 and DPY-27 ChIP-qPCR data is plotted as percent of endogenous recruitment . Dashed circle indicates ectopic site being assayed in ChIP-qPCR . CRISPR was used to insert rex-8 sequence 30 kb downstream and 50 kb upstream of the ectopic rex-1 sequence on chromosome II . ( D ) As in ( C ) , dashed circle indicates the ectopic site being assayed . ( E ) CRISPR was used to insert a motif cluster from chromosome V into the ectopic chromosome X site . Because the chromosome V site does not normally recruit the DCC , SDC-3 and DPY-27 ChIP-qPCR data is plotted as percent of endogenous rex-1 recruitment . ( F ) Boxplot indicates intrinsic nucleosome occupancy for a 150 bp window centered on either motif or recruitment site as indicated . Recruitment sites with no motifs ( light green , median 0 . 684 ) have significantly higher DNA-encoded nucleosome occupancy compared to recruitment sites containing motifs ( dark green , median 0 . 583 ) . Motif clusters on the X ( dark blue , median 0 . 604 ) have higher DNA-encoded nucleosome occupancy compared to motif clusters on the autosomes ( light blue , median 0 . 548 ) . Rex-1 has a nucleosome occupancy score of 0 . 732; the chromosome V motif cluster has a nucleosome occupancy score of 0 . 450 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 013 Interestingly , when we insert the same 400 bp rex-1 sequence in single-copy into the X chromosome ( at a site that does not normally recruit the DCC ) the sequence is able to robustly recruit the DCC , even exceeding endogenous levels ( Figure 6A ) . ChIP-seq data also revealed a novel DCC spreading peak just upstream of the rex-1 insertion site on the X ( Figure 6B ) . Although many DCC spreading sites are located in X-linked promoters , this new spreading site is in an un-annotated region . We do not yet understand what may be driving the DCC to spread to this sequence in the ectopic strain where it fails to do so on the wild-type chromosome . Single-copy rex-1 sequence fails to fully recruit the DCC to an autosome , but is sufficient to ( both ectopically and endogenously ) recruit the DCC to the X chromosome , suggesting X-dependency of rex-1 activity . DCC recruitment at weaker sites on the X may be dependent on cooperation between multiple recruitment sites . Alternatively , activation of weaker recruitment sites may strictly require the presence of a strong recruitment site . We distinguished between these possibilities by inserting multiple-copies of rex-1 sequence into two different autosomal loci . If rex-1 function is dependent on activation by strong recruitment sites , then increasing copy number should have no effect on ectopic recruitment . Alternatively , if multiple recruitment sites function to cooperatively recruit the DCC , then ectopic recruitment should improve upon increased copy-number . Indeed , inserting rex-1 sequence in three and eight copies respectively increased DPY-27 enrichment to 61 . 49% and 99 . 89% of endogenous rex-1 binding ( Figure 6A ) . SDC-3 enrichment similarly increased to 80 . 98% and 50 . 08% of endogenous levels ( Figure 6A ) . These results indicate that rex-1 sequence in multiple-copy is sufficient to cooperatively recruit the DCC in the absence of strong recruitment sites . This suggests that the presence of strong recruitment sites in the context of the X chromosome enhances and reinforces cooperativity . Given that the median distance between recruitment sites on the X is ~90 kb ( ranging from 1543 bp to 1 . 2 Mb ) , cooperation between sites is likely long-range . To test the ability of recruitment sites to cooperate over long-distances , we used CRISPR to sequentially insert 400 bp of rex-8 sequence ~30 kb downstream and ~50 kb upstream of the single ectopic rex-1 that failed to recruit by itself on chromosome II ( Figure 6C ) . Notably , we observe similar levels of ectopic recruitment to rex-8 sequence at three independent insertion loci , suggesting little position effect on recruitment ability ( Figure 6D ) . Upon addition of rex-8 at a single locus ~30 kb downstream , recruitment of DPY-27 to the ectopic rex-1 increased ~1 . 5 fold ( Figure 6C ) . Insertion of rex-8 at a second site ~50 kb upstream resulted in further increase in both SDC-3 and DPY-27 binding at the ectopic rex-1 locus ( Figure 6C ) . It is remarkable that , without prior knowledge of the mechanism of long-range cooperativity , insertion of rex-8 sequence at two distant loci resulted in increased ectopic recruitment to the single copy rex-1 sequence . These results suggest that strong and weak recruitment sites , dispersed across the length of the X chromosome , cooperatively define a recruiting environment that restricts DCC binding to the X chromosome . Recent HiC data indicated that interactions between recruitment sites are among the most prominent long-distance contacts on the X chromosome ( Crane et al . , 2015 ) . These interactions are largely DCC dependent as many are lost in an sdc-2 mutant . We used the same data to determine recruitment site interactions , limiting our analysis to interaction scores above a 99% quantile threshold . Accordingly , we found that 62 out of 64 recruitment sites interact with at least one other recruitment site in the wild-type worm ( Figure 5—figure supplement 2 ) . This level of site-to-site contact is highly significant ( p<0 . 001 , permutation test ) . In the sdc-2 mutant , the number of interactions is reduced: 35 out of 64 recruitment sites interact with at least one other site . In Drosophila , 3D interaction between high affinity sites facilitates spreading of the dosage compensation complex across the X chromosome ( Ramírez et al . , 2015 ) . One possible explanation for DCC recruitment to weaker sites , especially to those lacking a motif , is that the complex utilizes pre-established spatial proximity to spread from strong sites . Increased DCC binding might reinforce recruitment site interactions , helping to establish the observed X chromosome conformation ( Crane et al . , 2015 ) . Because most strong recruitment sites are composed of motif clusters , and because recruitment sites bind the DCC more effectively in the context of the X chromosome , we next tested sufficiency of a motif cluster to recruit the complex when inserted into the X chromosome . To this end , we inserted one of the five identified autosomal motif clusters ( containing two motifs ) into the X chromosome . We selected this autosomal motif cluster because it contained the strongest scoring motifs ( scores 9 . 5 and 6 . 25 ) separated by the shortest genomic distance ( 68 bp ) , and was most similar to motif clusters found at strong recruitment sites on the X . To eliminate any potential position effects , we inserted the motif cluster into the same X chromosomal locus where ectopic rex-1 robustly recruited the DCC ( Figure 6A ) . The autosomal motif cluster failed to strongly recruit the DCC to the X chromosome ( Figure 6E ) . This result suggests that motif clustering alone is not sufficient for DCC recruitment , even in the context of the X chromosomes . Because we found that strong recruitment sites containing motif clusters overlap with HOT sites and encode for high intrinsic nucleosome occupancy , we reasoned that flanking DNA sequence and/or chromatin structure is important for DCC recruitment . Compared to motif clusters on the X chromosome ( median 0 . 604 ) , autosomal motif clusters have lower intrinsic nucleosome occupancy ( median 0 . 548 ) ( Figure 6F ) . Interestingly , recruitment sites that do not contain motifs have significantly higher intrinsic nucleosome occupancy than recruitment sites that do contain motifs ( Figure 6F ) , suggesting that a specific DNA or chromatin structure is important for recruitment activity . We reasoned that , if recruitment sites act cooperatively to achieve robust and X-specific DCC localization , then deleting a single site on the X chromosome should reveal long-range effects on DCC binding . Using CRISPR , we deleted three recruitment sites on the X chromosome: two strong sites , rex-40 and rex-41 are respectively the left and the rightmost strong sites on the X , while the intermediate strength rex-1 lies ~4 Mb into the X chromosome ( Figure 7—figure supplement 1A ) . Deletions were confirmed by Sanger sequencing ( Figure 7—figure supplement 2 ) . Compared to wild-type , DPY-27 ChIP-seq data in the rex-41 deletion strain indicates that the DCC binding profile is largely unchanged along the length of the X chromosome ( Figure 7—figure supplement 1B , C ) . This suggests that DCC recruitment is a robust process; cooperation between the remaining recruitment sites is capable of countering the effect of a single recruitment site deletion . Although gross changes are not apparent , we observed a subtle reduction in DCC binding upon rex-41 deletion . We used a sliding window analysis , an assay more sensitive to slight alterations in DCC enrichment , to determine both the level and significance of change in DCC binding upon recruitment site deletion . Briefly , a 2 Mb window was stepped across each of the ~2500 DPY-27 binding sites on the X chromosome ( median step size of 4 . 8kb ) . At each step , the mean log2 ChIP-seq ratio between wild-type and recruitment site deletion strain was calculated from all binding sites within the window . To allow comparison across experiments , mean values were normalized using the log2 ratio at the rex-8 locus and percent deviation from wild-type was plotted . Significance of reduced DCC binding was calculated using students t-test , comparing log2 ratio of DPY-27 binding within a window to binding across the X chromosome . This analysis revealed a significant decrease of ~10–20% in DCC binding across ~1–2 Mb regions surrounding each individual deletion ( Figure 7A ) . Additional analyses using 1 Mb and 500 kb windows indicate similar results ( Figure 7—figure supplement 3 ) . DPY-27 ChIP-seq data from a set-4 mutant strain ( defective in dosage compensation but not in DCC localization ) revealed no localized reduction of DCC binding , validating the analysis and confirming that the effect on DCC binding is specific to the deletion strains ( Figure 7—figure supplement 4 ) . Additionally , mRNA-seq data from the rex-41 deletion strain indicated a significant increase in local gene expression in the affected region , consistent with reduced DCC binding ( Figure 7B ) . 10 . 7554/eLife . 23645 . 014Figure 7 . Multiple DCC recruitment sites establish DCC binding within defined X chromosomal domains . ( A ) DPY-27 ChIP-seq data was used to calculate percent deviation from wild-type in three different recruitment site deletion strains . Briefly , the log2 ratio between wild-type and deletion strain was calculated for all DPY-27 ChIP-seq peaks and compared to a positive control locus ( rex-8 , the strongest recruitment site ) . A variable-step sliding window ( window size of 2 Mb stepped across each DPY-27 peak ) was used to calculate average deviation from wild-type binding . Regions shown in pink have significantly decreased DPY-27 enrichment compared to the rest of the chromosome , p-value<0 . 05 , determined by one tailed students t-test , comparing log2 ratios in an individual window to log2 ratios across the whole X chromosome . In each experiment , the deleted recruitment site is indicated by a dashed line . Green lines ( darker with stronger insulation activity ) indicate TAD boundaries ( as described in [Crane et al . , 2015] ) . ( B ) mRNA-seq data comparing rex-41 deletion strain to wild-type is shown . Boxplot indicates the log2 ratio between deletion and wild-type for 500 kb windows tiled contiguously across the X chromosome . The rightmost window , shown in pink , contains a significant number of genes with increased transcription compared to wild-type ( fisher test , p-value=0 . 0126 ) . Asterisks mark windows with p-value<0 . 05 ) . ( C ) Summed distance between borders of deletion-affected regions from all three recruitment site deletions in ( A ) and nearest TAD boundary ( min: 54 kb , max: 230 kb ) is closer than expected by chance alone ( average: 313 kb ) . Shuffling the boundaries of deletion-affected regions ( 10000 permutations ) produced a normal distribution wherein the observed proximity to TAD borders is expected to occur less than 7% of the time by chance . Shown is the distribution of the summed total distance between permuted deletion boundaries and nearest TAD boundary expressed as a log10 value . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01410 . 7554/eLife . 23645 . 015Figure 7—figure supplement 1 . Schematic of recruitment site deletions and ChIP-seq data demonstrating DCC recruitment in the deletion strains . ( A ) Cartoon indicating location and size of the deleted recruitment sites in Figure 7 . Recruitment sites are shown as light blue boxes . Strong motifs are shown in pink , weak motifs are shown in blue . Deleted regions are indicated by dashed boxes . Rex-41 is part of an inverted repeat . The sgRNA used to delete rex-41 ( ranked #12 ) simultaneously deleted its sister recruitment site ( ranked #20 ) 1 kb downstream . ( B ) DPY-27 ChIP-seq enrichment for the right-most 1 Mb of the X chromosome . Data from wild-type , ∆rex-41 , ∆rex-1 , and ∆rex-40 is plotted . Dashed box around rex-41 shows lack of recruitment to this site is specific to the ∆rex-41 strain . ( C ) Overlap of SDC-3 and DPY-27 peaks in wild-type and ∆rex-41 indicate that the gross pattern of DCC binding across the X is not changed in the deletion strain . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01510 . 7554/eLife . 23645 . 016Figure 7—figure supplement 2 . Sanger sequencing results from the wild-type and recruitment site deletion strains . Sequence that is deleted is capitalized and in italics . Strong motifs are given in pink; weak motifs are given in blue . The rex-40 deletion bears a small insert of dpy-10 sequence , indicated in red . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01610 . 7554/eLife . 23645 . 017Figure 7—figure supplement 3 . Sliding-window analysis of DCC binding change across the X chromosome using different window sizes . As in Figure 7 , DPY-27 ChIP-seq data was used to calculate percent deviation from wild-type . Analysis was repeated for 2 Mb , 1 Mb , and 500 kb windows . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01710 . 7554/eLife . 23645 . 018Figure 7—figure supplement 4 . Control for sliding-window analysis of DCC binding change across the X chromosome . As in Figure 7 , DPY-27 ChIP-seq data was used to calculate percent deviation from wild-type . The set-4 ( n4600 ) mutation affects DCC function but not localization . TAD boundaries are shown in green . Regions with significantly decreased DPY-27 enrichment are shown in pink . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 01810 . 7554/eLife . 23645 . 019Figure 7—figure supplement 5 . 12-bp motif directionality and recruitment site interactions . . ( A ) Motif directionality for all strong motifs ( score ≥7 ) contained within annotated recruitment sites across the length of the X chromosome . The numbers in parenthesis indicate the rank of each recruitment site . Motifs on the plus strand are shown in yellow . Motifs on the minus strand are shown in blue . Locations of the recruitment sites are indicated . Strong recruitment sites are shown in pink , intermediate in blue , weak in grey . Insulation boundaries are plotted below the recruitment sites for reference . Insulation boundary strength is indicated: strongest boundaries are in dark green , weakest boundaries in light green . HiC data from ( Crane et al . , 2015 ) was used to plot the top 35% of contacts involving recruitment sites . ( B ) Table listing all recruitment sites , which contain at least one motif . Strong recruitment sites are highlighted . Motif directionality is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 23645 . 019 Interestingly , the edges of the regions with significant DCC depletion upon rex deletion align with published TAD boundaries on the X chromosome ( Crane et al . , 2015 ) ( Figure 7A ) . This observed proximity , calculated by summing the distance between the edges of the DCC depleted regions and the nearest TAD boundary for all three deletions , is expected by chance less than 7% of the time ( permutation test , Figure 7C ) . One possibility is that DCC spreading from recruitment sites is limited to sequences contained within the same TAD . TADs spanning 1–2 Mb have been observed in mammals ( Dixon et al . , 2012; Nora et al . , 2012 ) and C . elegans ( Crane et al . , 2015; Gabdank et al . , 2016 ) , while smaller domains have been identified in Drosophila ( 50–100 kb ) ( Hou et al . , 2012; Sexton et al . , 2012 ) . Recent work in C . elegans suggests a link between TAD formation and dosage compensation as X chromosome TADs are significantly altered in a dosage compensation mutant ( Crane et al . , 2015 ) . In addition , several recruitment sites coincide with annotated TAD boundaries on the C . elegans X , suggesting that strong recruitment at one site may prevent spreading across this site . Alternatively , condensin DC may be loaded along the entire length of the X chromosome before spreading , halting at recruitment sites bound by SDC proteins and resulting in condensin DC enrichment at these sites . This idea is reminiscent of current models of TAD formation wherein loop extrusion by complexes such as cohesin or condensin form progressively larger loops until stopped by TAD boundaries ( Sanborn et al . , 2015; Nasmyth , 2001; Goloborodko et al . , 2016; Fudenberg et al . , 2016 ) . In these models , boundary elements bound by CTCF block loop extrusion in an orientation dependent manner ( Fudenberg et al . , 2016; Rao et al . , 2014 ) . In humans , CTCF sites occur predominantly in a convergent orientation ( Rao et al . , 2014 ) . Though C . elegans lack CTCF , we reasoned that motif orientation at recruitment sites might play a similar role . To investigate this possibility , we plotted and listed ( Figure 7—figure supplement 5A and B ) the orientation of motifs contained within recruitment sites across the X chromosome , but saw no obvious pattern linking motif orientation to the location of the top HiC contacts . However , it is possible that not all the motifs within a recruitment site are functional . Therefore , given the negative data , we cannot comment on the potential importance of motif orientation in DCC recruitment-mediated TAD organization . Future studies analyzing DCC binding and chromosomal interactions in various recruitment site deletion strains will be needed to elucidate the conservation and diversification of the mechanisms that control condensin recruitment , condensin spreading , and the formation of TAD structures across different systems .
In the absence of SDC-3 , the hermaphrodite specific SDC-2 recognizes and binds only the strongest recruitment sites on the X chromosome ( Figure 2C ) . This observation , combined with evidence that DCC enrichment at ectopic rex-8 ( a strong recruitment site ) exceeds that of ectopic rex-1 ( an intermediate recruitment site ) ( Figure 6A ) , is highly suggestive of a recruitment site hierarchy . Recruitment site hierarchy was recently suggested as a mechanism to specify Drosophila MSL binding to the male X chromosome . Sites of initial recruitment , termed PionX , contain a novel motif specified by DNA sequence and shape that serve to distinguish the X chromosome from the autosomes ( Villa et al . , 2016 ) . Similarly , we find here that DCC binding on the C . elegans X is initiated at a subset of strong recruitment sites distinguished from weaker sites by the presence of 12 bp motif clusters , intrinsic nucleosome occupancy , and overlap with HOT sites . While each of these factors by themselves cannot distinguish the X chromosome from the autosomes , in combination they define a unique set of X-specific DCC recruitment sites . Interestingly , two strong recruitment sites lack motif sequence . It is possible that some yet-unidentified signal ( an alternative DNA motif , histone modification , non-coding RNA or long range interaction ) may be guiding the DCC to these sites . Our work indicates that SDC-2 is required for maintaining open chromatin at strong DCC recruitment sites ( Figure 3A ) . Additionally , the 12 bp motifs bound by the DCC on the X chromosome are predicted to have higher nucleosome occupancy than motifs that remain unbound ( Figure 4E ) . We reason that closed chromatin at the strong recruitment sites prevents the DCC binding prior to sdc-2 expression , thus ensuring hermaphrodite-specific recruitment of the dosage compensation machinery . This is similar to a recently described mechanism in Drosophila wherein the Zelda pioneer protein activates early enhancers by reducing intrinsic nucleosome occupancy and allowing TFs to bind ( Sun et al . , 2015 ) . It is not yet clear if SDC-2 acts as a pioneer factor , mediating DCC recruitment by increasing DNA accessibility . Alternatively , some yet unidentified factor ( s ) might work with or in advance of SDC-2 to recognize and evict nucleosomes . For instance , overlap of clustered 12 bp motifs with annotated HOT sites may reflect interaction between SDC-2 and various TFs in establishing open chromatin at recruitment sites . Recent work in fission yeast has proposed a role for transcriptional co-activators Gcn5 and RSC in evicting nucleosomes prior to condensin binding ( Toselli‐-Mollereau et al . , 2016 ) . Notably , it remains unclear how SDC-2 is able to recognize the 12 bp motif , as it contains no known DNA binding domains . Ectopic insertion of recruitment site sequence ( Figure 6A , C ) and deletion of recruitment sites on the X ( Figure 7A ) revealed an important feature of DCC recruitment: the cooperativity of recruitment sites over long distance . In single copy , both rex-1 and rex-8 recruited the DCC to an autosomal locus at a fraction of their activity on the X . While it is possible that reduced ectopic recruitment is a consequence of chromosomal location , our data shows little evidence for positional effects . We found that rex-1 sequence inserted in three copies at a secondary autosomal locus similarly showed reduced rex-1 activity compared to the endogenous site ( Figure 6A ) , and ectopic recruitment to rex-8 sequence was largely consistent between three chromosome II insertion sites ( Figure 6D ) . Additionally , others have recently reported that rex-32 ( a strong recruitment site ) recruits the DCC to multiple sites across the genome ( Wheeler et al . , 2016 ) . In light of these observations , lack of ectopic DCC recruitment to single-copy autosomal rex-1 sequence suggests that recruitment activity of this intermediate strength recruitment site is X-dependent , requiring the cooperation of nearby sites . The importance of recruitment site cooperativity is supported by previous observations that several X chromosome duplications containing recruitment sites do not recruit the DCC when detached from the X ( Csankovszki et al . , 2004; Blauwkamp and Csankovszki , 2009 ) . Here , we found that adding multiple recruitment site sequence either nearby ( Figure 6A ) or at a distance ( Figure 6C ) increased ectopic rex-1 recruitment , while recruitment site deletions resulted in reduced DCC binding across large chromosomal domains ( Figure 7A ) . Together these results demonstrate the functional importance of cooperation between multiple recruitment sites in establishing robust DCC binding to the X chromosome . Notably , our ChIP-qPCR data suggests a distinction between recruitment of the condensin DC core and the SDC subunits . In single-copy ectopic rex-1 and rex-8 insertions , SDC-3 was recruited better than DPY-27 ( Figure 6A ) . Adding additional recruitment sites nearby resulted in comparatively larger gains in DPY-27 enrichment , suggesting a greater role for cooperativity in the recruitment of condensin DC . Previous work has also demonstrated a difference in the X-chromosome binding pattern between condensin DC and the SDC proteins ( Ercan and Lieb , 2009 ) . By comparing DCC subunit binding between a well-defined site of spreading and a site of recruitment , it was shown that condensin DC spreads more effectively than both SDC-2 and SDC-3 ( Ercan and Lieb , 2009 ) . Lastly , our ChIP-seq data indicates the presence of the DCC recruitment module ( SDC-2 , SDC-3 , and DPY-30 ) at scattered sites across all autosomes that fail to recruit condensin DC . These differences , combined with a lack of biochemical evidence for an intact DCC , are highly suggestive of distinct mechanisms governing recruitment of the SDC proteins and of condensin DC . While SDC proteins recognize and target specific sites on the X and the autosomes , condensin DC binding is much more dependent on cooperativity of multiple recruitment sites on the X . Because recruitment sites are separated by a median distance of ~90 kb ( 544 bp – 1 . 2 Mb ) in the native X chromosome , we expect cooperation to act over long-distances . But what is the mechanism governing recruitment site cooperativity ? We can imagine several possibilities , none of which are mutually exclusive . First , cooperative 3D interactions between recruitment sites ( Crane et al . , 2015 ) might reinforce DCC binding by increasing the local concentration of condensin DC . Previous work indicates that DCC binding alters the nuclear positioning of the X chromosome ( Sharma et al . , 2014 ) and compacts it by ~40% ( Lau et al . , 2014 ) . It has been postulated that condensin complexes induce chromosome compaction by introducing and stabilizing long-range contacts between distant chromosomal loci ( Cuylen and Haering , 2011; Thadani et al . , 2012 ) . Indeed , condensin , similar to cohesin , can entrap DNA , mediating interactions between distant chromosomal sites ( Cuylen et al . , 2011 , Cuylen et al . , 2013 ) . Recent work in fission yeast demonstrated the ability of condensin to drive long-range associations across distances spanning 100 kb to 1 Mb ( Kim et al . , 2016 ) . Similarly , SDC-2 depletion , which results in lack of condensin DC recruitment to the X , reduced the number of 3D contacts between recruitment sites ( Crane et al . , 2015 ) . It is possible that DCC binding strengthens pre-existing X chromosome contacts while simultaneously inducing conformational changes that mediate interactions between recruitment sites and that these direct 3D interactions allow for the cooperative recruitment of the DCC . We envision that physical interaction between the DCC recruitment sites ( Crane et al . , 2015 ) paired with X chromosome compaction ( Lau et al . , 2014 ) may generate a nuclear compartment of high DCC concentration . Sequestration of the DCC to a compacted X chromosome territory may function in restricting re-loading of the DCC to the X chromosome . Potential recruitment sites on the autosomes would not benefit from this increased concentration of the DCC and would fail to maintain DCC binding . This model is reminiscent of transcription factories wherein stable nuclear compartments containing a high local concentration of Pol II lower the threshold for preinitiation complex formation for all nearby promoters ( Cook , 2010 ) . A surplus of DCC would reinforce complex binding at strong sites while simultaneously allowing for DCC recruitment to weaker recruitment sites . Clustering of recruitment sites in the nucleus to generate a compartment for DCC loading is an effective and elegant mechanism by which to cooperatively and specifically recruit the DCC to the X chromosomes . Second , it was shown that in vivo depletion of SUMO significantly disrupts DCC recruitment to the X chromosomes ( Pferdehirt and Meyer , 2013 ) . Although it is not clear if this effect is direct , western blot and mass-spectrometry analyses indicated that SDC-3 , DPY-27 , and DPY-28 are targets of SUMOylation ( Pferdehirt and Meyer , 2013 ) . DPY-28 , the non-SMC subunit that functions in both condensin I and condensin DC , is SUMOylated only in the context of the DCC ( Pferdehirt and Meyer , 2013 ) . It may be that successive SUMOylation of DCC subunits at multiple recruitment sites along the X somehow modulates DCC dynamics on the chromatin , perhaps increasing DCC affinity at recruitment sites , thereby reinforcing X specificity . Finally , recruitment site cooperativity may be mediated through condensin DC spreading . It has been proposed that after recruitment , the DCC spreads along the length of the chromosome ( Ercan et al . , 2009 ) . Spreading may be achieved by sliding of condensin DC ring along the chromatin . Alternatively , condensin DC may hop to distant chromosomal loci via long-range interactions between recruitment and spreading sites . The mechanism of DCC spreading remains unknown . Notably , the size of the region affected by a recruitment site deletion is similar to the distance that the DCC can spread from the X chromosome into autosomal sequence ( Ercan et al . , 2009 ) . In three X;A fusion chromosomes , levels of DCC enrichment on autosomal sequence was shown to be inversely proportional to the distance from the fused end of the X chromosome . Similarly , reduced DCC binding near a deleted recruitment site is also dependent on proximity to the deletion ( Figure 7A ) . The observed pattern of decreased DCC binding in the recruitment site deletion strains suggests that DCC spreading is restricted to TAD boundaries . In mouse embryonic cells , deletion of a boundary element within the Hox locus resulted in spreading of active chromatin across the border ( Narendra et al . , 2015 ) . Spreading of SMC complexes such as cohesin and condensin , as well as other chromatin regulating proteins may be limited by the TAD boundaries . The DCC is also implicated in the creation of the TAD boundaries themselves ( Crane et al . , 2015 ) , suggesting that the mechanism by which the DCC is loaded onto and spreads across the chromosome may influence boundary formation . Mechanistically , spreading may serve to both establish and reinforce DCC binding sites across the length of the X chromosome . It is not clear if cooperativity also contributes to the X-specific recruitment of the Drosophila MSL complex . Although the strategies are distinct , X-specific binding of the C . elegans DCC and the Drosophila MSL show significant parallels . In both , a DNA sequence motif is important for recruitment , but is not specific to the X chromosome ( Conrad and Akhtar , 2012; Alekseyenko et al . , 2012 , Alekseyenko et al . , 2008 ) . Recent work has indicated that GA repeat expansion around an 8 bp motif core helps specify MSL binding at high affinity sites , mediated through CLAMP zinc-finger binding ( Kuzu et al . , 2016 ) . This dinucleotide expansion may serve to distinguish X and autosomal motif sites ( Kuzu et al . , 2016 ) . However , just like the DCC recruiters ( SDC-2 , SDC-3 and DPY-30 ) , CLAMP binds to autosomal loci independent of MSL ( Soruco et al . , 2013 ) . Previous studies found that ectopic insertion of Drosophila high affinity MSL recruitment sites are sufficient to recruit the MSL complex to an autosome ( Kelley et al . , 1999; Alekseyenko et al . , 2008 ) . Although the level of this ectopic recruitment was not precisely quantified , it results in spreading to nearby active genes where the complex functions to up-regulate transcription ( Ramírez et al . , 2015 ) . It would be interesting to determine if the specific recruitment of Drosophila MSL to the X chromosome also involves cooperativity between high affinity sites on the X . Condensin complexes are conserved from yeast to humans , functioning in both chromosome condensation during cell division and gene regulation during interphase ( Hirano , 2016; Cobbe et al . , 2006; Hirano , 2012; Lau and Csankovszki , 2014 ) . Studies in several organisms have revealed high-resolution condensin binding patterns , demonstrating that condensin complexes bind to distinct intergenic sites that are enriched for promoters and enhancers ( Kranz et al . , 2013; Pferdehirt et al . , 2011; Kim et al . , 2013; Dowen et al . , 2013 ) . In yeast , the TATA box binding protein ( TBP ) , the histone acetyltransferase , Gcn5 , and the RSC remodeling complex have been shown to recruit condensin to its chromosomal binding sites ( Toselli‐-Mollereau et al . , 2016; Iwasaki et al . , 2015 ) . While the metazoan condensin recruiters remain largely uncharacterized ( Piazza et al . , 2013 ) , our previous work indicated that , similar to condensin DC , the C . elegans condensin II complex binds to DCC recruitment sites in an SDC-2 dependent manner ( Kranz et al . , 2013 ) . It remains unclear if , similar to condensin DC , recruitment is cooperative for canonical condensins . Interestingly , in yeast , tRNA genes that recruit condensin cluster in three-dimensional space and this clustering is , in part , mediated by condensin itself ( Haeusler et al . , 2008 ) . Furthermore , yeast ribosomal DNA ( rDNA ) repeat region robustly recruits condensin in a FOB1 dependent manner ( Johzuka et al . , 2006 ) . It is possible that the repetitive structure of the rDNA locus allows for cooperative recruitment through multiple tandem copies of the recruiting sequence . Notably , the weaker DCC recruitment sites are enriched at tRNA gene loci , suggesting that once the initial X-specific sites recruit the DCC , common mechanisms of condensin loading help to maintain DCC binding on the X chromosomes . Our model postulates that long-range cooperation between a hierarchical set of recruitment sites allows for the robust and specific targeting of the DCC to the X chromosomes without the necessity of evolving and maintaining domain-specific motifs at every recruitment site . This model has important implications for TF targeting and domain-level transcriptional regulation in eukaryotic genomes . Recent work in various organisms has begun to focus on the functional role of cooperation between regulatory elements . For example , in Drosophila , tandem gene duplications often exhibit higher than 2-fold increased transcription , potentially due to the additive effects of transcription factors binding identical sites on both gene copies ( Loehlin and Carroll , 2016 ) . In human B cells , long-range enhancer interactions help explain lineage-specific control of Tcrb recombination ( Proudhon et al . , 2016 ) . In yeast , cooperation between mating-type silencers leads to transcriptional repression ( Boscheron et al . , 1996 ) . Finally , similar to how weak recruitment sites bind the DCC only in multiple copies or in the context of the X chromosome , it has been proposed that some Drosophila Polycomb domains with weak binding sites may recruit the Polycomb group protein , PHO , through cooperation between high affinity sites ( Schuettengruber et al . , 2014 ) . From yeast , to flies , to humans , cooperative interactions in 3D space are increasingly being used to explain genome-targeting specificities of proteins important in processes ranging from recombination to transcriptional silencing . We add to this list a clear paradigm for studying the genetic mechanisms behind long-range cooperativity: recruitment of the dosage compensation complex to the C . elegans X chromosomes .
Wild-type ( N2 ) , ERC06 ( knuSi254[SNP400bprex-1 , unc-119 ( + ) ] II; unc-119 ( ed3 ) III ) , ERC08 ( knuIs6[pSE-02 ( 400bprex-1SNP ) , unc-119 ( + ) ) ] I; unc-119 ( ed3 ) III ) , ERC09 ( knuIs7[pSE-02 ( 400bprex-1SNP ) , unc-119 ( + ) ) ] III; unc-119 ( ed3 ) III ) , ERC38 ( ers30[delX:17544437–17544484 , delX:17545624–17545624] ) , ERC51 ( ersIs17[SNP400bp_rex1 , X:14373144] ) , ERC54 ( ers20[delX:4394846–4396180]; knuIs6[pSE-02 ( 400bprex-1SNP ) , unc-119 ( + ) ) ] I; unc-119 ( ed3 ) III ) , ERC64 ( ers28[delX:806628–806813]; knuIs6[pSE-02 ( 400bprex-1SNP ) , unc-119 ( + ) ) ] I; unc-119 ( ed3 ) III ) , ERC61 ( ersSi25[wt400bp_rex8 , cb-unc-119 ( + ) ] II; unc-119 ( ed3 ) III ) , ERC62 ( ersIs26[X:11093923-11094322[rex-8] , II: 8449965 ) ; knuSi254[SNP400bprex-1 , unc-119 ( + ) ] II; unc-119 ( ed3 ) III ) , ERC63 ( ersIs27[X:11093923-11094322[rex-8] , II:8371600 , II:8449968 ) ; knuSi254[SNP400bprex-1 , unc-119 ( + ) ] II; unc-119 ( ed3 ) III ) , ERC66 ( ersIs30[V:14040305-14040785[chrV_clusteredmotifs] , X:14373144] ) , TY2205 ( her-1 ( e1520 ) sdc-3 ( y126 ) V; xol-1 ( y9 ) X ) , TY1072 ( her-1 ( e1520 ) V; sdc-2 ( y74 ) X ) , MT14911 ( set-4 ( n4600 ) II ) . Information regarding strain names , genotype , and generation of all new strains , used in this study can be found in Supplementary file 1A . Briefly , for all insertion strains , 400 bp of either rex-1 ( chrX:4395400–4395799 , WS220 ) or rex-8 sequence ( chrX:11093923–11094322 , WS220 ) , ( centered at the DCC ChIP binding peak summit and containing recruitment motif sequence ) was integrated into an ectopic genomic location using MosSci ( Zeiser et al . , 2011 ) , bombardment , or CRISPR ( Dickinson and Goldstein , 2016 ) based techniques . In the bombardment strains ( ERC08 and ERC09 ) , insertion sites were mapped using genomic DNA paired end Illumina sequencing data . We isolated discordant read pairs that connected bombarded plasmid sequence to a specific region of the genome , eliminating spurious connections by comparing against paired-end sequencing data obtained from wild-type N2 genomic DNA . Insertion copy-numbers were calculated using qPCR analysis of genomic DNA , normalizing against a negative control region in the genome , and using the single-copy insertion strain ( ERC06 ) generated by MosSCI as a positive control reference . Deletion strains were generated using CRISPR; sgRNAs were designed flanking recruitment site sequence , ensuring deletion of all identified 12 bp motifs contained within . PCR followed by Sanger sequencing was used to validate insertions and deletions . Deleted sequences and the motifs contained within can be found in Figure 7—figure supplement 3 . Strains were maintained at 20°C on NGM agar plates using standard C . elegans growth methods . Mixed stage embryos were isolated from gravid adults by bleaching . For ChIP experiments , embryo samples were fixed by treating with 2% formaldehyde for 30 min . For RNA experiments , embryo samples were resuspended in 10 volumes of Trizol . Supplementary file 1B includes the list of the ChIP experiments and information on the antibodies . For ChIP , embryos were washed and dounce homogenized in FA buffer ( 50 mM HEPES/KOH pH 7 . 5 , 1 mM EDTA , 1% Triton X-100 , 0 . 1% sodium deoxycholate; 150 mM NaCl ) . Sarkosyl ( 0 . 1% sodium lauroyl sarcosinate ) was added before sonicating to obtain chromatin fragments between 200 and 800 bp in length . As in Ercan et al . , 2007 , 1 to 2 mg of embryo extract and 3 to 8 ug of antibody was used per ChIP . For library preparation , half of the ChIP DNA was ligated to Illumina TruSeq adapters and amplified by PCR . Library DNA between 250 and 500 bp was gel purified . Single-end 50 bp sequencing was performed using the Illumina HiSeq-2000 at the New York University Center for Genomics and Systems Biology , New York , NY . ChIP-qPCR was performed using 4–5% of each ChIP sample and the corresponding diluted input DNA , as described in Mukhopadhyay et al . ( 2008 ) . KAPA SYBR FAST 2X qPCR Master Mix ( Kapa Biosystems , MA ) was used in 20 μl reactions and performed on the Roche LightCycler 480 system . Data ( input and chip sample ) were used to calculate ∆∆Ct between experimental and control loci . Briefly , ∆∆Ct = ( experimental Ct –control Ct ) ChIP – ( experimental Ct –control Ct ) input . Enrichment over background for both endogenous and ectopic loci was calculated as 2-∆∆Ct . Percent endogenous recruitment was then calculated by taking the ratio of ectopic enrichment over endogenous enrichment and multiplying by 100 . Because background enrichments are variable between biological replicates , ∆∆Ct and percent endogenous were calculated individually for each replicate . DNA sequence for the qPCR primers and product sizes are given in Supplementary file 1C . We aligned 50 bp single-end reads to C . elegans genome version WS220 using bowtie version 1 . 0 . 1 ( Langmead et al . , 2009 ) , allowing two mismatches in the seed , returning only the best alignment , and restricting multiple alignments to , at most , four sites in the genome . Mapped reads from ChIP and input were used to call peaks and obtain read coverage per base using MACS version 1 . 4 . 2 ( Zhang et al . , 2008 ) with default parameters . Coverage per base was normalized to the genome-wide median coverage ( excluding the mitochondrial chromosome ) . Final ChIP enrichment scores per base were obtained by subtracting matching input coverage . We merged replicates by taking the average ChIP enrichment scores at every position in the genome . For the sdc-2 and sdc-3 null mutants ( each of which have altered X:A ratios ) , X and autosomal reads were processed separately and final ChIP enrichment scores were combined after normalization . Raw data files and wiggle tracks of ChIP enrichment per base pair are provided at NCBI’s Gene Expression Omnibus ( Edgar et al . , 2002 ) database under accession number [GEO: GSE87741] . Basic statistics and the GEO number for each ChIP-seq data are provided in Supplementary file 1B . To determine a set of binding sites for each protein , we imposed a majority rule using the biological replicates , as explained in Kranz et al . , 2013 . Briefly , reads from the replicates were combined using samtools version 0 . 1 . 19 ( Li et al . , 2009 ) and MACS was used to call peaks at a stringent p-value cutoff e-10 . Only those peaks from the combined set that were also present in the majority of the individual replicates were included in the final peak set . Peak summits were determined as the position with the maximum ChIP enrichment score in the combined average . To avoid penalization of long peaks with multiple summits , peaks were split into smaller peaks using PeakAnalyzer version 1 . 4 ( Salmon-Divon et al . , 2010 ) , with the minimum height being equal to the median coverage at all determined summits of the given data set and a separation float of 0 . 85 . The analysis scripts are written in R and Perl and are available as Source code 1 . Worms were collected and stored in ten volumes of Trizol ( Invitrogen ) . Samples were freeze-cracked five times and RNA purification was done according to the manufacturer protocol . Isolated RNA was cleaned up using the Qiagen RNeasy kit . The mRNA was purified using Sera-Mag Oligo ( dT ) beads ( Thermo Scientific ) from at least 1 ug of total RNA . Stranded mRNA-seq libraries were prepared based on incorporation of dUTPs during cDNA synthesis using a previously described protocol ( Parkhomchuk et al . , 2009 ) . Single-end 50 bp sequencing was performed using the Illumina HiSeq-2000 . Reads were aligned to genome version WS220 with Tophat version 2 . 0 . 0 ( Trapnell et al . , 2012 ) using default parameters . For each biological replicate , read numbers and mapping percentages ( which refer to the percentage of unique reads with at least one alignment ) can be found in Supplementary file 1B . Count data was calculated using HTSeq version 0 . 6 . 1 ( Anders et al . , 2015 ) and normalized using the R package DESeq ( Anders and Huber , 2010 ) . The raw reads and counts can be obtained from GEO under accession number [GEO: GSE87741] . To estimate the strongest binding sites ( foci [Ercan et al . , 2007] ) for SDC-2 , a mixture distribution with two components was assumed: one for the majority of binding sites with low coverage and one for binding sites with an extremely high coverage at the peak summits . To differentiate between binding sites with strong and low coverage , the coverage at the peak summits of all binding sites was used to estimate the parameters of a mixture distribution with the function normalmixEM from the mixtools package version 1 . 0 . 3 ( Benaglia et al . , 2009 ) in R version 3 . 2 . 1 ( R Core Team , 2015 ) . SDC-2 peaks overlapping with SDC-3 , DPY-30 , and DPY-27 but not overlapping with H3K4me3 peaks were successively determined using the command intersectBed of the BEDTools suite version 2 . 12 . 0 ( Quinlan and Hall , 2010 ) . Deeptools version 2 . 2 . 3 ( Ramírez et al . , 2016 ) was used to rank order recruitment sites based on SDC-2 , SDC-3 , DPY-30 , and DPY-27 ChIP enrichment for a 3 kb window centered on the peak summit . Recruitment sites were clustered into three categories ( strong , intermediate , and weak ) using k-means clustering ( n = 3 ) . Recruitment site coordinates and rankings can be found in Supplementary file 1D . DNA sequence of ±100 bp around the summit of the top 200 SDC-2 ChIP-seq binding peaks was used to identify potential binding motifs using MDScan ( Liu et al . , 2002 ) . The position weight matrix of the top motif was used to scan and score the whole genome using TRAP ( Thomas-Chollier et al . , 2011 ) . Resulting TRAP affinity scores were standardized to a 0 to 10 scale . Motif locations and scores can be found in Supplementary file 1E . | The DNA inside living cells is organized in structures called chromosomes . In many animals , females have two X chromosomes , whereas males have only one . To ensure that females do not end up with a double dose of the proteins encoded by the genes on the X chromosome , animals use a process called dosage compensation to correct this imbalance . The mechanisms underlying this process vary between species , but they typically involve a regulatory complex that binds to the X chromosomes of one sex to modify gene expression . Caenorhabditis elegans , for example , is a species of nematode worm in which individuals with two X chromosomes are hermaphrodites and those with one X chromosome are males . In C . elegans , a regulatory complex , called the dosage compensation complex , attaches to both X chromosomes of a hermaphrodite , and reduces the expression of the genes on each by half to match the level seen in the males . Previous research has shown that short DNA sequences , known as motifs , recruit the dosage compensation complex to the X chromosomes . However , these sequences are also found on the other chromosomes and , until now , it was not known why the complex was only recruited to the X chromosomes . Albritton et al . now show the X chromosomes have a ‘hierarchical’ recruitment system . A few sites on the X chromosomes contain clusters of a specific DNA motif , which initiate the process and attract the dosage compensation complex more strongly than other sites . These ‘strong’ recruitment sites are placed across the length of the X chromosomes and cooperate with several ‘weaker’ ones located in between . This way , multiple recruitment sites can cooperate over a long distance , while non-sex chromosomes , which have only one or two stronger recruitment sites , do not have thisadvantage . Hierarchy and cooperativity may be general features of gene expression , in which proteins are targeted to chromosomes without the need for having specific motifs at every recruitment site . The way DNA sequences are distributed across the genome may give us clues about their role . Thus , knowing how genomes are structured will help us identify disrupted areas in diseases such as cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2017 | Cooperation between a hierarchical set of recruitment sites targets the X chromosome for dosage compensation |
Concussion is associated with a myriad of deleterious immediate and long-term consequences . Yet the molecular mechanisms and genetic targets promoting the selective vulnerability of different neural subtypes to dysfunction and degeneration remain unclear . Translating experimental models of blunt force trauma in C . elegans to concussion in mice , we identify a conserved neuroprotective mechanism in which reduction of mitochondrial electron flux through complex IV suppresses trauma-induced degeneration of the highly vulnerable dopaminergic neurons . Reducing cytochrome C oxidase function elevates mitochondrial-derived reactive oxygen species , which signal through the cytosolic hypoxia inducing transcription factor , Hif1a , to promote hyperphosphorylation and inactivation of the pyruvate dehydrogenase , PDHE1α . This critical enzyme initiates the Warburg shunt , which drives energetic reallocation from mitochondrial respiration to astrocyte-mediated glycolysis in a neuroprotective manner . These studies demonstrate a conserved process in which glycolytic preconditioning suppresses Parkinson-like hypersensitivity of dopaminergic neurons to trauma-induced degeneration via redox signaling and the Warburg effect .
Selective vulnerability of different neuronal subtypes to dysfunction , degeneration , and death underlies all neurological diseases . Differential interplay between genetic and environmental factors contribute to this neuronal selectivity . Environmental stimuli have the capacity to trigger pathogenic cascades contingent on genetic factors . Such is the case for traumatic brain injury ( TBI ) and concussion , in which the initial biomechanical insult precipitates numerous cellular and systemic alterations with the potential to promote neurological dysfunction and progressive neurodegeneration . In particular , ionic imbalances , cytoskeletal disruption , aberrant proteostasis , genomic instability , and mitochondrial dysfunction within neurons and support cells initiate systemic abnormalities including metabolic impairment , vascular disruption , inflammation , blood-brain barrier defects , and reduced cerebral blood-flow ( Giza et al . , 2018 ) . Yet , the stochastic nature of the mechanical insult has complicated our ability to define the underlying mechanisms of neurodegeneration . It remains unclear whether the inherent chemical properties of different neuronal subtypes confer varying degrees of vulnerability to trauma-induced neurodegeneration . Additionally , the genetic contribution to this widespread disorder is poorly understood , genetic effectors are likely to amplify the cellular and systemic aberrations initiated by the mechanical insult . Thereby , we hypothesize that genetic targeting and modulation may be sufficient to suppress the ensuing neurodegeneration after concussive brain injury . Metabolic fluctuations in the brain occurring with age or injury have been linked with neurological health ( Camandola and Mattson , 2017; Marino et al . , 2007 ) . The neurometabolic cascade resulting from concussion is characterized by erratic fluctuations in metabolism , which are hypothesized to accommodate the energetic demand required for neuronal repolarization and repair ( Giza and Hovda , 2001 ) . Brain energetics relies heavily on two metabolic pathways including mitochondrial oxidative phosphorylation and cytosolic glycolysis ( Kasischke et al . , 2004; Pellerin and Magistretti , 1994 ) . The astrocyte-neuron lactate shuttle ( ANLS ) entails astrocytic production and transport of the glycolytic by-product , lactate , to neurons which preferentially oxidize it rather than glucose to meet their energetic demands ( Magistretti et al . , 1999 ) . Immediately following injury , rodent brains display signs of decreased oxidative phosphorylation that can persist for weeks ( Gilmer et al . , 2009; Xiong et al . , 1997 ) . Conversely , the injured brain transiently elevates normoxic glycolysis in humans and animal models ( Bergsneider et al . , 1997; Sokoloff et al . , 1977; Yoshino et al . , 1991 ) , suggesting an energetic reallocation through astrocytic means . This metabolic shift is a phenomena observed in cancer cells and is termed the Warburg effect , in which aerobic glycolysis is favored over oxidative phosphorylation ( Warburg , 1956 ) . Thus , the injured brain initially undergoes a Warburg-like response , but how this metabolic shift occurs within the astrocyte-neuron axis and its ability to impact neurological function and degeneration after concussion is not well understood .
In contrast to the cellular and network complexity of the human brain , the adult nematode , C . elegans , possesses 302 post-mitotic neurons ( White et al . , 1986 ) . A simple nervous system offers the opportunity to rapidly screen different neural subtypes and gene products in pursuit of uncovering neurodegenerative mechanisms that translate to the mammalian brain . To this end , we developed a collision-based , rapid deceleration model of trauma in which high-frequency , multidirectional agitation delivers a well-calibrated injury to a large population of worms ( Egge et al . , 2021 ) . To determine whether the inherent biochemical properties of different neural subtypes impart variable sensitivity to blunt force trauma , dopaminergic , GABAergic , glutamatergic , serotonergic , and cholinergic neurons were individually monitored using targeted GFP expression . Fluorescence retention in individual neurons housed in the central nerve ring was measured to assess viability at various time points post-trauma ( Nass et al . , 2002 ) . When compared to age-matched non-injured counterparts , dopaminergic neurons displayed the greatest reduction in fluorescence with a 42 . 1% ( +/−1 . 4% ) loss in signal intensity within 24 hr after injury , indicative of elevated sensitivity to degeneration ( Figure 1A ) . Microscopic examination of dopaminergic neurons revealed phenotypes characteristic of neuronal damage ( Gennarelli , 1996; Kilinc et al . , 2009 ) 24 hr after injury , including fluorescence beading in dendritic processes ( Figure 1B ) . To examine conservation of dopaminergic hypersensitivity to trauma-induced neurodegeneration , we administered a concussive close-head injury to mice . This injury impaired immediate righting reflex response , indicating temporary loss of consciousness ( Figure 1—figure supplement 1A ) . Seven days after injury , dopaminergic neurons of the substantia nigra exhibited signs of increased apoptosis as evidenced by enhanced immunofluorescence staining of cleaved caspase-3 , and a 59 . 06% ( +/−3 . 03% ) loss of tyrosine hydroxylase-positive neurons was observed 28 days post-trauma ( Figure 1C and D; Figure 1—figure supplement 1B–D ) . Open-head models of rodent brain injury consistently report necrotic tissue loss at or near the lesion site ( Hall et al . , 2005 ) . Yet , one study demonstrates neuronal loss within the substantia nigra ( Liu et al . , 2017 ) . In contrast , closed-head brain injury without cranial fracture revealed no significant cleaved caspase-3 staining in the hippocampus , thalamus and visual cortex seven days post-injury ( Figure 1—figure supplement 1E–I ) . Furthermore , neuroinflammation was not evident in the immunolabeling of microglia ( IBA1 ) with CD68-positive inflammatory cytoplasmic granules or in the transcriptional activation of known inflammatory targets ( Figure 1—figure supplement 1J–K ) . Dopaminergic neurons within the midbrain comprise the nigrostriatal circuit , which influences voluntary movement . In concussed mice , we observed reduced latency on the accelerating rotarod and decreased resting state ambulation coincident with dopaminergic degeneration ( Figure 1E and F ) . Thus , modeling blunt force trauma in C . elegans and concussive injury in mice demonstrate conserved hypersensitivity of dopaminergic neurons to biomechanical insult . To determine the physical strain on brain regions proximal and distal to the injury site , we developed a computational biophysical model of TBI . Specifications for our closed-head trauma device were incorporated into a detailed finite element model ( FEM ) with ~2 . 5 million 3D solid elements ( Figure 2—figure supplement 1A and B ) . To capture stress and strain across the entire brain , six distinct material properties were assigned to accurately represent varying brain regions ( Table 1 ) . Biophysical modeling revealed a stress wave propagating throughout the mouse brain due to linear impact ( Figure 2—figure supplement 1C-E; Video 1 ) . By FEM analysis , a peak effective stress of 22 . 8 kPa ( corresponding principal strain of 8 . 3% ) first appeared in the cortex region at 0 . 31 ms . Then , stress propagated to 28 . 62 kPa ( corresponding principal strain of 23 . 6% at the thalamus and 18 . 5% at the substantia nigra ) at 0 . 85 ms post-injury in the thalamus and substantia nigra region . After impacting , a stress wave caused another peak effective stress of around 30 kPa ( corresponding principal strain of 40 . 9% at the brainstem region and 27 . 0% at the thalamus region ) at 1 . 64 ms ( Figure 2A–C; Video 2 ) . Thus , dopaminergic neurons may be sensitized to biomechanical insults due to inherent physiological properties since non-invasive , linear impact generates a strain within the midbrain comparable to brain regions closer to the impact site that lack evidence of neuronal death . Longitudinal studies from the clinic show that traumatic brain injury with a loss of consciousness presents a significant risk factor for development of Parkinson’s disease ( PD ) , Parkinsonism , and Lewy body accumulation ( Crane et al . , 2016 ) . Indeed , the pathology and motor deficits that we observed after concussive injury in mice are similar to those of individuals with PD . Based on familial links and environmental risk factors , mitochondrial components including the electron transport chain ( ETC ) have emerged as prominent drivers of PD ( Perier and Vila , 2012 ) . Utilizing our C . elegans trauma model , we found that reducing the complex IV cytochrome C oxidase activity of the ETC through cox-5b RNAi promotes a dose-dependent survival of dopaminergic neurons after injury ( Figure 3A and B ) . Moreover , reduced expression of cox-5b exclusively in the nervous system was sufficient to promote survival of dopaminergic neurons after blunt force trauma ( Figure 3—figure supplement 1A ) . To examine the conservation of cytochrome C oxidase-related neuroprotection , we examined mice lacking the complex IV assembly factor surfeit locus protein 1 , SURF1 ( Dell'agnello et al . , 2007 ) . Surf1 mutations reduce rather than ablate cytochrome C oxidase activity , leading to a 53 . 9% reduction in function ( Figure 3—figure supplement 1B ) . Surf1-/- mice exhibited a 40% reduction in mortality rate following lethal head trauma and a 2 . 3-fold improvement in post-injury reflex recovery time compared to wild type littermates ( Figure 3—figure supplement 1C and D ) , suggesting immediate protection against TBI . Cleaved caspase-3 signal intensity within the midbrain was reduced by 55% ( +/−3 . 5% ) after 7 days in concussed Surf1-/- mice ( Figure 3C and D ) . Coincident with decreased neurodegeneration , Surf1-/-mice showed increased latency in high-speed rotarod analysis compared to wild-type littermates ( Figure 3E; Figure 3—figure supplement 1E ) . This was accompanied by a 25 . 93% ( +/−6 . 828% ) dopaminergic loss 28 days post-injury , compared to the 59 . 06% ( +/−3 . 03% ) observed in wild-type counterparts ( Figure 3—figure supplement 1F–H ) . Moreover , resting state ambulation was unaffected by trauma in Surf1-/- mice unlike deficits observed in wild-type littermates ( Figure 3F ) . Thus , suppressing trauma-induced degeneration of dopaminergic neurons by reduced cytochrome C oxidase activity is evolutionarily conserved between C . elegans and mice . To understand the cause of neurodegeneration in our worm-to-mouse translational trauma model , we examined similarities reported for concussive brain injury and Parkinson’s disease . Reactive oxygen species ( ROS ) are toxic by-products of brain injury ( Hall and Braughler , 1993 ) that originate in the mitochondrial inner membrane by partial reduction of oxygen . Upon accumulation , ROS initiate cellular damage and stress . In genetic and pharmacological models of Parkinson’s disease , mitochondrial dysfunction and ROS accumulation underlie dopaminergic neurodegeneration and subsequent disease progression ( Blesa et al . , 2015 ) . In C . elegans , we observed a five fold increase in ROS production immediately after blunt force trauma , which was suppressed upon cox-5b RNAi ( Figure 4A ) . Since cox-5b RNAi treatments in C . elegans are reported to reduce oxygen consumption ( Kaufman and Crowder , 2015 ) , we hypothesized that reduced reliance on mitochondrial respiration mitigates the likelihood of overwhelming respiratory output . Essentially , reducing the rate of electron transfer between ETC complexes reduces the rate of ROS generation and accumulation . In this model , cox-5b RNAi should fail to protect neurons against ectopically generated ROS . Indeed , elevating neuronal ROS via expression of the mitochondrial targeted oxidative photosensitizer , Tom20::KillerRed ( Wojtovich and Foster , 2014 ) , was sufficient to induce dopaminergic neurodegeneration in worms that could not be suppressed by cox-5b RNAi ( Figure 4B ) . Thus , intra-neuronal ROS production is sufficient to drive dopaminergic degeneration , while neuroprotection conferred by reduced cytochrome C oxidase activity appears to act upstream of this ROS-induced neurotoxicity . To evaluate the conservation of this neuroprotective mechanism in the mouse , we examined mitochondrial function and ROS production in the absence of Surf1 . Despite no significant change in mitochondrial membrane potential ( Figure 4—figure supplement 1A ) , Surf1-/- cells consumed two fold less oxygen and exhibited reduced reticular mitochondrial morphology ( Figure 4C and D; Figure 4—figure supplement 1B ) , strongly suggesting reductions in oxidative phosphorylation ( Galloway et al . , 2012 ) . Two hours after concussive injury in mice , a transient 69% increase of the ROS byproduct , superoxide , was observed in mitochondria isolated from brain tissue compared to a 6% increase in Surf1-/- mice ( Figure 4E ) . Moreover , metabolite analysis from brain homogenates shows a 61% increase in glutathione oxidation within 2 hr of injury in wild-type animals , which was not observed in Surf1-/- mutants ( Figure 4F ) . Thus , decreased cytochrome C oxidase activity suppresses transient trauma-induced ROS accumulation , a process conserved from worm to mouse . Notably , the basal mitochondrial superoxide levels and oxidized glutathione were significantly elevated in the Surf1-/- brains compared to wild-type littermates ( Figure 4—figure supplement 1C ) . Therefore , minimizing transient production and accumulation of oxidative species immediately after trauma parallels neuroprotection . Traumatic brain injury is a costly energetic process in which neurons work to re-establish resting membrane potential and facilitate cellular repair ( Giza and Hovda , 2001 ) . Despite the apparent decrease in oxidative phosphorylation ( Figure 4C ) , ATP levels in the Surf1-/- brain were not affected compared to wild type littermates ( Figure 5—figure supplement 1A ) . Furthermore , respiratory exchange rate , food intake , body temperature , and overall body weight were also unaffected in Surf1-/- mice ( Figure 5—figure supplement 1B–E ) . Thus , Surf1-/- mice likely accommodate their energetic needs through alternate means in response to suboptimal mitochondrial respiration . Post-trauma , the brain transiently increases glycolysis ( Yoshino et al . , 1991 ) and will readily metabolize the glycolytic by-product , lactate ( Glenn et al . , 2015 ) . Surf1-/- animals appeared to undergo this Warburg-like increase in glycolysis . Consistent with elevated blood lactate reported in Surf1-/- mice ( Pulliam et al . , 2014 ) , increased transcription of the lactate dehydrogenase , Ldhd , in the Surf1-/- brain correlated with the rapid acidification of growth media by Surf1-/- cells ( Figure 5A; Figure 5—figure supplement 1F ) . Similar transcriptional upregulation of the lactate dehydrogenase , ldh-1 , was also observed in worms treated with cox-5b RNAi ( Figure 5—figure supplement 1G ) . We hypothesized that neuroprotection by reduced cytochrome C oxidase is the result of preconditioning through metabolic reallocation . By preemptively shifting metabolism away from mitochondria respiration , the likelihood of overwhelming ETC function is reduced during the metabolic demand of TBI . To accommodate energetic needs , the brain relies more heavily on normoxic glycolysis in a process referred to as the Warburg effect ( Warburg , 1956 ) . In Surf1-/- brains , we observed transcriptional upregulation of several enzymes in the glycolytic pathway ( Figure 5A; Figure 5—figure supplement 1H ) . As a critical regulator of the Warburg shunt , which facilitates the shift from mitochondrial respiration to cytosolic glycolysis , the pyruvate dehydrogenase complex ( PDC ) is inactivated by pyruvate dehydrogenase kinases , PDK1-4 , through phosphorylation of its E1 alpha subunit ( PDHE1α ) ( Kolobova et al . , 2001; Patel and Korotchkina , 2006 ) . In the brains of Surf1 mutants , we observed transcriptional upregulation of the brain-enriched Pdk2 in conjunction with the repression of its complementary PDH phosphatase , Pdp1 ( Figure 5A; Figure 5—figure supplement 1H ) . PDHE1α phosphorylation has previously been reported after open-head brain injury and provides a potential molecular explanation for trauma-induced hyperglycolysis ( Xing et al . , 2009 ) . Consistent with transcriptional changes in Surf1-/- brains , we observed increased PDHE1α phosphorylation both before and after injury in Surf1-/- compared to wild-type littermates ( Figure 5B; Figure 5—figure supplement 2A ) , while total PDH levels are equivalent ( Figure 5—figure supplement 2B and C ) . Thus , Surf1 mutants are preconditioned prior to injury into a Warburg-like state , which likely mitigates the transient ROS production observed immediately after concussion . While Surf1 mutants possess a shift in their global transcriptome , it remained unclear how this neuroprotective response occurs at the cellular level within the astrocyte-neuron axis . To obtain this cellular resolution , we employed single nuclei RNA sequencing , which allowed transcriptomic analysis of individual cells from the midbrain and striatum ( Figure 5—figure supplement 1M–O ) . Consistent with the astrocyte-neuron lactate shuttle hypothesis , the brain-enriched Warburg shunt regulator , Pdk2 , as well as every activated glycolytic gene except for Gpi1 , was elevated in astrocytes ( Figure 5C ) . Indirect immunofluorescence within the midbrain confirmed that steady-state expression of the PDK2 protein was confined to astrocytes ( Figure 5—figure supplement 1P and Q ) . Moreover , glycolytic transcripts were elevated within Surf1 mutant astrocytes compared to wild-type littermates ( Figure 5D ) . Through cellular resolution of this glycolytic preconditioning mechanism , we provide evidence that the neuroprotective Warburg shift initiated by electron transport impairment originates within astrocytes . To understand the molecular mechanisms linking the Surf1 mutation with the Warburg effect , we examined the hypoxia inducible factor , Hif1a , which promotes aerobic glycolysis by activating expression of PDK and glycolytic genes ( Kim et al . , 2006 ) . Mitochondria act as oxygen sensors which can signal to cytosolic HIF1α through ROS production ( Guzy et al . , 2005 ) . We confirmed that elevated ROS production caused by paraquat treatment was sufficient to stabilize HIF1α in culture ( Figure 5—figure supplement 1J ) . We hypothesized that elevated basal ROS levels observed in Surf1-/- brains ( Figure 4—figure supplement 1C ) reduce mitochondrial respiration in favor of glycolysis through Hif1a activation . Pharmacological inhibition of trauma-induced Hif1a activation has previously been shown to enhance necrotic lesion formation after brain trauma ( Umschweif et al . , 2013 ) , suggesting a neuroprotective role for Hif1a activation in concussive injury . Consistent with previous studies in C . elegans ( Lee et al . , 2010 ) , reducing cytochrome C oxidase activity through cox-5b RNAi increased hif-1 activity as evidenced by induction of the established hif-1 transcriptional target in C . elegans , nhr-57 ( Figure 5—figure supplement 1L ) . Supporting this conserved response , we observed HIF1α protein stabilization in Surf1-/-mouse embryonic fibroblasts ( Figure 5—figure supplement 1K ) as well as transcriptional upregulation of established glycolytic Hif1a targets in Surf1-/- brains , including critical enzymes in the Warburg shunt ( Figure 5A ) . Our data indicates that mild elevation of ROS caused by reduced ETC function , promotes a Hif1a mediated Warburg-like effect in the brain , which reduces the chance of overwhelming mitochondrial respiratory capacity under the strong energetic demand of TBI . To examine the necessity of this Warburg shift in our neuroprotective model , we administered the pharmacological Warburg inhibitor , dichloroacetate ( DCA ) , and observed that Surf1 mutants were no longer protected against trauma-induced neurodegeneration ( Figure 5E and F; Figure 5—figure supplement 1R ) . Consistent with earlier reports of rodent brain injury ( Lazzarino et al . , 2019 ) and SURF1 mutants ( Figure 5A ) , pdp-1 expression was repressed in C . elegans after blunt force injury ( Figure 5—figure supplement 1I ) . To further validate the Warburg shift as a neuroprotective mechanism , we employed genetic means in C . elegans and observed that pdp-1 RNAi suppressed degeneration of dopaminergic neurons after trauma ( Figure 5G and H ) . Albeit to a lesser degree , reduced expression of pdp-1 exclusively in the nervous system was sufficient to promote survival of dopaminergic neurons after blunt force trauma ( Figure 5—figure supplement 2D ) . Thus , modulating PDC activity through the key regulatory factors is sufficient to suppress trauma-induced neurodegeneration in both mice and worms .
The molecular details of TBI and its immediate and long-term consequences can be discovered using a translational animal model approach . To this end , we describe a comparative animal model to identify conserved molecular mechanisms operating in TBI . The worm C . elegans is amenable to high throughput genetic , biochemical , and behavioral studies , while the mouse is used to further uncover and validate conserved genetic targets and molecular mechanisms . These complementary models allowed us to characterize an evolutionary conserved , hypersensitive neuronal subtype to physical insult . Rather than being a stochastic mechanism of progressive tissue degeneration dictated by the type and site of injury , the physiological properties and regulatory mechanisms inherent to different neuronal subtypes likely determine their vulnerability to trauma . This is evidenced by sensitivity of dopaminergic neurons to biomechanical insult in both our randomized , high-frequency trauma model in worms and concussive injury in mice . Furthermore , these complementary models were used to demonstrate that genetic modification of a single gene can suppress degeneration of this neuronal subtype following concussive injury . Pharmacological induction of Parkinsonian phenotypes , including death of dopaminergic neurons in the midbrain , requires highly oxidative drugs that irreversibly inhibit mitochondrial ETC complexes I and III ( Subramaniam and Chesselet , 2013 ) . Yet , our studies demonstrate that ETC complex IV insufficiency protects dopaminergic neurons from trauma-induced degeneration . We attribute these neuroprotective effects to a metabolic shift away from mitochondrial respiration , which minimizes or even abolishes transient oxidative stress in the neuron and suppresses dopaminergic degeneration . Although often thought of as unavoidable toxic byproducts of oxidative phosphorylation , ROS can have beneficial roles in the cell as a signaling molecule ( Guzy et al . , 2005 ) . In our complementary models of blunt force injury , we believe that both toxic and protective mechanisms of ROS are operating , which historically has been difficult to discern . Trauma-induced hyperglycolysis was reported decades ago yet its cellular resolution and pathophysiological role have remained unclear ( Bergsneider et al . , 1997; Carpenter et al . , 2015 ) . The astrocyte-neuron lactate shuttle is a crucial mechanism by which the brain maintains energetic homeostasis . Perturbations to this system have been reported in neurodegenerative disorders and contribute to disease progression ( Mason , 2017 ) . Our data suggests that preemptively repressing mitochondrial respiration initiates a Warburg-like response in astrocytes as an attempt to restore homeostatic energetics . Thus , the injured brain likely induces hyperglycolysis in astrocytes as a protective mechanism to mitigate the oxidative damage resulting from over-activated mitochondrial respiration . Cerebral metabolic preconditioning presents a plausible prophylactic method for individuals at higher risk for TBI . Interestingly , neuroinflammation does not correlate with the rapid emergence of dopaminergic neurodegeneration in our study . However , neuroinflammation and other factors likely contribute to the later stages of this neurodegenerative condition ( McKee et al . , 2015 ) . Therefore , longitudinal studies are required to further characterize disease progression , including neuroinflammation and proteotoxicity , after the initial loss of dopaminergic neurons in the midbrain .
Adult mice of at least 8 weeks of age were used in this study . Surf1+/+ and Surf1-/- ( RRID:MGI:3698949 ) were obtained from Massimo Zeviani ( University of Cambridge ) . Littermates were obtained from Surf1 heterozygous crossing and housed in groups of 5 regardless of phenotype with food and water ad libitum . All mouse studies were approved by the University of Texas Southwestern Medical Center Institutional Animal Care and Use Committee ( Protocol No . 2016–101750 ) and performed in accordance with institutional and federal guidelines . The closed-head traumatic brain injury device consisted of an upright , railed-guided weight drop composed of an aluminum body and equipped with a speed monitor device to measure the velocity of the weight drop when it lands on a brass impactor with a slightly concave nylon tip . The modular weight system consisted of a 50 ml polypropylene conical tube filled with lead buckshot , which weighted 220 g when filled . Before injury , mice were anesthetized with vaporized isoflurane/oxygen in a chamber connected to an inhalation anesthesia system ( Cat . No . 901810; VetEquip ) . Vaporizer settings were set to 3% USP grade isoflurane and 2 . 5 lpm USP grade oxygen . After being anesthetized , mice were placed on a thick memory foam cushion resting on a height-adjustable platform . Once the desired site of impact was localized on the head of mice , the impactor tip was placed directly in contact with that site and the weight was dropped by activating an electronic trigger . After mice were injured , they were placed on their backs to record their righting reflex recovery time and allowed to recover . After recovery , mice were placed back in their cage with food and water ad libitum . Control mice received only anesthesia and were not subjected to impact , but their righting reflex recovery time was also recorded in the same manner as the injured mice . Biomechanical responses of the mouse brain under impact were analyzed using the commercial finite element ( FE ) code ABAQUS ( Dessault Systemes Simulia Corp . , Providence RI ) . More specifically , a 2D geometric model of the sagittal section of the mouse brain tissue was first generated based on a partition scheme defined previously ( MacManus et al . , 2016 ) and the 2D mouse brain image obtained from the experiment . This model also contains a layer of skull of thickness of 1 . 5 mm . The 2D model was meshed with 5067 4-noded quadrilateral elements , which was then converted to 3D by extruding the 2D meshed plane orthogonally that gives a layer thickness of 16 mm . The 3D geometric model of the impactor was constructed based on the measured dimensions shown in ( Figure 2—figure supplement 1A ) , which is meshed with 40 , 320 8-noded brick elements ( C3D8R ) . The final 3D FE model of the brain was modeled with 2 , 546 , 048 8-noded brick elements ( C3D8R ) ( Figure 2—figure supplement 1B ) . The brain tissue was partitioned into six regions ( MacManus et al . , 2016 ) , and 6 sets of viscoelastic parameters from P56 mice ( MacManus et al . , 2017 ) were applied to model the different viscoelastic properties of corresponding regions . Viscoelastic parameters from adult rat ( Finan et al . , 2012 ) had to be used to define the hippocampus due to lack of characterization in the mouse . Values of these parameters are listed in Tables 1 and 2 . To start the simulation , the impactor was placed at 0 . 9 mm above the skull with initial velocity of 4 . 54 m/s . Dynamic responses of the mouse brain were computed with the standard explicit time integration scheme with a time step of 2 ms and total steps of 200 . Interaction between the impactor and skull was modeled by the frictionless and surface-to-surface contact algorithm . Contact between the skull and brain tissue was treated with tie constraint . Five points in cortex , thalamus , substantia nigra , cerebellum , and brainstem were selected to monitor the time history of the principal strain and effective stress during impact ( Figure 2—figure supplement 1B ) . Points for the cortex , thalamus , and substantia nigra were respectively located 2 . 4 mm , 4 . 1 mm , and 8 . 2 mm in vertical distance to the point of impact . Points for the cerebellum and brainstem were located 4 . 7 mm posterior to the three previous points , 4 . 1 mm and 8 . 2 mm in vertical distance to the point of impact . The impactor contained brass and nylon parts , both of which were modeled as isotropic linear elastic material . Material properties of the two are provided in Table 2 . The skull is also modeled as isotropic linear elastic material with Young’s Modulus of 1 GPa and Poisson’s ratio of 0 . 33 ( Unnikrishnan et al . , 2019 ) . Mass density skull is 1710 kg/m3 ( Hua et al . , 2015 ) . Mass density of the brain tissue is 1040 kg/m3 and it is modeled as a viscoelastic material ( Hua et al . , 2015 ) in which stress is evaluated through the introduction of the relaxation function G ( t ) : ( 1 ) τ ( t ) =∫0tG ( t−s ) γ˙ ( s ) ds in which γ is the shear strain and superimposed dot representing the time derivative . The relaxation function G ( t ) takes the form of the Prony series , given as ( 2 ) G ( t ) =G∞+∑i=1NGie−t/τiin which t is time , G∞ is the long-term relaxation modulus , τi is the characteristic relaxation time and Gi is the corresponding relaxation modulus . The number of the terms N included in this series depends on fitting of the model predictions to the experiments ( Finan et al . , 2012 ) . Mice were perfused with cold 1x PBS and 4% paraformaldehyde . Brains were harvested and sectioned to a thickness of 40–50 µm with the VF-310-0Z Compresstome Vibrating Microtome ( Precisionary Instruments , Greenville , NC ) . All staining was performed on free floating brain sections on a 24-well plate . Incubation in primary antibodies was performed in blocking solution ( 10% normal donkey serum ) and 2% triton at 4°C for 48 hr . Primary antibodies used were: rabbit anti-Iba1 ( Cat . No . 019–19741; Wako Chemicals; RRID:AB_839504 ) at 1:200 , rat anti-CD68 ( Cat . No . MCA1957; AbD Serotec; RRID:AB_322219 ) at 1:250 , chicken anti-tyrosine hydroxylase ( Cat . No . TYH; Aves Labs Inc; RRID:AB_10013440 ) at 1:1000; rabbit anti-cleaved caspase-3 ( Cat . No . 9661S; Cell Signaling Technologies; RRID:AB_2341188 ) at 1:400 . Species-specific secondary antibodies ( Jackson ImmunoResearch/Invitrogen ) were used at a concentration of 1:250 at 4°C for 24 hr . Nissl staining ( Neurotrace 530/615; Cat . No . N21482; Invitrogen ) was performed according to manufacturer’s instructions . The metabolic phenotyping experiments were run by the Metabolic Phenotyping Core of UTSW Medical Center in metabolic cages using a TSE Systems , Inc ( Chesterfield , MO ) indirect calorimetry system . Briefly , the day following a single concussive brain injury , mice were singly housed in shoebox-sized cages with wood chip bedding to acclimate for 5 days , followed by 5 days of experimental recording . O2 consumption and CO2 production were measured to determine energy expenditure and respiratory exchange ratio . Overall physical activity was determined through x , y , and z beam breaks . Body heat production , food intake , and water consumption were also measured . Data was normalized according to body weight . Metabolomics profiling was performed by the Metabolomics Facility at the Children’s Medical Center Research Institute at UT Southwestern . Briefly , after concussive brain injury , brains were extracted , midbrains were dissected and flash frozen in liquid nitrogen . Tissue was then homogenized in a Precellys homogenizer at 5000 rpm for 30 s ( 2x ) with a 20 s pause at 4°C . The metabolome was then extracted with an 80% ice cold solution of methanol and ddH2O . The metabolite containing supernatant was transferred into a clean centrifuge tube and desiccated for 12 hr . Metabolome profiling was done via LC-MS , normalized to total ion current and SIMCA analyzed . Tissue or cells were homogenized in a Precellys homogenizer at 4 °C with ceramic beads at 5 , 000 rpm for 30 s ( 2x ) with a 20 s pause using RIPA lysis buffer ( 150 mM NaCl , 5 mM EDTA , 50 mM Tris , 1% NP-40 , 0 . 5% sodium deoxycholate , 1% sodium dodecyl sulfate ( SDS ) , final pH 8 ) supplemented with cOmplete EDTA-free mini-protease inhibitor cocktail ( Cat . No . 11836170001; Roche ) . Extracts were created in the presence of the 1% SDS detergent to ensure protein linearization and inactivation of phosphatases ( Stinson , 1984 ) . Protein samples were centrifuged for 10 m at 10 , 000 rcf at 4 °C and the supernatant was used for protein quantification via Pierce BCA protein assay ( Cat . No . 23225; Thermo Fisher Scientific ) . All sample concentrations were standardized and diluted in sample buffer . Samples were boiled at 90°C for 10 m , resolved by SDS-PAGE , transferred to nitrocellulose membranes and subject to western blot analysis . SDS-PAGE gels ( 10% ) were prepared the night before and stored at 4°C . The EZ-Run pre-stained protein ladder ( Cat . No . BP3603500; Thermo Fisher Scientific ) was loaded and electrophoresis was performed at 100 V until the 40 kDa protein standard reached the bottom of the gel . All antibodies were prepared in 5% BSA/PBST . Mouse anti-αTubulin ( Cat . No . T6074; Sigma; RRID:AB_477582 ) was used at 1:10000 , rabbit anti-phospho-PDHEα1 ( Cat . No . 31866S; Cell Signaling Technologies; RRID:AB_2799014 ) was used at 1:1000 , and rabbit anti-HIF1α ( Cat . No . 3716S; Cell Signaling Technologies; RRID:AB_2116962 ) was used at 1:1000 . Western blots were quantified using Image Studio software ( LI-COR Biosciences , Lincoln , NE ) . Quantified bands of interest were standardized based on band signal intensity of αTubulin . ATP levels were measured with the CellTiter-Glo luminescence assay ( Cat . No . G7570; Promega ) according to manufacturer’s instructions . Brains were harvested , enzymatically dissociated with papain , and passed through a cell strainer ( 45 µm ) . Myelin was removed with the Debris Removal Solution ( Cat . No . 130-109-398; Miltenyi Biotec ) . Freshly dissociated cells ( 100 , 000 cells ) were resuspended in 100 µl of 1x PBS and transferred to a 96-well plate . A total of 100 µl of CellTiter-Glo reagent was added to make a final volume of 200 µl . The 96-well plate was placed on an orbital shaker for 2 m to induce cell lysis . After a 10 m incubation at room temperature , the plate was read for luminescence ( 0 . 25 s per well ) in a CLARIOstar Plus microplate reader ( BMG LABTECH , Ortenberg , Germany ) . Mitochondrial superoxide in mouse brain single cell suspension was measured with the MitoSOX Red superoxide indicator ( Cat . No . M36008; Invitrogen ) according to manufacturer’s instructions . Briefly , brains were harvested , enzymatically dissociated with papain , and passed through a cell strainer ( 45 μm ) . Myelin was removed with the Debris Removal Solution ( Cat . No . 130-109-398; Miltenyi Biotec ) . Freshly dissociated cells were then incubated in a 5 μM MitoSOX solution in a 96-well plate for 10 m at 37°C . Each well was measured for fluorescence ( Ex/Em: 510/580 nm ) . Reactive oxygen species were measured in C . elegans with the Amplex Red Hydrogen Peroxide/Peroxidase Assay Kit ( Cat . No . A22188; Invitrogen ) . Worms were grown to Day one adults , transferred to a 2 ml centrifuge tube containing 500 µl of M9 , and submitted to traumatic injury . Immediately after traumatic injury , equal number of worms ( 50 µl total M9 volume ) were transferred to a 96-well plate ( three technical repeats ) and 50 µl of 100 µM Amplex Red reagent/0 . 2 U/ml horseradish peroxidase was added to a final concentration of 50 µM Amplex Red reagent/0 . 1 U/ml horseradish peroxidase . Each well was measured for fluorescence ( Ex/Em: 530–560/~590 nm ) every 92 s for 1 . 5 hr on a CLARIOstar Plus microplate reader ( BMG LABTECH ) . C . elegans were maintained as previously described ( Brenner , 1974 ) . Worm strains were expanded on nematode growth media ( NGM ) plates supplemented with Escherichia coli OP50 . Worm strains that were used for experimental purposes were grown on NGM plates supplemented with HT115 E . coli at 20°C ( see RNAi Administration ) . For temperature sensitive strains , worms were grown at 25°C . The following C . elegans strains were generated in our laboratory and used in this study: PMD13 ( egIs1[dat-1p::GFP]; rrf-3 ( b26 ) ; fem-1 ( hc17 ) ) , PMD74 ( unc-119p::TOM20::KillerRed; egIs1[dat-1p::GFP]; rrf-3 ( b26 ) ; fem-1 ( hc17 ) ) , PMD63 ( egIs1[dat-1p::GFP]; rrf-3 ( b26 ) ; fem-1 ( hc17 ) ; uIs69 [pCFJ90 ( myo-2p::mCherry ) + unc-119p::sid-1]; sid-1 ( pk3321 ) ) . Trauma was administered to C . elegans as previously described ( Egge et al . , 2021 ) . Briefly , age-synchronized worms were grown on empty vector ( EV ) or the respective RNAi until day 1 of adulthood . Temperature-restrictive strains were grown at 25°C to avoid generation of progeny . Worms were rinsed off growth plates in liquid M9 buffer ( M9 ) and pelleted by centrifugation at 1000 x g for 30 s . 100 µl worm pellets were then transferred to 2 ml Precellys tubes ( Cat . No . 02-682-556; Thermo Fisher Scientific ) in a total volume of 500 µl of M9 . Worms were subject to high-frequency , multidirectional agitation in a Precellys Evolution homogenizer ( Cat . No . P000062-PEVO0-A . 0; Bertin Instruments ) for 16 s at 8600 rpm . Worms were pelleted at 1000 x g for 30 s and transferred to recovery plates containing the respective RNAi or EV on which they were cultured prior to trauma . Control , non-injured worms were suspended in M9 for comparable lengths of time and then transferred to recovery plates without having received trauma . Worms were paralyzed with 1 mM levamisole and mounted on microscope slides with M9 buffer . Brain sections were mounted in microscope slides with Fluoromount-G with DAPI ( Cat . No . 00-4959-52; Invitrogen ) . Confocal images were collected with a Leica SP8 confocal microscope equipped with one photomultiplier tube , two super-sensitive hybrid HyD detectors , and highly stable lasers ( UV/405 nm DMOD compact , 488 nm , 552 nm , 638 nm ) . The following Leica PL APO CS2 objectives were used to collect images: air 10x/0 . 40 NA , oil 40x/1 . 30 NA , oil 63x/1 . 40 NA . Super-resolution imaging was achieved with the Leica LIGHTING module built into the Leica LAS X software . Brain parenchyma images were collected as Z-stacks with 1 µm steps as follows: midbrain micrographs 20 µm/10x objective/5 tile Z-stacks , cortex micrographs 10 µm/40x objective/5 tile Z-stacks , hippocampus micrographs 20 µm/10x objective/8 tile Z-stacks , thalamus micrographs 5 µm/40x objective/2 tile Z-stacks . Number of tiles to be collected was determined by completely covering the brain structure of interest ( i . e . midbrain 1x5 tiles , hippocampus 2x4 tiles ) . Tiled images were stitched with the Leica LAS X software . Cell culture micrographs were collected from cells grown on glass coverslips with a total of 10 fields collected with a 40x objective for image processing and analysis . Brain parenchyma images are displayed as maximal intensity projections of Z-stacks . Cell culture images are displayed as single plane micrographs . All confocal images were processed and analyzed with ImageJ ( NIH , Bethesda , MD; RRID:SCR_003070 ) . Motor deficits in mice were measured by Rotarod ( Cat . No . 76–0770; Harvard Apparatus ) with a rod diameter of 3 cm and a rod height of 20 cm . Animals were not trained on a rotating rod but rather exposed to a static Rotarod machine 2 days before testing for 2 m , three times daily . For testing , animals were placed on the rod with accelerating speed ( 0–24 rpm in 120 s ) for all experiments . Latency to fall was recorded electronically in seconds by the apparatus and values were averaged per group for reporting . The clock was stopped if an animal held to the rod on two consecutive rotations and if the animal failed to fall after 120 s . Animals were returned to their cage after each trial . Ambulation was measured via laser beam breaks for 5 days in mice housed individually in cages equipped with laser beams and other probes meant to determine changes in metabolism . Four biological repeats of age-matched adult Surf1+/+ and Surf1-/- brains were collected for each of the following conditions: Surf1+/+ uninjured , Surf1+/+ injured , Surf1-/- uninjured , Surf1-/- injured at time-points 2 hr post-injury and 7 days post-injury . Mouse brains were harvested in TRIzol ( Cat . No . 15596018; Thermo Fisher Scientific ) , flash-frozen in liquid nitrogen , and stored at −80°C . Brains were triturated with syringes and freeze-thawed three times , followed by a chloroform/isopropanol extraction process . The RNA pellets were washed twice with ethanol , air-dried , and reconstituted in 20 μl ddH2O . RNA quality ( 260/280 , 260/230 ratio ) and concentration was determined using the DS-11 FX+ spectrophotometer ( DeNovix , Inc , Wilmington , DE ) . Quality control , mRNA purification , and paired-end 150 bp Illumina sequencing were performed by Novogene ( Sacramento , CA ) . mRNA was enriched using oligo ( dT ) beads , randomly fragmented in fragmentation buffer , and reverse transcribed to cDNA using random hexamers . Following first-strand synthesis , Illumina sysnthesis buffer was added with dNTPs , RNase H , and E . coli polymerase I to synthesize the second strand by nick-translation . The cDNA library was purified , underwent terminal repair , A-tailing , and ligation of adapters before PCR enrichment . The cDNA library concentration was quantified with a Qubit 2 . 0 fluorometer ( ThermoFisher Scientific ) and sized with an Agilent 2100 Bioanalyzer ( Agilent , Santa Clara , CA ) . RNAseq statistical analysis was performed using CLC software ( version 9 . 0 , CLC Bio , Aarhus , Denmark ) . Data is presented as reads per kilobase million ( RPKM ) with values normalized to control levels and made relative to 1 . Single nuclei isolation methods were adapted from 10x Genomics protocols CG000212 Rev B and CG000124 Rev D . Surf1+/+ and Surf1-/- brains were harvested under cold sterile 1x Dulbecco’s PBS ( without Ca2+ and Mg2+ ) . Immediately after harvesting , the cortex , olfactory bulbs and cerebellum were removed on an ice block and discarded . Brains were then triturated and lysed using the Nuclei PURE Prep Isolation Kit ( Cat . No . NUC-201 , Sigma ) . Briefly , brains were homogenized under cold Nuclei PURE Lysis Buffer and mechanically triturated with sterile pipettes and pipette tips of decreasing sizes until creating a uniform suspension . Suspensions were then filtered through 70 μm , 40 μm , and 30 μm cell strainers to remove cell debris . Myelin removal proceeded using the Debris Removal Solution ( Cat . No . 130-109-398 , Miltenyi ) according to manufacturer’s instructions . Single nuclei were isolated and purified via density gradient centrifugation with the Nuclei PURE sucrose cushion solution ( Cat . No . NUC-201 , Sigma ) at a centrifugation speed of 13 , 000 rcf . Nuclei was resuspended in 1x Dulbecco’s PBS supplemented with 1% bovine serum albumin and 20 U/μl RNase inhibitor ( Cat . No . AM2694 , Invitrogen ) . Single nuclei suspensions were submitted to the Next Generation Sequencing Core at UT Southwestern Medical Center for sequencing and library preparation . Briefly , nuclei were loaded with Single Cell 3’ Gel Beads into a Next GEM Chip G and run on the Chromium Controller . GEM emulsions were incubated and then broken . Silane magnetic beads were used to clean up GEM reaction mixture . Read one primer sequence was added during incubation and full-length , barcoded cDNA was then amplified by PCR after cleanup . Sample size was checked on the Agilent Tapestation 4200 using the DNAHS 5000 tape and concentration was determined by the Qubit 4 . 0 Fluorimeter ( ThermoFisher ) using the DNA HS assay . Samples were enzymatically fragmented and underwent size selection before proceeding to library construction . During library preparation , Read two primer sequence , sample index , and both Illumina adapter sequences were added . Subsequently , samples were cleaned up using Ampure XP beads and post library preparation quality control was performed using the DNA 1000 tape on the Agilent Tapestation 4200 . Final concentration was ascertained using the Qubit DNA HS assay . Samples were loaded at 1 . 6 pM and run on the Illumina NextSeq500 High Output Flowcell using V2 . 5 chemistry . Run configuration was 28x98x8 . Raw gene counts were obtained by aligning the FASTQ files to Mus musculus mm10 as reference genome using CellRanger Software ( v5 . 0 . 0 ) and then analyzed using Seurat-3 . 2 . 0 ( Stuart et al . , 2019 ) . Genes detected in <0 . 2% of the nuclei and from mitochondrial and sex chromosomes ( X and Y ) were filtered out . Nuclei that had >12 , 000 UMI , <300 genes , and/or >12% mitochondrial content were further excluded . After quality control , 5140 nuclei ( primary dataset ) were further analyzed for their gene expression profiles . Post filtering , the expression values were log-normalized , scaled with a factor of 10 , 000 and regressed to covariates ( percent mitochondrial content and number of genes per nuclei ) , with Seurat's SCTransform method and integrated . The nuclei were then assessed by Principal Component Analysis ( PCA ) dimensionality reduction , followed by shared nearest neighbor ( SNN ) modularity optimization ( Louvain algorithm ) based clustering algorithm to identify the clusters . The clusters were visualized using the Uniform Manifold Approximation and Projection ( UMAP ) dimensional reduction technique . Cell types were assigned to clusters based on enrichment of marker genes from top expressed genes as follows: Excitatory neurons ( Rbfox3 ) , GABAergic neurons ( Gad1 , Gad2 ) , Dendrocytes ( Cacng4 ) , Astrocytes ( Gaj1 , Aqp4 ) , and Dopaminergic neurons ( Slc6a3 , Slc18a2 , Ddc ) . Astrocyte and neuronal clusters were subclustered ( secondary dataset ) and then exclusively studied using the same approach as described above . Pairwise differential gene expression analysis tests were performed with Wilcox test within Seurat . The expression of specific genes within clusters were visualized using the DotPlot function in Seurat . The average expression of these genes in the dot plot was calculated by using the normalized gene counts , natural-log transforming them using log1p , and then scaling them using a z-transformation . Worm strains were grown on HT115 E . coli harboring RNAi constructs from either the Ahringer or Vidal RNAi libraries ( Rual et al . , 2004 ) . The L4440 empty vector ( EV ) RNAi construct was used for control treatments . RNAi strains were grown in small cultures before inoculating larger cultures in Terrific broth ( TB ) and grown for 15 hr on an orbital shaker at 37°C . After 15 hr , cultures were treated with 1 mM IPTG and incubated for an additional 4 hr at 37°C to induce expression . Cultures were then centrifuged at 4000 x g and bacterial pellets were re-suspended to specific concentrations in TB before being spread on 100 mm NGM plates containing a final concentration of 1 mM IPTG and 0 . 1 mg/ml carbenicillin . Optical density ( OD600 ) of RNAi-producing bacterial cultures was used to standardized and equalize cell number for multiple RNAi construct combinations . For titrated single RNAi construct treatments , induced HT115 bacteria expressing a single RNAi construct was diluted with L4440 empty vector HT115 bacteria . Large-particle flow cytometry was performed as previously described ( Egge et al . , 2019 ) . Briefly , flow cytometry of C . elegans was performed on a COPAS FP-250 flow cytometer ( Union Biometrica , Holliston , MA ) and an automated sample introduction system LP Sampler ( Union Biometrica ) fitted with 96-well plates . M9 buffer was utilized as the sample solution for worm flow . The COPAS GP sheath reagent ( PN: 200-5070-100 , Union Biometrica ) was used as sheath solution . Prior to every run , laser power and flow rates are calibrated by use of GP control particles ( Cat . No . 310-5071-001; COPAS Biosorter , Union Biometrica ) as recommended by the manufacturer . From experiment to experiment , we maintain a consistent PTM laser power for each respective fluorophore . Flow data was collected in FlowPilot software ( Union Biometrica ) . Details regarding analysis and normalization of data obtained from large-particle cytometry are found under the ‘Statistical analysis’ methods section . HEK293t cells were obtained from ATCC ( Cat . No . CRL-3216; RRID:CVCL_0063 ) and cultured in DMEM/F-12 ( Cat . No . 11320033; Gibco ) supplemented with 10% Fetal Bovine Serum and 1x Penicillin-Streptomycin ( Cat . No . P4333; Sigma ) . Mouse embryonic fibroblasts ( MEF ) were obtained via timed mating by detection of a vaginal plug was used to obtain embryonic day 13 . 5 embryos . Embryos were dissected in cold 1x PBS to remove the head and all inner organs , leaving only the carcass . After finely mincing the carcass in cold 1x PBS , the pieces were allowed to settle , and PBS was aspirated . The tissue was then enzymatically dissociated with 1 ml of 0 . 25% trypsin/EDTA at 37°C for 5 m . The tissue was then mechanically dissociated using a 1 ml pipette tip until homogeneous and passed through a nylon cell strainer ( 45 μm ) . Cells were cultured in DMEM-Hi glucose ( Sigma-Aldrich; 4 . 5 g/L glucose , supplemented with L-glutamine and sodium pyruvate ) , 10% fetal bovine serum ( heat inactivated ) , and supplemented with penicillin streptavidin . All cultures were confirmed to be negative for mycoplasma with the MycoAlert Mycoplasma Detection Kit ( Cat . No . LT07-218; Lonza ) . Surf1+/+ and Surf1-/- mouse embryonic fibroblasts were plated on a clear bottom 96-well plate ( 100 , 000 cells/well ) . Each well was treated with 100 ng/ml of phorbol myristate acetate ( PMA ) for 1 hr . The PMA containing medium was then replaced with fresh medium and the basal oxygen consumption rate was measured with the Cayman Chemical Oxygen Consumption Rate Assay Kit ( Cat . No . 600800 ) according to manufacturer’s instructions . Mouse embryonic fibroblasts ( MEF ) were cultured in a 96-well plate . Each well contained equal MEF numbers cultured in DMEM/F-12 ( Cat . No . 11320033; Gibco ) supplemented with 10% Fetal Bovine Serum . DMEM/F-12 used contains phenol red as a pH indicator with pH ranges of 6 . 8 ( yellow ) to 8 ( red ) . Cells were cultured for 5 d in a cell culture incubator , at 37°C with 5% CO2 . After 5 days , the Optical Density was immediately measured in a CLARIOstar Plus spectrometer ( BMG LABTECH ) by performing a spectral scan between 220 nm and 1000 nm with a 1 nm resolution . Data was normalized to 1x105 cells for analysis . The maximum LDH activity was measured with the Pierce LDH Cytotoxicity Assay Kit ( Cat . No . 88953; Thermo Fisher Scientific ) according to manufacturer instructions . Briefly , 20 , 000 MEF were cultured overnight in a flat bottom , 96-well plate at 37 °C , 5% CO2 . Cells were lysed with lysis buffer provided in the kit and 50 μl of supernatant was transferred to a fresh 96-well flat bottom plate . The kit’s proprietary reaction solution was added to the supernatant and incubated at room temperature for 30 m . After stopping the reaction , absorbance was measured at 490 nm and 680 nm in a CLARIOstar Plus microplate reader ( BMG LABTECH ) . LDH activity was determined by subtracting the 680 nm from the 490 nm absorbance values . All statistical analyses were performed using Prism eight software ( GraphPad , San Diego , CA ) unless noted . Student’s t-test was used to compare means between two normal populations . Mann-Whitney U test was used to compare differences in the dependent variable between two groups . Post-hoc analysis performed after ANOVA included Dunnett’s multiple comparison ( to compare means from several experimental groups against a control group mean ) . Tukey’s multiple comparisons test ( to compare all possible pairs of means ) . Statistical analysis of large-particle flow cytometry data was performed in Prism and Excel ( Egge et al . , 2021 ) . To evaluate large worm numbers across biological repeats , error propagation was performed according to the general formula:δR= ( ∂R∂X*δX ) 2+ ( ∂R∂Y*δY ) 2+…where δR , the total error within each independent repeat , is a function of each independent variable ( X , Y … ) . For the particular case of error propagation for the Dopaminergic GFP Index within each of multiple biological repeats , error is further added in quadrature and the function becomes:δR=∑X ( δX1-EV ) 2where δX represents the error and X contains all biological repeats . EV is a constant determined by the relative loss of fluorescence observed in injured worms grown on EV bacteria during that experiment . Lastly , the total error within each repeat was added to the error between repeats in quadrature . | Concussion is a type of traumatic brain injury that results from a sudden blow or jolt to the head . Symptoms can include a passing headache , dizziness , confusion or sensitivity to light , but experiencing multiple concussions can have drastic repercussions in later life . Studies of professional athletes have shown that those who experience one or more concussions are prone to developing Alzheimer’s and Parkinson’s disease , two well-known neurodegenerative diseases . Both conditions involve the progressive loss or breakdown of nerve cells , called neurons . But exactly how this so-called neurodegeneration of brain cells stems from the original , physical injury remains unclear . Head trauma may cause damage to the structural support of a cell or disrupt the flow of electrical impulses through neurons . Energy use and production in damaged cells could shift into overdrive to repair the damage . The chemical properties of different types of brain cells could also make some more vulnerable to trauma than others . Besides neurons , star-shaped support cells in the brain called astrocytes , which may have some protective ability , could also be affected . To investigate which cells may be more susceptible to traumatic injuries , Solano Fonseca et al . modelled the impacts of concussion-like head trauma in roundworms ( C . elegans ) and mice . In both animals , one type of neuron was extremely vulnerable to cell death after trauma . Neurons that release dopamine , a chemical involved in cell-to-cell communication and the brain’s reward system , showed signs of cell damage and deteriorated after injury . Dopaminergic cells , as these cells are called , are involved in motor coordination , and the loss of dopaminergic cells has been linked to both Alzheimer’s and Parkinson’s disease . Astrocytes , however , had a role in reducing the death of dopaminergic neurons after trauma . In experiments , astrocytes appeared to restore the balance of energy production to meet the increased energy demands of impacted neurons . Single-cell analyses showed that genes involved in metabolism were switched on in astrocytes to produce energy via an alternative pathway . This energetic shift facilitated via astrocytes may help mitigate against some damage to dopamine-producing neurons after trauma , reducing cell death . This work furthers our understanding of cellular changes in the concussed brain . More research will be required to better characterise how this immediate trauma to cells , and the subsequent loss of dopaminergic neurons , impacts brain health long-term . Efforts to design effective therapies to slow or reverse these changes could then follow . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2021 | Glycolytic preconditioning in astrocytes mitigates trauma-induced neurodegeneration |
Restoring somatosensory feedback to people with limb amputations is crucial to improve prosthetic control . Multiple studies have demonstrated that peripheral nerve stimulation and targeted reinnervation can provide somatotopically relevant sensory feedback . While effective , the surgical procedures required for these techniques remain a major barrier to translatability . Here , we demonstrate in four people with upper-limb amputation that epidural spinal cord stimulation ( SCS ) , a common clinical technique to treat pain , evoked somatosensory percepts that were perceived as emanating from the missing arm and hand . Over up to 29 days , stimulation evoked sensory percepts in consistent locations in the missing hand regardless of time since amputation or level of amputation . Evoked sensations were occasionally described as naturalistic ( e . g . touch or pressure ) , but were often paresthesias . Increasing stimulus amplitude increased the perceived intensity linearly , without increasing area of the sensations . These results demonstrate the potential of SCS as a tool to restore somatosensation after amputations .
Individuals with amputations consistently state that the lack of somatosensory feedback from their prosthetic device is a significant problem that limits its utility ( Cordella et al . , 2016 ) and is often a primary cause of prosthesis abandonment ( Biddiss and Chau , 2007; Wijk and Carlsson , 2015 ) . In the case of upper-limb amputations , the absence of somatosensory feedback particularly affects the ability to generate the finely controlled movements that are required for object manipulation ( Cordella et al . , 2016; Lundborg et al . , 1998; Pylatiuk et al . , 2007; Wijk and Carlsson , 2015 ) . Although sophisticated myoelectric prostheses with multiple degrees of freedom ( Belter et al . , 2013 ) are becoming increasingly available , their potential is limited because they provide little or no somatosensory feedback ( Biddiss et al . , 2007; Biddiss and Chau , 2007; FAAOP et al . , 2015; Peerdeman et al . , 2011 ) . In fact , body-powered devices are often preferred because of the feedback they provide through their harness and cable system ( Huang et al . , 2001; Stark and LeBlanc , 2004; Uellendahl , 2000; Williams , 2011 ) . Partially addressing this limitation , advanced robotic prosthetic arms have been designed with embedded sensors that could be harnessed to provide somatosensory signals to a neural interface ( Cipriani et al . , 2011; Perry et al . , 2018; Saudabayev and Varol , 2015 ) . Thus , developing a robust and intuitive means to deliver somatosensory information to the nervous system is an important endeavor to ensure the adoption and use of the latest advancements in prosthetics . Several research groups have explored the potential of peripheral nerve stimulation to provide sensory feedback to people with amputation and examined the effects of feedback on prosthetic control . Sensory restoration has been achieved using a variety of neural interfaces including epineural cuff electrodes like the spiral cuff ( Ortiz-Catalan et al . , 2020; Ortiz-Catalan et al . , 2014 ) and flat interface nerve electrode ( Tan et al . , 2014 ) or microelectrodes that penetrate the epineurium , such as the longitudinal intrafascicular electrode ( Horch et al . , 2011 ) , transverse intrafascicular multichannel electrode ( Raspopovic et al . , 2014 ) , or Utah slant array ( Davis et al . , 2016 ) . Targeted sensory reinnervation is another approach that can allow vibrotactile or electrotactile feedback on the residual limb to be perceived as emanating from the missing limb ( Marasco et al . , 2011; Marasco et al . , 2009 ) . This is achieved by first surgically redirecting nerves that formerly innervated the missing limb to patches of skin on the residual limb or elsewhere , and then providing electrical or mechanical stimulation to the newly innervated site ( Kuiken et al . , 2007a; Kuiken et al . , 2007b ) . These approaches can evoke focal sensations that are perceived to emanate from the upper-limb , even decades after injury , and can improve the control of prosthetic limbs . However , all of these approaches involve specialized electrodes and/or surgeries that are not part of common surgical practice . Further , these approaches often target nerves in the distal limb , which could limit their use in people with proximal amputations such as shoulder disarticulations . Spinal cord stimulation ( SCS ) systems are an FDA-approved , commercially available technology that could potentially be used to restore somatosensation . SCS leads are currently implanted in approximately 50 , 000 patients every year in the USA to treat chronic back and limb pain ( Kumar and Rizvi , 2014 ) . The standard clinical approach begins with a week-long trial phase with temporarily implanted leads , and if patients experience pain relief , permanent implantation occurs during an hour-long follow-up procedure . For the trial phase , SCS leads are inserted percutaneously into the epidural space on the dorsal side of the spinal cord via a minimally invasive , outpatient procedure ( Kinfe et al . , 2014 ) . Clinically effective stimulation parameters typically evoke paresthesias ( i . e . sensation of electrical buzzing ) that are perceived to be co-located with the region of pain . SCS leads are usually placed over the dorsal columns along the midline of the spinal cord which limits the evoked paresthesias to the proximal areas of the trunk and limbs . However , recent studies have demonstrated that stimulation of lateral structures in the spinal cord and spinal roots can evoke paresthesias that selectively emanate from the distal regions of the body ( Deer et al . , 2013; Harrison et al . , 2018; Liem et al . , 2013; Lynch et al . , 2011 ) , likely by stimulating the same sensory afferent neurons that are targeted by peripheral nerve stimulation for prosthetic applications ( Capogrosso et al . , 2013 ) . As such , these devices provide an attractive option for widespread deployment of a neuroprosthesis than can evoke somatosensory percepts from distal aspects of the amputated limb , including the hand and fingers . In this study , we implanted percutaneous SCS leads into the lateral epidural space of four people with upper-limb amputations and characterized the sensations evoked when the cervical spinal cord and spinal roots were stimulated . The goals of the study were to demonstrate the feasibility of lateral SCS to restore somatosensation and to guide technical development for future studies that will include full implantation of SCS leads and stimulators . In all subjects , lateral SCS evoked sensations that were perceived to emanate from the missing limb , including focal regions in the hand , regardless of the level of amputation ( trans-radial to shoulder disarticulation ) . These sensations were stable throughout the 29 day testing period and showed only minor changes in area and location . Additionally , in some cases , it was possible to evoke naturalistic , rather than paresthetic sensations , though the incidence of naturalistic sensations varied by subject . Considering these results along with the extensive clinical use of SCS , this approach to somatosensory restoration could be one that is beneficial to a diverse population of amputees , including those with proximal amputations . Further , these percutaneously implanted SCS devices are a useful tool for the development of somatosensory neuroprosthetic systems , especially for research projects that focus on advanced prosthetic control but have not developed their own technologies and techniques for restoring sensory feedback .
Three SCS leads were implanted in the cervical epidural space in each of four individuals with upper-limb amputation ( Table 1 ) . The percutaneous implant was maintained for the full 29 day duration of the study for all subjects except Subject 2 , who requested removal of the leads after two weeks due to personal factors and discomfort from caudal migration of one of the leads . We stimulated with both monopolar and multipolar electrode configurations . Stimulus amplitudes , frequencies , and pulse widths ranged 0–6 mA , 1–300 Hz , and 50–1000 µs , respectively . In all four subjects , epidural SCS evoked sensory percepts in distinct regions of the missing limb including the fingers , palm , and forearm . While some sensory percepts were diffuse and covered the entire missing limb , other percepts were localized to a very specific area , such as the ulnar region of the palm or wrist , or individual fingers . Figure 1 shows select representative percepts for all subjects; for an interactive visualization of all evoked percepts localized to the missing limb , see Supplementary file 1 . In Subjects 1 and 2 , only multipolar stimulation evoked sensory percepts that were localized to focal regions of the missing hand and fingers ( Figure 1—figure supplement 1 ) . In Subjects 2 and 3 , most percepts were accompanied by a sensation on the residual limb . This was the case even when there was a percept that was focally restricted to a distal region of the missing limb , such as a finger ( e . g . purple thumb/shoulder sensation in Subject 2 , Figure 1 ) . These additional proximal sensations emanated predominantly from the end of the residual limb . The incidence rate of such simultaneous sensations varied from 0% and 8% for Subjects 1 and 4 to 92% and 98% for Subjects 2 and 3 . There were also a subset of mono- and multi-polar electrodes that evoked sensations bilaterally or only in the contralateral intact limb ( 14 . 3% and 15 . 4% of all electrodes that generated a sensation across all subjects; n = 447 ) . While these sensations might be useful in a neuroprosthesis for people with bilateral amputation , they were not a focus of this study and were not included in any of the analyses presented here . We sought to determine if stimulation of specific regions of the spinal cord consistently evoked sensations that were perceived to emanate from specific regions of the arm and hard across subjects . We hypothesized that the location of the perceived sensation would be driven by the location of the cathodic electrode with respect to the spinal cord according to expected dermatomes . Figure 2 shows the proportion of sensory percepts in a specific anatomical region ( dashed lines , Figure 2A ) evoked by electrodes situated at each spinal level ( Figure 2B , C ) . There were notable similarities between the perceived locations and dermatomes ( Foerster , 1933; Lee et al . , 2008 ) , however there was considerable inter-subject variability and sensations were not evoked in all regions of the hand in all subjects ( Figure 2C ) . For example , sensations reported in the thumb were predominantly evoked by electrodes located near the C6 root in Subjects 2 and 4 ( 0% , 67% , 26% , and 50% for Subjects 1–4 respectively ) . Similarly , a high proportion of the percepts localized to the 2nd and 3rd digits were evoked by electrodes near the C7 root in Subjects 2 and 3 ( 0% , 50% , 66% , and 23% , for Subjects 1–4 , respectively ) . However , sensations in 4th and 5th digits ( within the C8 dermatome ) were evoked predominantly by electrodes near the C7 root in Subjects 2 and 3 ( 0% , 75% , 78% , and 23% in Subjects 1–4 , respectively ) . Interestingly , for Subject 4 , electrodes near the C6 root produced most of the percepts in the hand ( 2nd and 3rd digits: 52% , 4th and 5th digits: 45% ) . Moreover , almost all the electrodes in Subject 1 , including those that evoked focal percepts in the fingers and palm , were located near the T1 root . Overall , these results demonstrate that , while there was some consistency between the locations of stimulation and dermatomes , there was considerable inter-subject variability in many of the evoked sensations . We asked the subjects to describe the evoked sensations using a set of words provided in a predefined list ( Table 2 ) . This allowed us to standardize the descriptions of the percepts across subjects and put them in context of previous research ( Heming et al . , 2010; Lenz et al . , 1993 ) . Subjects could report more than one modality simultaneously . All sensations that had an ‘electrical tingle’ , ’ pins and needles’ , ‘sharp’ , or ‘tickle’ component were considered paresthetic . If these sensations also included descriptors for mechanical , movement , or temperature modalities , they were considered mixed modality sensations . Sensations that did not include any paresthetic descriptors were considered naturalistic . The unique combinations of percept descriptors used by each subject along with the fraction of naturalistic , paresthetic , and mixed modality sensations are shown in Figure 3 . For Subjects 1 , 2 , and 4 , most sensory percepts were either paresthetic or of mixed modality ( 90 . 2% , 75 . 2% , and 96 . 5% , respectively ) . Subject 1 reported 74 . 2% of these percepts as purely paresthetic , whereas only a small fraction of these percepts were reported as purely paresthetic by Subjects 2 and 4 ( 4 . 1% and 0 . 3% ) . For Subject 2 the evoked percept was most frequently described as tingle-pressure and for Subject 4 the evoked percept was most frequently described as tingle-pressure-vibration . Subject 3 predominantly reported naturalistic sensations ( 79 . 9% ) with most of those percepts described as pure vibration . In fact , for this subject , 80% of all evoked percepts contained a ‘vibration’ component , and most mixed modality percepts ( 97 . 8% ) were described as tingle-vibration with only one instance of a purely paresthetic percept . More naturalistic modalities , like ‘touch’ and ‘pressure’ , were elicited to varying degrees among the subjects ( 0 . 5% , 60 . 5% , 25 . 2% and 75 . 8% of unique stimulation parameter combinations for Subjects 1–4 , respectively ) . Interestingly , sensations described as purely touch or pressure were reported in 8 . 25% and 19 . 5% of all evoked percepts in Subjects 2 and 3 , respectively . Otherwise , these naturalistic sensations were commonly accompanied by a paresthesia , as particularly seen in Subject 4 . Percepts containing a dynamic ( ‘movement’ ) component that may be described as proprioceptive were evoked at least once in all subjects . Subjects were able to describe distinct proprioceptive sensations in the phantom limb such as opening and closing of the hand , movement of the thumb , and flexing of the elbow . However , unlike the tactile percepts that were predominantly stable across days , these proprioceptive sensations could be repeatedly evoked only for a few minutes even with consistent stimulus parameters . Only in the case of Subject 4 ( the subject with trans-radial amputation ) , were we able to evoke sensations of thumb and wrist movement reliably over longer time courses , spanning multiple days and weeks . Interestingly , these proprioceptive percepts were elicited by a set of three closely situated electrodes over a narrow range of stimulus parameters ( stimulus amplitude = 3 mA , stimulus frequency = 1–5 Hz ) . Varying the stimulation frequency influenced the modality of the evoked sensation in Subject 3 , but not in the other subjects . For Subject 3 , the sensory percepts that were described as ‘touch’ or ‘pressure’ occurred in up to 90% of trials at low stimulation frequencies ( below 20 Hz ) while stimulation frequencies above 50 Hz evoked percepts that were always characterized as mixed modality . Subject 1 never reported these naturalistic sensations , which could be because we never stimulated at frequencies below 20 Hz , while Subjects 2 and 4 reported naturalistic , mixed , and paresthetic sensations independent of the stimulus frequency . We quantified the detection threshold for sensations in the missing limb in all four subjects using a two-alternative forced-choice paradigm . Because Subjects 2 and 3 frequently experienced co-evoked sensations in the phantom and on the residual limb , for psychophysical assessments , we asked them to focus only on the distal phantom percept whenever stimulation co-evoked a sensation in the residual limb . In this task , the subject reported which of two intervals contained the stimulus train . With a randomized presentation of various stimulation amplitudes , we measured the detection threshold as the minimum amplitude at which the subject could correctly report the interval containing the stimulation train with 75% accuracy ( Figure 4A ) . Mean detection thresholds ( Figure 4B ) for Subjects 1–4 were 3 . 75 mA ( n = 2 electrodes ) , 1 . 25 ± 0 . 36 mA ( n = 5 electrodes ) , 1 . 58 ± 0 . 39 mA ( n = 14 electrodes ) and 1 . 94 ± 0 . 27 mA ( n = 14 electrodes ) , respectively . We measured just-noticeable differences ( JND ) in stimulation amplitude with a two-alternative forced choice task in Subjects 3 and 4 . We evaluated the goodness-of-fit using the probability of transformed likelihood ratio ( pTLR ) , which spans 0–1 with a higher value signifying a better fit and values below 0 . 05 signifying an unacceptable fit . In Subject 3 , for one electrode , the subject could perceive a change of 86 µA ( slope , β = 0 . 045 , pTLR = 0 . 58 ) at 75% accuracy when the standard amplitude was 2 . 5 mA , and a higher standard amplitude of 4 mA increased the JND to 280 µA ( slope , β = 0 . 073 , pTLR = 0 . 83; Figure 4C ) . In Subject 4 , the JNDs showed a similar dependence on standard amplitude with mean JND2 . 5=60 ± 21 µA ( median slope , β = 0 . 040 , median pTLR = 0 . 79 ) and mean JND4 . 0=338 ± 98 µA ( median slope , β = 0 . 005 , median pTLR = 0 . 40 , n = 5 electrodes; Figure 4D and Figure 4—figure supplement 1 ) . To put these numbers in context , for Subject 4 , with mean threshold at approximately 2 mA and maximum stimulation amplitude at 6 mA , the JNDs represent 1 . 3% ( at standard amplitude of 2 . 5 mA ) and 9% ( at standard amplitude of 4 mA ) of the available stimulation range . To measure the relationship between stimulation amplitude and sensation intensity , subjects performed a free magnitude estimation task , in which they were instructed to rate perceived intensity on an open-ended numerical scale as stimulation amplitude was varied randomly . They were instructed to scale their response such that a doubling in perceived intensity was reported as a doubling in the numerical response . To control for variability across different electrodes and across different testing sessions , we normalized each electrode to the mean of its response . We observed that as stimulation amplitude was increased , the perceived intensity of the sensory percept increased linearly for all subjects; an effect that was consistent across repetitions of the task on multiple days ( Figure 4E ) . A linear fit was determined to be better than or at least as good as a sigmoid or logarithmic fit based on adjusted R2 values , and all electrodes had a significant linear relationship between stimulus amplitude and perceived intensity , ( pint <0 . 01 , F-test , where pint is the two-sided p-value for the null hypothesis that the slope of the regression line was zero ) . This linear relationship between amplitude and intensity was maintained even though different electrodes were tested with different pulse widths and frequencies . Supplementary file 2 shows a complete list of stimulation parameters used for these experiments . These results taken together show that subjects should be able to perceive graded sensory feedback for linearly spaced gradations greater than the JND for each electrode . The number of gradations in stimulation determines the number of discrete targets ( such as identifying three different levels of force ) that can be represented in a functional task . To identify the optimal gradation for functionally relevant sensory feedback via SCS , we partitioned stimulation amplitudes and subject responses during the free magnitude estimation task into three or five discrete linearly spaced ranges . The partitioned data were used to estimate how reliably subjects can distinguish sensations for each of these amplitude ranges . Figure 4F shows the distribution of the normalized subject responses for a 3-target task where the overall accuracy was 72% . All subjects reported sensations in the low and high range of stimulation with a high accuracy , ( 79% and 95% accuracy , respectively ) with medium targets having an accuracy of 54% . When the data were partitioned into five discrete ranges , the overall accuracy was 46% ( Figure 4—figure supplement 2 ) . In the context of clinical translation , these results indicate that it may be possible for the user to discriminate three specific intensity levels based on stimulation amplitude alone . Since we found a consistent linear relationship between percept intensity and stimulation amplitude , we quantified the changes in percept area that occurred as the stimulation amplitude was increased . In a prosthetic device , being able to modulate the percept intensity independent of the area is critical to deliver graded feedback that remains focal . Figure 5A shows an example of a percept where the area and centroid remain stable as the stimulation amplitude is increased . To examine the effect of stimulation amplitude on the area and intensity of the evoked percept , we computed the least-squares regression line for area versus stimulation amplitude and obtained the two-sided p-value ( parea ) for the null hypothesis that the slope of the regression line was zero . We also compared the slope of this regression line with the slope of the linear fit between stimulation amplitude and reported intensity ( obtained from magnitude estimation trials , Supplementary file 2 ) for each electrode to identify whether percept area and intensity were modulated independently ( Figure 5B ) . Across all electrodes ( n = 24 ) , the slope of the linear fit for area was less than the slope for intensity . For three electrodes , both area and intensity were modulated by stimulation amplitude ( parea <0 . 01 , median β = 0 . 25 and pint <0 . 01 median β = 1 . 36 ) . For the remaining 21 electrodes , only intensity , but not area , was modulated by stimulation amplitude ( parea >0 . 05 , median β = 0 . 01 and pint <0 . 01 , median β = 1 . 33 ) . This indicates that for most electrodes , it is possible to modulate percept intensity independent of percept area . Lead migration is a common clinical complication for SCS , with an incidence rate as high as 15–20% ( Cameron , 2004; Kinfe et al . , 2014; Mekhail et al . , 2011; Mironer et al . , 2004 ) . Lead migration would change the location and modality of evoked sensations , which could limit the long-term viability of SCS and would also complicate the scientific utility of percutaneous SCS as a testbed for novel neuroprosthetic techniques . We performed weekly X-rays that allowed us to monitor the position of the leads and quantify migration over the duration of the implant . Superimposing the intraoperative fluoroscopy image and the final X-ray ( Figure 6A ) revealed that lead migration was largely restricted to the rostro-caudal axis . In all subjects , the largest caudal migration was observed when comparing the intraoperative fluoroscopy image with the X-ray at the end of first week ( Figure 6B ) . One of the leads in Subject 2 almost completely migrated out of the epidural space in this post-operative period ( Figure 6B ) , rendering it unusable for stimulation experiments . In contrast to the migration that occurred during the first week , X-rays from the first and last week of testing showed minimal lead migration ( Figure 6B ) . In the weeks following the initial migration , the median migration in the rostro-caudal direction across the three leads in any subject never exceeded 5 mm . For all subjects , the initial placement of the leads rostral to the target cervical levels prevented loss of coverage of those spinal levels following the caudal migration of the leads . We assessed the stability of each evoked percept throughout the duration of the study ( e . g . Figure 7A ) in terms of the threshold charge ( Figure 7B ) for evoking a percept in the missing hand . A one-way ANOVA performed for each subject confirmed that there was no significant difference in the threshold charge for each week for Subjects 1 , 2 , and 3 ( p>0 . 01 , F = 2 . 3 , 1 . 1 , 1 . 7 respectively ) . For Subject 4 , there was a significant change in threshold after weeks one and three ( p<0 . 01 , F = 9 . 0 ) . A post-hoc multiple pairwise comparison analysis using the Tukey HSD test confirmed that there was a significant increase in the thresholds between weeks one and two ( p<0 . 01 ) and a significant decrease between weeks three and four ( p<0 . 01 ) . We also characterized stability in terms of the size ( area ) and location ( centroid ) of percepts evoked in the missing hand . The centroid and area were calculated for all percepts evoked at the minimum stimulus amplitude that was tested at least once each week during the implant . If no stimulus amplitude was tested during every week of testing , the lowest stimulus amplitude that was tested for the next highest number of weeks ( for at least two weeks ) was chosen . We quantified the migration of the mean centroid location across all stimulus repetitions for each week with respect to the mean centroid location of the previous week for each electrode ( Figure 7C ) . Across all subjects , the evoked percepts exhibited a median migration of 25 . 2 mm between weeks 1 and 2 , 11 . 6 mm between weeks 2 and 3 , and 9 . 9 mm between weeks 3 and 4 . This week-to-week decrease in centroid migration follows the trend of decreased week-to-week vertical electrode migration ( Figure 7B ) . Similarly , the change in area for each week was calculated with respect to the mean area of the previous week for each electrode ( Figure 7D ) . Across all subjects , the median change in area of percepts evoked in the missing hand was 8 . 1 cm2 between weeks 1 and 2 , 0 . 14 cm2 between weeks 2 and 3 , and 1 . 1 cm2 between weeks 3 and 4 . We constructed two separate auto-regressive time series model to examine the changes in distributions of area and centroid distance over time , adjusting for autocorrelations in the data . We found a significant decrease in area over time across all weeks , ( β = −0 . 201 , p<0 . 01 ) . For centroid migration , there was a decrease during weeks 2 ( β = −23 . 224 , p<0 . 05 ) and 3 ( β = −40 . 585 , p<0 . 01 ) .
SCS-evoked sensory percepts were perceived to emanate from the missing limb in all subjects . Some percepts were highly localized to a single finger or focal region of the palm , while others were diffuse , covering large regions of the limb . In our second and third subjects , distal sensations were often accompanied by a secondary sensation at the residual limb . It is unclear whether these secondary sensations are a result of neuroplastic changes in the representation of the amputated hand in the cortex or are a limitation of the selectivity of the SCS electrodes used in this study . Future studies in people with intact limbs undergoing lateral SCS may help to differentiate these effects , since those subjects would not have similar neuroplastic changes . Factors that may limit stimulation selectivity with our approach include the thickness of the cerebrospinal fluid in the subdural space and the relatively large size of the contacts on the SCS leads . Consequently , the sensory percepts evoked in this study were sometimes more diffuse than those reported in other studies using peripheral neurostimulation approaches ( Charkhkar et al . , 2018; Davis et al . , 2016; Raspopovic et al . , 2014; Tan et al . , 2015 ) . Importantly though , they are comparable in focality to those used to effectively deliver sensory feedback during recent long-term take-home studies of bidirectional prosthesis using peripheral nerve stimulation ( Cuberovic et al . , 2019 ) or targeted reinnervation ( Schofield et al . , 2020 ) . In all except Subject 4 , monopolar stimulation primarily evoked sensations in the forearm and upper arm , whereas multipolar stimulation allowed us to evoke sensations that were localized to distal regions of the missing hand and wrist . We could not identify any difference ( e . g . in surgical technique ) that led to this difference in focality of monopolar stimulation for Subject 4 as compared to all other subjects . In all subjects , the leads were steered toward the lateral spinal cord and spinal roots , ipsilateral to the amputation . At this location , the dorsal rootlets fan out under the dura before entering the spinal cord at the dorsal root entry zone . In the cervical spinal cord , the rootlets are each approximately 0 . 4–1 . 3 mm in diameter and densely packed with few spaces between them ( Alleyne et al . , 1998; Karatas et al . , 2005; Tanaka et al . , 2000 ) . This arrangement , superficially resembling the flattened peripheral nerve cross-section achieved by the flat interface nerve electrode ( Charkhkar et al . , 2018; Tan et al . , 2015 ) , may lend itself to a higher degree of selective activation than could be achieved with stimulation of more traditional SCS targets such as the dorsal columns or the dorsal root ganglia , although this may require development of new SCS devices with more optimal electrode sizing and spacing . The relationship between the locations of the electrodes and that of the evoked percepts showed marked inter-subject variability and deviation from expected dermatomes . For example , all electrodes in Subject 1 were in the T1 region , but the reported sensations were in the missing hand , a region covered by the C6–C8 dermatomes . Recently , it has been recognized that dermatomes inadequately reflect inter-individual variability in dermatome coverage and overlap , suggesting that the variability observed in our study may reflect natural inter-subject differences ( Lee et al . , 2008 ) . However , a limitation of this study is that we did not directly image the spinal cord or dorsal roots . As such , we could not determine the exact spatial arrangement of the implanted SCS electrodes relative to target neural structures . Several research groups have developed highly detailed computational modeling techniques to study how the electric fields generated in SCS interact with neural structures ( Capogrosso et al . , 2013; Greiner et al . , 2020; Lempka et al . , 2015 ) . These techniques could potentially help illuminate the specific neural targets and pathways that were activated in this study . These observations combined with simulation studies could also inform the design of stimulation schemes and novel electrodes to improve the selectivity of our somatosensory neuroprosthesis . With respect to the perceptual qualities of evoked sensations in this study , we observed a robust relationship between stimulus amplitude and percept intensity . Every electrode tested across all four subjects demonstrated a statistically significant linear relationship between stimulation amplitude and perceived intensity . This is similar to what has been observed with peripheral nerve stimulation ( Graczyk et al . , 2016; Petrini et al . , 2019b ) . Interestingly , we observed JNDs to be proportional to the stimulation amplitude with higher stimulation amplitudes resulting in larger JNDs . Such a relationship between JNDs and stimulus amplitude is consistent with Weber’s law which governs the behavior of most peripheral sensory receptors ( Gös , 1959 ) . We also observed that subject responses could be separated into three , but not five , intensity categories ( i . e . low , medium , and high ) based on the stimulation amplitude , which suggests that they would be able to successfully perform a 3-level discrimination task based only on perceived intensity , such as identifying three different force levels exerted by objects of different stiffnesses . These expected performance levels are similar to the success rates for object stiffness experiments demonstrated by others using peripheral nerve stimulation to restore somatosensation ( Raspopovic et al . , 2014 ) , and it is possible that they could improve with time , training , continuous ( rather than discrete ) modulation of amplitude , and the addition of active efferent control of a prosthesis . Future work should focus on demonstration of such closed-loop control with sensory feedback using lateral SCS . An increase in stimulus amplitude is thought to increase perceived intensity by recruiting a larger volume of somatosensory afferent neurons ( Graczyk et al . , 2016 ) . An increase in the volume of recruited neurons could also result in an increase in percept area . However , we observed little to no effect of increasing stimulation amplitude on percept area . It is possible that the anatomical distance between adjacent spinal roots reduces this effect ( Greiner et al . , 2020 ) . Additionally , it is currently unknown whether there is strong somatotopic organization of the fanned-out dorsal rootlets where they enter the spinal cord ( i . e . whether neighboring rootlets innervate neighboring patches of skin ) . The minor changes in focality of sensation as amplitude increases may be due to the presence of this somatotopic organization and recruitment of neurons in adjacent rootlets . A primary aim of providing artificial somatosensory feedback has been to evoke naturalistic sensations , particularly those described as touch or pressure . Most of the percepts reported in this and previous studies of somatosensory neuroprostheses have been described as paresthesias . However , for Subjects 2 and 3 , we report that 8 . 25% and 19 . 5% of all evoked percepts were described as touch or pressure alone . Studies that relied on peripheral nerve stimulation to restore somatosensory feedback have reported similar proportions of naturalistic percepts , e . g . 8–30% as touch-like percepts ( Petrini et al . , 2019b; Strauss et al . , 2019; Tan et al . , 2014 ) and 2–29% as pressure-like percepts ( D’Anna et al . , 2019; George et al . , 2019; Petrini et al . , 2019b; Strauss et al . , 2019 ) . Continuous modulation of stimulus parameters , such as modulating pulse width ( Charkhkar et al . , 2018; Tan et al . , 2014 ) or varying charge density [Charkhkar et al . , 2018] have been proposed to evoke more naturalistic cutaneous or proprioceptive sensations . However , a recent study demonstrated that patterned stimulation did not reliably change paresthetic sensations to more naturalistic ones ( Ortiz-Catalan et al . , 2019 ) . Additionally , biomimetic stimulus trains [George et al . , 2019; Okorokova et al . , 2018; Valle et al . , 2018] have been proposed to evoke more naturalistic sensations , though none of these approaches have established a stimulation paradigm that reliably elicits naturalistic sensations across subjects . We too , did not uncover a reliable way to evoke naturalistic sensation during the course of this study . Thus , the choice of electrodes and stimulation parameters would have to be optimized for each individual user to evoke percepts with the desired modalities ( Ortiz-Catalan et al . , 2019 ) . Only one subject ( Subject 4 , trans-radial amputation ) reported proprioceptive percepts that were repeatedly evoked over more than a few minutes . This result aligns with other studies , which often report only limited examples of proprioceptive percepts ( George et al . , 2019 ) , or which describe proprioceptive sensations that result directly from muscle contractions in the residual limb ( Petrini et al . , 2019a ) . While SCS did not evoke overt reflexive movements of the residual limb in any subject at the stimulation amplitudes used in this study , it is possible that these proprioceptive percepts result from small reflexive contractions of residual limb muscles which themselves activate muscle spindle afferents . Complex coordination of activation of muscle spindle and cutaneous ( e . g . slowly adapting type II ) afferents may be required for directly evoking realistic kinesthetic percepts . Future work should explore the downstream effects of stimulation of proprioceptive and cutaneous afferents on perception of kinesthesia . Regardless , we propose that even though we evoked primarily paresthetic sensations , the ability to evoke these percepts via a clinically translatable approach in individuals with high-level amputations establishes the promise of this approach towards restoring sensation . The location of the implanted SCS electrodes and the corresponding evoked percepts showed only minor migration across the duration of implantation . In clinical practice , SCS lead migration is a common complication , occurring in as many as 15–20% of cases ( Cameron , 2004; Kinfe et al . , 2014; Mekhail et al . , 2011; Mironer et al . , 2004 ) , and is typically classified by a complete loss of paresthetic coverage of the region of interest . Repeated monitoring of both the physical location of the SCS leads and the evoked sensations demonstrated that there was some migration immediately after implantation , but minimal movement thereafter . As a preemptive measure against loss of coverage due to the initial migration , we opted to use 16-contact leads in Subjects 2–4 . By placing the leads such that the most rostral contacts were above the target spinal levels , we ensured continued coverage even in the case of caudal migration . It is worth noting that we did not anchor these leads to any bony structures or nearby tissue . Future permanently implanted systems for restoring sensation using SCS can utilize these anchoring techniques and thereby reduce or eliminate lead migration ( Mekhail et al . , 2011 ) . The stability in the electrodes is reflected in the stability of the evoked percepts . In the hand region , we observed a migration of evoked percepts of 10–25 mm , which is similar to the shift reported in peripheral stimulation approaches ( Tan et al . , 2015 ) . Moreover , given that the spatial acuity in the palm region is approximately 8–10 mm ( Catley et al . , 2013; Craig and Lyle , 2001; Solomonow et al . , 1977; Tong et al . , 2013 ) , the scale of migration observed is within the range that would not likely be detectable by the user . The techniques described in this study have both important advantages and disadvantages that should be considered when selecting an approach for restoring sensation after upper-limb amputation . A major advantage of the percutaneous approach described here is that there is a relatively low barrier to initiating clinical studies because the electrodes are commercially available from multiple manufacturers and the surgical procedures are commonly performed at most major medical centers . However , this reliance on commercially available electrodes also likely limited selectivity and focality . Regardless , a great deal of technical and scientific development can be achieved with this approach before moving on to more complex studies involving custom electrodes and implantable stimulators . Another major advantage of the approach is its viability for people with high-level amputations , in which the peripheral nerve has been amputated . In this population , the spinal cord and roots typically remain intact , and we have demonstrated that stimulation of those structures can produce focal sensations in the missing hand . Currently , the only other viable neuroprosthetic techniques for restoring sensation after proximal amputation are invasive approaches such as targeted reinnervation or stimulation of structures in the central nervous system . As compared to other techniques that focus on peripheral nerve stimulation , such as epineural stimulation with cuff electrodes or penetrating stimulation with Utah arrays or longitudinal intrafascicular electrodes , our results demonstrate substantially less focal percepts and less consistent coverage of each individual digit across subjects . Further , sensations in the hand were often accompanied by a sensation on the residual limb . It is currently unclear to what degree this is a limitation of the relatively large size of the electrodes we used here , as opposed to a fundamental limitation of the selectivity of epidural SCS . Future work will focus on computational studies to explore this question and design new electrodes that can more selectively target the sensory afferents in the dorsal rootlets to maximize the selectivity and focality of our approach . With respect to clinical applications , the 3-fold dynamic range afforded by the stimulus amplitude is similar to those reported previously ( Petrini et al . , 2019b ) . Though the absolute current values we used are an order of magnitude higher than those required for peripheral nerve stimulation , epidural stimulation systems are widely used in a clinical setting and also in patient homes . This suggests that a neuroprosthetic device based on this approach can be effectively utilized in clinical or home setting . An important limitation of this study is that we focused on characterizing the sensations evoked by SCS but did not demonstrate that those sensations could be used as part of a closed-loop neuroprosthetic system . While we demonstrate that many of the qualities of the evoked sensations are similar to those reported by others ( e . g . sensation intensity modulates linearly with stimulation amplitude ) , it will be critical to demonstrate that sensations remain stable and are useful during closed-loop prosthetic applications . For example , while we did not control subject posture during any of our experiments , it will be important to demonstrate that sensations remain stable during intentional movements of the neck , shoulders , and arms . Certainly , future work will focus on achieving these demonstrations and characterizing the effects of sensory restoration via SCS on dexterous control of prosthetic limbs . Since this approach targets proximal neural pathways , SCS-mediated sensory restoration lends itself to use for a wide range of populations , such as individuals with proximal amputations and those with peripheral neuropathies in which stimulation of peripheral nerves may be difficult or impossible . Provided that the injury does not affect the dorsal roots and spinal cord , our results suggest that these techniques can be effective in restoring sensation , regardless of the level of limb loss . Moreover , the widespread clinical use of SCS and the well-understood risk profile provide a potential pathway towards clinical adoption of these techniques for a somatosensory neuroprosthesis .
The aim of this study was to investigate whether electrical stimulation of lateral structures in the cervical spinal cord could evoke sensations that are consistently perceived to emanate from the missing hand and arm . We also aimed to characterize those sensations and establish the relationship between stimulation parameters and the perceptual quality of evoked sensory percepts . Four subjects with upper-limb amputations ( three females , one male; Table 1 ) were recruited for this study . Three amputations were between the elbow and shoulder and one was below the elbow . The time since amputation ranged from 2 to 16 years . All procedures and experiments were approved by the University of Pittsburgh and Army Research Labs Institutional Review Boards and subjects provided informed consent before participation . SCS leads were implanted through a minimally invasive , outpatient procedure performed under local anesthesia . With the subject in a prone position , three 8- or 16-contact SCS leads ( Infinion , Boston Scientific ) were percutaneously inserted into the epidural space on the dorsal side of the C5–C8 spinal cord through a 14-gauge Tuohy needle . Contacts were 3 mm long , with 1 mm inter-contact spacing . Leads were steered via a stylet under fluoroscopic guidance , and electrode placement was iteratively adjusted based on the subjects’ report of the location of sensations evoked by intraoperative stimulation . The entire procedure usually took approximately 3–4 hr . The leads were maintained for up to 29 days and subsequently explanted by gently pulling on the external portion of the lead . Subjects attended testing sessions 3–4 days per week during the implantation period . The testing sessions lasted up to a maximum of 8 hr . Lead location and migration were monitored via weekly coronal and sagittal X-rays throughout the duration of implant . During testing sessions , stimulation was delivered using three 32-channel stimulators ( Nano 2+Stim; Ripple , Inc ) . The maximum current output for these stimulators was 1 . 5 mA per channel . In order to achieve the higher current amplitudes required for SCS , a custom-built circuit board was used to short together the output of groups of four channels , thereby increasing the maximum possible output to 6 mA per channel resulting in a total of 8 effective channels per stimulator . Custom adapters were used to connect each stimulator to eight contacts on each of the implanted leads . Custom software in MATLAB was used to trigger and control stimulation . Stimulation pulse trains were charge-balanced square pulses , with either asymmetric or symmetric cathodic and anodic phases . For Subjects 1-3 , the first phase of stimulation was cathodic , while for Subject 4 , an error in the stimulation control code caused the first phase of stimulation to be anodic . For asymmetric pulses , the second phase was twice the duration and half the amplitude of the first phase . Stimulation was performed either in a monopolar configuration , with the ground electrode placed at a distant location such as on the skin at the shoulder or hip , or in a multipolar configuration with one or more local SCS contacts acting as the return path . Stimulation frequencies and pulse widths ranged from 1 to 300 Hz and 50–1000 µs , respectively . The interphase interval was 60 µs . All stimulus amplitudes reported in this manuscript refer to the first phase amplitude . The first few sessions of testing were primarily devoted to recording the location and perceptual quality of sensory percepts evoked with various stimulation configurations . An auditory cue was provided to denote the onset of stimulation . At the offset of each stimulation train , the subject used a touchscreen interface developed in Python ( Figure 1—figure supplement 2 ) to document the location and perceptual quality of the evoked sensation . This interface can be downloaded from Nanivadekar et al . , 2020 https://github . com/pitt-rnel/perceptmapper . The location of the sensory percept was recorded by the subject using a free-hand drawing indicating the outline of the evoked percept on an image of the appropriate body segment ( i . e . , hand , arm , or torso ) . The percept quality was recorded using several descriptors: mechanical ( touch , pressure , or sharp ) , tingle ( electrical , tickle , itch , or pins and needles ) , movement ( vibration , movement across skin , or movement of body/limb/joint ) , temperature , pain due to stimulation , and phantom limb pain . Each descriptor had an associated scale ranging from 0 to 10 to record the corresponding perceived intensity . Additionally , the subject was instructed to rate the naturalness ( 0–10 ) and the depth of the perceived location of the percept ( on or below the skin , or both ) . This set of descriptors have been used previously to characterize evoked sensory percepts ( Heming et al . , 2010; Lenz et al . , 1993 ) . The order of stimulation electrodes and amplitudes was randomized to prevent subjects from predicting the location and perceptual qualities of sensations from previous trials . All percepts that were localized ipsilateral to the amputation were included for analysis in this work . In Figure 1 , only those percepts which show less than 70% area overlap ( as in Charkhkar et al . , 2018 ) with any other percept are shown for clarity . Supplementary file 1 visualizes all the evoked percepts in an interactive fashion . For each trial , subjects were allowed to report more than one descriptor simultaneously . Each unique combination of ‘mechanical’ , ‘movement’ , and ‘tingle’ descriptors was considered a separate modality for the evoked percept . All percepts that contained a descriptor for tingle ( ‘electric current’ , ‘tickle’ , ‘sharp’ , ‘pins and needles’ ) were considered paresthetic and were grouped together . A sunburst plot was constructed for each subject to analyze the fraction of paresthetic and non-paresthetic percepts that contained mechanical or movement components . Therefore , all unique modalities were divided in to three groups: paresthetic percepts that had ‘tingle’ but no ‘mechanical’ or ‘movement’ component ( Figure 3 , teal sectors ) , mixed percepts that had a ‘mechanical’ and/or ‘movement’ component and ‘tingle’ ( Figure 3 , grey sectors ) , and non-paresthetic percepts that only had ‘mechanical’ and/or ‘movement’ components ( Figure 3 red sectors ) . For each sunburst plot , the inner , middle , and outer annuli represent ‘tingle’ , ‘mechanical’ , and ‘movement’ modality descriptors , respectively . Each sector represents a unique descriptor and the size of each sector represents the fraction of all percepts that contained the corresponding descriptor . This allows us to identify the distribution of unique modalities , such that , for a given sector in the tingle annulus ( for example , n = 761 for Subject 2 , ‘tingle’ ) , we can identify the fraction of percepts that had a specific mechanical descriptor ( e . g . 'sharp’ n = 245 ) and the fraction of these percepts that had a specific movement descriptor ( e . g . ‘vibration’ n = 104 ) . An interactive version of Figure 3 with expandable sectors for each descriptor is available in Supplementary file 3 . The spinal cord segment targeted by stimulation through each electrode was inferred from the X-ray images . We used the pedicles of each vertebra to mark the boundaries that separated each spinal root ( Figure 2B ) . These boundaries provided an anatomical marker to establish where each electrode was located , in the rostrocaudal axis . Similarly , boundaries were drawn on the body segment outline images to divide them into seven anatomical segments ( Figure 2A ) including thumb , D2–D3 , D4–D5 , wrist , forearm , elbow , and upper arm . The sensory percepts were categorized as being associated with one of the seven anatomical segments based on which segment contained the maximal area of the perceived sensation . For this analysis , only those electrodes that evoked a sensory percept ipsilateral ( n = 315 ) to the amputation were included . Electrodes that only evoked bilateral ( n = 64 ) and contralateral ( n = 68 ) sensations at threshold would not be useful for neuroprosthetic applications for people with unilateral amputation and were excluded . Dermatome maps were generated per subject , by determining the proportion of electrodes situated at each spinal level that evoked a sensation in a specific anatomical region . The intraoperative fluoroscopy image , superimposed over the X-rays from the first and last week of testing , gave an indication of gross movements of the leads . Using bony landmarks , the X-ray from the first week was aligned to the intraoperative fluoroscopy image , and each subsequent X-ray was aligned to the X-ray from the previous week using an affine transformation method in MATLAB . The SCS contact that appeared to be most parallel to the plane of imaging was used to determine the scale length for the image ( SCS contacts are 3 mm in length ) . For each lead , the distance between the rostral tips of the electrodes as seen in the aligned image pairs ( Figure 6 ) was measured to determine the rostro-caudal migration . Positive values signified caudal migration and negative values signified rostral migration . For all electrodes that evoked a percept in the missing hand , the threshold charge was calculated for each week . A one-way ANOVA was performed for each subject to test for differences in thresholds across weeks . For subjects where a significant difference was reported , a post-hoc multiple pairwise comparison analysis using the Tukey HSD was performed to identify the pairs of consecutive weeks with a significant difference in thresholds . To quantify migration of perceived sensations , we measured the change in the position of the centroid and the change in area of each percept that was localized to the hand . For sensations that included a percept outside the hand , we only used the hand percept in these calculations , as this is the most relevant location for a somatosensory neuroprosthesis . We chose the minimum stimulus amplitude that was tested at least once per week for the highest number of weeks during the implant . We quantified the migration of the mean percept centroid for each week , with respect to the mean percept centroid for the previous week . This analysis was repeated for all electrodes . Similarly , to quantify the change in percept area , the mean area of the percept for each week was compared to the mean area for the previous week . The distances were converted to millimeters using the average hand length of 189 mm ( as measured from the tip of the middle finger to the wrist ) and average palmar area of 75 cm2 of a human male ( Agarwal and Sahu , 2010; Ilayperuma et al . , 2009; Kono et al . , 2014; Martin and Nguyen , 2004; Rhodes et al . , 2013; Zafar et al . , 2017 ) . All electrodes that were tested in at least two of the weeks were included in the analysis . We also constructed separate auto-regressive time series models to examine the changes in distributions for both area and centroid migration over time , adjusting for autocorrelations in the data . The AUTOREG procedure in SAS estimates and forecasts linear regression models for time series data when the errors are autocorrelated or heteroscedastic . If the error term is autocorrelated ( which occurs with time series data ) , the efficiency of ordinary least-squares ( OLS ) parameter estimates is adversely affected and standard error estimates are biased , thus the autoregressive error model corrects for serial correlation . For models with time-dependent regressors , the , AUTOREG procedure performs the Durbin t-test and the Durbin h-test for first-order autocorrelation and reports marginal significance levels . A two-alternative forced choice task was used to determine detection thresholds . The subject was instructed to focus on a fixation cross on a screen . Two one-second-long windows , separated by a variable delay period , were presented and indicated by a change in the color of the fixation cross . Stimulation was randomly assigned to one of the two windows . After the second of the two windows , the fixation cross disappeared , and the participant was asked to report which window contained the stimulus . The stimulus amplitude for each trial was varied using a threshold tracking method ( Leek , 2001; Levitt , 1971 ) with a ‘one-up , three-down’ design . In this design , an incorrect answer resulted in an increase in stimulus amplitude for the next trial while three consecutive correct trials were required before the stimulus amplitude was decreased . Stimulus amplitude was always changed by a factor of 2 dB . Five changes in direction of the stimulus amplitude , either increasing to decreasing or vice versa , signaled the end of the task . Using this task design , the detection threshold was determined online as the average of the last 10 trials before the fifth change in direction . A detection threshold calculated this way corresponds approximately to correctly identifying the window containing the stimulus 75% of the time ( García-Pérez , 1998 ) . To get a finer estimate of the detection threshold we also used a non-adaptive design in a subset of trials for Subject 4 , where we presented a predetermined set of stimulus amplitudes . This block of stimulus amplitudes was repeated up to eight times and the presentation sequence was randomized within each block . A cumulative-normal psychometric curve was fit to both types of detection experiments post-hoc using the Palamedes toolbox ( Kingdom and Prins , 2016 ) with the guessing rate γ and lapse rate λ held fixed at 0 . 5 and 0 respectively . The detection threshold was calculated as the stimulus amplitude at the 75% accuracy level . Tasks in which accuracy levels for all stimulus amplitudes were <0 . 6 or>0 . 9 were omitted from this analysis . We carried out a goodness-of-fit analysis with 1000 simulations using the Palamedes toolbox and discarded any fit with probability of transformed likelihood ratio ( pTLR ) less than 0 . 05 . pTLR signifies the proportion of simulated likelihood ratios that were smaller than the likelihood ratio obtained from the data and it spans 0–1 with a higher value signifying a better fit and values below 0 . 05 signifying an unacceptable fit . Thresholds calculated for the same electrodes on different days were averaged together to obtain a mean detection threshold for each electrode , with all other stimulus parameters ( e . g . frequency , pulse width ) held constant . A similar two-alternative forced choice task was used to determine just-noticeable differences ( JND ) for stimulation amplitude . The design of the task was identical to the detection task except stimulation was provided in both the windows and the subject was instructed to choose the window where the stimulus was perceived as being at a higher intensity . One of the stimulation amplitudes in every trial was held constant while the other was chosen randomly from a list of stimulus amplitudes constituting a block . The constant amplitude was either fixed at 2 . 5 mA for the lower standard amplitude or at 4 . 0 mA for the higher standard amplitude . The windows in which standard and the test amplitude were administered was randomized as well . This block of stimulus amplitudes was repeated up to eight times and the presentation sequence was randomized within each block . A cumulative-normal psychometric curve was fit to the data post-hoc using the Palamedes toolbox ( Kingdom and Prins , 2016 ) with the guessing rate γ and lapse rate λ held fixed at 0 . 5 and 0 respectively . The JND was calculated as the stimulus amplitude at the 75% accuracy level . Tasks in which accuracy levels for all stimulus amplitudes were <0 . 6 or>0 . 9 were omitted from this analysis . We carried out a goodness-of-fit analysis with 1000 simulations using the Palamedes toolbox and discarded any fit with pTLR <0 . 05 . To determine average JNDs at the two different standard amplitudes , we included data from only those electrodes for which testing at both standard amplitudes were carried out in the same session . JNDs calculated for the same standard amplitude on different electrodes were averaged together to obtain a mean JND for each standard amplitude . As JNDs were expected to be highly subject-specific , data from different subjects were not pooled together . A free magnitude estimation task was used to determine the relationship between stimulus amplitude and perceived intensity of the evoked sensations ( Ellermeier et al . , 1991; Stevens , 1986; Stevens , 1956; Verrillo et al . , 1969 ) . In this task , subjects were instructed to rate the perceived intensity on an open-ended numerical scale as stimulation amplitude was varied randomly . A block of stimulus amplitudes consisted of 6–10 values linearly spaced between the detection threshold of the electrode being tested and the highest value that did not evoke a painful percept up to 6 mA . This block of chosen amplitudes was presented six times and the presentation sequence was randomized within each block . The subject was instructed to scale the response appropriately such that a doubling in perceived intensity was reported as a doubling in the numerical response . Zero was used to denote that no sensation was perceived in response to the stimulus . During the first block , the subject experienced the full range of stimulation amplitudes while establishing their subjective scale , so data from this block were not included in the analysis . Data across electrodes or across different testing sessions were compared after normalizing each electrode to its mean response . We performed a post-hoc analysis to determine the maximum number of intensities a subject would likely be able to discriminate . For each electrode , stimulation amplitude was normalized to the maximum amplitude tested and the range of stimulation amplitude was partitioned into three or five linearly spaced discrete values . Similarly , the perceived intensities reported by the subjects were normalized to the maximum reported intensity and partitioned into three or five discrete linearly spaced ranges . Across all subjects , the distribution of the binned reported intensity for each discretized stimulation level was used to estimate how reliably subjects would be able to distinguish feedback at three or five different amplitudes ( Figure 4F and Figure 4—figure supplement 2 ) . To determine whether stimulation amplitude had differential effects on the area and intensity of evoked percepts , we computed the least-squares regression line for the relationship between stimulation amplitude and percept intensity and from magnitude estimation trials and stimulation amplitude and percept area from percept mapping trials . The two-side p-values ( pint and parea , respectively ) for each line were obtained for the null hypothesis that the slope of the regression line was zero and the slopes of the two lines were compared for each electrode . Instances where the slopes of each line were significantly different indicate electrodes where stimulation amplitude can modulate percept intensity independent of the area of the percept . | Even some of the most advanced prosthetic arms lack an important feature: the ability to relay information about touch or pressure to the wearer . In fact , many people prefer to use simpler prostheses whose cables and harnesses pass on information about tension . However , recent studies suggest that electrical stimulation might give prosthesis users more sensation and better control . After an amputation , the nerves that used to deliver sensory information from the hand still exist above the injury . Stimulating these nerves can help to recreate sensations in the missing limb and improve the control of the prosthesis . Still , this stimulation requires complicated surgical interventions to implant electrodes in or around the nerves . Spinal cord stimulation – a technique where a small electrical device is inserted near the spinal cord to stimulate nerves – may be an easier alternative . This approach only requires a simple outpatient procedure , and it is routinely used to treat chronic pain conditions . Now , Chandrasekaran , Nanivadekar et al . show that spinal cord stimulation can produce the feeling of sensations in a person’s missing hand or arm . In the experiments , four people who had an arm amputation underwent spinal cord stimulation over 29 days . During the stimulation , the participants reported feeling electrical buzzing , vibration , or pressure in their missing limb . Changing the strength of the electric signals delivered to the spinal cord altered the intensity of these sensations . The experiments are a step toward developing better prosthetics that restore some sensation . Further studies are now needed to determine whether spinal cord stimulation would allow people to perform sensory tasks with a prosthetic , for example handling an object that they cannot see . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2020 | Sensory restoration by epidural stimulation of the lateral spinal cord in upper-limb amputees |
The two partners required for sexual reproduction are rarely the same . This pattern extends to species which lack sexual dimorphism yet possess self-incompatible gametes determined at mating-type regions of suppressed recombination , likely precursors of sex chromosomes . Here we investigate the role of cellular signaling in the evolution of mating-types . We develop a model of ligand-receptor dynamics , and identify factors that determine the capacity of cells to send and receive signals . The model specifies conditions favoring the evolution of gametes producing ligand and receptor asymmetrically and shows how these are affected by recombination . When the recombination rate evolves , the conditions favoring asymmetric signaling also favor tight linkage of ligand and receptor loci in distinct linkage groups . These results suggest that selection for asymmetric gamete signaling could be the first step in the evolution of non-recombinant mating-type loci , paving the road for the evolution of anisogamy and sexes .
Consider a population where cells encounter one another at random and can mate when in physical contact . Interactions between cells leading to successful mating are dictated by a ligand-receptor pair . Population wide effects may emerge if the ligand is highly diffusible ( Youk and Lim , 2014; Hadjivasiliou et al . , 2015 ) . The employment of membrane bound ligands during sexual signaling is universal , whereas diffusible signals are not ( Hadjivasiliou and Pomiankowski , 2016 ) . In this work we therefore assume that the ligand-receptor interactions only operate locally . Receptors remain bound to the cell surface and ligands only undergo localized diffusion ( Figure 1 ) as is the case in several yeast and other unicellular eukaryotes ( Cappellaro et al . , 1991; Wilson et al . , 1999; Phadke and Zufall , 2009; Merlini et al . , 2013 ) . The following equations describe the concentration of free ligand L , free receptor R and bound ligand LR within a single cell , ( 1 ) d[L]dt=νL−k+[R][L]+k−[LR]−γL[L] , ( 2 ) d[R]dt=νR−k+[R][L]+k−[LR]−γR[R] , ( 3 ) d[LR]dt=k+[R][L]−k−[LR]−γLR[LR] . νL and νR describe the rate of production of the ligand and receptor respectively . γL , γR , and γLR , are the degradation rate of the ligand , receptor and bound complex respectively . The terms k+ and k- are the binding and unbinding rates that determine the affinity of the ligand to its receptor within a single cell . We can solve Equations ( 1-3 ) by setting the dynamics to zero to obtain the amount of free ligand , free receptor ( [L]* , [R]* ) and bound complex at steady state ( [LR]* ) , ( 4 ) [L]*=k+γLR ( νL-νR ) -k-γLγR-γLγRγLR+Δ2k+γLγLR ( 5 ) [R]*=k+γLR ( νR-νL ) -k-γLγR-γLγRγLR+Δ2k+γRγLR , ( 6 ) [LR]*=k+γLR ( νR+νL ) +k-γLγR+γLγRγLR-Δ2k+γLR2 , Where Δ is given by , ( 7 ) Δ= ( k-γLγR+γLR ( γLγR+k+γLR ( νR+νL ) ) ) 2+4k+γLγRγLR ( k-+γLR ) νR . We assume that the rates of ligand and receptor production and degradation are associated to timescales that are much shorter than the timescale of interactions between cells . Hence the concentrations of [L] , [R] and [LR] in individual cells will be at steady state when two cells meet . The likelihood of a successful mating between two cells depends not just on partner signaling levels but also on how accurately the cells can compute the signal produced by their partner . Binding of ligand and receptor originating from the same cell can obstruct this interaction . To capture this , we define the strength of the incoming signal for cell1 when it interacts with cell2 as , ( 8 ) W12=kb[L2]*[R1]* ( 1-[LR1]*[LR1]*+kb[L2]*[R1]* ) n , where subscripts denote concentrations in cell1 and cell2 , and the parameter kb determines the affinity of the ligand and receptor between cells . If kb is the same as the affinity of receptor and ligand within cells , then kb=k+k- . We also consider cases where kb≠k+k- , for example , when ligand interacts differently with receptors on the same as opposed to a different cell ( LeBon et al . , 2014; Hadjivasiliou et al . , 2016a ) . The cost of self-signaling is determined by n . When n=0 , W12 reduces to kb[R1]*[L2]* with the incoming signal dependent on the concentration of ligand produced by cell2 and receptor produced by cell1 . This corresponds to a case where self-binding does not lead to activation but only causes an indirect cost through the depletion of available ligand and receptor molecules . When n≥1 , binding within a cell leads to some form of activation that interferes with between cell signaling , imposing a cost in evaluating the incoming signal . Higher values for n correspond to more severe costs due to self-binding . The likelihood that two cells successfully mate ( P ) depends on the quality of their interaction given by , ( 9 ) P=W12W21K+W12W21 . Equation ( 9 ) transforms the signaling interaction into a mating probability . For the analysis that follows , we choose large values of K so that P is far from saturation and depends almost linearly on the product W12W21 . In summary , the probability that two cells mate is defined by the production and degradation rates of the ligand and receptor molecules , and the binding affinities between and within cells . To explore the evolution of signaling roles , we simplify the model by assuming that the degradation rates γL , γR , γLR are constant and equal to γ , and investigate mutations that quantitatively modify the ligand and receptor production rates . We consider a finite population of N haploid cells and set N=1000 throughout the analysis unless otherwise stated . Ligand and receptor production are controlled by independent loci with infinite alleles ( Tajima , 1996 ) . The ligand and receptor production rates of celli is denoted by ( νLi , νRi ) . We also consider different versions of the ligand and its receptor . Cells have two ligand-receptor pairs , ( L , R ) and ( l , r ) which are mutually incompatible , so the binding affinity is zero between l and R , and between L and r . Each cell has a ( L , R ) and ( l , r ) state , which are subject to mutational and evolutionary pressure as described below . W12 is re-defined as the summation of the interactions of these two ligand-receptor pairs , ( 10 ) W12=kb[L2]*[R1]* ( 1-[LR1]*[LR1]*+kb[L2]*[R1]* ) n+kb[l2]*[r1]* ( 1-[lr1]*[lr1]*+kb[l2]*[r1]* ) n . Again for the sake of simplicity , the ligand-receptor affinities are set to be the same between and within cells for each ligand-receptor pair ( i . e . k+ , k- and kb are the same for L-R and l-r interactions ) . A cell undergoes recurrent mutation that changes the production rate for the ligand L so that νLi′=νLi+ϵ with ϵ∼N ( 0 , σ ) with probability μ . The same mutational process occurs for all ligand and receptor production rates . We assume that mutation occurs independently at different loci and that there is a maximum capacity for ligand and receptor production , so that νL+νl<1 and νR+νr<1 . It follows that the production rates of the two ligand genes are not independent of one another and similarly for the two receptor genes . We also consider cases where νL+νl<α and νR+νr<α for α≠1 to reflect the relative synergy ( α>1 ) or relative competition ( α<1 ) between the production of the two ligands ( or receptors ) . For example , synergy between two ligands ( or receptors ) could reflect reduced energy expenditure for the cell if the same machinery is used to produce the two molecules . Competition on the other hand could reflect additional costs due to the production of two different ligands ( or receptors ) . Selection on ligand-receptor production rates is governed by the likelihood that cells pair and produce offspring . We assume that cells enter the sexual phase of their life cycle in synchrony , as is the case in the majority of unicellular eukaryotes ( Hadjivasiliou and Pomiankowski , 2016 ) . Pairs of cells are randomly sampled ( to reflect random encounters ) and mate with probability P defined in Equation ( 9 ) . Cells failing to mate are returned to the pool of unmated individuals . The process is repeated until M cells have mated , giving rise to M/2 mated pairs ( we set M<N , so only some cells mate ) . Each mated pair produces 2 haploid offspring so the population size shrinks from N to M . The population size is restored back to N by sampling with replacement . It follows that Equations ( 9 ) and ( 10 ) together provide a proxy for fitness according to the ligand and receptor production rates of individual cells . Initially , recombination is not allowed between the genes controlling ligand and receptor production but then is considered in a later section .
The strength of an incoming signal W12 depends on the concentration of free receptor in cell1 and free ligand in cell2 , and the cost of self-binding ( n ) ( Equation ( 10 ) ) . The steady state concentration of [L] , [R] and [LR] are governed by different production rates ( Figure 2—figure supplement 1; details of the derivation can be found in the Materials and methods section ) . For low degradation rates ( γ small ) , the removal of available molecules is dominated by self-binding ( k+ ) ( Equations ( 1 ) and ( 2 ) and Figure 2a , b ) . At the same time , a lower degradation rate leads to higher levels of ligand and receptor ( Figure 2a ) even if the relative drop of free ligand and receptor is steeper as k+ increases ( Figure 2b ) . As a consequence , the ability of a cell to generate a strong signal and read incoming signals can change drastically when the pair of interacting cells produce the ligand and receptor in a symmetric manner ( e . g . ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) for both cells ) rather than in an asymmetric manner ( e . g . ( νL1 , νR1 , νl1 , νr1 ) = ( 1 , 0 , 0 , 1 ) and ( νL2 , νR2 , νl2 , νr2 ) = ( 0 , 1 , 1 , 0 ) ) . The fold-increase in W12 is large even when self-binding confers no cost ( n=0 ) , while larger values for n ramp up the costs ( Figure 2c ) . If cells produce the ligand and receptor asymmetrically , self-binding ceases to be a problem in receiving incoming signals . Although the strength of the signaling interaction between two cells ( W12W21 ) may improve when the interacting cells produce the ligand and receptor asymmetrically , this need not be the case . Consider the interaction of a resident cell with production rates ( νL , νR , νl , νr ) res= ( 1 , 1 , 0 , 0 ) with itself and a mutant cell with production rates given by ( νL , νR , νl , νr ) mut= ( 1-dx , 1-dy , dx , dy ) . For all values of dx and dy , [W12W21]res+mut-[W12W21]res+res < 0 ( Figure 3a ) . It follows that ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) cannot be invaded by any single mutant . However , if the resident is already slightly asymmetric , for example ( νL , νR , νl , νr ) res= ( 1 , 0 . 9 , 0 , 0 . 1 ) , then a mutant conferring an asymmetry in the opposite direction can be better at interacting with the resident ( Figure 3b ) . When the resident produces both ligand and receptor equally ( e . g . ( νL , νR , νl , νr ) res= ( 0 . 5 , 0 . 5 , 0 . 5 , 0 . 5 ) ; Figure 3c ) , then most mutants conferring an asymmetry in either ligand or receptor production are favored . The strongest interaction occurs with mutants that produce the ligand or receptor fully asymmetrically ( i . e . ( νL , νR , νl , νr ) mut= ( 1 , 0 , 0 , 1 ) or ( 0 , 1 , 1 , 0 ) ; ( Figure 3c ) ) . Finally , when the resident production rates are already strongly asymmetric given by ( νL , νR , νl , νr ) res= ( 1 , 0 , 0 , 1 ) , a mutant with an asymmetry in the opposite direction is most strongly favored ( Figure 3d ) . Note that a population composed only of cells with production rates at ( νL , νR , νl , νr ) res= ( 1 , 0 , 0 , 1 ) is not viable since the probability that two such cells mate is zero . However , this analysis provides insight about how asymmetry in signaling evolves . To explore the evolution of signaling asymmetry , we follow mutations that alter the relative production of two mutually incompatible types of ligand and receptor ( L , R ) and ( l , r ) . To ease understanding , the population symmetry s in the production of ligand and receptor is measured , ( 11 ) s=1-12N∑i=1N ( |νLi-νRi|+|νli-νri| ) . The population is symmetric ( s=1 ) if cells produce ligand and receptor equally , for both types ( i . e . ( νR , νL , νr , νl ) = ( a , a , 1-a , 1-a ) , for constant a ) , and fully asymmetric ( s=0 ) when cells adopt polarized roles ( i . e . ( νL , νR , νl , νr ) = ( 1 , 0 , 0 , 1 ) and ( 0 , 1 , 1 , 0 ) ) . Starting from a population where all cells are symmetric producers of only one ligand and receptor , ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) , the population evolves to one of two equilibria ( Figure 4a ) . E1 where s*≈1 and all cells produce the ligand and receptor symmetrically ( νL , νR , νl , νr ) ≈ ( 1 , 1 , 0 , 0 ) or E2 where s*≈0 and the population is divided into ligand and receptor producing cells , with equal frequencies of ( νL , νR , νl , νr ) ≈ ( 1 , 0 , 0 , 1 ) and ( νL , νR , νl , νr ) ≈ ( 0 , 1 , 1 , 0 ) ( Figure 4b , c ) . Equilibria with intermediate values of s* are not found . The exact production rates at E1 and E2 exhibit some degree of noise due to mutation and finite population size ( Figure 4b , c ) . At E2 , individual cells with high νR ( and low νr ) have low νL ( and high νl ) , confirming that s*≈0 captures a fully asymmetric steady state ( Figure 4b , c ) . Whether E2 is reached from E1 depends on key parameters that determine the strength of self-binding and signaling interactions between cells . E1 persists and no asymmetry evolves when k+ ( the intracellular ligand-receptor binding coefficient ) is small ( Figure 4d ) . In this case , the concentration of self-bound ligand-receptor complex is small ( Equation ( 6 ) ) and there is little cost of self-signaling ( Equation ( 8 ) ) , so there is weak selection in favor of asymmetry . When the population is at E1 , asymmetric mutants are slightly deleterious on their own ( Figure 3a ) . They are therefore more likely to be lost when k+ is small and selection for asymmetric signaling is weak ( Figure 4d ) . The opposite is true for larger values of k+ , as self-binding now dominates and restricts between cell signaling , promoting the evolution of asymmetry ( Figure 4d ) . The transition from E1 to E2 occurs at a smaller value of k+ when the degradation rate ( γ ) is decreased ( Figure 4d ) , as the effective removal of free ligand and receptor depends more strongly on intercellular binding ( Figure 2a , b ) . Furthermore , the mutation rate affects the value of k+ at which the transition from E1 to E2 occurs . The transition from E1 to E2 when mutation rates are smaller occurs at larger k+ ( Figure 4—figure supplement 1 ) . We further explore the role of the mutational process below . Another important consideration is the relative strength of signaling within and between cells , given by k+/k- and kb respectively . For example , the threshold value of the within cell binding rate beyond which symmetric signaling ( E1 ) evolves to asymmetric signaling ( E2 , Figure 4a ) increases when kb becomes much larger than k+/k- ( Figure 4e ) . Furthermore , this threshold value is smaller for larger values of n indicating that asymmetric signaling is more likely to evolve when the cost for self-signaling is higher ( larger n , Figure 4e ) . However , asymmetric signaling can evolve even when self-binding carries no cost ( n=0 ) as high rates of self-binding can restrict the number of ligand and receptor molecules free for between cell interactions ( Figure 4e ) . We also wondered how the relative synergy or competition between the two ligands ( or receptors ) could affect our results . When the two ligands ( or receptors ) exhibit synergy so that νL+νl<α and νR+νr<α for α>1 , a signaling asymmetry evolves more easily ( for smaller values of k+ , Figure 4—figure supplement 2 ) . Now the second ligand ( or receptor ) begins to evolve without imposing a cost on the preexisting ligand ( or receptor ) and can therefore remain present in the population longer until an asymmetry in the opposite direction evolves in other cells . The reverse dynamics are observed when the two ligands ( or receptors ) compete with one another ( νL+νl<α and νR+νr<α for α<1 ) ( Figure 4—figure supplement 2 ) . The observations above suggest that both E1 and E2 are evolutionary stable states and the transition from E1 to E2 depends on the mutational process , drift and the parameters that determine signaling interactions . To explore this we investigated the stability of E1 in response to rare mutations in the receptor and ligand production rates . We assume the population is initially at E1 ( i . e . ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) ) , introduce mutations in the receptor and ligand loci ( νL , νR , νl , νr ) = ( 1-dx , 1 , dx , 0 ) and ( νL , νR , νl , νr ) = ( 1 , 1-dy , 0 , dy ) at frequency p , and calculate the population symmetry at steady state for different values of dx and dy ( Figure 5 ) . Single mutations never spread ( i . e . if dx=0 no value of dy allows mutants to spread and vice versa ) . This is in agreement with the analytical predictions presented in the previous section ( Figure 3a ) . When both dx and dy are nonzero the population may evolve to E2 , where the two mutants reach equal frequencies at ~0 . 5 and replace the resident . The basin of attraction for E2 ( and so asymmetric signaling roles ) is larger when k+ and p are high and γ is small ( Figure 5a–d ) , as predicted analytically ( Figures 2 and 3 ) and in accordance with our findings when mutations were continuous ( Figure 4 ) . Note that the initial mutation frequency ( p ) matters in our system . Single mutations are slightly deleterious on their own as predicted analytically ( Figure 3a ) and seen here when dx=0 or dy=0 ( Figure 5 ) . The two mutants , however , can be favored when they are asymmetric in opposite directions ( i . e . dx>0 and dy>0; Figure 5 ) . When mutants are introduced at a lower frequency ( compare Figure 5a–b ) , the probability that they meet one another before they are lost by drift increases . This explains why smaller values of p result in narrower basins of attraction for E2 ( Figure 5a–b ) . We next investigated how mutations invade when the resident already signals asymmetrically ( i . e . produces both ligands ) . The resident was set to ( νL , νR , νl , νr ) res= ( 1-dx , 1 , dx , 0 ) and a mutant able to produce both receptors ( νL , νR , νl , νr ) mut= ( 1 , 1-dy , 0 , dy ) was introduced . If dx>0 , a mutant conveying a small asymmetry in receptor production ( i . e . dy>0 ) increases in frequency until the population reaches a polymorphic state with the resident and mutant at 50% ( Figure 6a ) . If dx>0 but the mutant only produces one receptor ( i . e . dy=0 ) , the mutant invades , reaching a low frequency when dx is small and replaces the resident when dx is large . It follows that an asymmetry in both ligand and receptor production is necessary for the evolution of a signaling asymmetry as predicted analytically ( Figure 3a ) . We also consider a resident type that produces both ligands and both receptors with some degree of asymmetry in ligand production ( i . e . ( νL , νR , νl , νr ) res= ( 0 . 5-dx , 0 . 5 , 0 . 5+dx , 0 . 5 ) ) and map the spread of a mutant with asymmetry is receptor production ( νL , νR , νl , νr ) mut= ( 0 . 5 , 0 . 5-dy , 0 . 5 , 0 . 5+dy ) . The pairwise invasibility plots for values of dx and dy show that signaling asymmetries in opposite directions are favored . These evolve to a polymorphic state with equal frequencies of cells at dx=dy=-0 . 5 and dx=dy=0 . 5 ( Figure 6b ) . These findings together illustrate how the asymmetric state E2 evolves from the symmetric state E1 . The results above assume that the loci controlling ligand and receptor production are tightly linked which prevents the production of deleterious combinations following meiosis . Recombination is a minor problem at the E1 equilibrium which is monomorphic ( except for mutational variation ) . But it is likely to be a problem at the polymorphic E2 equilibrium . For example , at E2 mating between ( νL , νR , νl , νr ) = ( 1 , 0 , 0 , 1 ) and ( 0 , 1 , 1 , 0 ) cells generates non-asymmetric recombinant ligand-receptor combinations , either ( 1 , 1 , 0 , 0 ) or ( 0 , 0 , 1 , 1 ) . To implement recombination we assume that the two ligands are tightly linked in a single locus and are inherited as a pair ( likewise the two receptors ) , and investigate the effects of recombination between the ligand locus and the receptor locus . Note that if we allow recombination between ligands ( or receptors ) , this would be expected to generate combinations with a similar deleterious impact . Consider the effect of recombination on a population at E1 . As before , the population either stays at E1 or evolves to E2 dependent on parameter values ( Figure 7a ) . When the population evolves to E2 , s* becomes larger as the recombination rate , ( ρ ) , increases ( Figure 7b ) . For low recombination rates ( ρ≤0 . 1 ) , the population largely consists of equal frequencies of ( 1 , 0 , 0 , 1 ) and ( 0 , 1 , 1 , 0 ) cells , producing the ligand and receptor asymmetrically . A small percentage of recombinant cells produce conspecific pairs of ligand and receptor ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) and ( 0 , 0 , 1 , 1 ) ( Figure 7b , c ) . Recombination in this case creates ‘macromutations’ where production rates that were 0 become 1 and vice versa . As the recombination rate rises ( ρ≥0 . 2 ) , the two leading cell types diverge from ( νL , νR , νl , νr ) = ( 1 , 0 , 0 , 1 ) and ( 0 , 1 , 1 , 0 ) towards ( 1−ϵ1 , ϵ2 , ϵ3 , 1−ϵ4 ) and ( ϵ5 , 1−ϵ6 , 1−ϵ7 , ϵ8 ) where the ϵi are below 0 . 5 but greater than zero Figure 7d ) . Higher recombination rates ( ρ≥0 . 3 ) push s*=0 . 5 at E2 ( Figure 7b ) . Ηere , there is a predominance of ( νL , νR , νl , νr ) = ( 1 , 0 . 5 , 0 , 0 . 5 ) and ( 0 , 0 . 5 , 1 , 0 . 5 ) cells at equal frequencies ( or ( 0 . 5 , 1 , 0 . 5 , 0 ) and ( 0 . 5 , 0 , 0 . 5 , 1 ) by symmetry ) . This arrangement is robust to recombination since the receptor locus is fixed at ( νR , νr ) = ( 0 . 5 , 0 . 5 ) and the ligand locus is either at ( νL , νl ) = ( 1 , 0 ) or ( 0 , 1 ) ( or ( νL , νl ) = ( 0 . 5 , 0 . 5 ) ) and the receptor is either at ( νR , νr ) = ( 1 , 0 ) or ( 0 , 1 ) ) . So pairing between these two cell types results in ( 1 , 0 . 5 , 0 , 0 . 5 ) and ( 0 , 0 . 5 , 1 , 0 . 5 ) offspring , whether recombination occurs or not . Note that this arrangement maintains some degree of asymmetry even with free recombination ( ρ=0 . 5 ) . Even though both cell types produce both receptors , they produce the ligand asymmetrically ( or vice versa ) . Cells on average are more likely to mate successfully between rather than within the two types of cells . Similar to the case of no recombination , the invasion of E1 by E2 depends on the mutational process and parameter values . Figure 7f shows the steady state symmetry measure in a population initially at ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) when two mutations ( 1-dx , 1 , dx , 0 ) and ( 1 , 1-dy , 0 , dy ) are introduced at low frequencies . Whether or not the mutants invade depends on the magnitude of the mutation in a similar way as in the case of no recombination ( Figure 5d versus Figure 7f ) . However , the value of s* now diverges from zero reflecting the nonzero rate of recombination . In the analysis above , recombination between the ligand and receptor loci is fixed . However , the recombination rate itself can evolve . To investigate this , we let the recombination rate ρ undergo recurrent mutation with probability μρ so that the mutant recombination rate becomes ρ′=ρ+ερ with ερ∼N ( 0 , σρ ) . In a diploid zygote , the rate of recombination is given by the average of the two recombination alleles , ρ1 and ρ2 , carried by the mating cells . In this way , the recombination rate evolves together with the ligand and receptor production rates . We start with maximal recombination rate ρ=0 . 5 and ( νL , νR , νl , νr ) = ( 1 , 1 , 0 , 0 ) for all cells and allow the recombination rate to evolve by drift for 1000 generation before we introduce mutation in the ligand and receptor loci . The recombination rate evolves to ρ*=0 whenever E2 was reached from E1 in the no recombination analysis . Under these conditions , tight linkage between receptor and ligand genes is favored ( Figure 8a ) . Furthermore , asymmetric signaling roles coevolve together with the recombination rate . The evolved trajectories of s and ρ depend on the strength of selection for asymmetric signaling . For example , when k+ is large ( k+=10 ) , signal asymmetry rapidly evolves; s moves away from one and this is followed by a sharp drop in the recombination rate ( Figure 8b ) . Eventually the population evolves asymmetric signaling roles ( s in orange , Figure 8b ) and tight linkage ( ρ in blue , Figure 8b ) . These dynamics are similar when k+ is smaller ( k+=3 , Figure 8c ) and selection for asymmetry is weaker . However , it now takes longer for the asymmetric types to co-evolve ( Figure 8c ) . When selection for asymmetric signaling is even weaker ( k+=1 , Figure 8d ) , no asymmetry evolves ( s remains at 1 ) and the recombination rate fluctuates randomly between its minimum and maximum value as one would expect in the case of a neutral allele .
Explaining the evolution of mating types in isogamous organisms constitutes a major milestone in understanding the evolution of anisogamy and sexes ( Randerson and Hurst , 2001; Lehtonen et al . , 2016 ) . Mating type identity is determined by a number of genes that reside in regions of suppressed recombination and code for ligands and receptors that guide partner attraction and recognition , as well as genes that orchestrate cell fusion and postzygotic events ( Billiard et al . , 2011; Perrin , 2012; Hadjivasiliou and Pomiankowski , 2016; Branco et al . , 2018 ) . In this work we show that an asymmetry in ligand and receptor production evolves as a response to selection for swift gamete communication and mating . Furthermore , the same conditions favoring asymmetric signaling select for tight linkage between the receptor and ligand genes . Our findings indicate that selection for asymmetric signaling roles could have played an important role in the early evolution of gamete differentiation and identity . We investigated the evolution of mating type roles by considering two types of ligand and receptor in individual cells . Gene duplication followed by mutation is a well established route to novelty evolution ( Susumu , 1970; Zhang , 2003; Magadum et al . , 2013 ) and could explain the co-existence of two pairs of ligand and receptor in our system . Alternatively , individual cells could produce multiple ligands and receptors which evolve independently , as is the case in some basidiomycete fungi ( Fowler and Vaillancourt , 2007 ) . The production rate of the two types of ligand ( and receptor ) in our system is subject to mutation using an assumption of infinite alleles ( Tajima , 1996 ) , so that the amount of expressed ligand ( and receptor ) of each kind is modulated quantitatively . In this way we were able to explicitly express the likelihood of mating as a function of the amount of free and bound molecules on the cell membrane and the ability of cells to accurately read their partner’s signal . This framework allowed us to follow the evolution of the quantitative production of ligand and receptor in mating cells for the first time . We found that the ligand-receptor binding rate within a cell ( k+ ) is key in the evolution of asymmetric signaling roles ( Figures 3 and 4 ) . k+ holds an important role because it dictates the rate at which free ligand and receptor molecules are removed from the cell surface . In addition , k+ determines the amount of intracellular signal that interferes with the ability of cells to interpret incoming signal . Although in theory cells could avoid self-binding ( by reducing k+ to zero ) , there is likely to be a strong association of the within-cell and between-cell binding affinities . So reductions in k+ are likely to have knock-on costs in reducing kb as well . An extreme example is the case of locally diffusible signals ( Figure 1 ) , such as those used by ciliates and yeasts to stimulate and coordinate fusion ( Sugiura et al . , 2010; Merlini et al . , 2013 ) . Here binding affinities between and within cells are inevitably identical ( since the ligand is not membrane bound ) . Work in yeast cells has shown that secreted ligands utilized for intercellular signaling during sex are poorly read by cells that both send and receive the same ligand ( Youk and Lim , 2014 ) . In the case of strictly membrane bound molecules avoiding self-binding could also be an issue as it requires a ligand and receptor pair that bind poorly within a cell without compromising intercellular binding . For example , choosy budding yeast gametes ( which are better at discriminating between species ) take longer to mate ( Rogers et al . , 2015 ) . It would be interesting to further study these trade-offs experimentally . We never observed the co-existence of a symmetric ‘pansexual’ type with asymmetric self-incompatible types . The two steady states consist of either a pansexual type alone or two mating types with asymmetric signaling roles . This could explain why the co-existence of mating types with pansexuals is rare in natural populations ( Billiard et al . , 2011; Billiard et al . , 2012 ) . This is in contrast to previous models where pansexual types were very hard to eliminate due to negative frequency dependent selection ( Hoekstra , 1982; Czárán and Hoekstra , 2004; Hadjivasiliou et al . , 2013 ) . For example , in the case of the mitochondrial inheritance model , uniparental inheritance raises fitness not only in individuals that carry genes for uniparental inheritance but also for pansexual individuals ( benefits ‘leak’ to biparental individuals ) ( Hadjivasiliou et al . , 2013; Christie and Beekman , 2017b ) . A similar pattern is seen with inbreeding avoidance because the spread of self-incompatibility reduces the population mutation load , and so reduces the need for inbreeding avoidance ( Czárán and Hoekstra , 2004 ) . These dynamics are reversed in the present model where there is positive frequency dependent selection . The spread of asymmetric signalers generates stronger selection for further asymmetry ( Figures 3 and 4 ) . This also occurs when there is recombination ( Figures 7 and 8 ) . Even though recombination between the two asymmetric types generates symmetric recombinant offspring , these are disfavored and eliminated by selection . These observations suggest that the mitochondrial inheritance and inbreeding avoidance models are unlikely to generate strong selection for suppressed recombination which is the hallmark of mating types . Finally , it would be interesting to explore how the reinstatement of recombination could be a route back to homothallism which is a state derived from species with mating types ( Billiard et al . , 2011 ) . Mating type identity in unicellular eukaryotes is determined by mating type loci that typically carry a number of genes ( Billiard et al . , 2012; Hadjivasiliou and Pomiankowski , 2016 ) . Suppressed recombination at the mating type locus is a common feature across the evolutionary tree ( Branco et al . , 2018 ) . Our work predicts the co-evolution of mating type specific signaling roles and suppressed recombination with selection favoring linkage between loci responsible for signaling and an asymmetry in signaling roles . This finding suggests that selection for asymmetric signaling could be the very first step in the evolution of tight linkage between genes that control mating type identity . In yeasts , the only genes in the mating type locus code for the production of ligand and receptor molecules ( Merlini et al . , 2013 ) . These then trigger a cascade of other signals downstream that also operate asymmetrically . Evidence across species suggests that mating type loci with suppressed recombination are precursors to sex chromosomes ( Menkis et al . , 2008; Geng et al . , 2014 ) . In this way our work provides crucial insights about the origin of sex chromosomes . The framework developed here could be used together with recent efforts to understand numerous features of mating type evolution . For example , opposite mating type gametes often utilize diffusible signals to attract partners ( Luporini et al . , 1995; Tsuchikane et al . , 2005 ) . The inclusion of long range signals such as those used in sexual chemotaxis will provide further benefits for asymmetric signaling roles and mating types ( Hadjivasiliou et al . , 2015 ) . Furthermore the number of mating types varies greatly across species and is likely to depend on the frequency of sexual reproduction and mutation rates ( Constable and Kokko , 2018 ) . Signaling interactions between gametes could also play a role in determining the number of mating types and reducing their number to only two in many species ( Hadjivasiliou et al . , 2016b ) . It would be interesting to use the framework developed here to study the evolution of additional ligands and receptors and their role in reaching an optimal number of mating types . Other important features such as the mechanism of mating type determination ( Billiard et al . , 2011; Vuilleumier et al . , 2013 ) and stochasticity in mating type identity ( Hadjivasiliou et al . , 2016b; Nieuwenhuis and Immler , 2016; Nieuwenhuis et al . , 2018 ) could also be understood in light of this work . Our analysis revealed that the evolution of asymmetric gamete signaling and mating types is contingent upon the mutation rate . Single mutants that exhibit an asymmetry are initially slightly disadvantageous . When further mutations emerge that are asymmetric in the opposite direction , a positive interaction between these mutants occurs that can lead to the evolution of distinct mating types . When the population size is small and mutation rates are low , there is a low probability that individuals carrying asymmetric mutations in opposite directions are segregating at the same time . Increasing the population size or the mutation rate would enhance the probability of co-segregation , making the evolution of asymmetric signaling more likely . In an infinite population the evolution of signaling asymmetry should be independent of the mutation rate . Finally , it is worth noting that unicellular eukaryotes undergo several rounds of asexual growth ( tens to thousands ) between each sexual reproduction ( Hadjivasiliou et al . , 2016b; Constable and Kokko , 2018 ) . It follows that the effective mutation rate between sexual rounds will end up being orders of magnitude higher than the mutation rates at each vegetative step . Taken together our findings suggest that selection for swift and robust signaling interactions between mating cells can lead to the evolution of self-incompatible mating types determined at non-recombinant mating type loci . We conclude that the fundamental selection for asymmetric signaling between mating cells could be the very first step in the evolution of sexual asymmetry , paving the way for the evolution of anisogamy , sex chromosomes and sexes .
We model N cells so that each cell is individually characterized by a ligand locus ℒ and a receptor locus ℛ . Two ligand genes at the locus ℒ determine the production rates for two ligand types l and L given by νl and νL . Similarly , two receptor genes at the locus ℛ determine the production rates for the two receptor types r and R given by νr and νL . The two ligand and receptor genes in our model could could arise from duplication followed by mutation that leaves two closely linked genes that code for different molecules . In our computational set-up each cell is associated with production rates νl , νL , νr and νR where we assume a normalized upper bound so that νl+νL<1 and νr+νR<1 . The steady state concentrations for L , R , and LR are computed by setting d[L]dt=d[R]dt=d[LR]dt=0 in Equations ( 1-3 ) and solving the resulting quadratic equations . This leads two solutions only one of which gives positive concentrations . It follows that there is a unique physical solution to our system , which is what we use to define the probability of mating in our numerical simulations . The program is initiated with νL=νR=1 and νl=νr=0 for all cells ( unless otherwise stated , see Section 5 . 4 ) . We introduce mutation so that the ligand and receptor production rates of individual cells mutate independently with probability μ . A mutation event at a production gene changes the production rate by an increment ϵ where ϵ∼N ( 0 , σ ) . Mutation events at the different genes l , L , r and R are independent of one another . If νl+νL>1 or νl+νL>1 the production rates are renormalized so their sum is capped at 1 . If a mutation leads to a production rate below 0 or above one it is ignored and the production rate does not change . We implement mating by randomly sampling individual cells . The probability that two cells mate is determined by their ligand and receptor production rates as defined in Equation ( 9 ) in the main text . We assume that K takes a large value relative to W12W21 so that P is linear in W12W21 . Because the absolute value for W12W21 varies greatly between parameter sets , and what we are interested in is the relative change in W12W21 when signaling levels change , we chose K to be equal to the maximum value W12W21 can take for a given choice of γ , k+ , k- and kb . Sampled cells that do not mate are returned to the pool of unmated cells . This process is repeated until M=N/2 cells have successfully mated . This produces N/4 pairs of cells each of which gives rise to two offspring . These are sampled with replacement until the population returns to size N . We assume that a mutation-selection balance has been reached when the absolute change in s , defined in Equation ( 10 ) in the main text , between time steps t1 and t2 is below ϵ=10−5 across t2-t1=100 . Certain parameter sets resulted in noisy steady states and were terminated following 105 generations . The numerical code keeps track of all production rates for individual cells over time . We model adaptive dynamics by initiating the entire population at state ( νL , νR , νl , νr ) res and introducing a mutant ( νL , νR , νl , νr ) mut at low frequency p . We allow the population to evolve according to the life cycle introduced in the main text and record the frequency of the resident and mutant type when a steady state is reached . For the purposes of Figure 5 , the resident type is set to ( νL , νR , νl , νr ) res and two mutants ( νL , νR , νl , νr ) mut1 and ( νL , νR , νl , νr ) mut2 are introduced both at frequency p . In this case we track the frequencies of the resident and both mutants until steady state is reached . We define steady state as the point where the average value of s in the population between time steps t1 and t2 is below ϵ=10-7 across t2-t1=100 . The population always reached steady state . We implement recombination by considering a modifier ℳ that lies between the ligand and receptor loci ℒ and ℛ . That is , we assume that the two ligand genes and two receptor genes are tightly linked on the ligand and receptor locus ℒ and ℛ respectively , and only model recombination between the two loci . For simplicity , we assume that ℳ determines the physical distance between ℒ and ℛ so that the distances ℒ-ℳ and ℛ-ℳ are the same . The modifier ℳ determines the rate of recombination between the ligand and receptor loci quantitatively by determining ρM , the probability of a single recombination event following mating . Consider for example two individuals whose ligand and receptor production rates and recombination rates are determined by the triplets R1-M1-L1 and R2-M2-L2 , the possible offspring resulting from such a mating are given by , where ρM1 , 2=12 ( ρM1+ρM2 ) is the joint recombination rate when cell1 and cell2 with recombination rates ρM1 and ρM2 respectively mate . We allow mutation at the recombination locus at rate μρ independently of the ligand and receptor loci . A mutation event leads to a new recombination rate so that ρM′=ρM-ϵ for ϵ∼N ( 0 , σρ ) . We assume that the mutation-selection balance has been reached when the absolute change in s , defined in Equation ( 10 ) in the main text , and the change in the average recombination rate between time steps t1 and t2 is below ϵ=10-5 across t2-t1=100 . | Sexual reproduction , from birds to bees , relies on distinct classes of sex cells , known as gametes , fusing together . Most single cell organisms , rather than producing eggs and sperm , have similar sized gametes that fall into distinct ‘mating types’ . However , only sex cells belonging to different mating types can fuse together and sexually reproduce . At first glance , it seems illogical that cells from the same mating type cannot reproduce with each other , as this restricts eligible partners within a population and makes finding a mate more difficult . Yet the typical pattern amongst single cell organisms is still two distinct classes of sex cells , just as in birds and bees . How did the obsession with mating between two different types become favored during evolution ? One possibility is that cells with different mating types can recognize and communicate with each other more easily . Cells communicate by releasing proteins known as ligands , which bind to specific receptors found on the cell’s surface . Using mathematical modelling , Hadjivasiliou and Pomiankowski showed that natural selection typically favors ‘asymmetric’ signaling , whereby cells evolve to either produce receptor A with ligand B , or have the reverse pattern and produce receptor B with ligand A . These asymmetric mutants are favored because they avoid producing ligands that clog or activate the receptors on their own surface . As a result , different types of cells are better at recognizing each other and mating more quickly . When cells sexually reproduce they exchange genetic material with each other to produce offspring with a combination of genes that differ to their own . However , if the genes coding for ligand and receptor pairs were constantly being ‘swapped’ , this could lead to new combinations , and a loss of asymmetric signaling . Hadjivasiliou and Pomiankowski showed that for asymmetric signaling to evolve , natural selection favors the genes encoding these non-compatible ligand and receptor pairs to be closely linked within the genome . This ensures that the mis-matching ligand and receptor are inherited together , preventing cells from producing pairs which can bind to themselves . This study provides an original way to address an evolutionary question which has long puzzled biologists . These findings raise further questions about how gametes evolved to become the sperm and egg , and how factors such as signaling between cells can determine the sex of more complex organisms , such as ourselves . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2019 | Evolution of asymmetric gamete signaling and suppressed recombination at the mating type locus |
The anterior thalamic nucleus ( ATN ) is thought to play an important role in a brain network involving the hippocampus and neocortex , which enables human memories to be formed . However , its small size and location deep within the brain have impeded direct investigation in humans with non-invasive techniques . Here we provide direct evidence for a functional role for the ATN in memory formation from rare simultaneous human intrathalamic and scalp electroencephalogram ( EEG ) recordings from eight volunteering patients receiving intrathalamic electrodes implanted for the treatment of epilepsy , demonstrating real-time communication between neocortex and ATN during successful memory encoding . Neocortical-ATN theta oscillatory phase synchrony of local field potentials and neocortical-theta-to-ATN-gamma cross-frequency coupling during presentation of complex photographic scenes predicted later memory for the scenes , demonstrating a key role for the ATN in human memory encoding .
The anterior thalamic nuclei ( ATN ) are thought to play an important role in an extended hippocampal network central to memory formation ( encoding ) and novelty processing , which coordinates synaptic changes involving widespread neocortical areas , enabling life events to be recorded and later reinstated ( Knight , 1996; Aggleton et al . , 2010; Nyhus and Curran , 2011; Ritchey et al . , 2013; Schott et al . , 2013 ) . While corticothalamic interactions are well-known to be crucial for adaptive behavior ( Saalmann et al . , 2012 ) , the functional role of the ATN in humans has resisted investigation with non-invasive techniques owing both to its depth and small size . Here we had the rare opportunity to record electrophysiological activity during memory encoding directly from the ATN and dorsomedial thalamic nuclei ( DMTN ) of eight epileptic human volunteers with electrodes implanted for epilepsy treatment , as well as from frontal ( and in two cases , parietal ) scalp electrodes , reflecting neocortical activity ( Figure 1 ) . The DMTN are thought to be involved in executive control during memory retrieval ( Van der Werf et al . , 2003 ) and were hypothesized to play a lesser role than the ATN during encoding . 10 . 7554/eLife . 05352 . 003Figure 1 . Intracranial electrode location in Participant 1 . Left: Reconstruction of intrathalamic contact location using intraoperative X-ray image coordinates , superimposed on preoperative MRI scan . Dorsomedial thalamic nucleus ( DMTN ) : blue ( localized using masks from Wake Forest University Pick Atlas [http://fmri . wfubmc . edu/software/PickAtlas ( Maldjian et al . , 2003 ) ] warped into participant's brain space ) . Anterior thalamic nucleus ( ATN ) contacts: green ( left ) , red ( right ) . Middle: Most superficial contacts ( upper panel ) clearly lie in the ATN , by reference to Schaltenbrand atlas ( Schaltenbrand and Wahren , 1977 ) ( lower panel: A . pr . = nucleus anterior principalis ) . Right: Scalp electrode locations for this participant . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 003© 1977 Thieme Medical Publishers . All Rights Reserved . 1977Thieme Medical PublishersFigure 1 , lower panel , is reproduced from ( Schaltenbrand and Wahren , 1977 ) , Atlas for Stereotaxy of the Human Brain with permission . Indication that the ATN has a role in memory processing comes from human lesion and animal studies . Damage to the human ATN results in amnesia for new episodes ( Harding et al . , 2000; Van der Werf et al . , 2003; Aggleton et al . , 2010 ) , and reciprocal frontal and parietal connections with the ATN ( Aggleton , 2012 ) have led to the hypothesis that ATN-hippocampal connections play a regulatory role in encoding ( Vertes et al . , 2001; Aggleton et al . , 2010 ) . Notably , 75% of ATN oscillatory power in non-human animal studies is in the theta ( 4–8 Hz ) range , the dominant hippocampal rhythm ( Vertes et al . , 2001 ) , which is also implicated in ATN-hippocampal communication ( Vertes et al . , 2001; Aggleton et al . , 2010 ) . Neural communication and synaptic plasticity rely on long-range phase–phase and phase–amplitude synchrony of neural oscillations ( Lachaux et al . , 1999; Lisman and Jensen , 2013 ) . Fronto-hippocampal theta synchrony accompanies memory formation ( Benchenane et al . , 2010; Nyhus and Curran , 2011 ) . Furthermore , theta synchrony is reported to bind medial temporal ( MT ) and parietal areas during associative encoding ( Crespo-Garcia et al . , 2010 ) , and MT with frontal and parietal cortex during successful retrieval ( Watrous et al . , 2013 ) . Neocortical and hippocampal gamma ( >30 Hz ) oscillations appear to reflect local processing related to activation and maintenance of neuronal object representations ( Jensen et al . , 2007 ) , binding diverse perceptual and contextual information ( Nyhus and Curran , 2011 ) . Spatially separate gamma oscillations have been found to be locked to the phase of the theta oscillation , supporting binding of the coherent ensemble underlying a given memory trace ( Mizuhara and Yamaguchi , 2011 ) . Indeed such gamma-power-to-theta-phase cross-frequency coupling ( CFC ) has been identified within human neocortex during word recognition ( Canolty et al . , 2006 ) , working memory maintenance ( Axmacher et al . , 2010 ) , and in successful long-term memory encoding ( Friese et al . , 2012 ) , as well as within the rat hippocampus ( Colgin et al . , 2009 ) , and has been proposed as a mechanism for transiently coupling distributed cortical activity ( Canolty et al . , 2006; Friese et al . , 2012; Lisman and Jensen , 2013 ) . Together , these findings suggest that neocortical-ATN communication might be related to local ATN processing during encoding . We hypothesized that neocortical-ATN theta phase synchrony , and the relationship between theta phase and local ATN gamma amplitudes , would be critical for memory encoding .
We assessed the role of the ATN in memory encoding by contrasting electrophysiological activity during successful compared with unsuccessful encoding of serially presented photographs of 200 complex indoor and outdoor scenes . Participants judged whether each photograph depicted an indoor or an outdoor scene . Successful encoding was defined as correct recognition of a scene as old on a subsequent recognition memory test combining photographs of old and similar new scenes . All eight participants were able to discriminate old from new scenes ( Table 1 ) . Mean successful encoding across participants was 55% , and mean unsuccessful encoding was 45% , resulting in means of 101 and 87 observations per category , respectively , for EEG analysis . 10 . 7554/eLife . 05352 . 004Table 1 . Behavioral resultsDOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 004PtSE % ( NE ) SE-C %UE % ( NE ) FA % ( NE ) FA-C%CR % ( NE ) SE-FA %SE-C–FA-C %174 ( 74 ) 7426 ( 26 ) 29 ( 29 ) 2571 ( 71 ) 4549266 ( 132 ) 39 . 534 ( 68 ) 36 ( 72 ) 6 . 564 ( 128 ) 3033342 . 5 ( 85 ) 35 . 557 . 5 ( 115 ) 5 ( 10 ) 1 . 595 ( 190 ) 37 . 534436 . 5 ( 73 ) 27 . 563 . 5 ( 127 ) 21 ( 42 ) 6 . 579 ( 158 ) 15 . 521557 . 5 ( 115 ) 3842 . 5 ( 85 ) 47 ( 94 ) 13 . 553 ( 106 ) 10 . 524 . 5638 . 5 ( 77 ) 36 . 561 . 5 ( 123 ) 10 ( 20 ) 490 ( 180 ) 28 . 532 . 5781 . 5 ( 163 ) 75 . 518 . 5 ( 37 ) 63 ( 126 ) 28 . 537 ( 74 ) 18 . 547843 . 5 ( 87 ) 41 . 556 . 5 ( 113 ) 25 ( 50 ) 1075 ( 150 ) 18 . 531 . 5Mean55 ( 101 ) 4645 ( 87 ) 29 . 511 . 970 . 525 . 534 . 1SD17 . 318 . 217 . 319 . 19 . 919 . 111 . 89 . 3Pt = Participant . SE = successful encoding ( hits ) . NE = number of epochs . SE-C = correctly judged ‘old’ with high confidence . UE = unsuccessful encoding ( misses ) . FA = false alarms . FA-C = incorrectly judged ‘old’ with high confidence . CR = correct rejections . SD = standard deviation . SE-FA = index of ability to discriminate between old and new test items ( i . e . , hits minus false alarms ) . SE-C–FA-C = discrimination index for items confidently judged old . All significance tests reported here were two-tailed . Response times for the indoor/outdoor judgment at encoding were compared between successfully ( group mean = 1 . 04 s ) and unsuccessfully ( group mean = 1 . 14 s ) encoded scenes for the seven participants for whom the paradigm was identical ( see ‘Materials and methods’ ) , and no difference was detected ( paired T-test: T = 1 . 81 , p = 0 . 12 , with 6° of freedom , DF ) . For each participant , the difference between their expected probability of sequential successful encoding , calculated according to their overall rate of successful encoding , and their observed probability of sequential successful encoding , was calculated . The mean difference between these probabilities across the group did not significantly differ from zero ( one-sample T-test: T = 1 . 35 , p = 0 . 22 , with 7 DF ) . Both these behavioral findings suggest that the neural findings to be reported did not reflect simple global attentional fluctuations ( see ‘Discussion’ ) . We contrasted key oscillatory features of the EEG ( 512 Hz sampling frequency ) during successful compared with unsuccessful encoding . To assess long-range communication , we calculated an amplitude-independent measure of phase synchrony ( Lachaux et al . , 1999 ) of oscillations in local field potentials recorded from thalamus and neocortex . We then employed CFC to assess the relationship between long-range communication and local neural processing , and Granger causality ( GC ) to assess likely direction of influence . Mean fronto-thalamic phase-locking values ( PLVs ) across all eight participants during encoding are plotted against time and frequency in Figure 2 . Corticothalamic theta synchrony differences in a late ( 0 . 5–1 . 5 s ) time window were predicted , given the theoretical importance of theta oscillations in memory , and the typical late timing of encoding-related differences in frontal and parietal event-related potentials ( ERPs ) ( Schott et al . , 2002 ) and in post-stimulus MT theta oscillatory power ( Hanslmayr and Staudigl , 2014 ) . Frontal-right-ATN ( RATN ) theta synchrony was indeed greater during successful than unsuccessful encoding at 5–6 Hz between 0 . 5 and 1 . 5 s after picture presentation ( permutation tests , PT: p = 0 . 001; paired T-tests , TT: T = 9 . 9 , p = 0 . 000022; Figure 2 , Figure 2—figure supplement 1 ) . Conservative false discovery rate correction ( PT; 57 frequencies , 0–100 Hz; 1024 data-points , 0–2 s ) yielded a threshold of p = 0 . 0019 ( overall criterion p = 0 . 05 ) . Cluster-size PT ( CSPT ) on binary TT outcomes ( significant/nonsignificant at p = 0 . 05 ) showed the second cluster of this late theta synchrony to be significant ( ∼1 . 0–1 . 5 s; p = 0 . 016; observed contiguous cluster 437 pixels; criterial cluster 285 pixels for overall p = 0 . 05 ) . Synchrony at 5 . 2 Hz , averaged from 0 . 5 to 1 . 5 s for each participant , was greater during successful than unsuccessful encoding in seven of eight participants ( Figure 2E ) , yielding a group difference in median synchrony ( Wilcoxon test: p = 0 . 038 ) . Participant 4 ( Figure 2E ) also showed the difference from 1 to 1 . 5 s ( Figure 2F , Figure 2—figure supplement 2 ) , so that all eight participants showed a difference in this time-window . The time course of theta synchrony is shown in Figure 2—figure supplement 3 . Compared with surrogate data ( Theiler et al . , 1992 ) , the PLV was significant ( criterion p = 0 . 05 ) for 230 ms , which is more than a complete theta cycle at 5 Hz . The timescale is of the order of that over which synchrony is commonly detected ( Varela et al . , 2001 ) . Note from Figure 2F and Figure 2—supplement 2 that theta synchrony was greater during successful than unsuccessful encoding in all eight participants from 1 . 25–1 . 5 s post-stimulus , and the appearance of two separate episodes at the group level is likely to have arisen due to inter-participant differences in the timing of the synchrony episode over the 1 s time window from 0 . 5–1 . 5 s post-stimulus . 10 . 7554/eLife . 05352 . 005Figure 2 . Frontal-right anterior thalamic nucleus ( RATN ) phase synchrony . PLV = phase-locking value . ( A ) Successful encoding . ( B ) Unsuccessful encoding . ( C ) Successful minus unsuccessful encoding . ( D ) Permutation tests: Successful minus unsuccessful encoding . ( E ) Mean theta ( 5 . 2 Hz ) PLVs for successful encoding and unsuccessful encoding averaged from 0 . 5 to 1 . 5 s for the eight individual participants . ( F ) Number of participants showing greater theta synchrony during successful compared with unsuccessful encoding in four sub-time-windows from 0 . 5 to 1 . 5 s . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 00510 . 7554/eLife . 05352 . 006Figure 2—figure supplement 1 . Significance of phase synchrony between frontal neocortex and right anterior thalamic nucleus ( RATN ) . ( A ) Statistical significance ( uncorrected ) of the difference between phase-locking values ( PLVs ) during successful vs unsuccessful encoding using paired T-tests . The pattern of significance in early upper theta and alpha from 0 to 0 . 2 s ( T = 5 . 6 , p = 0 . 00085 ) and theta from 0 . 5 to 1 . 5 s ( T = 9 . 9 , p = 0 . 000022 ) resembles that detected by permutation testing ( see Figure 2D ) . The second theta cluster in the later time window was significant on cluster-size permutation testing ( p = 0 . 016 ) . ( B ) The time-frequency of PLVs that differed significantly between successful and unsuccessful encoding after false discovery rate correction of the permutation-test results shown in Figure 2D , which yielded p = 0 . 0019 for significance at criterion p = 0 . 05 . 1 = significant after correction . 0 = not significant after correction . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 00610 . 7554/eLife . 05352 . 007Figure 2—figure supplement 2 . Mean theta ( 5 . 2 Hz ) phase-locking value ( PLV ) between frontal neocortex and right anterior thalamic nucleus ( RATN ) over four consecutive time windows from 0 . 5 to 1 . 5 s . ( A ) Theta synchrony was greater during successful than during unsuccessful encoding in seven of eight participants . ( B ) Synchrony was greater during successful encoding in six of eight participants . ( C ) Synchrony was greater during successful encoding in seven of eight participants . ( D ) Synchrony was greater during successful encoding in all eight participants . ( Participant 4 showed greater synchrony during successful encoding from 1 to 1 . 5 s , but the overall difference as shown in Figure 2E was absent due to lesser synchrony from 0 . 5 to 1 s , as shown in Figure 2F ) . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 00710 . 7554/eLife . 05352 . 008Figure 2—figure supplement 3 . Time course of theta phase synchrony . Group mean theta ( 4–8 Hz ) phase-locking values ( PLVs ) during successful memory formation . Synchrony was significant ( criterion p = 0 . 05 ) compared with phase-scattered surrogate data in two episodes between 0 . 5 and 1 . 5 s post-stimulus . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 00810 . 7554/eLife . 05352 . 009Figure 2—figure supplement 4 . Frontothalamic synchrony involving other thalamic nuclei . Average difference in phase synchrony during successful compared with unsuccessful encoding . ( A ) Phase-locking values ( PLVs ) between frontal neocortex and right dorsomedial thalamic nucleus ( DMTN; seven participants: one participant had no DMTN contact ) . ( B ) PLVs between frontal neocortex and left anterior thalamic nucleus ( ATN ) . Significant synchrony patterns ( Figure 2—figure supplement 1 ) were observed only in the right ATN . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 00910 . 7554/eLife . 05352 . 010Figure 2—figure supplement 5 . Synchrony using 1-cycle wavelets to enhance time resolution . PLV = phase-locking value . Average phase synchrony between frontal neocortex and right anterior thalamic nucleus ( RATN ) for all eight participants during successful encoding . The early upper theta and alpha synchrony ( Figure 2—figure supplement 1 ) was post-stimulus . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01010 . 7554/eLife . 05352 . 011Figure 2—figure supplement 6 . Corticothalamic phase synchrony in Participant 1 . PLV = phase-locking value . Left: Phase synchrony between frontal and parietal neocortex and right anterior thalamic nucleus ( RATN ) during successful encoding . Right: Significance of difference in synchrony between successful and unsuccessful encoding ( uncorrected p values ) , calculated using permutation tests . ( A ) Synchrony with Fz ( frontal neocortex ) . ( B ) Significance of difference for Fz ( early p = 0 . 008; late p = 0 . 035 ) . ( C ) Synchrony for FCz ( frontocentral neocortex ) . ( D ) Significance of difference for FCz ( early: p = 0 . 037; late p = 0 . 018 ) . ( E ) Synchrony with P4 ( right parietal neocortex ) . ( F ) Significance of difference for P4 ( early: p = 0 . 019; late: p = 0 . 008 ) . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01110 . 7554/eLife . 05352 . 012Figure 2—figure supplement 7 . Corticothalamic phase synchrony in Participant 2 . PLV = phase-locking value . Left: Phase synchrony between frontal and parietal neocortex and right anterior thalamic nucleus ( RATN ) during successful encoding . Right: Significance of difference in synchrony between successful and unsuccessful encoding ( uncorrected p values ) , calculated using permutation tests . ( A ) Synchrony with AFz ( anterior frontal neocortex ) . ( B ) Significance of difference for AFz ( early: p = 0 . 025; late: p = 0 . 004 ) . ( C ) Synchrony with Fz ( frontal neocortex ) . ( D ) Significance of difference for Fz ( early: p = 0 . 046; late: p = 0 . 019 ) . ( E ) Synchrony with Pz ( central parietal neocortex ) . ( F ) Significance of difference for Pz ( early: p = 0 . 046; late: p = 0 . 007 ) . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01210 . 7554/eLife . 05352 . 013Figure 2—figure supplement 8 . Power difference between successful and unsuccessful encoding . Power during successful minus unsuccessful encoding , and significance of the difference ( uncorrected p values ) between mean power spectra calculated using paired T-tests . ( A ) Mean frontal theta power during successful minus during unsuccessful encoding . ( B ) Frontal theta power was somewhat greater during successful than during unsuccessful encoding in the time-window in which theta phase synchrony significantly differed ( 0 . 5–1 . 5 s post-stimulus ) , but the difference was not significant ( T = 1 . 90 , p = 0 . 10 ) . ( C ) Mean right anterior thalamic nucleus ( RATN ) theta power during successful minus during unsuccessful encoding . ( D ) RATN theta power differed significantly ( T = 4 . 01 , p = 0 . 0051 ) only briefly and at the end of the time-window in which theta phase synchrony significantly differed ( 0 . 5–1 . 5 s post-stimulus ) . The absence of frontal and RATN theta power differences at the time that memory-related frontal-RATN theta synchrony occurred confirms that this synchrony reflected timing rather than amplitude information . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01310 . 7554/eLife . 05352 . 014Figure 2—figure supplement 9 . Synchrony with differing cortical sites . The mean difference between frontal-right anterior thalamic nucleus phase-locking values ( PLVs ) during successful minus during unsuccessful encoding across participants . ( A ) Across the six participants with electrodes at Fpz . ( B ) Across the two participants with electrodes at AFz and Fz . ( C ) Across the two participants with electrodes at Pz . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 014 To further test frequency-specificity , a two-way repeated measures analysis of variance was applied to frontal-RATN PLVs obtained using theta ( 5 . 2 Hz ) and beta ( 17 . 5 Hz ) wavelets averaged from 0 . 5 to 1 . 5 s for each participant during successful and unsuccessful encoding . The mean PLVs showed a significant interaction between frequency and encoding success ( p < 0 . 001 ) . This interaction remained significant when taking adjacent theta ( 4 . 9 Hz and 5 . 5 Hz ) and beta ( 16 . 5 Hz and 18 . 6 Hz ) wavelets , and taking theta ( 5 . 2 Hz ) and alpha ( 11 . 7 Hz ) wavelets ( all ps < 0 . 002 ) . In all cases the advantage for successful over unsuccessful encoding for theta was greater than for alpha or beta , which showed negligible differences . The interaction was not significant when taking two theta wavelets ( 4 . 9 Hz and 5 . 2 Hz: p > 0 . 30; 5 . 2 Hz and 5 . 5 Hz: p > 0 . 60 ) . No significant difference was observed between delta ( 2–4 Hz ) phase synchrony during successful compared with unsuccessful encoding . We did not observe corresponding significant synchrony differences for DMTN or left ATN ( Figure 2—figure supplement 4 ) . Indeed , on direct comparison , the difference in theta ( 5 . 2 Hz ) PLVs between successful and unsuccessful encoding in the RATN 0 . 5–1 . 5 s post-stimulus was greater than the difference in the right DMTN ( Wilcoxon test: p = 0 . 018 ) , supporting nucleus specificity . A direct comparison of the relevant difference in theta PLVs between RATN and LATN only approached significance ( Wilcoxon test: p = 0 . 130 ) . While we do not wish to make strong claims about laterality , the finding that the synchrony difference in RATN was significant , whereas that in LATN was not ( Figure 2—figure supplement 4 ) , accords well with the non-verbal scene stimuli used ( Maillard et al . , 2011 ) ( see also ‘Discussion’ ) . There was additional early upper theta and alpha synchrony during successful compared with unsuccessful encoding ( ∼8–12 Hz; 0–0 . 2 s; PT: p = 0 . 002; TT: T = 5 . 6 , p = 0 . 00085; CSPT: p = 0 . 029; observed contiguous cluster 342 pixels; Figure 2 , Figure 2—figure supplement 1 ) , which might reflect enhanced item-specific attention and perception during successful encoding ( Düzel et al . , 2005 ) . All synchrony patterns were post-stimulus ( Figure 2—figure supplement 5 ) . There was a significantly greater synchrony difference between successful and unsuccessful encoding in upper theta/alpha post-stimulus ( 0–0 . 2 s ) than in the same frequency range during the 1 s pre-stimulus period ( Wilcoxon test: p = 0 . 039 ) . Furthermore , the difference between pre-stimulus theta synchrony preceding successful compared with unsuccessful encoding was not significant ( Wilcoxon test: p = 0 . 46 ) , whereas post-stimulus , the synchrony difference between successful and unsuccessful encoding was significant ( Wilcoxon test: p = 0 . 039 ) . Because parietal scalp signals could be recorded from only two participants due to post-operative dressing placement , these were also analyzed as individual cases ( Figure 2—figure supplements 6–7 ) . Parietal-RATN and frontal-RATN ( consistent with the group data ) theta synchrony 0 . 5–1 . 5 s post-stimulus were significantly enhanced individually during successful compared with unsuccessful encoding ( permutation tests on individual epochs within participants: ps < 0 . 05 ) . Because the main synchrony findings occurred in theta between 0 . 5 and 1 . 5 s , further analyses focussed on this time-frequency range . CFC was greater during successful than unsuccessful encoding between frontal theta phase-troughs and RATN gamma ( ∼40–50 Hz ) amplitude-peaks ( TT , T = 4 . 9 , p = 0 . 0018; CSPT: p = 0 . 038; observed contiguous cluster 14 pixels; criterial cluster 12 pixels for overall p = 0 . 05; Figure 3 ) . The difference was absent for thalamic high gamma ( up to 256 Hz ) ( Canolty et al . , 2006 ) . 10 . 7554/eLife . 05352 . 015Figure 3 . Frontal-right anterior thalamic nucleus ( RATN ) cross-frequency coupling ( CFC ) . ( A ) Successful encoding . ( B ) Unsuccessful encoding . ( C ) Paired T-tests: Successful minus unsuccessful encoding . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01510 . 7554/eLife . 05352 . 016Figure 3—figure supplement 1 . Cross-frequency coupling ( CFC ) within right anterior thalamic nucleus ( RATN ) and within frontal neocortex . ( A ) Successful encoding: within-RATN coupling . ( B ) Unsuccessful encoding: within-RATN coupling over wider gamma range . ( C ) Successful minus unsuccessful encoding: paired TT-tests ( TT ) : T = 3 . 6 , p = 0 . 0086 ( uncorrected p values ) , on which a cluster-size permutation test ( CSPT ) was performed ( CSPT: p = 0 . 045; observed contiguous cluster 15 pixels; criterial cluster 14 pixels for overall p = 0 . 05 ) . Note: the blue portion of the scale indicates significantly greater coupling during unsuccessful than successful encoding at higher gamma frequencies . ( D ) Successful encoding: greater within-frontal lower-frequency-theta to gamma coupling ( trough-to-peak ) ; higher-frequency-theta to gamma coupling ( peak to peak ) . ( E ) Unsuccessful encoding: higher-frequency-theta to gamma coupling ( peak-to-peak ) . ( F ) Successful minus unsuccessful encoding: TT: T = 3 . 6 , p = 0 . 0087 ( uncorrected p values ) , on which CSPT was performed ( CSPT: p = 0 . 017; observed contiguous cluster 19 pixels; criterial cluster 11 pixels for overall p = 0 . 05 ) . Note: the red portion of the scale indicates significantly greater coupling during successful than unsuccessful encoding . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01610 . 7554/eLife . 05352 . 017Figure 3—figure supplement 2 . Cross frequency coupling ( CFC ) during successful encoding in Participant 7 . All epochs are aligned to the first theta trough in the 0 . 5–1 . 5 s window , rendering time arbitrary , because the first theta trough in the window occurred at a different time point for each epoch . The power time series of each epoch was normalized to allow inter-frequency comparison . ( A and B ) Right anterior thalamic nucleus ( RATN ) gamma power . ( C ) Frontal theta troughs coupled with RATN gamma peaks . ( D ) RATN theta peaks coupled with RATN gamma peaks . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 01710 . 7554/eLife . 05352 . 018Figure 3—figure supplement 3 . Cross-frequency coupling ( CFC ) with differing frontal cortical sites . The mean difference between fronto-RATN theta-gamma CFC during successful minus during unsuccessful encoding across participants . ( A ) Across the six participants with electrodes at Fpz . ( B ) Across the two participants with electrodes at AFz and Fz . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 018 In contrast to the frontal-RATN findings , within the RATN , theta phase-peaks were coupled with gamma amplitude-peaks . Of note is that the gamma range was narrower ( ∼40–50 Hz compared with ∼40–70 Hz ) during successful than during unsuccessful encoding , resulting in significantly greater within-RATN CFC involving higher gamma-frequency amplitudes during unsuccessful compared with during successful encoding ( paired T-test , TT , T = 3 . 6 , p = 0 . 0086; cluster-size permutation test , CSPT: p = 0 . 045; observed contiguous cluster 15 pixels; criterial cluster 14 pixels for overall p = 0 . 05; Figure 3—figure supplement 1A–C ) . The CFC patterns during successful and unsuccessful encoding were then compared directly . Within-RATN CFC during successful encoding was compared with a distribution of 1000 phase-scattered surrogates , and the theta-gamma ( 3 . 8–8 . 1 Hz , 30 . 5–64 . 6 Hz ) cross-frequency points at which CFC during successful encoding exceeded a threshold with criterion p = 0 . 05 were identified . The same analysis was performed for unsuccessful encoding . The two resulting CFC patterns differed significantly ( two-dimensional Kolmogorov–Smirnov test—2-D KS test: d = 0 . 81 , p = 0 . 048 ) , with a wider gamma range during unsuccessful encoding . Assuming that an assembly of synchronously firing neurons is associated with a particular memory trace , the narrower RATN gamma range coupled with theta phase during successful memory formation could be interpreted as reflecting firing of only relevant neural assemblies , thus reflecting neural specificity during encoding ( Desimone , 1996; Düzel et al . , 2005; Schott et al . , 2006 ) . We correspondingly postulate that the corticothalamic coupling may coordinate the firing of particular thalamic neural assemblies underpinning the memory to be encoded , facilitating synaptic strengthening and relevant memory formation . Indeed , ongoing CFC has been detected in the centromedian thalamic nucleus during cognitive task performance ( Fitzgerald et al . , 2013 ) . Again , there was no encoding-related CFC difference apparent for thalamic high gamma ( Canolty et al . , 2006 ) . CFC findings within the frontal cortex ( Figure 3—figure supplement 1D–F ) were consistent with the literature . Theta-to-gamma CFC was peak-to-peak at higher theta frequencies , and occurred for both successful and unsuccessful encoding ( phase-scattered surrogate-data tests for successful and for unsuccessful encoding: p = 0 . 001 ) . There was no significant relationship to encoding success ( TT , p > 0 . 05; CSPT: p > 0 . 05; Figure 3—figure supplement 1D–F ) , consistent with hippocampal CFC patterns during working memory ( Axmacher et al . , 2010 ) . By contrast , theta-to-gamma CFC was trough-to-peak at lower theta frequencies , and was greater during successful than unsuccessful encoding ( TT , T = 3 . 6 , p = 0 . 0087; CSPT: p = 0 . 017; observed contiguous cluster 19 pixels; criterial cluster 11 pixels for overall p = 0 . 05 ) , consistent with ( Canolty et al . , 2006 ) . Theta-gamma CFC patterns within RATN and within frontal neocortex differed from the frontal-RATN pattern ( Figure 3—figure supplements 1–2 ) . Frontal-RATN theta-gamma ( 3 . 8–8 . 1 Hz , 30 . 5–64 . 6 Hz ) coupling during successful vs unsuccessful encoding was compared using the above paired T-tests , to provide the cross-frequency pattern , and the same was performed for within-RATN CFC . These CFC patterns differed significantly ( 2-D KS test: d = 0 . 83 , p = 0 . 024 ) . Frontal-RATN and within-frontal CFC patterns also differed significantly ( 2-D KS test: d = 0 . 98 , p = 0 . 011 ) . GC revealed that frontal theta better predicted RATN theta than vice-versa ( Figure 4 ) . In GC analyses , the peak model order provides an indication of how far into the past one signal provides information about another , and may thus be interpreted as indicating the approximate delay in transfer of information ( Staudigl et al . , 2012 ) . We found that frontal theta prediction of RATN theta peaked at a model-order of 32 , which corresponds to 63 ms , or one third of a theta cycle phase-lag between frontal and RATN theta . Such a delay is broadly consistent with frontal-RATN theta-to-gamma CFC being trough-to-peak ( Figure 3 ) , and intra-RATN theta-to-gamma CFC being peak-to-peak ( Figure 3—figure supplement 1A–C ) , illustrated in Figure 3—figure supplement 2 . Together , the CFC and GC findings suggest that frontal theta modulates RATN gamma via frontal-ATN theta synchrony during successful encoding . 10 . 7554/eLife . 05352 . 019Figure 4 . Granger causality ( GC ) in the theta frequency range . ( A ) During successful encoding , frontal theta activity predicted right anterior thalamic nucleus ( RATN ) activity , peaking at model order 32 , corresponding with a 63 ms phase lag ( i . e . , one third of a theta cycle ) . ( B ) During successful encoding , RATN activity was significantly ( paired T-tests: T = 3 . 2 , p = 0 . 014 ) less predictive of frontal activity . ( C ) During unsuccessful encoding , frontal theta activity predicted RATN activity . ( D ) During unsuccessful encoding , RATN was significantly ( paired T-tests: T = 3 . 0 , p = 0 . 018 ) less predictive of frontal activity . DOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 019
We demonstrate increased corticothalamic synchrony during successful memory encoding , recording directly from the two most memory-relevant thalamic nuclei ( ATN and DMTN ) . Our amplitude-independent phase synchrony measurements ( Lachaux et al . , 1999 ) show that timing of post-stimulus ATN theta activity alone ( see also Figure 2—figure supplement 8 ) , and its relation to local ATN processing as indexed by gamma amplitudes , is critical in successful memory encoding , providing the first electrophysiological evidence concerning the role of ATN in human memory formation . The absence of neocortical-DMTN synchrony differences , together with their recent detection during verbal memory retrieval ( Staudigl et al . , 2012 ) , fits with evidence suggesting ATN specialization for encoding ( Harding et al . , 2000; Van der Werf et al . , 2003 ) , and DMTN for retrieval ( Harding et al . , 2000; Van der Werf et al . , 2003; Aggleton et al . , 2010; Aggleton , 2012 ) . RATN involvement is moreover consistent with the non-verbal scene stimuli employed ( Maillard et al . , 2011 ) , and with evidence that left thalamus lesions produce more severe memory deficits for verbal than non-verbal material ( Squire et al . , 1989 ) . The late theta synchrony is in accord with the timing of differences in ERPs ( Schott et al . , 2002 ) and post-stimulus MT theta power ( Hanslmayr and Staudigl , 2014 ) during successful compared with unsuccessful encoding of episodic memories . While oscillations in the delta frequency range have recently been found to show a memory-related difference in the hippocampus ( Watrous et al . , 2011; Lega et al . , 2014 ) , we did not detect a significant difference between corticothalamic delta phase synchrony during successful compared with unsuccessful encoding . While we provide here novel evidence that the coordination of theta and gamma oscillations involving the ATN plays a critical role in trial-by-trial memory ability , it should be noted that memory encoding is recognized to include not only content processing and information storage , but also attention ( reviewed by Kim , 2011 ) . Item-specific attention indeed has a well-recognized effect on whether an item is successfully encoded ( Raz and Buhle , 2006; Muzzio et al . , 2009; Kim , 2011; Burke et al . , 2014 ) . Attention is not only necessary for optimal memory encoding , but multiple brain structures involved in memory formation may be subject to attentional modulation , including the hippocampus itself ( Muzzio et al . , 2009 ) . There are several indicators , however , that our findings reflect memory encoding processes beyond simply global arousal or attention fluctuations . Firstly , the timing of the key memory differences here at around 1 s post-stimulus is strongly suggestive of a memory-related difference . Neural encoding differences related to later memory have been identified using ERPs and cortical theta/gamma oscillations , generally peaking at around 1 s after stimulus presentation , continuing up to 1 . 5 s post-stimulus ( Paller et al . , 1987; Schott et al . , 2002; Sederberg et al . , 2003; Osipova et al . , 2006; Lega et al . , 2012; Hanslmayr and Staudigl , 2014; Long et al . , 2014 ) . Secondly , whereas oscillatory dynamics between the pulvinar thalamic nucleus and the parietal cortex support visual attention in macaques , consistent with attention deficits following focal pulvinar lesions ( Saalmann et al . , 2012 ) , our main findings pertain to the ATN , for which lesion and animal studies suggest a role in memory . Moreover , based on current knowledge about the ATN and the DMTN ( Van der Werf et al . , 2003 ) one would rather expect attentional differences to be reflected in DMTN activity , which we did not find . Most critically , if global attentional fluctuations were responsible for our theta synchrony findings , one would expect longer response times during the encoding phase for later forgotten compared with later remembered scenes , due to lack of attention to the encoding task . However , response times did not differ . Finally , global attentional fluctuations over time would also imply a dependency between the probability of successful encoding on successive trials during the study phase of the experiment , and we found no dependency . We note , however , that variability in factors such as attention and the emotional valence of stimuli are fundamental to the study design in subsequent memory paradigms , affecting the probability of successfully encoding each item and thus enabling comparison of successful with unsuccessful encoding . Indeed , the early upper theta and alpha synchrony immediately following stimulus presentation during successful memory encoding is likely to reflect enhanced item-specific attention and perception ( Düzel et al . , 2005 ) , with the later theta synchrony timing fitting well with previous findings relating to memory encoding ( Schott et al . , 2002; Hanslmayr and Staudigl , 2014 ) . The ATN is the target of stimulation in the treatment of focal epilepsy on the basis that seizure activity starting focally is propagated through this site to widespread cortical areas ( Lega et al . , 2010 ) . Regional cortical specificity in frontal-RATN synchrony would support our argument that the differences that we have identified are memory-specific . While the phase synchrony pattern we report is discernible in all three electrode placements ( frontopolar , other frontal , and parietal ) , it is indeed not identical across regions ( Figure 2—figure supplement 9 ) . We also show the difference between fronto-RATN theta-gamma CFC involving the different frontal electrode placements ( Figure 3—figure supplement 3 ) . Again , the patterns are similar but not identical . The limited available scalp electrode coverage in our participants , however , precludes drawing strong conclusions in this regard . We note also , though , that both frontal and parietal cortices are well-recognized as being involved in memory processing ( Weiss and Rappelsberger , 2000; Otten et al . , 2002; Sauseng et al . , 2005; Uncapher and Wagner , 2009; Friese et al . , 2012; Sweeney-Reed et al . , 2012 ) . The encoding and retrieval tasks were chosen for their simplicity , because recording was only possible at the bedside in the few days following intracranial surgery , and because we did not wish to lose data owing to failure of these rare participants to succeed at the tasks . Under these circumstances , we were able to achieve large and comparable behavioral trial numbers for successful and unsuccessful encoding , which could then be submitted to oscillatory analysis . Pilot testing revealed that one participant ( not included in the current cohort ) could not adequately perform an objective source-memory task under these post-operative conditions , and we thus additionally judged it unlikely that the participants could properly implement the more complex ‘remember/know/guess’ instructions ( Gardiner and Richardson-Klavehn , 2000 ) necessary to obtain reports of subjective recollection and familiarity during recognition . Moreover , given the nature of the data ultimately obtained from our eight participants , we elected to collapse the data across the levels of response certainty to maximize trial numbers in each category for electrophysiological synchrony analysis , in a pragmatic trade-off between the power of the electrophysiological analyses and behavioral/psychological resolution . Thus , a limitation of the study is that we cannot conclusively link our results to recollection , and therefore episodic memory formation , separately from familiarity . Our findings concerning ATN electrophysiological involvement in human memory formation are , however , consistent with the theoretical proposal that the ATN is a part of an extended hippocampal system supporting episodic recollection ( Aggleton , 2012 ) , whether during encoding or retrieval , and with the competing theoretical proposal that the ATN is specialized for encoding rather than retrieval , whether the information involved is episodic ( recollection ) or semantic ( Van der Werf et al . , 2003 ) . The latter proposal is consistent with human lesion data , showing that sufferers from Korsakoff's syndrome with lesions to the ATN fail to acquire new semantic as well as new episodic information ( for example , Harding et al . , 2000 ) and is thus particularly consistent with our new data . We note , furthermore , that whether recollection and familiarity are separate processes ( Eichenbaum et al . , 2008 ) , or reflect a single process , with familiarity and recollection reflecting different degrees of memory strength ( Wixted and Squire , 2008 ) , is an area of continuing debate ( see also Gardiner and Richardson-Klavehn , 2000 and Yonelinas , 2002 ) for relevant information regarding humans . Despite this lack of behavioral/psychological resolution , our data nevertheless provide novel evidence concerning the real-time role of the human ATN in memory formation . It should also be noted that memory processing may be divided into different subsystems , with different components of memory processing involving different thalamic nuclei ( Mennemeier et al . , 1992 ) . For example , whereas human lesion and animal studies suggest a regulatory role for the ATN in memory encoding ( Harding et al . , 2000; Vertes et al . , 2001; Van der Werf et al . , 2003; Aggleton et al . , 2010; Aggleton , 2012 ) and for the DMTN in retrieval ( Van der Werf et al . , 2003; Staudigl et al . , 2012 ) , other thalamic nuclei have also been found to be involved in different aspects of memory processing . For example , evidence supports a role for the nucleus reuniens in fear conditioning and memory generalization ( Vertes et al . , 2007; Xu and Südhof , 2013 ) , as well as spatial processing ( Jankowski et al . , 2014 ) , for the pulvinar nucleus in attention ( Saalmann et al . , 2012 ) and nonverbal memory processing ( Johnson and Ojemann , 2000 ) , and for the left ventro-lateral thalamus in verbal memory encoding ( Johnson and Ojemann , 2000 ) . Our focus on the ATN and DMTN is based on the rare availability of human electrophysiological data from these sites , which was determined by clinical requirements , and is consistent with extant data concerning the amnesic effects of lesions in these thalamic areas in humans ( Harding et al . , 2000; Van der Werf et al . , 2003 ) . In summary , our findings shed new light on real-time interaction between the ATN and the neocortex , thus broadening understanding of the brain structures involved in memory formation from a focus on the hippocampus and neocortex to recognition of a pivotal role for the ATN . More generally , the CFC theta-gamma findings , together with the amnesic effects of lesions to the human ATN , provide evidence that the ATN plays an active role in encoding , instead of simply relaying cortical or hippocampal signals ( Aggleton et al . , 2010 ) .
Intrathalamic data were recorded from 1 . 5 mm platinum-iridium electrodes implanted bilaterally ( four contacts each ) in the thalamus for stimulation therapy for multiple pharmacoresistant focal epilepsy in eight adult participants , all of whom were not suitable candidates for resective surgery . A minimum sample size was not set , because memory-dependent phase synchrony has been detected on an individual level in intracranial recordings ( Fell and Axmacher , 2011; Staudigl et al . , 2012 ) ( see also Figure 2 , Figure 2—figure supplements 6–7 ) . The final number of participants was determined by the number of patients available in the days just following implantation during the approximately 2 years of the relevant therapeutic program , and also willing to participate in the study . This sample size was considerably greater than that usually available for intracranial studies in humans ( Canolty et al . , 2006; Fitzgerald et al . , 2013; Bonini et al . , 2014 ) , and in non-human primates ( Saalmann et al . , 2012 ) . The mean age of the participants was 37 . 5 years ( range 28–52 years , standard deviation 8 . 2 years ) , and four participants were female . Stimulation via intracranial electrodes did not occur during the data recording . The measurements were approved by the Ethics Commission of the Medical Faculty of the Otto-von-Guericke University , Magdeburg ( application number 0308 ) , and all participants gave written informed consent in accordance with the Helsinki Declaration of 1975 , as revised in 2000 and 2008 . Consent to participate in our study , as well as for publication of results in an anonymized format , was obtained by the neurosurgeon at the same time as consent was obtained for the surgical procedure . Contacts were located in the ATN for all eight participants and in the DMTN for seven of the participants ( with Participant 5 not having DMTN contacts ) . Placement of the thalamic electrodes was performed stereotactically . The angle of entry through the skull and the depth of each electrode was calculated based on MRI images of each patient's brain pre-operatively . An intra-operative X-ray and postoperative CT-scans were carried out in order to confirm correct localization of each electrode , by reference to the Schaltenbrand and Pick Atlases ( Schaltenbrand and Wahren , 1977; Maldjian et al . , 2003 ) . Scalp EEG data were simultaneously collected from frontal electrodes ( Fz , AFz , or Fpz ) in all eight participants , and from parietal electrodes only in Participant 1 ( P3 , Pz , P4; Figure 1 , Figure 2—figure supplement 6 ) and Participant 2 ( Pz only; Figure 2—figure supplement 7 ) . Positioning of post-operative dressings meant that parietal electrodes could not be placed for the other six participants . Frontal electrodes were centered over Fz ( frontal ) , AFz ( anterior frontal ) , and Fpz ( frontopolar ) , with the data from the most frontal available electrode from each participant included in the group analyses . These scalp sites reflect underlying cortical activity from midline frontal cortices ( Paller et al . , 1987 ) . Data from the two participants with parietal electrodes were analyzed as single cases ( Figure 2—figure supplements 6–7 ) , as well as contributing to the group analyses for frontal electrodes . Nose-referenced voltages were amplified with a Walter Graphtek ( Lübeck , Germany ) EEG amplifier and recorded with a 512 Hz sampling frequency . Offline re-referencing of thalamic voltages to the next deepest contact rendered six bipolar channels . Radiological intracranial electrode localization was performed for all participants intra- and postoperatively by the neurosurgeons and a physicist . Electrode location is provided in detail for Participant 1 for illustration purposes ( Figure 1 ) . The deepest two thalamic contacts were located in the DMTN , and the most superficial in the ATN . The second most superficial was at the ATN/DMTN border . The critical results from this participant ( Figure 2—figure supplement 6 ) were obtained from a bipolar recording from the two most superficial right-sided thalamic contacts . We confirmed that the synchrony patterns were not observed when the second most superficial contact was referenced to either of the two deeper contacts , suggesting that the critical signals originated from the most superficial contact , which was clearly located in ATN , and not from its reference contact . Note that no measureable temporal phase-shifts that simply reflect the physical ( e . g . , capacitative ) properties of the brain-to-scalp-electrode interface have been detected at typical EEG frequencies ( Nunez and Srinivasan , 2006 ) . Any phase-lags between the scalp and intrathalamic recordings should , therefore , reflect genuine neural conduction delays between neocortex and thalamus . Such phase-lags should not in any case affect the corticothalamic phase synchrony measurements detailed below , which were independent of phase-lag as well as amplitude , but were taken into account in the Granger causality calculations detailed below . During encoding , the participants viewed a series of 200 ( 100 for Participant 1 ) photographs of unfamiliar real-world scenes on a computer screen . Each scene was shown for 2 . 4 s , followed by a fixation cross for 1 . 4 s , and the participant judged whether the depicted scene was indoors or outdoors ( average response time around 1 s after scene presentation ) . The responses were made using left and right index fingers , and the response hand was counterbalanced across participants . Recognition testing occurred after a short distraction break to ensure retrieval based on long-term memory . Each participant viewed all 200 scenes from the encoding phase in a different random order , randomly interspersed with 100 similar but new scenes . The total pool of 300 scenes was randomly assigned into 3 groups of 100 . Which of these 3 groups formed the old and new scenes for a particular participant was systematically counterbalanced across participants by rotation . A short practice session with both encoding and recognition test phases was provided for each participant , rendering the experiment itself an intentional encoding paradigm , because participants knew during encoding that they would later be tested . Nevertheless , the focus of the encoding task was on deciding whether each depicted scene was indoors or outdoors . In the recognition test , each scene was first shown for 1 . 25 s , then a 6-point scale was superimposed on the scene for 2 . 8 s , along which a marker was moved by pressing one of two keyboard buttons , with the index finger of each hand , to indicate degrees of confidence as to whether the scene was old or new ( direction and response hand counterbalanced across participants ) . A fixation cross then appeared , jittered between 0 . 75 and 1 . 25 s . Behavioral data from the test phase are shown in Table 1 . The data presented here were collapsed across the three scale points indicating ‘old’ , and the three scale points indicating ‘new’ at test , to obtain binary ‘old’/‘new’ judgments . All eight participants showed a greater percentage of hits ( correct ‘old’ responses to old scenes ) than of false alarms ( incorrect ‘old’ responses to new scenes ) , demonstrating that they had formed memories for scenes from the encoding phase . The encoding data ( electrophysiological data and response times ) were then sorted , according to test phase responses , into epochs with successful encoding ( later correct ‘old’ judgments to old scenes at test , hits ) and epochs with unsuccessful encoding ( later incorrect ‘new’ judgments to old scenes at test , misses ) ( Paller and Wagner , 2002 ) . The electrophysiological encoding data were segmented into epochs 1 s pre-stimulus ( i . e . , scene presentation at encoding ) to 2 s post-stimulus , because work on memory encoding measuring ERPs and theta/gamma oscillations has demonstrated differences in electrophysiological activity related to later memory up to 1 . 5 s post-stimulus , generally peaking at around 1 s after stimulus presentation ( Paller et al . , 1987; Fell et al . , 2001; Schott et al . , 2002; Sederberg et al . , 2003; Osipova et al . , 2006; Lega et al . , 2012; Long et al . , 2014 ) . Data were also recorded during retrieval , but we focus on the encoding data here . Note that the time axes in all figures ( except in Figure 3—figure supplement 2 , where time is arbitrary ) are set such that the stimulus was shown at time = 0 s . Recording of intracranial signals was performed at the bedside in the few days following electrode implantation , before the electrodes were attached to a stimulator under the skin over the chest wall for epilepsy treatment , in a second operation approximately 1 week later . No seizures took place during the testing sessions , and all patients were fully alert and cooperative throughout . Epochs were individually visually inspected and cleaned of ocular and other artifacts using temporal-decorrelation-separation independent component analysis ( Sweeney-Reed et al . , 2012 ) . Spikes and spike-waves were also removed using this approach , to maximize the number of epochs for analysis . We deemed the differences in electrophysiological activity detected between epochs recorded during successful compared with unsuccessful encoding in our simple task to be unlikely to result from global attentional fluctuations due to epileptiform activity for several reasons . Firstly , accuracy was close to ceiling for the indoor vs outdoor judgment during encoding . Secondly , there was no difference in response times during encoding for successful compared with unsuccessful memory formation . Thirdly , successful and unsuccessful encoding of successive scenes in the series showed no sequential dependency . Furthermore , while the cortical site of the epileptic focus differed across participants ( Table 2 ) , the reported results were highly consistent across the group ( Figure 2E–F , Figure 2—figure supplements 2 , 6–7 ) . Spikes or spike-waves were visible in the data from three of the eight participants ( Participant 2: 10 . 3% during successful and 7 . 3% during unsuccessful encoding; Participant 6: 29 . 9% and 24 . 7% respectively; Participant 7: 2 . 7% and 5 . 4% respectively ) . Across these three participants , the mean difference in these percentages between successful and unsuccessful encoding was only 1 . 83% ( i . e . , slightly greater for successful encoding ) , suggesting that concerns about spikes and spike-waves being confounded with the two data categories of interest are minimal . Nevertheless , the analyses were also performed excluding epochs with spikes and spike-waves , and the findings reported remained statistically significant . 10 . 7554/eLife . 05352 . 020Table 2 . Clinical informationDOI:http://dx . doi . org/10 . 7554/eLife . 05352 . 020PtM/FAge at surgeryAge at first seizureEpilepsy syndromeSeizure originCurrent medication1F4213PLEbilateralLTG 200 mgLCM 400 mg2F5233TLEbilateralLCM 400 mgLTG 200 mg3M3426TLErightLEV 4000 mgESL 1200 mg4F2916TLEbilateralLTG 400 mgRTG 600 mg5M399TLEleftLTG 400 mgLCM 400 mg6M321TLEfrontal and right temporalSTP 4500 mgOXC 900 mgCLB 5 mg7M4129FLEbilateralLTG 400 mgZNS 400 mg8F4414TLEleftCBZ 1200 mgPt = participant . M/F = male/female . PLE = parietal lobe epilepsy . TLE = temporal lobe epilepsy . FLE = frontal lobe epilepsy . LTG = lamotrigine . LCM = lacosamide . LEV = levetiracetam . ESL = esilcarbazepine . RTG = retigabine . STP = striripentol . OXC = oxcarbazepine . CLB = clobazepam . ZNS = zonegran . CBZ = carbamazepine . Corticothalamic phase synchrony was calculated following wavelet time-frequency decomposition of each epoch , using 6-cycle Morlet wavelets , yielding 57 logarithmically spaced frequencies between 1 and 100 Hz . A logarithmic scale was used to take account of the frequency resolution of the Morlet wavelet ( Düzel et al . , 2005 ) , such that wavelet spacing became sparser as frequency increased . After wavelet transformation , a phase series was extracted for each epoch , and the phase differences between scalp and thalamic channels were calculated ( for details of this well known method ( see Lachaux et al . , 1999 and Sweeney-Reed et al . , 2012 ) . Phase-locking values ( PLVs ) could vary between 0 ( no PL ) and 1 ( complete PL ) . In order to enhance time resolution ( at the expense of frequency resolution ) , we also applied 1-cycle wavelets , confirming that the synchrony that we report ( Figure 2 , Figure 2—figure supplement 1 ) took place post-stimulus ( Figure 2—figure supplement 5 ) . Statistical analysis thus focused on the 2 s post-stimulus period of 1024 time-points . It should be noted that the calculation of PLVs may be influenced by a common reference ( Vinck et al . , 2011 ) , and we addressed this issue by using bipolar referencing in our thalamic recordings ( see also Staudigl et al . , 2012 ) . Volume conduction may also influence PLVs ( Vinck et al . , 2011 ) , but in the present study , phase synchrony is calculated between the frontal cortex and the RATN , whose spatial separation should exclude this influence . All statistical tests were two-tailed . After PLVs were calculated for each epoch in each category ( successful encoding and unsuccessful encoding ) , permutation tests , which are conservative in that they do not make parametric assumptions , were initially used to evaluate the significance of differences between successful and unsuccessful encoding for each pixel within the 2-dimensional time-frequency space of 57 frequencies ( 0–100 Hz ) and 1024 post-stimulus time-points ( 0–2 s ) . A mean PLV was calculated for successful and for unsuccessful encoding for each of the eight participants , and the 16 PLVs were pooled , reassigned to two artificial categories 1000 times , and a PLV difference between categories calculated , in order to obtain a two-tailed error distribution of differences against which the observed mean difference between successful and unsuccessful encoding was tested . A statistical comparison of mean PLVs across successful and unsuccessful encoding necessitates an equal number of epochs per category , so that , prior to calculating the mean PLVs for each participant , epochs were randomly selected from the larger category to match the size of the smaller ( Table 1 ) . Figure 2A–C shows the mean PLVs calculated from the equal number of epochs per category that were included in the statistical analysis . Figure 2D shows the results of the group permutation tests . We also confirmed the key corticothalamic synchrony findings with paired T-tests ( with mean PLVs calculated for each participant as just described for permutation testing ) . The assumptions of paired T-tests were satisfied ( i . e . , an approximately normal distribution of the differences between PLVs for successful and unsuccessful encoding across participants , and a positive correlation between PLVs for successful and unsuccessful encoding across participants ) . Mean PLVs differed significantly , in a pattern similar to that found using the permutation tests ( Figure 2—figure supplement 1 ) . All paired T-tests had 7° of freedom except where otherwise noted . The electrophysiological literature on memory formation clearly suggests a focus on theta ( 4–8 Hz ) oscillations , a range covered by 14 frequencies ( i . e . , wavelets ) . However , in view of the scarcity and novelty of memory-related intrathalamic data in humans , a highly conservative approach to evaluating these observed uncorrected p values was taken . For the permutation tests , a false discovery rate correction ( Canolty et al . , 2006 ) was applied for the 57 frequency ( 0–100 Hz ) and 1024 time-point ( 0–2 s ) comparisons , which was especially conservative given the dependency between adjacent time and frequency points ( Figure 2—figure supplement 1 ) . For the T-tests , a cluster-size permutation test ( Maris and Oostenveld , 2007 ) was performed , in which a cluster was defined as adjacent significant ( criterion: p = 0 . 05 by paired T-test ) time-frequency points . Mean PLV matrices for each participant for each condition were randomly assigned to two groups 1000 times , and T-tests were performed across time and frequency for each permutation . T-test outcomes were rendered binary ( 1 = significant , 0 = nonsignificant ) for each of the 57 frequencies and 1024 time-points , and the maximum cluster size ( as defined above ) emerging randomly in each iteration was calculated in order to provide a distribution against which to determine the significance of the observed cluster size . The same approach was taken for cluster-size significance in the cross-frequency coupling analysis described below . Finally , we evaluated the consistency of the results across the eight participants by applying a nonparametric ( and thus conservative ) Wilcoxon test of the differences between theta ( 5 . 2 Hz ) PLVs for successful and unsuccessful encoding in the 0 . 5 to 1 . 5 s time-window in which theta differences had been revealed by the above-described conservative methods ( see also Figure 2 , Figure 2—figure supplement 2 ) . The significance of the difference between PLVs for successful vs unsuccessful encoding was additionally calculated on an individual case basis for Participants 1 and 2 , because they were the only participants with parietal electrodes . Participant 1 had only 26 unsuccessful encoding epochs , so 26 successful encoding epochs were randomly selected from the 74 available and the mean PLV calculated . In order to use all available data to calculate the PLV difference between successful and unsuccessful encoding , this selection of 26 epochs and mean PLV calculation was carried out 200 times and an overall observed mean PLV was then calculated for successful encoding . For significance testing by permutation tests , this difference was compared to an error distribution , which was obtained by randomly assigning 52 epochs , irrespective of encoding success , to two artificial categories 1000 times ( each time randomly selecting 26 of the 74 successful encoding epochs , and using all 26 unsuccessful encoding epochs ) and calculating the mean PLV difference each time . The observed mean PLV difference was compared to this error distribution . A test with criterion p = 0 . 05 was applied ( Figure 2—figure supplement 6 , right ) . For descriptive purposes , the mean PLV across all 74 epochs of successful encoding is shown ( Figure 2—figure supplement 6 , left ) . Permutation of epochs was also performed for Participant 2 , with all 68 unsuccessful encoding epochs being used , and 68 encoding epochs being randomly selected from the 132 successful encoding epochs . The same procedure was then followed , except that given the larger number of trials in the smaller category compared with Participant 1 , the observed mean PLV for successful encoding was calculated only once . For descriptive purposes , the mean PLV across all 132 epochs of successful encoding is shown ( Figure 2—figure supplement 7 , left ) . Cross-frequency coupling ( CFC ) was calculated as per the work of Canolty et al . ( 2006 ) and Axmacher et al . ( 2010 ) . The frontal and RATN signals were wavelet transformed , then the theta phase and the gamma amplitude time series were extracted from the frontal and thalamic signals for each epoch . New complex signals were created by combining phases ( ranging from 4–16 Hz , to include frequencies in the theta and alpha range ) with gamma amplitudes ( 30–256 Hz ) from the channels between which coupling was assessed , and the complex signals were averaged over time for each epoch . The phase was then extracted from the new complex value for each epoch by taking the arctangent of the imaginary divided by the real part . The average complex value across epochs was then also calculated , and from this value , the modulatory phase was determined by taking the arctangent of the imaginary over real parts , thus quantifying the average phase of the lower frequency oscillation at which the amplitude of the high frequency oscillation was highest . The theta amplitudes were then shifted by minus modulatory phase , and the correlation coefficient ( CC ) between the shifted theta oscillations and the gamma amplitudes was found . The CC was Fisher-Z transformed , then a mean was taken over epochs for each participant to provide the modulation index , separately for successful and unsuccessful encoding . Paired T-tests comparing successful with unsuccessful encoding were then performed for each pixel within the 37 theta-phase frequency by 21 gamma-amplitude frequency matrix , comparing the degree of theta-phase with gamma-amplitude coupling . The resulting p values , as shown in Figure 3 , were then subjected to a cluster-based correction via permutation testing as described above for synchrony differences . To illustrate the coupling in a single participant , the frontal and RATN theta phases and RATN gamma power are shown for Participant 7 in Figure 3—figure supplement 2 , averaged across epochs during successful encoding . Note that to enable comparison across frequencies , the temporal mean was subtracted from the power values , and they were then divided by the temporal standard deviation of power ( Canolty et al . , 2006 ) . Gamma power and theta troughs were aligned to the first theta trough in the 0 . 5 to 1 . 5 s post-stimulus window for each epoch ( Staudigl et al . , 2012 ) . The absence of an encoding-related difference in frontal CFC between upper theta peaks and gamma amplitudes is consistent with working memory findings ( Axmacher et al . , 2010 ) . In order to confirm the presence of this CFC for both successful and unsuccessful encoding , levels of CFC were also compared with a distribution of CFC indices generated from 1000 phase-scattered surrogate data sets . Granger causality ( GC ) was used to investigate information flow direction ( Seth , 2010; Staudigl et al . , 2012 ) at time-frequency locations ( 4–12 Hz; 0 . 5–1 . 5 s ) encompassing the significantly greater theta phase synchrony during successful compared with unsuccessful encoding . GC uses multivariate autoregressive modeling to ascertain whether time series A may be more accurately predicted from time series B , with a certain time lag , than B from A . If incorporating values from B in the regression of A allows better prediction of A than vice versa , B is said to influence A . The data were first detrended and rendered zero mean across epochs to remove nonstationarity . The variances of the prediction errors of the autoregressive models were then used to assess likely information flow direction . The Akaike and Bayesian information criteria ( AIC , BIC ) were employed to calculate a model order balancing overparameterization and adequate spectral resolution ( Seth , 2010; Staudigl et al . , 2012 ) . When a minimum is not reached and the BIC/AIC does not show substantial decreases at higher orders , as here , model orders for EEG are usually chosen over a range ( Brovelli et al . , 2004; Staudigl et al . , 2012 ) . We applied model orders from 26 to 36 , identifying a peak GC corresponding to a third of a theta cycle . Model coefficients were interpreted in the frequency domain . Significance of the difference between frontal-RATN and RATN-frontal GC in successful and in unsuccessful encoding was assessed by paired T-tests . The flow of time is commonly used to make inferences from time series data regarding directional causal influences ( Bollimunta et al . , 2008 ) , and indeed GC is commonly equated with directions of information flow in neural circuits ( Ding et al . , 2006; Bollimunta et al . , 2008; Anderson et al . , 2010 ) . The model order specifies the number of time lagged observations made ( Bressler and Seth , 2011 ) , and the order at which GC peaks provides an indication of how long into the past one signal provides information about subsequent activity in the other signal . We interpreted the model order corresponding with peak causality as providing an indication of the delay in transfer of information ( Staudigl et al . , 2012 ) , which has been measured directly between the frontal cortex and hippocampus in rats and found to be consistent with the delays suggested by the present data ( Siapas et al . , 2005; Benchenane et al . , 2010 ) . | Memories , both the mundane and the significant , play an integral role in our daily lives . Scientists have long sought to establish exactly how our memories are formed; how does an experience , with its sights , sounds and feelings , become a mental representation stored within our brain ? One way to investigate this question is to look at the activity of different parts of the brain . Brain imaging techniques have helped researchers identify two key brain regions that are involved in the process of memory formation: the neocortex and the hippocampus . The neocortex forms the outer layer of the brain , and performs complex tasks such as decision-making and language comprehension . The hippocampus , which sits deeper within the brain , deals primarily with memory and navigation . Research has shown that memory formation depends on communication between the neocortex and the hippocampus . However , scientists suspected that additional structures located beneath the neocortex—among them , the anterior thalamic nuclei ( ATN ) —are also crucial for forming memories . This has been difficult to confirm as the small size of the ATN , and their location deep within the brain , make their activity almost impossible to monitor using standard brain imaging techniques . One way reliable data can be recorded from the ATN is by inserting electrodes into the brain . Brain surgery of course cannot be carried out on healthy human participants , but occasionally an opportunity arises to study the brain activity of patients who have electrodes inserted for therapeutic purposes . For example , in cases where a patient's epilepsy does not respond to conventional treatments , electrodes may be implanted to electrically stimulate the ATN in an attempt to improve their symptoms . Sweeney-Reed et al . asked eight volunteers to perform a memory task , and monitored the activity of each volunteer's ATN via electrodes that had already been implanted in their brain to treat epilepsy . Simultaneously , electrodes attached to the scalps of the volunteers recorded the activity of the neocortex . When a memory was successfully stored in the brain , the activity of the two regions became synchronized . This suggests that successful memory formation depends upon communication between the ATN and the neocortex . While the involvement of the ATN in human memory formation has long been a topic of speculation , Sweeney-Reed et al . now provide direct biological evidence for its crucial role in the process . Consequently , future research into memory formation should focus upon the ATN in addition to the more familiar structures of the neocortex and the hippocampus . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Corticothalamic phase synchrony and cross-frequency coupling predict human memory formation |
Vaccination of cattle against bovine Tuberculosis ( bTB ) has been a long-term policy objective for countries where disease continues to persist despite costly test-and-slaughter programs . The potential use of vaccination within the European Union has been linked to a need for field evaluation of any prospective vaccine and the impact of vaccination on the rate of transmission of bTB . We calculate that estimation of the direct protection of BCG could be achieved with 100 herds , but over 500 herds would be necessary to demonstrate an economic benefit for farmers whose costs are dominated by testing and associated herd restrictions . However , the low and variable attack rate in GB herds means field trials are unlikely to be able to discern any impact of vaccination on transmission . In contrast , experimental natural transmission studies could provide robust evaluation of both the efficacy and mode of action of vaccination using as few as 200 animals .
While the ultimate goal of the locally devolved strategies in Great Britain is to eliminate bovine tuberculosis from domestic cattle herds , the more economically important goal is to achieve officially TB-free ( OTF ) status . OTF status is defined by the EU in terms of demonstrating a long-term herd level prevalence of confirmed bTB of less than 0 . 1% ( Council Directive 64/432/EEC ) . While Scotland has already achieved this goal , herd level prevalence in England and Wales continues to rise despite intensifying control measures . The current Welsh ( Welsh Government , 2012 ) and English ( DEFRA , 2014 ) strategies for achieving OTF status , and international trade regulations , depend on the continued use of tuberculin testing and compulsory removal of test positive animals from herds . The only viable candidate vaccine for use in cattle at this time is the Bacillus Calmette-Guérin ( BCG ) vaccine which sensitizes vaccinated animals to tuberculin and dramatically increases the likelihood of false-positive tests . The practical and economic benefits of cattle vaccination therefore hinge on the performance of a new diagnostic test that can accurately Differentiate Infected from Vaccinated Animals ( a so-called DIVA test ) as much as on the efficacy of vaccination . DIVA tests for BCG have already been developed in the form of a interferon-gamma blood test ( Vordermeier et al . , 2011 ) and a skin test based on defined antigens ( Whelan et al . , 2010 ) . However , both these tests must still be validated and , in the case of the skin test , approved by regulatory authorities . From the perspective of maintaining the security of international trade , and highlighted by EFSA ( EFSA , 2013 ) , the most important requirement for validation is that the sensitivity of any proposed DIVA test is at least as good as the existing tuberculin test . However , under the intensive schedule of testing in Great Britain , where affected herds are repeatedly tested until clear , it is diagnostic specificity that provides the greatest barrier to delivery of an economic benefit of vaccination ( Conlan et al . , 2015 ) . To explore this issue of DIVA specificity , and the more general costs and benefits of cattle-based control measures , we previously developed and fitted dynamic herd-level transmission models that mimic the sequence of testing in GB herds ( Conlan et al . , 2012 , 2015 ) . We compared two basic models for bTB transmission , which are distinguished by different assumed relationships between epidemiological and diagnostic latency and described fully in Appendix 1 . For our purposes in this study , the SOR ( susceptible , occult , reactive ) and SORI ( susceptible , occult , reactive and infectious ) models can be considered as plausible upper and lower bounds on the transmission potential of Mycobacterium bovis in Great Britain . Such dynamic transmission models are essential to predict the effectiveness of vaccination within populations due to the indirect benefits of vaccination on transmission ( Anderson and May 1992 ) . When some individuals in the population are directly protected from infection by vaccination ( all-or nothing vaccine effect ) , they can no longer contribute to transmission; this leads to a further , indirect reduction in the potential spread of a disease within the population . In order for a vaccine to be useful , it does not necessarily have to provide sterilising immunity to infection ( Smith et al . , 1984 ) . ‘Leaky’ vaccines that reduce , but do not eliminate , the risk of infection of vaccinates can still control the spread of disease , particularly if the vaccine also reduces the infectiousness of vaccinated individuals . This distinction between the direct and indirect modes of action of vaccination is particularly relevant for BCG . Evidence from challenge ( Hope et al . , 2005 ) and natural transmission studies ( Ameni et al . , 2010 ) argues more strongly for a reduction in the rate of progression , with a larger proportion of vaccinated animals demonstrating a reduction in the extent of lesions than presenting with sterilizing immunity . For this reason , EFSA specified that field trial designs for the evaluation of BCG in cattle should be able to directly estimate the impact of vaccination on transmission ( EFSA , 2013 ) . Experimental transmission studies can be designed such that the impact of vaccination on transmission can be directly estimated , but achieving this in the field and within an ongoing test-and-slaughter program is considerably more challenging . The UK bTB control program is complex and dynamic , with the scheduling and interpretation of tests linked to the ( apparent ) burden of infection within herds ( Conlan et al . , 2012 ) . Furthermore , the removal of test-positive animals from herds as soon as they are disclosed means that the force of infection , which drives statistical power , is dependent on unobserved infection within cattle , wildlife and the wider environment . Exact likelihood-based methods of inference which can deal with this missing information , such as data-augmented MCMC ( Jewell et al . , 2009 ) have so far proven to be computationally intractable for bTB . As a result , published estimates of transmission rates in Great Brita have all used approximate methods of inference , depending on aggregating data at the population level from large numbers of herds ( Conlan et al . , 2012; O'Hare et al . , 2014; Brooks-Pollock et al . , 2014 ) . Given the scale of data required for these advanced methods and the need for results of any field trial to be transparent and easily communicated to stakeholders , we consider them inappropriate as a framework to design field trials . Instead , we focus on the use of classical relative risk measures of vaccine efficacy ( Smith et al . , 1984; Halloran et al . , 1991 ) , commonly used in field trial design for human vaccines , to quantify the likely impact of BCG vaccination on transmission . The basic requirement to estimate the indirect benefit of vaccination from either field ( Halloran et al . , 1991 ) or experimental trial designs ( Velthuis et al . 2007 ) is the inclusion of at least two groups with differing levels of vaccine coverage . By comparing the relative risk of transmission for unvaccinated individuals within herds that contain different proportions of vaccinated animals , the reduction in infectiousness of vaccinates that subsequently become infected can be estimated . For such designs , three separate vaccine efficacy measures can be defined ( Figure 1 ) . Direct efficacy quantifies the protection of individuals from infection and compares the risk of infection of vaccinated animals relative to unvaccinated animals either within the same herd or a control herd . Indirect efficacy compares the risk of infection of unvaccinated animals within a partially vaccinated herd to the risk of unvaccinated animals in an unvaccinated control herd . Finally , Total Efficacy compares the risk of infection of all animals within a partially vaccinated herd to that in unvaccinated control herds . We define the end point for calculation of these risk ratios as evidence of visible lesions or culture confirmation of all animals that are removed over the course of a trial due to a positive test reaction or natural turnover . Following EFSA recommendations , this controls for the impact that the imperfect specificity of bTB diagnostic tests may have on estimates of vaccine efficacy ( EFSA , 2013 ) . Of equal importance to the efficacy of the vaccination , and essential to quantify the potential costs-and-benefits of a cattle vaccination program , is the population level effectiveness of vaccination within the existing surveillance system . The herd level effectiveness of vaccination strategies can be assessed by comparing vaccinated , or partially vaccinated , herds to whole herd controls . As the lion’s share of costs associated with bovine tuberculosis , for both farmers and government , are incurred from testing and compensation , we choose to measure effectiveness of vaccination through statistical measures of within-herd persistence . Specifically , we consider the risk of breakdown ( herd level incidence ) , duration of breakdowns and the probability of recurrence . Note that despite similar definitions , these measures are not directly comparable to published estimates of within-herd persistence in Great Britain ( Karolemeas et al . , 2010; Karolemeas et al . , 2011; Conlan et al . , 2012 ) due to differences in the scheduling of testing during the proposed trials and the replacement of tuberculin with DIVA testing for both vaccinated and unvaccinated herds ( see Appendix 1 ) . We define the herd level incidence as the proportion of study herds that have a breakdown over the fixed time horizon of the trial design ( 3 years ) ; prolonged breakdowns as the proportion of herds that require more than 1 DIVA test in addition to the disclosing test to clear restrictions and recurrence as the proportion of herds that experience a breakdown and subsequently see a second incident with the time horizon of the trial . As previously discussed , estimating the indirect efficacy of vaccination requires at least two groups with different levels of vaccination coverage . By selecting one of these groups to be a set of unvaccinated controls , a two-phase design can also be used to estimate the herd level effectiveness of vaccination . We use our herd-level simulation models to predicted the effectiveness of vaccination and calculate appropriate sample sizes for different measures of vaccine efficacy . Throughout , we aim for an 80% statistical power , defined as the probability of failing to detect a given effect at the 97 . 5% significance level . Vaccinated and control animals are tested at 60-day intervals throughout the trial , with slaughter of DIVA test positive animals . Within the vaccinated group , the statistical power to estimate the direct efficacy of vaccination depends on our ability to estimate the attack rate in both the vaccinated and unvaccinated sub-populations . As such , for within-herd controls a balanced design where the target coverage is 50% will be optimal . This balanced design will also be optimal for estimating indirect efficacy for a vaccine that halves the rate of transmission . Estimates of the indirect efficacy depend on comparing the attack rate in the unvaccinated controls within the partially vaccinated group , to that in unvaccinated herds . In the case of indirect efficacy , we must balance our ability to estimate the attack rate in the within-herd controls against the effect size generated by the presence of vaccinated animals within the group . A priori , for an efficacious vaccine , we would expect rates of transmission to be lower in the vaccinated herds and thus we weight the design to vaccinate 75% of recruited herds , retaining 25% as unvaccinated controls . This design places a greater importance on our ability to measure the direct effect of vaccination , while still allowing for the estimation of a relatively large impact of vaccination on transmission should it exist .
In this conceptual design , Direct Efficacy can be estimated relative to either within-herd ( WH ) or between-herd ( BH ) control animals ( Figure 1 , 2 ) . For an assumed direct protection ( εS ) of 90% , and average duration of immunity of 1 year , the power calculations are relatively insensitive to this design choice and the assumed effect of vaccination on infectiousness ( εI ) . For this baseline assumed effect size of 90% ( Figure 2E ) , which corresponds to an effective efficacy ( ~ 60% ) comparable with existing experimental and field estimates for BCG ( Hope et al . , 2005; Ameni et al . , 2010; Lopez-Valencia et al . , 2010 ) , 100 randomly selected herds in GB would comfortably provide > 90% power to estimate a positive direct efficacy for both the alternative SORI ( Figure 2E , F ) and SOR models ( Figure 2—figure supplement 1 E , F ) . This lack of sensitivity of statistical power to the choice of controls extends to lower levels of protection ( εS=30% ) ) . However , in this scenario there is an increased sensitivity to the effect of vaccination on infectiousness ( εI ) and > 300 herds would be necessary to achieve the target of 80% power ( Figure 2A , B ) . Alternative designs with a single target level of vaccination ( distributed between or within-herds ) can mitigate this reduction in power and achieve the same statistical power with 100 herds ( results not shown , Triveritas , 2014 ) . The necessitity for designs to directly estimate indirect effects of vaccination therefore has a very real impact on the necessary scale of trials and the statistical power to estimate the basic individual level protection afforded by the vaccine . In contrast to direct efficacy , estimates of the indirect efficacy are more sensitive to the choice of model with an indirect efficacy of ~0 predicted by the SORI model ( Figure 2 ) and a positive indirect efficacy of up to 10% from the SOR model ( Figure 2—figure supplement 2 ) . This is a consequence of the different assumptions , discussed in detail in Appendix 1 , concerning the time from infection to infectiousness . For the SORI model , estimates of transmission rates are higher than for the SOR model; however , animals must pass through a period of latency before becoming infectious . For the SOR model , animals have lower estimated transmission rates but are immediately infectious upon infection . As a result , the SOR model is more sensitive to the impact of vaccination on infectiousness and predicts a greater indirect benefit of vaccination . Nonetheless , both models predict such a small indirect efficacy that there is a high probability of estimating a negative vaccine efficacy - implying an increase in infectiousness in vaccinated animals - even when a true protective effect exists ( Figure 2—figure supplements 1 , 2 and 3 ) . The magnitude of indirect protection for both models is constrained considerably by the removal of infectious animals as soon as they become DIVA positive and by the extrinsic rate of infection that captures the risk of both animal movements and the unobserved environmental reservoir . These within-herd models assume that vaccination has no impact on the reservoir of infection , hence the small magnitude of the predicted indirect benefits of vaccination . As a consequence of these two factors , the total efficacy is estimated to be approximately half that of the direct efficacy and the number of herds required to power a trial based on the total efficacy are correspondingly larger . Both models suggest that an 80% power of estimating a positive total efficacy would require >300 herds even for a true direct effect of vaccination on susceptibility of εS= 90% ( Figure 2 , Figure 2—figure supplement 2 ) . The underestimate of the ( instantaneous ) efficacy of vaccination ( εS , εI ) through ( cumulative ) relative risk measures is the natural consequence of the limited duration of immunity and dynamics of transmission within-herds . To explore how this systematic underestimate of vaccine efficacy through relative risk measures depends on trial duration and design , we examine the posterior predictive distributions for the within-herd prevalence of infection ( Figure 3 ) . We define within-herd prevalence as the proportion of the total at-risk population during a trial that is found to be either test-positive or culture confirmed at slaughter . These distributions reveal a high-frequency mode of singleton ( or very few ) reactor TB incidents - even for trial durations extending up to 9 years . This right skewed distribution of within-herd prevalence is consistent with the distribution of reactor animals seen within UK herds where less than half of bTB breakdowns have more than one reactor animal disclosed . The consequence of this low predicted attack rate in the majority of trial herds is a systematic underestimate of vaccine efficacy through relative risk measures . The power to discriminate between the attack rate in vaccinated and unvaccinated animal’s rests almost entirely with the relatively few herds that experience a high attack rate ( Figure 3 ) . The origins of this variability are multi-factorial including systematic differences in the within-herd reproduction ratio resulting from the demographic structure of herds , parametric uncertainty and variability in the extrinsic ( environmental ) risk of infection between herds even within the same risk areas . Herd size and the residence time of animals within a herd could in theory be used to target herds with a greater potential for transmission . However , the practicality of such targeting is limited by the relative infrequency of such herds , the necessity that participation in any trial would be voluntary and the additional requirement from the EU/EFSA that the study population for field evaluation should be representative of European production systems ( EFSA , 2013 ) . Targeting herds with a greater environmental risk of infection is impractical due to the lack of useful data or robust methodology to quantify these risks . Perhaps , the most natural step to increase the risk of transmission would be to retain , rather than cull , test-positive animals for the duration of any trial . However , such action has been ruled out by policy makers due to the legal and ethical issues of leaving animals known to be infected and may pose a risk of transmission to farm workers or researchers . Nonetheless , it is important to consider what effect this may have on the likely success of field trials . To this end , we explore the effect that retaining reactor animals has on the posterior predictive distribution for within-herd prevalence for trial durations of 3 , 6 and 9 years ( Figure 3 , Figure 3—figure supplement 1 ) . We see that for a 3-year trial this , unpalatable option for policy makers , would make no difference to the predicted attack rate in unvaccinated control herds due to the relatively low cattle-to-cattle transmission rates and long generation time of bTB in cattle ( Figure 3 , panel A ) . Even for impractically long trials of up to 9 years , retaining reactors only serves to thicken the long tail of herds that experience a relatively high rate of transmission ( Figure 3 , panel A ) . To maintain consistency with our design for efficacy , we consider the predicted impact of vaccination for all three of our herd level measures at a target vaccination coverage of 50% and compare with an alternative design with 100% whole herd vaccination . For all three herd level measures , the impacts of vaccination predicted by the SORI model at a baseline efficacy of εS=90% ( corresponding to a predicted effective direct efficacy of vaccination of ~ 75% ) are modest and variable , with an average improvement of between 10 and 20% for whole herd vaccination , halving to between 5 and 10% for a target coverage of 50% ( Figures 4 , 5 and 6 ) . As with the measures of vaccine efficacy , the predictive distributions for persistence measures manifest considerable variability with a substantial probability of observing a negative effect of vaccination even when a true protective effect exists ( Figure 4—figure supplement 1 , Figure 5—figure supplement 1 , Figure 6—figure supplement 1 ) . As a consequence of this variation , achieving the target statistical power of 80% , would require a study population of at least 500 herds for whole herd vaccination and in excess of 2000 herds ( the upper limit considered ) for the target coverage of 50% . The SOR model predicts a similar , but more variable effect size ( Figure 4—figure supplement 2 , Figure 5—figure supplement 2 , Figure 6—figure supplement 2 ) , that is more sensitive to the effect of vaccination on infectiousness ( Figure 5—figure supplement 3 , Figure 6—figure supplement 3 , Figure 6—figure supplement 3 ) .
We have used within-herd transmission models , with parameter distributions estimated from field data in Great Britain , to calculate indicative sample sizes for field trials of cattle vaccination with BCG as a supplement to an ongoing test-and-slaughter program for bTB . Our models suggest that evaluation of the direct protective effect of BCG in the field would be viable in the UK . A three year trial with 100 herds should provide an 80% power of estimating an individual protective efficacy of at least 30% . The scale of such a trial is affected by the requirement that test-positive animals are removed from trial herds , but is driven by the heterogeneity in within-herd prevalence of bTB in Great Britain . At the most basic level , demonstrating the efficacy of a vaccine depends on achieving sufficient exposure of vaccinated and control animals . This is a fundamental challenge for the managed cattle herds in Great Britain where high attack rates are limited to a very small proportion of affected herds . The relatively rarity of these herds and dependence on ( unmeasurable ) confounding factors , such as the environmental risk of infection , makes targeting this sub-population of herds impractical and biases estimates of vaccine efficacy through relative risk ratios . This distribution of disease has a bigger implication for the potential of field trials to measure the indirect efficacy of vaccination on transmission . For all of the conceptual trial designs and model scenarios considered in this paper , the Indirect Efficacy estimated from relative risk ratios would be essentially zero , with a high probability of estimating a negative efficacy in underpowered trials even when a considerable individual level reduction in infectiousness exists . Given the slow time-scale of bTB transmission , even the controversial step of retaining test-positive ( reactor ) animals within trial herds would not reduce the risk of a trial failing . Should BCG be licensed for use in cattle , at least in the UK , vaccination will be at the discretion of individual farmers who will be expected to bear the costs of vaccination . In the UK , the major economic costs for farmers accrue with respect to the frequency of testing and the period of time under restrictions . The individual efficacy of vaccination is therefore of far less interest to farmers than the herd level effects in terms of the impact of the surveillance and testing regime on their business ( Bennett and Balcombe , 2012 ) . Our models predict relatively modest improvements for farmers who would choose to vaccinate , with at most a 15% predicted reduction in the risk of a recurrent or prolonged breakdown . Part of the reason for this modest estimated effectiveness of vaccination in these model scenarios is that unvaccinated trial herds benefit from the likely benefits of the prospective DIVA test . The limited data available from challenge studies suggests that DIVA tests ( Conlan et al . , 2015 ) will have a higher sensitivity than tuberculin testing . The overall benefit of vaccination and DIVA testing together would be expected to be larger than the effect of vaccination or tuberculin testing alone . Another factor likely to limit the effectiveness of vaccination in our models is the constant extrinsic rate of infection , estimated from data , that is unaffected by the level of infection within the herd . This is a pragmatic modelling assumption , taken due to the complete lack of information routinely collected on the burden of disease within environmental and wildlife reservoirs . More complex dynamic models of the reservoir could , and have been , constructed in national level models ( Brooks-Pollock et al . , 2014 ) . However , no model can account for our basic data gap of the balance of transmission between cattle and wildlife populations that will ultimately determine the long-term outcome of vaccination ( Brooks-Pollock and Wood , 2015 ) . The appropriateness of our assumption of a constant reservoir will depend on the extent to which the rate of extrinsic infection into herds varies over the course of simulation . For the purposes of trial evaluation , this should be considered as a worst case scenario as vaccination will have no direct impact on reducing the infection risk within the static reservoir . However , over the ( relatively ) short timescale of a trial we believe this will be a reasonable approximation . The extent to which the impact of vaccination over longer time-frames will be greater depends critically on the relative rates of infection to and from the environmental reservoir ( Woodroffe et al . , 2016 ) and between species ( Brooks-Pollock and Wood , 2015 ) , the magnitude of which are highly uncertain and difficult to quantify . A consequence of this modest predicted benefit of vaccination is that herd level effectiveness would be exceptionally difficult to estimate from partially vaccinated herds , requiring a sample size in excess of 2000 herds . This highlights once more the devastating impact including partially vaccinated herds in the design , required to estimate the indirect effect of vaccination , has on the necessary scale of trials . The number of herds required could be reduced by a three arm design which includes fully vaccinated , partially vaccinated and unvaccinated control herds . However , such a design would still require of the order of 500 fully vaccinated herds and controls , compared to 100 to evaluate the direct protection , and still have a high risk of failing to provide actionable information on the impact of vaccination on transmission . On advice from Defra and informed by the results of this paper , the Triveritas consortium proposed an alternative to a three arm design with a phased series of trials to first evaluate vaccine efficacy , and then proceed to larger scale trials to quantify herd level effectiveness ( Triveritas 2014 ) . Such an approach to mitigating risk is implicit in the established standards for evaluation of human vaccines , a comparison that warrants further discussion . For human vaccines , evaluation of the population level effectiveness and indirect protection of a vaccine is typically reserved for Phase IV trials , carried out after the licensing and deployment of a vaccine at scale . In this light , the EU requirement that the impact of BCG on rates of transmission should be demonstrated before a vaccine can be licensed is notable . Although unusual , there are important biological reasons that motivated this requirement for cattle vaccination for bTB . It is possible that the use of ineffective vaccine in combination with a less sensitive DIVA test could lead to a perverse consequence of vaccination and increase the rates of silent transmission of infection . This important question must be addressed before the widespread deployment of BCG , but we would argue that field trials are not the most effective way to achieve this . In Appendix 2 , we illustrate that an natural transmission experiment involving as few as 200 animals over a two-year period could provide a greater power to not only estimate the efficacy of BCG , but also the mode of action in terms of the impact on susceptibility to infection and the infectiousness . Equivalent to a Phase II trial of a human vaccine , a successful experimental transmission study could provide the confidence to go ahead with field evaluation of the efficacy and effectiveness of vaccination without the necessity to compromise the power of trials with the inclusion of partially vaccinated herds . Our calculated sample sizes for natural transmission studies presented in Appendix 2 depend on estimates of transmission rates from field data where transmission rates scale with herd size ( discussed in Appendix 3 ) . The validity of density dependent scaling of transmission rates for small group settings is debatable , as the empirical relationship may be the consequence of husbandry factors that correlate with herd size rather than a true dependence on group size . For this reason , we suspect that field estimates of transmission may underestimate the transmission potential in small groups allowing for a shorter contact time . Nonetheless , the experience of previous transmission experiments with reactor animals from Great Britain ( Khatri et al . , 2012 ) would caution against committing to large , and potentially expensive , natural transmission study in the absence of more encouraging pilot data . Our proposed design recommends a group size of 52 animals and a contact period of 1 year in line with the more optimistic model scenarios . Endemically infected countries where the feasibility of natural transmission models has already been demonstrated ( Ameni et al . , 2010 ) are more promising locations for such experiments than Great Britain . However , the two-phase design of our natural transmission trial allows for a stop/go point , where phase I can be continued for an additional year if insufficient transmission is seen within the unvaccinated control animals . In this way , a trial could still provide key information on the direct efficacy of vaccination , even if low rates of transmission rule out evaluation of the indirect effects . Experimental trials for vaccine efficacy have the advantage - and disadvantage - that extrinsic sources of infection from wildlife and the environment can be eliminated and controlled for . Such experimental designs would provide more precise information on the efficacy and mode of action of vaccination for predicting the potential impact than could realistically be achieved in a field setting . They would not satisfy the current EC requirement , and EFSA recommendation , that trials should be carried out under European production conditions ( EFSA , 2013 ) or convince farmers about the practicality of cattle vaccination alongside an unmanaged wildlife reservoir . Natural transmission studies should therefore be considered as an initial screening step for any prospective vaccine before larger , more expensive and riskier trials in the field . Such field trials could ( or should ) be based on modelling of transmission in the cattle-wildlife system using among others parameter estimates from these transmission studies . From challenge data , we already know that BCG has the potential to provide a protective benefit to cattle . However , our results highlight the enormous scale of trials that would be necessary to evaluate BCG alongside continuing testing in the field . The scale of such trials could be dramatically reduced by addressing the mode of action of vaccination through smaller scale natural transmission studies . Based on our current knowledge of the likely efficacy of BCG , our models do not predict a substantial benefit of vaccination at the herd level when used as a supplement to ongoing test-and-slaughter . Indeed , the primary benefits predicted by our model come from the likely increase in diagnostic sensitivity provided by a replacement DIVA test rather than vaccination in itself . The format of the tuberculin skin test used in Great Britain – the Single Intradermal Comparative Cervical Tuberculin test ( SICCT ) prioritises diagnostic specificity over sensitivity . This is in contrast to countries who have successfully achieved TB-free status based on the use of the more sensitive Single Intradermal Test ( SIT ) . Although not the primary focus our study , our results reinforce the benefits for management of bTB that would come from routine use of a more sensitive and equally specific test . Likewise , our results highlight that ruling out the use of vaccination as a replacement , rather than a supplement , to test-and-slaughter will inevitably limit the effectiveness and perceived benefits for farmers . Reconsidering this policy option would revolutionise the economic case for the deployment of an effective vaccine , not only in Great Britain but in developing countries which can not afford to adopt expensive test-and-slaughter programmes .
For each vaccination scenario , defined by a unique level of vaccination coverage and assumed efficacy of vaccination , we simulate 5000 trials with from a sample of herds representative of the range of herd sizes and demography seen in Great Britain . Model parameters for each simulation are sampled from approximate Bayesian posterior distributions estimated for the relevant model , as described in Appendix 1 . Sensitivity to model parameters is thus implicit in our analysis , with simulations used to generate predictive posterior distributions for the statistical measure under consideration . We use the median value of these predictive distributions to quantify the expected effect size of vaccination and the full distribution to estimate the statistical power for each measure of vaccine efficacy . Sensitivity of our results to model structure is explored by comparing the two alternatives within herd transmission models ( SOR and SORI ) described in full in Appendix 1 . Simulations are initiated with no infection within herds and an extrinsic force of infection as estimated from breakdown herds in high incidence ( historic annual testing ) areas . Herds are initialised with no infection within the herd , and become infected at this extrinsic infection rate . Our simulated study population will therefore contain both affected ( breakdown ) and unaffected herds . Thus , estimated sample sizes correspond to the total number of herds that must be recruited rather than breakdowns . As the model is fitted to breakdown herds only , this background rate of infection should only be considered as representative of herds with a past history of bTB . Herds with no previous history of bTB might be expected to experience a lower rate of challenge from the outside of the herd and increase the calculated sample sizes . Relative risk measures of vaccine efficacy compare the attack rate in unvaccinated and vaccinated groups within a defined population as illustrated in Figure 1 . The attack rate within each group is calculated as the ratio of the number of cases divided by the total at risk population . For our purposes the at-risk population is defined as the total population of animals removed from herds over the duration of a trial and cases can either be culture confirmed test-positive animals or TB lesioned animals found at routine slaughter . Direct Efficacy compares the attack rate in vaccinated animals ( ARV ) against unvaccinated control animals ( ARU ) and is calculated as:1-ARVARUwhere ARV and ARU are calculated for each scenario using 10 , 000 independent samples from a pool of 5000 model simulations as described above . Indirect Efficacy can only be measured within designs with whole herd controls and vaccinated herds with target vaccination coverage of < 100% . Indirect efficacy compares the attack rate in unvaccinated animals ( ARUV ) within a vaccinated herd and that from unvaccinated control herds ( ARU ) and is calculated as:1-ARUVARU Total Efficacy can also only be measured within designs with whole herd controls and compares the attack rate in all animals on a partially vaccinated herd ( AR ) to that within unvaccinated control herds ( ARU ) and is calculated as:1-ARARU We base our power calculations for field trial designs upon a classical hypothesis test on the relative risk of infection ( RR ) in vaccinated compared to unvaccinated animals ( Kirkwood and Sterne , 2003 ) . We test against a null hypothesis of no difference between the two populations ( RR = 1 ) . To account for the high probability of estimating a negative efficacy , even when a protective efficacy exists , we use a one-sided test with alternative hypothesis H1: RR < 1 . The hypothesis test takes the form of a z-test with z = log ( RR ) /s . e . ( log ( RR ) ) , where the standard error of the relative risk is calculated using the standard result based upon the numbers of cases and at-risk animals in the vaccinated and unvaccinated groups . Power is then estimated based upon the empirical distribution of RR generated by sampling 10 , 000 independent outcomes from our pool of 5000 model simulations generated for each scenario . For each simulation , we calculate the z-statistic as described above and estimate the proportion of simulations where z is less than the critical value ( zcr ) defining the 95% level ( p=0 . 025 for 1-sided test ) . The power , defined as the probability of observing a significantly protective effect when it exists , is then calculated as the proportion of simulations where z < zcr . The effectiveness of vaccination at the herd level can be quantified in terms of the risk of breakdown ( herd level incidence ) , duration of breakdowns and the probability of recurrence . Note that due to the differences in the scheduling of testing during the proposed trials these measures are not directly comparable to those previously used to quantify within-herd persistence under the current statutory regime of testing . Quantifying these herd level measures requires a design with both vaccinated and unvaccinated herds subject to the same ( DIVA ) testing protocol . We consider three complementary measures of the potential effectiveness of cattle vaccination: | Bovine tuberculosis is an infectious disease of livestock and wildlife in many parts of the world . It also can spread to humans . In the United Kingdom ( UK ) , infected cattle and badgers contribute to its spread . To control bovine tuberculosis , cattle are tested and infected animals are slaughtered . Badgers in areas near cattle are killed to keep their populations small and reduce the likelihood of them infecting cattle . These control strategies are very controversial . Testing and slaughtering cattle is expensive and many people object to badger culling . Developing a vaccine that would protect cattle against bovine tuberculosis is a potential alternative approach being investigated by the UK government . But such a vaccination is currently illegal in Europe because vaccinated animals may test positive for infection , creating confusion . Tests for bovine tuberculosis exist , but these DIVA ( short for “Differentiates Infected from Vaccinated Animals” ) tests are not yet licensed for use in the UK . The European Union ( EU ) said it would consider relaxing its laws against bovine tuberculosis vaccination if the UK government is able to prove a vaccine is effective on farms . Now , Conlan et al . show that the specific field trials recommended by the EU would have to be extremely large to show a benefit of vaccination . Mathematical models were used to calculate how many cattle herds a bovine tuberculosis vaccine study would need to show that it protects cattle from infection , reduces transmission of the disease , and saves farmers money . Conlan et al . show that a study including 100 herds would be large enough to prove the vaccine protected individual animals . But a trial would have to include 500 herds to show that vaccination saves farmers money . Because transmission of bovine tuberculosis is slow in the UK , trials on working farms are unlikely to be able to measure whether vaccination reduces the spread of the disease . Instead , Conlan et al . show that smaller , less expensive experiments in controlled settings would be able to estimate the effects of bovine tuberculosis vaccination on transmission . These results informed the UK government decision to delay farm-based studies of a bovine tuberculosis vaccine until a DIVA test is available . If vaccination and the use of a DIVA test can be proven to be effective enough to replace test and slaughter policies it could be a huge economic boon to farmers , particularly those in lower income countries . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"computational",
"and",
"systems",
"biology"
] | 2018 | The intractable challenge of evaluating cattle vaccination as a control for bovine Tuberculosis |
Keratinocytes , the predominant cell type of the epidermis , migrate to reinstate the epithelial barrier during wound healing . Mechanical cues are known to regulate keratinocyte re-epithelialization and wound healing; however , the underlying molecular transducers and biophysical mechanisms remain elusive . Here , we show through molecular , cellular , and organismal studies that the mechanically activated ion channel PIEZO1 regulates keratinocyte migration and wound healing . Epidermal-specific Piezo1 knockout mice exhibited faster wound closure while gain-of-function mice displayed slower wound closure compared to littermate controls . By imaging the spatiotemporal localization dynamics of endogenous PIEZO1 channels , we find that channel enrichment at some regions of the wound edge induces a localized cellular retraction that slows keratinocyte collective migration . In migrating single keratinocytes , PIEZO1 is enriched at the rear of the cell , where maximal retraction occurs , and we find that chemical activation of PIEZO1 enhances retraction during single as well as collective migration . Our findings uncover novel molecular mechanisms underlying single and collective keratinocyte migration that may suggest a potential pharmacological target for wound treatment . More broadly , we show that nanoscale spatiotemporal dynamics of Piezo1 channels can control tissue-scale events , a finding with implications beyond wound healing to processes as diverse as development , homeostasis , disease , and repair .
The skin , the largest organ of the body , serves as a barrier against a myriad of external insults while also performing important sensory and homeostatic functions . Cutaneous wounds interfere with all these functions and expose the body to an increased risk of infection , disease , and scar formation ( Evans et al . , 2013 ) . During the repair of wounded skin , the migration of keratinocytes from the wound edge into the wound bed plays an essential step in re-establishing the epithelial barrier and restoring its protective functions ( Kirfel and Herzog , 2004; Gantwerker and Hom , 2011 ) . Accumulating evidence has shown that mechanical cues and cell-generated traction forces in keratinocytes play an important role in regulating the healing process and wound closure ( Evans et al . , 2013; Rosińczuk et al . , 2016; Brugués , 2014; Hiroyasu et al . , 2016; Huang et al . , 2017; Ladoux and Mège , 2017 ) . However , the molecular identity of keratinocyte mechanotransducers that control re-epithelialization remains unknown . Cells are able to sense and detect mechanical forces , converting them into biochemical signals through the process of mechanotransduction . One class of mechanosensors utilized by cells are mechanically activated ion channels which offer the unique ability for cells to rapidly detect and transduce mechanical forces into electrochemical signals ( Nourse and Pathak , 2017; Murthy et al . , 2017 ) . The Piezo1 ion channel has been shown to play an important role in a variety of cell types , and it regulates several key biological processes including vascular and lymphatic development , red blood cell volume regulation , stem cell fate , the baroreceptor response , cardiovascular homeostasis , cartilage mechanics , and others ( Li et al . , 2014; Ranade et al . , 2014; Pathak et al . , 2014; Rocio Servin-Vences et al . , 2017; Zeng et al . , 2018; Nonomura et al . , 2018; Cahalan et al . , 2015; Lee et al . , 2014; Rode et al . , 2017 ) . Previous studies in MDCK cells and in zebrafish larvae have demonstrated the importance of the channel in homeostatic regulation of epithelial cell numbers ( Gudipaty et al . , 2017; Eisenhoffer et al . , 2012 ) . As yet , the role of Piezo1 in skin wound healing , an important epithelial function , has not been investigated . We asked whether PIEZO1 may function as a mechanosensor regulating keratinocyte re-epithelialization during the wound healing process . Here , we show that PIEZO1 activity increases cellular retraction , reducing the efficiency of keratinocyte migration and wound healing , and that inhibition of PIEZO1 results in faster wound healing in vitro and in vivo . The channel exhibits dynamic changes in its subcellular localization , concentrating at areas of the wound edge and causing local retraction at these regions .
Analysis of Piezo channel mRNA expression in mouse tissues has previously shown that Piezo1 is highly expressed in skin , while Piezo2 is less abundant ( Coste et al . , 2010 ) . To characterize PIEZO1 expression profile in skin , we used a reporter mouse expressing a promoter-less β-geo ( β-gal and neomycin phosphotransferase ) in-frame with a portion of the PIEZO1 channel ( Ranade et al . , 2014 ) . LacZ staining of skin tissue from these reporter mice revealed a high expression of PIEZO1 in the epidermal layer of keratinocytes as well as in hair follicles ( Figure 1A ) . Since the global knockout of Piezo1 is embryonically lethal ( Li et al . , 2014; Ranade et al . , 2014 ) , we generated an epidermal-specific knockout mouse to investigate whether PIEZO1 plays a role in cutaneous wound healing . The Krt14Cre mouse line was crossed with Piezo1fl/fl mice ( Cahalan et al . , 2015 ) to generate Krt14Cre;Piezo1fl/fl mice ( hereafter referred to as conditional knockout , cKO ) which are viable , develop normally , and feature normal skin sections ( Figure 1—figure supplement 1 ) , consistent with observations by Moehring et al . , 2020 . qRT-PCR analysis using keratinocytes harvested from Piezo1-cKO and littermate control animals confirmed expression of Piezo1 and not Piezo2 in control mice ( Figure 1—figure supplement 2 ) , and showed that Piezo1 mRNA expression is efficiently abrogated in cells from cKO animals ( Figure 1B ) . Furthermore , we also generated a Piezo1 gain-of-function ( GoF ) mouse line ( Piezo1-GoF ) which expresses the gain of function ( GoF ) Piezo1 mutation , R2482H ( Ma et al . , 2018 ) , in keratinocytes . To confirm functional change to PIEZO1 in mutant keratinocytes , we performed Ca2+ imaging using total internal reflection fluorescence ( TIRF ) microscopy of keratinocytes isolated from Piezo1 cKO , GoF , and their respective control ( Cre- ) littermates . We previously reported that in adherent cells Piezo1 produces Ca2+ flickers in response to cell-generated forces in the absence of external mechanical stimulation ( Pathak et al . , 2014; Ellefsen et al . , 2019 ) . Compared to littermate control ( ControlcKO ) cells , keratinocytes from Piezo1-cKO mice showed a 63% reduction in Ca2+ flickers , indicating that a majority of Ca2+ flickers arise from cell-generated activation of the PIEZO1 channel ( Figure 1C , D , E , Figure 1—video 1 ) . Piezo1-GoF cells displayed a nearly threefold increase in the frequency of Ca2+ flickers relative to littermate controls ( ControlGoF ) ( Figure 1F , G , H . Figure 1—video 2 ) , further supporting PIEZO1 as a key source of Ca2+ flickers . A difference in the frequency of Ca2+ flickers between the ControlcKO and ControlGoF cells was observed , likely arising from different genetic backgrounds of the two strains . For this reason , in all subsequent experiments , mutant keratinocytes are compared to littermate control cells of the same genetic background . To investigate the function of PIEZO1 in keratinocytes in vivo , we generated full-thickness wounds on the dorsal skin of Piezo1 cKO , GoF , and their respective control littermates and assessed wound closure ( Figure 1I ) . Six days post wounding , Piezo1-cKO mice displayed significantly smaller wound areas relative to their control littermates , while Piezo1-GoF mice showed larger wound areas , suggesting that increased channel activity leads to impaired rates of wound closure ( Figure 1J ) . To determine whether the effect on wound healing was caused by changes to the rate of keratinocyte re-epithelialization , we mimicked the in vivo wound healing paradigm in vitro . We generated scratch wounds in keratinocyte monolayers to trigger the re-epithelialization process and allowed the monolayers to migrate toward each other ( Figure 1K ) . Scratches in monolayers of Piezo1-cKO keratinocytes closed faster than those from littermate ControlcKO cells ( Figure 1L , left ) . Conversely , scratch closure in monolayers of Piezo1-GoF keratinocytes was significantly slower ( Figure 1L , middle ) . Correspondingly , when the PIEZO1 agonist Yoda1 was added to healing ControlcKO monolayers at concentrations greater than 2 μM , scratch wound closure was also significantly impaired ( Figure 1L , right , Figure 1—figure supplement 3A ) , further supporting PIEZO1 involvement in re-epithelialization . No effect on wound closure was observed when Piezo1-cKO monolayers were treated with Yoda1 indicating that inhibition of scratch closure is the result of PIEZO1 activity ( Figure 1—figure supplement 3B ) . Collectively , our in vitro and in vivo data demonstrate that the PIEZO1 ion channel plays an important role in wound healing , with Piezo1 knockout accelerating the healing process ( Figure 1M ) . To determine whether the differences in wound closure rates arise due to PIEZO1’s effect on keratinocyte motility during the re-epithelialization process , we captured migration dynamics of dissociated single keratinocytes from Piezo1-cKO mice . Isolated cells were sparsely seeded onto fibronectin-coated glass-bottom dishes and imaged over several hours using differential interference contrast ( DIC ) time-lapse imaging ( Figure 2A , Figure 2—video 1 ) . We tracked the position of individual cells in the acquired movies and analyzed the extracted cell migration trajectories using an open-source algorithm , DiPer ( Gorelik and Gautreau , 2014 ) . The time-lapse images and corresponding cell migration trajectories ( Figure 2B; Figure 2—figure supplement 1 ) revealed that the migration patterns of Piezo1-cKO keratinocytes are distinct from their littermate control cells . To quantify cellular migration , we generated mean squared displacement ( MSD ) plots which provide a measure of the surface area explored by the cells , and is an indication of the overall efficiency of migration . Interestingly , Piezo1-cKO keratinocytes explored a larger area compared to littermate ControlcKO cells ( Figure 2C ) . The MSD of a migrating cell is determined by two parameters: directional persistence ( propensity of the cell to move in a straight line ) and displacement rate ( speed ) . To assess directional persistence , we performed direction autocorrelation analysis , a robust measure of migration directionality that , unlike the more commonly used directionality ratio analysis , is not confounded by differences in migration speed ( Gorelik and Gautreau , 2014 ) . The direction autocorrelation function for trajectories from Piezo1-cKO keratinocytes decayed slower than for littermate ControlcKO cells , indicative of fewer turns and a straighter trajectory ( Figure 2D ) . The average instantaneous speed calculated by DiPer analysis was higher for Piezo1-cKO cells relative to littermate ControlcKO cells ( Figure 2E ) . Thus , Piezo1-cKO keratinocytes migrate significantly faster and straighter . Similarly , we extracted cell migration trajectories from single migrating keratinocytes harvested from Piezo1-GoF and littermate ControlGoF mice ( Figure 2—figure supplement 2 , Figure 2—video 2 ) . We observed no difference in the MSD plots of Piezo1-GoF keratinocytes and littermate ControlGoF cells ( Figure 2—figure supplement 3 ) . However , separating the data into directionality and speed indicated that Piezo1-GoF cells moved straighter ( Figure 2—figure supplement 3 ) and slower ( Figure 2E ) . Overall , our data demonstrate that PIEZO1 regulates keratinocyte migration , with channel activity resulting in slower migration speed . The effects on directionality were more complex , with both PIEZO1 knockout and a GoF mutation resulting in straighter trajectories . To gain insights into how PIEZO1 may regulate cell migration , we visualized localization of endogenous PIEZO1 in single keratinocytes harvested from a Piezo1-tdTomato fusion knock-in reporter mouse ( Ranade et al . , 2014 ) . Using this model , we previously reported punctate membrane localization of endogenous PIEZO1-tdTomato channels in neural stem/progenitor cells and mouse embryonic fibroblasts ( Ellefsen et al . , 2019 ) . Here , we directly imaged endogenous PIEZO1-tdTomato’s subcellular localization in individual live , migrating keratinocytes using TIRF imaging and noticed higher PIEZO1 levels at the rear end of the cell ( Figure 3A , Figure 3—video 1 ) . This observation of PIEZO1-tdTomato enrichment at the rear of single migrating cells suggests that PIEZO1 may underlie cell polarization during migration . To determine whether PIEZO1 may be responsible for generating the polarized shape , we performed cellular morphometrics on the images obtained from the above time-lapse imaging of single cell migration . We used visually aided morpho-phenotyping image recognition ( VAMPIRE ) ( Phillip et al . , 2021 ) , a high-throughput machine-learning algorithm that analyzes the morphology of individual cells in a population by quantifying shape modes of segmented cells and showing the level of correlation between the shape modes through a dendrogram ( Figure 3B and C ) . VAMPIRE classification of the Piezo1-cKO and littermate ControlcKO keratinocytes into 20 shape modes revealed that Piezo1-cKO reduced the proportion of highly polarized shapes and increased the proportion of weakly polarized shapes relative to littermate ControlcKO keratinocytes ( Figure 3D and E ) . On the other hand , the GoF mutation increased the frequency of polarized and hyper-polarized shapes at the expense of unpolarized or weakly polarized cell shapes ( Figure 3F and G ) . Taken together , these results indicate that PIEZO1 activity promotes cell polarization . Based on imaging the localization of endogenous PIEZO1 channels in migrating cells , it appears that this may be mediated by regulation of the channel’s subcellular localization . We then examined a role for PIEZO1 localization in wounded cell monolayers . We generated a scratch wound in a confluent monolayer of Piezo1-tdTomato keratinocytes and imaged spatiotemporal dynamics of PIEZO1-tdTomato localization at the cell-substrate interface using TIRFM imaging together with DIC imaging over a period of several hours . We found that at some regions along the wound margin , PIEZO1-tdTomato was enriched in band-like structures ( Figure 4A ) . Interestingly , this enrichment , which was observed a few hours after scratch generation ( compare Figure 4—figure supplement 1 and Figure 4A ) , was also highly dynamic , such that it ebbed and flowed over the course of imaging ( Figure 4—video 1 ) . We asked whether regions displaying PIEZO1-tdTomato enrichment migrate differently from regions without enrichment . To systematically assess the relationship between PIEZO1-tdTomato enrichment and wound edge dynamics , we used kymographs to graphically represent PIEZO1-tdTomato position over the imaging period from regions that displayed PIEZO1-tdTomato enrichment at the wound edge ( Figure 4C , E , Figure 4—video 1 ) and compared them to control fields of view that showed no such channel enrichment throughout the videos ( Figure 4B , D , Figure 4—video 2 ) . PIEZO1-tdTomato enrichment events at the wound edge appeared as linear streaks in the kymographs ( Figure 4E , left panel ) which could be objectively identified by Kymobutler , a deep-learning-based kymograph analysis software ( Jakobs et al . , 2019; Figure 4E , middle and right panels ) . For the kymographs from control fields of view ( Figure 4B , Figure 4—video 2 ) , Kymobutler did not detect any tracks and we did not observe retraction of the wound edge ( Figure 4D ) . In fields of view that exhibited PIEZO1-tdTomato puncta enrichment at the wound edge , this channel enrichment was followed by a localized retraction of the wound edge ( Figure 4C , Figure 4E , Figure 4—video 1 ) . Kymobutler analysis of PIEZO1-tdTomato tracks overlaid on the DIC kymographs allowed examination of the migration dynamics of the wound edge in relation to Piezo1 enrichment ( Figure 4E ) ; 72% of these PIEZO1-tdTomato enrichment tracks displayed a negative slope corresponding to cell edge retraction and aligned with retraction events ( Figure 4G ) . In some fields of view , PIEZO1 enrichment and the accompanying retraction lasted for shorter periods of time , and the periods without channel enrichment were accompanied by wound edge protrusion . In other cases , channel enrichment was maintained for several hours and was accompanied by a sustained and overt retraction of the wound edge throughout that period ( Figure 4F , Figure 4—video 3 ) . Thus , enrichment of PIEZO1-tdTomato puncta resulted in wound edge retraction ( Figure 4H ) . Importantly , the rear end of single migrating cells , where PIEZO1 is found to localize ( Figure 3A ) , is also the site of cellular retraction ( Petrie et al . , 2009; Yam , 2007 ) , suggesting a general relationship between PIEZO1 localization and retraction . To examine the relationship between PIEZO1 activity and cellular retraction , we examined the effect of chemical activation of PIEZO1 by Yoda1 . We first focused on cellular dynamics of single , migrating ControlcKO keratinocytes by imaging at a high spatiotemporal resolution . Using DIC imaging , we monitored migrating keratinocytes at 5 second intervals under control conditions and after Yoda1 treatment ( Figure 5A , Figure 5—video 1 ) . Kymographs were used to visualize changes in the cell edge position over time . We observed that under control conditions , the cell edge displayed cycles of protrusion and retraction which was expected since cell migration is known to progress by iterative cycles of protrusion and retraction ( Giannone et al . , 2004; Giannone et al . , 2007 ) . PIEZO1 activation by 4 μM Yoda1 greatly affected these cycles and resulted in an extremely dynamic cell edge ( Figure 5A , Figure 5—video 1 ) . The frequency as well as the speed of cell edge retractions and protrusions increased upon Yoda1 treatment but resulted in a net cellular retraction over time ( Figure 5A , Figure 5—figure supplement 1C ) , with some cells demonstrating drastic retraction with Yoda1 treatment ( Figure 5—video 2 ) . Piezo1-cKO keratinocytes did not show an increase in retraction upon treatment with 4 μM Yoda1 ( Figure 5B and , Figure 5—video 3 , Figure 5—figure supplement 1 ) , demonstrating that Yoda1-induced increase in retraction is mediated by PIEZO1 . Additionally , kymographs of Piezo1-GoF keratinocytes also showed an increase in cell edge dynamics compared to littermate ControlGoF cells further supporting PIEZO1’s role in retraction ( Figure 5—figure supplement 1 and Figure 5—video 4 ) . Kymograph-based quantitation is limited to one point along the cell edge . To more objectively investigate the effect that PIEZO1 activation has on cell morphodynamics , cells were segmented for each frame of a DIC time-lapse series for the following conditions: ControlcKO cells before and after Yoda1 addition , Piezo1-cKO cells , ControlGoF cells , and Piezo1-GoF keratinocytes ( Figure 5C ) . By comparing segmented outlines between frames , we could obtain the velocity of the cell edge at every position along the detected boundaries for protrusion events ( positive velocities ) and retraction events ( negative velocities ) . Yoda1 treatment resulted in a significant increase in edge velocities compared to DMSO-treated ControlcKO cells ( Figure 5D ) . Yoda1 , which is expected to globally activate PIEZO1 channels , resulted in an increase of cell edge velocity during both protrusion and retraction events , though the increase in retraction velocity was greater . In contrast , Piezo1-cKO keratinocytes showed a reduction in edge velocities relative to littermate ControlcKO cells . Heatmaps of cell edge velocity illustrate the robustness of this response ( Figure 5E ) . Consistent with our observations of Yoda1 treatment , Piezo1-GoF keratinocytes also showed a significant increase in edge velocities relative to littermate ControlGoF ( Figure 5D , Figure 5—figure supplement 2 ) . Interestingly , there was a clear increase in the proportion of retracting positions relative to protruding positions in GoF keratinocytes . These results reveal that PIEZO1 activity regulates cell edge dynamics and further support our observations that PIEZO1 activity increases cellular retraction . We then asked whether PIEZO1-mediated retraction events observed in single cells are relevant in the context of a wounded cell monolayer . We generated scratch wounds in keratinocyte monolayers and performed DIC time-lapse imaging of the healing monolayer in the presence and absence of 4 μM Yoda1 . Monolayers of control keratinocytes advanced forward into the cell-free space under DMSO-treated control conditions ( Figure 5F , Figure 5—video 5 ) , while the presence of 4 μM Yoda1 increased retraction events which prevented the monolayer from advancing far into the wound bed ( Figure 5G , Figure 5—video 5 ) . Remarkably , in 35% of fields of view monitored in scratch assays we observed that Yoda1 treatment resulted in an increase in scratch area instead of wound closure ( Figure 1L ) . In Piezo1-cKO monolayers we observed cells protrude forward into the cell-free space to close scratch wounds to a greater extent than controls ( Figure 5H , Figure 5—video 6 ) , while GoF monolayers did so to a lower extent ( Figure 5—figure supplement 3 , Figure 5—video 7 ) . Additionally , no effect of Yoda1 addition was seen on the rate of advancement in kymographs taken at the wound edge of Piezo1-cKO monolayers ( Figure 5—figure supplement 4 ) . Collectively , our results show that PIEZO1 activity increases cellular retraction in keratinocytes , both in single cells and monolayers , which has a net effect on cell migration . Taken together , our results demonstrate that PIEZO1 induces cellular retraction to slow single and collective cell migration and thus causes delayed wound healing . We propose that dynamic enrichment of the channel protein serves to locally amplify channel activity and the downstream retraction events . In wound healing monolayers of keratinocytes , PIEZO1 enrichment and the subsequent wound edge retraction provide a molecular mechanism for how PIEZO1 slows wound healing , while absence of the channel accelerates wound healing .
Our findings demonstrate that epidermal-specific Piezo1 knockout resulted in faster wound healing in mice , and conversely a Piezo1-GoF mutation slowed wound healing . We recapitulate this observation in vitro , and show through a combination of orthogonal assays in single cells and in monolayers that PIEZO1 activity modulated by dynamic spatial reorganization of the channel protein slows keratinocyte cell migration during re-epithelialization . These findings provide physiological evidence for the role of a mechanically activated ion channel in wound healing and suggest potential therapeutics through a targeted inhibition of PIEZO1 , perhaps applied topically , that could help speed wound healing , potentially reducing risk of infection . However , further in-depth analysis regarding the quality of wound healing following PIEZO1 inhibition is required . Given that speedy wound healing affords an evolutionary advantage and that PIEZO1 activity slows wound healing , a puzzling question arises regarding the role of PIEZO1 expression in keratinocytes . Perhaps , there is an advantage to slower healing in the presence of PIEZO1 , or the channel is important for other functions in keratinocytes . Consistent with the latter idea , a recent study by Moehring et al . reports that keratinocyte PIEZO1 is critical for sensory afferent firing and behavioral responses to innocuous and noxious mechanical stimulation ( Moehring et al . , 2020 ) . As such , it would also be important to determine that inhibiting Piezo1 to speed wound healing does not have detrimental effects on normal mechanosensation . Ca2+ signals control many aspects of cell migration , including lamellipodial dynamics , traction force generation , rear retraction , focal adhesion turnover , and migration directionality ( Wei et al . , 2012; Tsai et al . , 2015; Canales et al . , 2019 ) . While mechanically activated ion channels were proposed to contribute to Ca2+ signaling in single cell migration in vitro as early as 20 years ago ( Lee et al . , 1999; Doyle et al . , 2004; Wei et al . , 2009; Tsai and Meyer , 2012; Patkunarajah et al . , 2020 ) , many important questions have remained unanswered , including those related to channel identity , functional effects in collective cell migration , and physiological contribution during wound healing . Using chemical activation as well as genetic modulation of PIEZO1 , we provide evidence for its involvement in regulating cellular retraction events in single cells as well as in collective cell migration during keratinocyte re-epithelialization . Notably , the effects we observed on scratch wound closure and cell edge retraction speeds following Yoda1 activation of PIEZO1 were consistently larger than the effects of Piezo1-GoF mutation . This is not surprising as the GoF mutation increases ion flux through PIEZO1 channels , without significantly affecting channel activation ( Glogowska et al . , 2017; Bae et al . , 2013 ) . Thus , channel activation in GoF keratinocytes is expected to occur based on subcellular localization of PIEZO1 and of the cellular forces that activate them . In contrast , Yoda1 treatment would globally activate PIEZO1 channels in the plasma membrane , leading to a larger effect . One of the most surprising findings to emerge from our studies is the highly dynamic nature of the spatial localization of PIEZO1 channels in migrating cells . Based on this finding , we propose a novel mechanism regulating cell migration wherein spatiotemporal enrichment of PIEZO1 channels serves to localize and amplify channel activity , and regulate contractile forces , to spatially control cellular retraction events . Piezo1 has been implicated in cell migration in different cell types in vitro ( Maneshi et al . , 2018; McHugh et al . , 2012; Chubinskiy-Nadezhdin et al . , 2019; Hung et al . , 2016; Li et al . , 2015; Yu et al . , 2020 ) ; however , reports of the effect of the channel on migration have varied in the literature , with channel activity supporting migration in some cell types and inhibiting migration in others . Perhaps , a determining factor of the channel’s impact on cell migration is how spatiotemporal localization of the channel is regulated in a given cell type . Our observations spark several new questions regarding the regulation and functional impacts of PIEZO1’s localization and clustering dynamics . Interestingly , molecular dynamics simulations of PIEZO1 suggest that interactions between neighboring PIEZO1 channels may enable cooperative gating between channels ( Jiang et al . , 2021 ) . Clustering of the bacterial mechanosensitive channel MscL has also been reported , but computational simulations predict these clusters in fact decrease channel open probability , providing a defence against unwanted channel gating that may cause osmotic shock ( Paraschiv et al . , 2020; Grage et al . , 2011 ) . Experimental evidence for functional interaction between PIEZO1 channels remains pending and two recent patch clamp studies examining this question come to divergent conclusions ( Lewis and Grandl , 2021; Wijerathne et al . , 2021 ) . It is well established that retraction during cell migration occurs due to force generated by myosin II ( Cramer , 2013; Ridley , 2011; Aguilar-Cuenca et al . , 2014 ) , and we previously showed that myosin II-mediated cellular traction forces elicit localized Piezo1 Ca2+ flickers ( Ellefsen et al . , 2019 ) . Since myosin II activation is enhanced by intracellular Ca2+ ( Somlyo and Somlyo , 2003 ) , we speculate that Piezo1 may induce cellular retraction through a feedforward loop between Piezo1 and myosin II: traction force generation by myosin II cause Piezo1-mediated Ca2+ influx , which in turn may increase myosin II phosphorylation and force generation through the Ca2+-regulated Myosin Light Chain Kinase . Enrichment of Piezo1 in subcellular regions would amplify this effect and result in a localized retraction . Supporting this model , Piezo1-mediated Ca2+ events were recently found to elicit retraction of developing endothelial tip cells during vascular pathfinding ( Liu et al . , 2020 ) . Efficient migration requires protrusion of the cell lamellipodia which is stabilized via the formation of focal adhesions . Without stabilization , lamellipodia protrusions retract backward causing membrane ruffling and reducing migration speed ( Borm et al . , 2005 ) . We observe an increase in cell edge velocity in keratinocytes with increased PIEZO1 activity ( in both Piezo1-GoF and Yoda1-treated cells ) indicating that PIEZO1-mediated effects on cell edge velocity may be the predominant mechanism contributing to inefficient migration . An intriguing paradox observed in our data is that both Piezo1-cKO and Piezo1-GoF keratinocytes exhibit straighter migration trajectories . We also observe that PIEZO1 localization and increased activity appears linked to cell polarization; thus one possible explanation that warrants further study is that the GoF mutation stabilizes PIEZO1 clusters so that once the channels localize to the rear of the cell , the cell retains its polarization and migration directionality . Cell migration involves a complex orchestration of events , including sub-cellular dynamics in which cytoskeletal processes in different compartments of the cell need to be implemented in a precise spatiotemporal order . How this is achieved remains an open question . Our findings suggest that spatiotemporal enrichment dynamics of Piezo1 play a role in this coordination . More broadly , our findings provide a mechanism by which nanoscale spatial dynamics of Piezo1 channels can control tissue-scale events , a finding with implications beyond wound healing to processes as diverse as development , homeostasis , disease , and repair .
All studies were approved by the Institutional Animal Care and Use Committee of University of California at Irvine and The Scripps Research Institute , as appropriate , and performed in accordance with their guidelines . Piezo1 LacZ reporter mice ( JAX stock 026948 ) and Piezo1-tdTomato reporter mice , expressing a C-terminal fusion of Piezo1 with tdTomato ( Piezo1-tdTomato; JAX stock 029214 ) , were generated in a previous study ( Ranade et al . , 2014 ) . Skin-specific Piezo1-cKO mice were generated by breeding Piezo1fl/fl mice ( Cahalan et al . , 2015 ) ( Jax stock 029213 ) with K14Cre ( The Jackson Laboratory , stock 004782 ) . Skin-specific Piezo1-GoF mice were generated by breeding mice with conditional GoF Piezo1 allelle ( Piezo1cx/cx mice [Ma et al . , 2018] ) with K14Cre mice . Piezo1fl/fl mice were generated in C57BL/6 background and Piezo1cx/cx mice were initially generated in BALB/c background and then maintained in C57BL/6 for >10 generations . K14Cre mice were in the C57BL/6 background . P0–P5 mice were anesthetized with ice prior to decapitation . Bodies were placed in 10% povidone for 1 min , rinsed with sterile PBS , prior to soaking in 70% ethanol for a further minute , and rinsed again with sterile PBS . Subsequently , the entire upper dorsal skin above the abdomen was separated from the body . Dorsal skin was left to dissociate in either 0 . 25% trypsin/EDTA ( Gibco ) for 1 hr at 37°C or 1× dispase solution ( CellnTec CnT-DNP-10 ) at 4°C overnight for 15–18 hr . After incubation , the epidermis was gently separated from the dermis , laid flat , dorsal side down in Accutase ( CellnTec CnT-Accutase-100 ) and incubated for 30 min at room temperature . The epidermis was then transferred to a dish of either CnT-02 or CnT-Pr media ( CellnTec ) , supplemented with 10% FBS and 1% penicillin/streptomycin . The epidermis was cut into small pieces with scissors prior to agitation on a stir plate for 30 min . Cells were then filtered through a 70 µm cell strainer ( Falcon ) and spun down at 1200 rpm for 5 min . The pellet was resuspended in CnT-Pr media ( CellnTec ) supplemented with ISO-50 ( 1:1000 ) ( CellnTec ) and Gentamicin ( 50 µg/ml ) ( Thermo Fisher ) , prior to counting and plating . For live-cell imaging , primary keratinocytes were plated on #1 . 5 glass-bottom dishes ( Mat-Tek Corporation ) coated with 10 µg/ml fibronectin ( Fisher Scientific , CB-40008A ) . For monolayer experiments , cells were plated at 1 . 5 × 105 cells/dish in the 14 mm glass region of dishes . For sparse cell migration experiments and for Ca2+ imaging experiments , cells were plated at 1 . 5 × 104 cells/dish in the 14 mm glass region of dishes . Keratinocytes were imaged following at least 2 days in Cnt-Pr-D ( CellnTec ) differentiation media . After initial keratinocyte isolation and filtering through a 70 µm cell strainer , cells were filtered again through a 40 µm strainer . The filtered solution was spun down and a cell pellet was obtained for RNA isolation . Total RNA was isolated using the RNeasy kit ( Qiagen ) , following which cDNA was synthesized using Superscript III ( Invitrogen ) and was used for subsequent qPCR experiments ( ABI 7900HT fast real-time system ) . qPCR Taqman probes ( Thermo Fisher ) used were Piezo1: Assay ID Mm01241570_g1; Piezo2; Assay ID Mm01262433_m1 , Krt14; Assay ID Mm00516876_m1 and Gapdh; Assay ID Mm99999915_g1 . PCR reactions were run in triplicate . Dorsal skin was harvested as described above , cryopreserved in OCT , and sectioned into 8 µm thick slices . Skin cryosections were allowed to completely dry prior to being fixed in ‘fix buffer’ composed of 1× PBS , 5 mM EGTA ( Sigma Cat#E4378 ) , 2 mM MgCl2 , 0 . 2% glutaraldehyde ( Sigma Cat#G-7776 ) , pH 7 . 4 for 15 min at room temperature . Next they were washed with ‘wash buffer’ composed of 1× PBS , 2 mM MgCl2 twice for 5 min each . X-gal staining buffer composed of 1× PBS , 2 mM MgCl2 , 5 mM potassium ferrocyanide [K4Fe ( CN ) 6·3H20] ( Sigma Cat# P-9287 ) , 5 mM potassium ferricyanide [K3Fe ( CN ) 6] ( Sigma Cat#P-8131 ) , and 1 mg/ml X-gal [5-bromo-4-chloro-3-indolyl-β-D-galactoside] was made fresh . Then , tissue slides were incubated overnight at 37°C in the ‘X-gal staining buffer’ inside a humidified chamber . The following day , slides were rinsed with 1× PBS and counterstained with Nuclear Fast Red . Slides were then fixed with 4% PFA for longer preservation . For immunostaining of skin sections in Figure 1—figure supplement 1 , dorsal skin was prepared and sectioned as for X-Gal staining . Skin cryosections were fixed for 10 min in cold acetone , washed twice in 1× PBS prior to blocking for 30 min in 10% normal goat serum at room temperature . Primary antibodies used were Rabbit anti-Keratin 14 ( Covance , Cat#PRB-155P ) , 1:1000 ( 1 µg/ml ) , and Rabbit anti-Keratin 10 ( Covance , Cat#PRB-159P ) , 1:1000 ( 1 µg/ml ) . Secondary antibody used was Goat anti-Rabbit Alexa Fluor 488 ( Invitrogen , Cat#A11008 ) , 1:1000 . Nuclei were stained by DAPI ( Invitrogen , Cat#D1306 ) , 1:50 , 000 . All antibody incubations were performed at room temperature , for 1 hr in 1% BSA in PBS . Slides were mounted in gelvatol containing DAPI . Adult ( 3–4 months ) male and female mice were anesthetized with isoflurane and placed on a heated blanket . The dorsal hair was shaved and further removed by hair-removal cream . Two full-thickness wounds were created in the upper dorsal skin above the abdomen using a 4 mm wide dermal biopsy punch ( Integra LifeSciences Corporation ) . Wounded areas were patched with medical dressing , Tegaderm ( 3 M ) . Wound sizes were measured with a scale loupe ( Peak Optics , #1975 ) at day 6 to compare healing progress . Both the short ( dS ) and long ( dL ) diameters of the oval-shaped wounds were measured and used to calculate an overall wound area using the equation: dS × dL × π . Sample sizes are indicated in corresponding figures . Cumming estimation plots were generated and Cohen’s d in all plots ( except Figure 5D ) was calculated using an online estimation stats tool ( https://www . estimationstats . com ) ( Ho et al . , 2019 ) . Estimation plots show the raw data plotted on the upper axes with bars beside each group denoting the sample mean ± s . d . ; the mean difference and Cohen’s d effect size is plotted on the lower axes . On the lower plot , the mean difference is depicted as a dot; the 95% confidence interval is indicated by the ends of the bold vertical error bar . OriginPro 2020 ( OriginLab Corporation ) was used for calculating p values ( for all figures except for Figure 5D ) and generating plots used in Figure 1B , Figure 2B , C , D , Figure 3D , E , F , G , Figure 4G , Figure 5D1—5 , Figure 1—figure supplement 2 , and Figure 2—figure supplements 1–3 . All plots generated in OriginLab are presented as the mean ± SEM . Statistical tests used to calculate p values are indicated in figure legends . For Figure 5D , a Python script was used for calculating Cohen’s d and p values . The datasets plotted in Figure 1E , H , J , L , Figure 2B , C , D , E , Figure 3D , E , F , G , Figure 4G , Figure 5E , Figure 1—figure supplements 2–3 , Figure 2—figure supplements 1–3 , and Figure 5—figure supplements 1C and 2 have been uploaded as source data files . Source data files for Figure 5C and D have been uploaded to Dryad ( doi:10 . 5061/dryad . hdr7sqvjr ) . Code used to analyze ADAPT heatmaps and provide statistical analysis for velocity violin plots shown in Figure 5D has been supplied as a supplementary datafile . This code has also been made publicly available as a jupyter notebook uploaded to Github . | The skin is the largest organ of the body . It enables touch sensation and protects against external insults . Wounding of the skin exposes the body to an increased risk of infection , disease and scar formation . During wound healing , the cells in the topmost layer of the skin , called keratinocytes , move in from the edges of the wound to close the gap . This helps to restore the skin barrier . Previous research has shown that the mechanical forces experienced by keratinocytes play a role in wound closure . Several proteins , called mechanosensors , perceive these forces and instruct the cells what to do . Until now , it was unclear what kind of mechanosensors control wound healing . To find out more , Holt et al . studied a recently discovered mechanosensor ( for which co-author Ardem Pataputian received the Nobel Prize in 2021 ) , called Piezo1 , using genetically engineered mice . The experiments revealed that skin wounds in mice without Piezo1 in their keratinocytes healed faster than mice with normal levels of Piezo1 . In contrast , skin wounds of mice with increased levels of Piezo1 in their keratinocytes healed slower than mice with normal levels of Piezo1 . The same pattern held true for keratinocytes grown in the laboratory that had been treated with chemicals to increase the activity of Piezo1 . To better understand how Piezo1 slows wound healing , Holt et al . tracked its location inside the keratinocytes . This revealed that the position of Piezo1 changes over time . It builds up near the edge of the wound in some places , and at those regions makes the cells move backwards rather than forwards . In extreme cases , an increased activity of Piezo1 can cause an opening of the wound instead of closing it . These findings have the potential to guide research into new wound treatments . But first , scientists must confirm that blocking Piezo1 would not cause side effects , like reducing the sensation of touch . Moreover , it would be interesting to see if Piezo1 also plays a role in other important processes , such as development or certain diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2021 | Spatiotemporal dynamics of PIEZO1 localization controls keratinocyte migration during wound healing |
Replicative aging in yeast is asymmetric–mother cells age but their daughter cells are rejuvenated . Here we identify an asymmetry in pH between mother and daughter cells that underlies aging and rejuvenation . Cytosolic pH increases in aging mother cells , but is more acidic in daughter cells . This is due to the asymmetric distribution of the major regulator of cytosolic pH , the plasma membrane proton ATPase ( Pma1 ) . Pma1 accumulates in aging mother cells , but is largely absent from nascent daughter cells . We previously found that acidity of the vacuole declines in aging mother cells and limits lifespan , but that daughter cell vacuoles re-acidify . We find that Pma1 activity antagonizes mother cell vacuole acidity by reducing cytosolic protons . However , the inherent asymmetry of Pma1 increases cytosolic proton availability in daughter cells and facilitates vacuole re-acidification and rejuvenation .
During replicative aging in budding yeast , mother cells produce a finite number of daughter cells before arresting ( Mortimer and Johnston , 1959 ) . Because replicative aging is asymmetric , the process of aging occurs in mother cells but is absent in daughter cells ( Egilmez and Jazwinski , 1989; Kennedy et al . , 1994 ) . Several asymmetric phenotypes have been identified and proposed to contribute to mother cell decline ( Sinclair and Guarente , 1997; Lai et al . , 2002; Aguilaniu et al . , 2003; Erjavec et al . , 2007; Eldakak et al . , 2010; McFaline-Figueroa et al . , 2011; Nystrom and Liu , 2014 ) . We recently found that the acidity of the yeast lysosome-like vacuole is asymmetric between mother and daughter cells . Vacuole acidity declines in mother cells in early age and limits lifespan , but daughter cells have acidic vacuoles ( Hughes and Gottschling , 2012 ) . To identify what reduces vacuole acidity and how vacuole acidity is regenerated in daughter cells , we further characterized vacuole pH asymmetry . Cells were aged using a genetic system ( Lindstrom and Gottschling , 2009 ) and vacuole acidity was monitored by staining cells with quinacrine , a fluorescent dye that accumulates in the acidic vacuole ( Weisman et al . , 1987 ) . We observed bright vacuolar quinacrine staining indicative of acidic pH in a high percentage of buds ( nascent daughter cells ) regardless of mother cell age , whereas staining was diminished or undetectable in mother cell vacuoles ( Hughes and Gottschling , 2012 ) ( Figure 1A ) . Thus , throughout their lifespan mother cells produce daughter cells capable of regenerating vacuole acidity . 10 . 7554/eLife . 03504 . 003Figure 1 . Vacuole acidity regenerates in daughter cells throughout mother cell aging and reacidification occurs prior to cytokinesis . ( A and B ) Age ( # of cell divisions ) is shown in the second row and represents exact age determined by calcofluor staining of bud scars . Representative images are shown . n ≥30 cells per timepoint . Arrowheads indicate the daughter cell . DIC , differential interference contrast . ( A ) Vacuole acidity indicated by quinacrine staining of aged cells expressing Vph1-mCherry ( vacuole membrane marker ) . ( B ) Vacuole acidity of cells expressing Vph1-mCherry and arrested prior to cytokinesis by nocodazole treatment . ( C ) Cells with septin morphology indicated by Cdc10-mCherry were quinacrine stained and vacuole acidity was examined before or after cytokinesis ( one septin ring or two rings ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 00310 . 7554/eLife . 03504 . 004Figure 1—figure supplement 1 . Subunits of the V-ATPase are not asymmetric between mother cells and buds . ( A ) Cells in their first division expressing Vph1-GFP ( V0 domain ) or ( B ) Vma2-GFP ( V1 domain ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 004 To further characterize vacuole pH asymmetry , the timing of re-acidification of the bud vacuole was examined . Vacuole acidity was asymmetric in cells treated with nocodazole ( Figure 1B ) , suggesting that the bud vacuole re-acidifies prior to cytokinesis . We confirmed that re-acidification occurred before cytokinesis by examining cells containing the septin marker Cdc10-mCherry ( Figure 1C ) . A single septin ring at the bud neck transitions to two rings during cytokinesis ( Lippincott et al . , 2001 ) . We observed high vacuole acidity in buds when there was a single septin ring , further supporting that vacuole acidity regenerates prior to cytokinesis . Thus , throughout their lifespan , mother cells produce daughter cells that regenerate vacuole acidity prior to cytokinesis , when mother and daughter cells share a common cytosol . Thus , whatever causes the asymmetry of vacuole pH must also be asymmetric between mother and daughter cells prior to cytokinesis . One of the major points of regulation of vacuole acidity is assembly of the vacuolar proton ATPase ( V-ATPase ) , a multi-subunit complex that pumps protons from the cytosol into the vacuole . The V-ATPase consists of the integral membrane V0 complex and the V1 complex that associates with the V0 ( Li and Kane , 2009 ) . We examined whether there was a difference in vacuole-associated V1 or V0 between mother and daughter cells by visualizing green fluorescent protein ( GFP ) tagged subunits of each domain and found no evidence of asymmetry ( Figure 1—figure supplement 1 ) . Thus , no obvious difference in V-ATPase assembly can account for vacuole pH asymmetry between mother and bud . In a screen to identify proteins asymmetrically retained in mother cells throughout aging , we identified the plasma membrane proton ATPase , Pma1 ( Thayer et al . , 2014 ) . Pma1 is the major regulator of cytosolic pH ( Ferreira et al . , 2001; Serrano et al . , 1986 ) , and has similar activity to the V-ATPase , in that they both translocate cytosolic protons across membranes . Pma1 pumps protons from the cytosol out of the cell , whereas the V-ATPase pumps cytosolic protons into the vacuole ( Orij et al . , 2011 ) . Because Pma1 regulates cytosolic pH , we hypothesized that it could antagonize vacuole acidity during aging and underlie vacuole pH asymmetry . As a first step in testing our hypothesis , we analyzed Pma1 protein localization . There are conflicting reports on Pma1 asymmetry ( Smardon et al . , 2013; Khmelinskii et al . , 2012; Malínská et al . , 2003 ) , however we found that Pma1 was asymmetric between mother and daughter cells . Pma1 levels at the plasma membrane were higher in mother cells than daughter cells as indicated by indirect immunofluorescence with antibody to Pma1 ( Figure 2A ) . Similarly , Pma1 was more abundant in mother cells than buds when visualized with either Pma1-GFP ( Figure 2A ) or Pma1-mCherry fusion protein ( Figure 2B ) . We also detected mCherry and GFP fluorescence in the vacuole , which likely represents misfolded protein directed to the vacuole for degradation ( Chang and Fink , 1995 ) . Importantly , we found that asymmetry of Pma1 at the plasma membrane was maintained through at least 18 mother cell divisions and that Pma1 was asymmetric prior to cytokinesis ( Figure 2B ) , paralleling the asymmetry of vacuole pH . 10 . 7554/eLife . 03504 . 005Figure 2 . Plasma membrane Pma1 levels are asymmetric between mother cells and buds , accumulate with age , and are inversely correlated with vacuole acidity . ( A ) Top panel: Indirect immunofluorescence imaging of Pma1 with anti-Pma1 antibody in untagged young cell . Bottom panel: Pma1-GFP localization in young cell . ( B ) Newborn daughter cells and aged mother cells expressing Pma1-mCherry were quinacrine stained . Arrowheads indicate the vacuoles of interest . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 005 When we examined Pma1 distribution during mother cell aging , we found that Pma1 increased at the plasma membrane in early age ( Figure 2B ) . Pma1 levels were very low in newborn daughter cells , increased as daughter cells became mothers , and continued to increase over the first three mother cell divisions . This pattern of Pma1 abundance in daughters and aging mother cells inversely correlated with vacuole acidity . When Pma1 was very low in buds and newborn cells , the vacuole was acidic . In contrast , vacuole acidity was reduced in mother cells that have high Pma1 levels . The inverse correlation between Pma1 levels and vacuole acidity suggested that Pma1 could cause vacuole pH asymmetry by antagonizing V-ATPase activity in mother cells . We first tested whether high levels of Pma1 could reduce vacuole acidity by overexpressing an extra copy of PMA1 in newborn daughter cells from an inducible promoter ( Gao and Pinkham , 2000; Veatch et al . , 2009 ) . Overexpression of PMA1-mCherry increased Pma1 at the plasma membrane of newborn cells ( Figure 3—figure supplement 1 ) . Without excess Pma1 , 87% of newborn cells had highly acidic vacuoles , whereas vacuole acidity was only high in 13% of cells upon PMA1 overexpression ( Figure 3A ) . To further test whether Pma1 antagonized vacuole acidity , we reduced Pma1 activity and examined vacuole acidity in aging mother cells . PMA1 is an essential gene and cannot be deleted ( Serrano et al . , 1986 ) , so we reduced its activity by 65% using the pma1-105 allele that has a mutation in the catalytic domain ( McCusker et al . , 1987; Perlin et al . , 1989 ) . In contrast to wild-type cells where vacuole acidity was reduced in more than 80% of cells in the third and subsequent mother cell divisions ( Figure 3B ) , pma1-105 cells retained high vacuole acidity after 3 divisions and up to at least 18 divisions ( 84% and 79% respectively , Figure 3B ) . These results suggest that Pma1 activity antagonizes vacuole acidification and , combined with the expression pattern of Pma1 , support the idea that increased Pma1 in aged mother cells causes the reduction of vacuole acidity . 10 . 7554/eLife . 03504 . 006Figure 3 . Pma1 antagonizes vacuole acidity and its absence facilitates regeneration of vacuole acidity in buds . ( A ) PMA1 was overexpressed in newborn daughter cells expressing Vph1-mCherry using a β-estradiol inducible system where a GAL4-Estrogen binding domain-VP16 ( GEV ) fusion protein drives GAL1 promoter expression of an extra copy of PMA1 . ( n ≥ 30 cells per condition ) . ( B ) Wild-type and pma1-105 cells expressing Vph1-mCherry were aged and quinacrine stained ( n ≥ 30 cells per timepoint ) . White arrowheads indicate mother cell vacuoles with reduced acidity . Orange arrowheads indicate acidic mother-cell vacuoles . ( C ) Replicative lifespan of wild-type , pma1-105 , vma2 , and vma2 pma1-105 cells by micromanipulation . Median lifespan is indicated . For the difference between wild-type and pma1-105 , p < 0 . 0001 , one-tailed logrank test . ( n = 114 cells for PMA1 , n = 119 for pma1-105 , n = 36 for vma2 , and n = 39 for vma2 pma1-105 ) . ( D ) PMA1-mCherry was overexpressed in cells undergoing their first division that expressed endogenous Pma1-mCherry and that were treated with β-estradiol and then with β-estradiol plus nocodazole ( Noc ) . ( E ) As in D , cells that expressed Vph1-mCherry were induced to overexpress PMA1 and were quinacrine stained . ( n ≥ 30 cells per condition ) . Arrowheads indicate the vacuoles of interest . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 00610 . 7554/eLife . 03504 . 007Figure 3—figure supplement 1 . Overexpression increases Pma1 levels at the plasma membrane . PMA1-mCherry was overexpressed in newborn daughter cells using a β-estradiol inducible system where a GAL4-Estrogen binding domain-VP16 ( GEV ) fusion protein drives GAL1 promoter expression of an extra copy of PMA1-mCherry in cells that also expressed endogenous Pma1-mCherry . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 007 We previously found that delaying the reduction of vacuole acidity during aging by increasing V-ATPase levels extends replicative lifespan ( Hughes and Gottschling , 2012 ) . Given the evidence presented above that Pma1 levels antagonize vacuolar acidity , we asked whether reduced Pma1 activity also affected lifespan . Indeed , the pma1-105 allele increased median replicative lifespan by ∼30% ( Figure 3C ) , comparable to well-characterized lifespan-extending mutations ( Delaney et al . , 2011 ) . The slope of the pma1-105 lifespan curve is similar to the slope of the wild-type curve . This suggests that instead of influencing the rate of aging throughout lifespan , the pma1-105 allele delays the onset of the normal aging process . To ascertain whether lifespan extension by the pma1-105 allele occurred entirely via increased vacuolar acidity , we examined the lifespan of pma1-105 cells that lacked V-ATPase function . Cells lacking the V-ATPase subunit Vma2 had a short median lifespan of 2 divisions , as previously reported ( Hughes and Gottschling , 2012 ) . The lifespan of cells that had reduced Pma1 activity and that were devoid of V-ATPase function ( vma2Δ , pma1-105 ) was much shorter than wild-type lifespan ( median 7 divisions ) , and more similar to cells lacking V-ATPase function . This suggests that most of the lifespan extension imparted by the pma1-105 allele requires V-ATPase function , but that the mechanism of lifespan extension is not limited to increased vacuolar acidification . Taken together these results support the idea that high Pma1 levels on mother cells impair vacuole acidification and limit lifespan . In addition to Pma1 antagonizing mother cell vacuole acidity with age , we also hypothesized that the inherent asymmetry of Pma1 , and thus low levels on buds , allows for re-acidification of the vacuole in buds . To test this idea , we asked whether expressing Pma1 in buds reduced vacuole acidity . We induced overexpression of PMA1-mCherry in cells arrested prior to cytokinesis with nocodazole and in untreated cells ( Figure 3D ) . In untreated cells , overexpression increased mother cell Pma1 levels but maintained mother-bud asymmetry . However , in nocodazole-arrested cells , PMA1-mCherry became equivalently high in mother cells and buds . At least 80% of buds had acidic vacuoles without PMA1 induction or when PMA1 was induced in the absence of nocodazole ( Figure 3E ) . In contrast , only 13% of buds had acidic vacuoles when Pma1 levels were high in buds . Because increased Pma1 levels in buds impaired re-acidification of the vacuole , we conclude that the inherent asymmetry of Pma1 is required for regeneration of vacuole acidity prior to cytokinesis . We speculate that regeneration of vacuole acidity is required for daughter cell rejuvenation and that if high levels of Pma1 were induced in the buds of aging mother cells , daughter cells would not rejuvenate . We wondered how Pma1 antagonizes vacuole acidity and how low Pma1 levels in buds permit vacuole reacidification . Given that Pma1 pumps protons out of the cell , we hypothesized that increased Pma1 activity antagonizes vacuole acidity by reducing cytosolic protons available to the V-ATPase . A prediction of this hypothesis is that cytosolic pH may become more basic with age , and may differ between mother and daughter cells . To test this hypothesis , we examined cytosolic pH with ratiometric pHluorin ( a pH-sensitive GFP ) ( Miesenbock et al . , 1998 ) fused to a plasma membrane targeting sequence ( residues 1–28 of the Psr1 protein ) ( Siniossoglou et al . , 2000 ) . With this reagent , cytosolic pH at the cell cortex was visualized ( Figure 4A ) and quantified in mother cells of varying ages and in their buds . Newborn daughter cells or mother cells that had undergone 1 or 2 divisions had a mean cytosolic pH of ∼7 . 1 , similar to previous measurements of bulk log phase cultures ( Orij et al . , 2009 ) ( Figure 4B ) . However , as mother cells aged , cortical cytosolic pH increased as much as ∼0 . 5 pH units ( Figure 4B ) . Moreover , when we examined mother cells ( on average 3 or 18 divisions old ) and their attached buds , cortical pH was ∼0 . 2 or ∼0 . 1 pH units lower in buds than mother cells ( Figure 4C ) . All together , these results indicate that cortical cytosolic pH increases during replicative aging and is asymmetric between mother and daughter cells . Mother-daughter asymmetry of cytosolic pH might be surprising given the rapid diffusion of protons ( Wraight , 2006 ) . However , local cytosolic pH differences have been observed in tumor cell invadopodia ( Magalhaes et al . , 2011 ) and cytosolic pH gradients can form during polarized growth ( Feijo et al . , 1999; Gibbon and Kropf , 1994 ) . 10 . 7554/eLife . 03504 . 008Figure 4 . Cortex-proximal cytosolic pH increases with age and is asymmetric between mother cells and buds . ( A ) Visualization of cortical pH of newborn and aged mother cells and their buds as indicated with a plasma membrane-anchored ratiometric pHluorin using its bimodal excitation spectrum . Age is indicated by wheatgerm agglutinin-Alexa 594 staining of budscars , which also detects birth scars on newborn cells . ( B ) As in A , measurement of cortical pH of newborn and aged mother cells was made at the plasma membrane . n ≥13 cells per timepoint . Mean cortical pH is significantly increased in mother cells undergoing their third division and thereafter compared to newborn cells ( p ≤ 0 . 014 , one-tailed unpaired t test ) . Error bars represent SEM . ( C ) Difference of the cortical pH of mother cells and their buds . Bud pH was lower ( more acidic ) than mother cell pH ( p = 0 . 003 , n = 17 cells at 3 divisions and p = 0 . 04 , n = 16 cells at 18 divisions , one-tailed paired t tests ) and was subtracted from mother cell pH . ( D ) Model of the effect of Pma1 asymmetry and increased Pma1 levels during aging on the magnitude of proton translocation out of the cytosol and into the vacuole . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 00810 . 7554/eLife . 03504 . 009Figure 4—figure supplement 1 . Pma1 asymmetry mediates cortical cytosolic pH asymmetry . ( A ) PMA1 was overexpressed as in Figure 3 and cortical pH measured in mother cells undergoing their first division and their buds as in Figure 4 . Cells were treated with β-estradiol for 2 hr to induce PMA1 and with Nocodazole ( Noc ) and β-estradiol for an additional 1 . 5 hr . For all mother-bud pairs , bud pH was significantly more acidic than mother pH ( p < 0 . 05 , one-tailed paired t test , n ≥ 14 cell per condition ) , except when Pma1 levels in buds were high due to the combination of PMA1 overexpression and nocodazole treatment . The difference between mother and bud pH was significantly greater when Pma1 was asymmetric ( PMA1 induced without nocodazole ) than when Pma1 was symmetric ( PMA1 induced plus nocodazole ) ( p < 0 . 0001 , one-tailed unpaired t test ) . Mean cortical pH is indicated , error bars represent SEM . ( B ) Cortical pH of wild-type and pma1-105 mother cells in their first cell division ( p < 0 . 0005 , t test , n = 12 cells for PMA1 and n = 13 for pma1-105 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03504 . 009 To test whether Pma1 asymmetry was required for cortical cytosolic pH asymmetry , we modulated Pma1 levels and activity in mother cells and buds and monitored cytosolic pH . We overexpressed PMA1 , which increased mother cell levels but maintained mother-bud asymmetry ( as in Figure 3D ) , and we overexpressed PMA1 in nocodazole treated cells to generate equivalent high levels of Pma1 in mother cells and buds . Overexpression of PMA1 in mother cells alone elevated mother cell cytosolic pH by ∼0 . 7 pH units and increased the mother-bud difference in cytosolic pH by ∼0 . 5 pH units ( Figure 4—figure supplement 1A ) . However , when Pma1 levels in buds were elevated to mother cell levels , bud pH increased , abrogating cytosolic pH asymmetry . Moreover , decreasing Pma1 activity with the pma1-105 allele decreased cortical cytosolic pH by ∼0 . 7 pH units ( Figure 4—figure supplement 1B ) . Taken together our results suggest that the inherent asymmetry of Pma1 creates mother-daughter cytosolic pH asymmetry prior to cytokinesis . Our findings support the idea that Pma1 activity antagonizes vacuole acidification via a competition with the V-ATPase for limited cytosolic protons ( at pH 7 there are ∼3000 free protons per yeast cell , ∼106 Pma1 molecules and ∼105 V-ATPase V0 subunits ) ( Orij et al . , 2011; Huh et al . , 2003 ) . While other modes of regulation may also be involved , our results can be explained by high Pma1 activity in aged mother cells translocating a sufficient number of protons out of the cytosol to restrict proton availability for the V-ATPase and reduce vacuole acidity ( Figure 4D ) . Conversely , lower Pma1 activity in buds leads to more cytosolic protons and higher vacuole acidification . Our findings identify increased cytosolic pH as an early contributing step to the aging process in budding yeast and suggest that cytosolic pH asymmetry facilitates daughter cell rejuvenation . Interestingly , this same discontinuity of cytosolic pH in plant and algae cells ( Feijo et al . , 1999; Gibbon and Kropf , 1994 ) and of Pma1 orthologs in fission yeast and pollen tubes ( Minc and Chang , 2010; Certal et al . , 2008 ) is conserved during polarized cell growth . We speculate that non-uniform cytosolic pH is generally important in polarized growth and has been co-opted in cell types that undergo asymmetric divisions to regenerate full cellular capacity .
Yeast strains are listed in Supplementary file 1 . All strains are derivatives of Saccharomyces cerevisiae S288c ( BY ) ( Brachmann et al . , 1998 ) . One-step PCR-mediated gene replacement and epitope tagging were performed using standard techniques , with template plasmids pRS306 , pRS400 , pKT127 and pKTmCherryKanMX ( Sheff and Thorn , 2004; Sikorski and Hieter , 1989 ) . Oligonucleotides for gene replacement , tagging , and cloning are listed in Supplementary file 1 . Strains expressing Vma2–GFP were derived from the yeast GFP collection ( Huh et al . , 2003 ) . The Gal4-Estrogen binding domain-VP16 ( GEV ) fusion protein ( Veatch et al . , 2009 ) was integrated into the leu2Δ0 allele by transforming PmeI-linearized pAGL plasmid . Strains expressing PMA1 or PMA1-mCherry from a GAL promoter were constructed by transformation of GEV yeast strains with NotI-digested pAG306-GAL-PMA1 chr1 or pAG306-GAL-PMA1-mCherry chr1 , which integrated them into an empty region of chromosome 1 ( 199456–199457 ) ( Hughes and Gottschling , 2012 ) . The pma1-105 mutant was derived from the heterozygous yeast deletion collection strain , pma1Δ::KanMX4/PMA1 ( Winzeler et al . , 1999 ) . The strain was transformed with a linear fragment ( derived from NotI and SacII digestion of pma1-105-URA3 plasmids ) containing pma1-105 marked with URA3 . We chose transformants that replaced the pma1Δ::KanMX4 allele similar to a previously described strategy ( Harris et al . , 1994 ) . This heterozygote was sporulated to obtain pma1-105 haploids , which were sequenced to verify the mutation . The PMA1 URA3 variant was created from the pma1Δ::KanMX4/PMA1 strain; the ura3Δ0 mutation was converted to URA3 via PCR amplification of URA3 from pRS306 and transformation . This resulting diploid strain was sporulated to obtain a PMA1 URA3 haploid . All PMA1 or pma1-105 strains that expressed other markers were created by backcrossing to these original haploids . Strains carrying the cortical pHluorin were constructed by amplifying plasmid pADH1pr-PSR1-RMP with primers ( URA3-tTA-intChr1F and R ) that allowed insertion into a second empty region of chromosome 1 ( 17068–17161 ) . pKTmCherryKanMX ( a kind gift from W Shou ) was obtained by digestion of pKT127 ( Sheff and Thorn , 2004 ) with PacI and BglII and insertion of the mCherry containing fragment from similarly digested pBS34 . pBS34 was obtained from the Yeast Resource Center at the University of Washington with permission from R . Tsien ( Shaner et al . , 2004 ) . The GEV plasmid was previously described ( Veatch et al . , 2009 ) . pAG306-GAL-PMA1chr1 , was generated in two steps . First , we created pAG306-GAL-ccdBchr1 , a plasmid for gene expression from the GAL promoter that can be integrated into chromosome 1 ( 199456–199457 ) after NotI digestion . We generated pAG306-GAL-ccdBchr1 by ligation of a SmaI-digested fusion PCR product that contained two ∼500-base-pair regions of chromosome 1 flanking a NotI site into AatII-digested pAG306-GAL-ccdB ( Addgene plasmid 14139 ) ( Alberti et al . , 2007 ) . We generated the fusion PCR product using oligonucleotides ChrI PartB SmaI F and ChrI PartA SmaI R to amplify two templates generated by PCR of yeast genomic DNA using oligonucleotide pairs ChrI PartA NotI F and ChrI PartA SmaI R , and ChrI PartB SmaI F and ChrI PartB NotI R , respectively . Second , we inserted PMA1 into pAG306-GAL-ccdBchr1 from donor Gateway plasmid pDONR221-PMA1 ( Harvard Institute of Proteomics [HIP] accession ScCD00008895 ) ( Hu et al . , 2007 ) , using LR Clonase according to the manufacturer's instructions ( Invitrogen , Carlsbad , CA ) . pAG306-GAL-PMA1-mCherryChr1 was generated by Gibson Assembly according to the manufacturer's instructions ( New England Biolabs , Ipswich , MA ) . First PMA1-mCherry was amplified from genomic DNA from strain UCC9645 using primers GibRXNpAGdest_Pma1ChryF and GibRXNpAGdest_Pma1ChryR and assembled with a PCR product amplified from pAG306-GAL-ccdBChr1 with primers GibsonRXNpAGdstnF-2 and R-2 . pADH1pr-PSR1-RMP was derived from pADH1pr-RMP using Quikchange Site-Directed Mutagenesis ( Stratagene , La Jolla , CA ) to insert the first 28 amino acids of PSR1 between the ADH1 promoter and the N-terminus of RMP ( ratiometric pHluorin ) using primers PSR1-28-RMpHluorinF and R . pADH1pr-RMP was generated in four steps . First , pKT127-RMP was created by removing GFP from pKT127 by restriction digestion with PacI-AscI and replacing it with similarly digested RMP generated by PCR of template plasmid pGM1 ( Miesenbock et al . , 1998 ) using oligonucleotides SEP PacIF and SEP AscIR . pADH1pr-RMP was created when RMP and the ADH1 terminator were amplified with primers UPGFP/pHluorin F and R from pKT127-RMP , digested with EcoRI-EagI , and ligated into similarly digested backbone of COX4-dsREDURA3int . COX4-dsREDURA3int was created by ligating the XhoI-NotI fragment of pHS12 ( Bevis and Glick , 2002 ) containing the ADH1 promoter and COX4 mitochondrial presequence fused to dsRED . T4 into similarly digested pRG919 . pRG919 was created when a SacI-SacII PCR fragment containing a URA3 targeting construct was inserted between the SacI-SacII sites in pRS406 ( Christianson et al . , 1992 ) . The PMA1-URA plasmid was created in two steps . First , the URA3 gene was inserted downstream of the previously characterized PMA1 transcriptional termination sites ( Nagalakshmi et al . , 2008; Yassour et al . , 2009 ) on chrVII ( 479252–479253 ) with primers MRKdownPma1F and MRKdownPma1R to create UCC9656 . Genomic DNA from this strain was amplified with primers Pma1_1kbupNotIF and Pma1_1kbdownSacIIR to acquire a fragment containing the entire PMA1 locus plus 1 kb upstream and 1 kb downstream and URA3 . This fragment was digested with NotI and SacII and ligated into similarly digested pBluescript SK+ ( Stratagene ) . The pma1-105-URA plasmid was generated by Quikchange Site-Directed Mutagenesis ( Stratagene ) of the PMA1-URA plasmid using primers pma1-S368FF and pma1-S368FR . Cells were cultured in YEPD ( 1% yeast extract , 2% peptone , 2% glucose ) and maintained in exponential growth for 15 hr to a maximum density of 5 × 106 cells ml−1 before initiating experiments . Where indicated , cells were treated with nocodazole ( Sigma , St . Louis , MO ) at 10 μg ml−1 for 1 . 5 hr or with 5 μM β-estradiol ( Sigma ) for 2 hr to induce PMA1 or PMA1-mCherry overexpression . Cells were cultured , biotin labeled , aged , and purified for quinacrine staining as previously described ( Hughes and Gottschling , 2012 ) . For cortical pH analysis of aged cells , cell labeling and purification were performed as described ( Hughes and Gottschling , 2012 ) except incubation with streptavidin-coated magnetic beads ( MicroMACS , Miltenyi Biotec , Bergisch Gladbach , Germany ) and purification took place in YEPD depleted of biotin . This was achieved by overnight incubation at 4°C of 45 ml YEP with 300 μl Avidin-Agarose beads ( Sigma ) . Glucose was added to 2% . Cells were recovered for 1 hr in YEPD prior to imaging . Pma1 was detected by indirect immunofluorescence as described ( Burke et al . , 2000 ) using the 40B7 monoclonal antibody ( Abcam , Cambridge , England ) followed by Alexa Fluor 488-conjugated goat anti-mouse secondary ( Invitrogen ) . Quinacrine ( Sigma ) staining was performed as previously described ( Hughes and Gottschling , 2012 ) . In most experiments , age was determined by calcofluor ( Sigma ) staining of bud scars by including 5 μg ml−1 calcofluor in the last wash step before imaging . Calcoflour staining reveals bud scars ( Pringle , 1991 ) and facilitates identification of newborn cells and the replicative age of mother cells . Calcofluor also stains the mother cell birth scar ( Pringle , 1991 ) and allowed identification of the new mother cell during nocodazole arrest when mother cells and buds are similar in size . Cells were imaged under ×60 oil magnification using a Nikon Eclipse E800 ( Nikon , Tokyo , Japan ) with the appropriate filter set: UV-2E/C DAPI for calcofluor; FITC-HYQ for quinacrine , GFP and Alexa Fluor 488; and G-2E/C TRITC for mCherry . Images were acquired with a CoolSNAP HQ2 CCD camera ( Photometrics , Tucson , AZ ) and Metamorph version 7 . 1 . 1 . 0 imaging software ( Molecular Devices , Sunnyvale , CA ) . Replicative lifespan was measured by micromanipulation as previously described ( Hughes and Gottschling , 2012 ) . Cells were cultured in YEPD , but transferred to low fluorescence medium ( Orij et al . , 2009 ) for pH measurement after rinsing them in an equal volume of low fluorescence medium . Calibration curves were created as previously described ( Orij et al . , 2009 ) except that that cells were permeabilized prior to pH equilibration by treatment in 5 μg ml−1 digitonin ( Sigma ) in 1× PBS for 5 min . To quantify replicative age , 1 × 107 cells were stained in YEPD for 5 min with 10 μg ml−1 Wheat Germ Agglutinin-Alexa fluor 594 conjugate ( Molecular Probes , Eugene , OR ) , washed once with an equal volume of YEPD , once with low fluorescence medium , and transferred to low fluorescence medium for 20 min prior to imaging at a density of 1 × 107 cells ml−1 . To quantify cortical pH of live single cells and to generate pH calibration curves , cells were imaged with a Leica DMI6000 B under ×63 oil magnification . Images were acquired with a Leica DFC365 FX camera and Leica Application Suite Advanced Fluorescence software ( Leica , Wetzlar , Germany ) . TRITC excitation and emission filters ( Ex525/25 , Em605/52 ) were used to image bud scars and the combination of FITC Excitation/FITC Emission ( Ex490/20 , Em525/36 ) and DAPI Excitation/FITC Emission ( Ex402/15 , Em525/36 ) filters were used to derive cortical pH using the bimodal excitation spectrum of ratiometric pHluorin ( Miesenbock et al . , 1998 ) to calculate 402/490-nm excitation ratios . Image analysis was performed using ImageJ ( Version 1 . 47m , NIH ) to quantify mean pHluorin intensity from the identical regions of images acquired at 402 and 490 nm excitation with a 2 pixel-wide freehand line tool traced along the majority of the length of the mother cell or attached bud plasma membrane . Local background was calculated from a cell-free region one cell diameter away and subtracted from all membrane intensity measurements prior to calculation of 402/490-nm excitation ratios that were fitted to calibration curves to derive cortical pH . Statistical analyses were performed using GraphPad Prism version 4 . 0a software ( GraphPad , La Jolla , CA ) . We and others ( Pineda Rodo et al . , 2012 ) note that repeated imaging differentially affected signal intensity captured at 402 and 490 nm excitation wavelengths , which altered the excitation ratios of sequential images in a pH-independent manner . Therefore we captured a single set of images per cell and never exposed cells to excitation wavelengths prior to pHluorin imaging . | Aging is a part of life—but its biological basis and , in particular , how aged cells give rise to young offspring ( or progeny ) has not been clearly established for any organism . Budding yeast is a microorganism that is a valuable model to understand aging in more complex organisms like humans . Budding yeast cells undergo a process called ‘replicative aging’ , meaning that each yeast mother cell produces a set number of daughter cells in her lifetime . However , when daughter cells arise from an aging mother cell , the daughter's age is ‘reset to zero’ . How mother cells age , and how their daughters are rejuvenated , are questions that have been studied for decades . Previously , researchers discovered that a mother cell's vacuole ( an acidic compartment that stores important molecules that can become toxic ) becomes less acidic as the mother cell ages . Daughter cells , on the other hand , have very acidic vacuoles , which was linked to their renewed lifespans . However , the mechanism behind this difference in the acidity of the vacuole between mother and daughter cells was unknown . Now , Henderson et al . have found that a protein ( called Pma1 ) , which is found at the cell surface and pumps protons out of the cell , is present in mother cells but not in their newly-formed daughter cells . Furthermore , the Pma1 protein also accumulates as mother cells age . By pumping protons out of the cell , Pma1 effectively reduces the number of protons available to acidify the vacuole in the mother cell . However , because at first the daughter does not have Pma1 , there are still plenty of protons inside the cell to acidify the vacuole . When Henderson et al . reduced the activity of Pma1 in mother cells , the entire cell became more acidic , and so did their vacuoles . Conversely daughter cells engineered to have more Pma1 were less acidic and had less acidic vacuoles than normal . Henderson et al . next asked whether reducing Pma1 activity to create a more acidic cell , could extend the lifespan of cells , and found that indeed cells with less Pma1 activity lived longer . As such , these findings indicate that an asymmetry in the acidity of the cell—caused by unequal levels of the Pma1 protein—contributes to reducing the lifespan of the mother cell and to rejuvenating the daughter cell . Thus Henderson et al . have identified one of the earliest events in the cellular aging process in budding yeast . Their findings suggest that an imbalance in an activity that is normally essential for cell survival ( in this case , the activity of Pma1 ) can have long-term consequences for the cell that lead to aging . | [
"Abstract",
"Main",
"text",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"short",
"report",
"cell",
"biology"
] | 2014 | Mother-daughter asymmetry of pH underlies aging and rejuvenation in yeast |
The directed evolution of biomolecules to improve or change their activity is central to many engineering and synthetic biology efforts . However , selecting improved variants from gene libraries in living cells requires plasmid expression systems that suffer from variable copy number effects , or the use of complex marker-dependent chromosomal integration strategies . We developed quantitative gene assembly and DNA library insertion into the Saccharomyces cerevisiae genome by optimizing an efficient single-step and marker-free genome editing system using CRISPR-Cas9 . With this Multiplex CRISPR ( CRISPRm ) system , we selected an improved cellobiose utilization pathway in diploid yeast in a single round of mutagenesis and selection , which increased cellobiose fermentation rates by over 10-fold . Mutations recovered in the best cellodextrin transporters reveal synergy between substrate binding and transporter dynamics , and demonstrate the power of CRISPRm to accelerate selection experiments and discoveries of the molecular determinants that enhance biomolecule function .
Directed evolution using living systems allows selections for improved biomolecule function to be directly coupled to phenotype . However , present directed evolution systems require the use of extrachromosomal gene libraries encoded in bacteriophage or plasmids , which suffer from high levels of copy number variation ( Yokobayashi et al . , 2002; Turner , 2009; Esvelt et al . , 2011 ) . For example , in the yeast Saccharomyces cerevisiae , common yeast centromere containing ( CEN ) plasmids vary widely in copy number per cell resulting in highly variable expression levels from cell to cell ( Figure 1 ) . Ideally , the selective pressure applied to evolve improved or new biomolecule function should be limited to the DNA sequence level and not to gene copy number in a given cell ( Zhou et al . , 2012 ) . To overcome copy number variation in yeast , the standard genetic tool of integrating linear DNA into the genome by homologous recombination is too inefficient to insert DNA libraries without multiple steps that rely on dominant selectable markers for chromosomal integrations ( Wingler and Cornish , 2011 ) or specialized strain backgrounds ( DiCarlo et al . , 2013a ) . 10 . 7554/eLife . 03703 . 003Figure 1 . Comparing noise in plasmid and genomic expression of two fluorescent proteins . Proteins mRuby2 ( false-colored red ) and Venus ( false-colored green ) were expressed in the same cell with a strong , PTDH3 promoter . ( A ) Fluorescent proteins expressed on two separate CEN plasmids; ( B ) in tandem on a single CEN plasmid; ( C ) integrated into two different loci ( LEU2 and URA3 ) in the S . cerevisiae genome; and ( D ) integrated in tandem at a single locus ( URA3 ) in the genome . Each image was auto-exposed for both red and green channels , with yellow showing co-expression of mRuby2 and Venus . The variability in the relative expression of the two fluorescent proteins is reduced by moving from two plasmids to one , and the variability in total expression is reduced by moving to the genome . The difference between integrating into one or two loci in the genome is minimal . DOI: http://dx . doi . org/10 . 7554/eLife . 03703 . 003 Bacterial type II CRISPR-Cas9 genome editing has been used successfully in several eukaryotic organisms ( Cong et al . , 2013; DiCarlo et al . , 2013b; Mali et al . , 2013a , 2013b; Shalem et al . , 2014 ) but has not been adapted for selection experiments . Genome editing CRISPR-Cas systems require a Cas9 endonuclease that is targeted to specific DNA sequences by a non-coding single guide RNA ( sgRNA ) ( Jinek , et al . , 2012; Jinek et al . , 2014 ) . The Cas9-sgRNA ribonucleoprotein complex precisely generates double-strand breaks ( DSBs ) in eukaryotic genomes at sites specified by a 20-nucleotide guide sequence at the 5′ end of the sgRNA that base pairs with the protospacer DNA sequence preceding a genomic protospacer adjacent motif ( PAM ) ( Sternberg et al . , 2014 ) . The presence of the Cas9-produced DSB in genomic DNA can increase the rate of homology-directed repair ( HDR ) with linear DNA at the DSB locus by several 1000-fold ( Storici et al . , 2003; DiCarlo et al . , 2013b ) potentially enabling high-throughput experimental methods . CRISPR-Cas9 could therefore be useful in probing industrial S . cerevisiae strains , which are more robust compared to laboratory strains ( Kerr and Service , 2005; Farrell et al . , 2006; Rubin , 2008 ) , but for which few genetic tools are available , due to the fact that these strains are often polyploid with low-efficiency mating and sporulation .
To optimize the efficiency of HDR in S . cerevisiae mediated by Cas9 , we first constructed a plasmid expressing Streptococcus pyogenes Cas9 fused to a nuclear localization signal ( Cas9-NLS ) that efficiently localizes to the nucleus of yeast cells ( Figure 2A , B; Jinek et al . , 2012 ) . We used an intermediate strength promoter , PRNR2 , to express Cas9 for genome targeting experiments , because Cas9 ( Cas9-NLS-His8x ) expression from the PRNR2 promoter resulted in yeast strains with near wild type fitness whereas Cas9 expressed using strong yeast promoters such as PTDH3 ( Lee et al . , 2013 ) reduced yeast fitness relative to wild type cells ( Figure 2C ) . Second , we designed a new sgRNA architecture in order to improve its expression and function . Cellular levels of sgRNAs correlate with the efficiency of Cas9-mediated genome targeting in mammalian cells ( Hsu et al . , 2013 ) . Controlled expression of nuclear-localized RNAs and the correlation between sgRNA levels and Cas9-editing in yeast has not been explored . To better control cellular levels of correctly folded sgRNA , we developed a modular design by fusing the sgRNA ( Mali et al . , 2013a ) to the 3′ end of the self-cleaving hepatitis delta virus ( HDV ) ribozyme ( Ke et al . , 2007; Webb et al . , 2009; Figure 2D ) . We reasoned that the structured ribozyme would protect the 5′ end of the sgRNA from 5′ exonucleases ( Houseley and Tollervey , 2009 ) . Additionally , the HDV ribozyme would cleave the RNA immediately 5′ of the ribozyme , removing extraneous RNA sequences and allowing flexibility in which promoters can be used , such as tRNAs whose DNA sequences also serve as RNA polymerase III promoters ( Marck et al . , 2006 ) . Ribozyme removal of the tRNA would then separate transcription initiation from within the tRNA sequence ( Marck et al . , 2006 ) and tRNA nuclear export ( Köhler and Hurt , 2007 ) from the process of forming functional sgRNA complexes with Cas9 . 10 . 7554/eLife . 03703 . 004Figure 2 . The engineered CRISPRm platform . ( A ) Cas9 construct used for nuclear localization experiments . The S . pyogenes Cas9 protein was tagged with a C-terminal nuclear localization motif and a green fluorescent protein ( GFP ) ( green ) . For genome editing experiments , the S . pyogenes Cas9 protein was tagged with a C-terminal nuclear localization motif and a Histidine affinity tag ( gray ) . ( B ) Cellular localization of Cas9-GFP in exponentially growing S . cerevisiae cells . The Cas9-GFP protein was expressed from the TDH3 promoter in this experiment . Left , bright field image; right fluorescence microscopy . ( C ) Growth profiles of yeast expressing Cas9 from a strong promoter PTDH3 ( black ) or a weaker promoter PRNR2 ( blue ) relative to wild type ( red ) . ( D ) The mature sgRNA contains a 5′ hepatitis delta virus ( HDV ) ribozyme ( δR , brown ) , 20mer target sequence ( green ) , sgRNA ( black ) and RNA polymerase III terminator ( red ) . The RNA polymerase III promoter tRNA ( blue ) is catalytically removed by the HDV ribozyme ( red arrow ) . ( E ) Schematic of yeast Cas9 targeting . Cas9 and sgRNA are expressed from a single plasmid , form a complex , and cleave targeted genomic DNA , which is repaired using a barcoded oligonucleotide . ( F ) The linear barcoded repair DNA molecule . Each repair DNA contains a 5′ TAA stop codon ( gold ) , a forward primer sequence ( green ) , a unique molecular barcode ( gray ) , and a 3′ reverse primer ( green ) ( Giaever et al . , 2002 ) . Barcodes are amplified using a forward primer that contains 50 bp of homology ( blue ) to the genome targeting site and a reverse primer that contains 50 bp of homology to the genome targeting site . The 50 bp of genomic targeting sequence are each 10 bp proximal to the PAM motif , resulting in a 20 nt deletion and barcode oligonucleotide integration . DOI: http://dx . doi . org/10 . 7554/eLife . 03703 . 004 To quantify the efficiency and specificity of HDR with the new Cas9/sgRNA format , phenotype screens were performed by the co-transformation of a single plasmid that expresses the Cas9 protein and one or more HDV ribozyme-sgRNAs , containing a 20-nucleotide guide sequence that matches a specific site in the yeast genome , along with a linear double-stranded DNA ( Figure 2E ) . The linear dsDNA , which contains a unique 20-mer barcode ( Giaever et al . , 2002 ) flanked by common primer sequences and 50 base pairs of DNA homologous to the regions proximal to the PAM motif ( Figure 2F ) , acts as a template for DNA repair by HDR , resulting in a unique markerless and barcoded insertion allele . With a single sgRNA , we found tRNA sequences used as RNA polymerase III promoters resulted in nearly 100% efficient barcode insertion in diploid yeast at the URA3 locus , resulting in 5-fluororotic acid resistance ( Boeke et al . , 1984 ) whereas the non-tRNA promoters were mostly ineffective ( Figure 3A; Supplementary file 1A ) . We also assessed target sequence bias by targeting 11 unlinked yeast genes in diploid S288C yeast cells using tRNATyr as the RNA polymerase III promoter ( Supplementary file 2A ) . We found that targeting was highly efficient for every genomic target except LEU2 , which at first was weak but was restored to 100% by selecting a different guide sequence ( LEU2-2 ) as the sgRNA ( Figure 3B ) . We also verified that our CRISPR-Cas9/sgRNA system did not result in non-specific genome targeting by sequencing 9 of the targeted strains ( Supplementary file 1B ) . We also tested efficacy of our approach in the polyploid industrial S . cerevisiae strain ATCC4124 that has superior tolerance and productivity phenotypes ( Ness et al . , 1993 ) . The non-tRNA promoter previously used in haploid yeast ( DiCarlo et al . , 2013b ) , PSNR52 , functioned in diploid S288C yeast but failed to result in targeting in the polyploid industrial yeast ATCC4124; in contrast , the tRNAPro promoter enabled highly-efficient barcode insertion into all copies of the URA3 locus in ATCC4124 , which we term cis-multiplexing , that is introducing a double-stranded DNA break at a single genomic locus across all chromosomes ( Figure 3C; Supplementary file 1A ) . Notably , tRNATyr was inefficient at targeting in ATCC4124 though it was effective in S288C diploid ( Supplementary file 1A ) , indicating that the choice of tRNA used for the expression of the HDV-sgRNA impacts multiplexing in a strain-dependent manner . 10 . 7554/eLife . 03703 . 005Figure 3 . CRISPRm barcode insertion in yeast cells . ( A ) Targeting efficiency at the URA3 locus in diploid S288C yeast using RNA polymerase III promoters ( x axis ) to drive the expression of the sgRNA . ( B ) Targeting efficiency measured across 11 loci in 10 genes in diploid cells of yeast strain S288C , using the tRNATyr promoter . ( C ) Single fragment barcode integration ( blue ) and three-fragment NatR cassette integration ( yellow ) efficiency in S288C diploid and ATCC4124 polyploid strains . For each experiment the promoter and strain are indicated as promoter-strain ( e . g . , SNR52-S288C ) . ( D ) Efficiency of multiplex insertion of barcoded DNA in diploid yeast cells with the 5′ HDV ribozyme ( black ) and without the 5′ HDV ribozyme ( red ) . Triplex targeting without the 5′ HDV ribozyme was not tested . The tRNATyr promoter was used in these experiments . ( E ) The addition of a 5′ HDV ribozyme increases the intracellular levels of sgRNA by sixfold . The RT-qPCR experiments were carried out in biological triplicate , with the mean and standard deviation shown . ( F ) Efficiency of targeting one ( URA3 ) , two ( URA3 , LYP1 ) and three ( URA3 , LYP1 and COX10 ) in haploid S288C yeast cells . The tRNATyr promoter was used for sgRNA expression . DOI: http://dx . doi . org/10 . 7554/eLife . 03703 . 005 Since the efficiency of cis-multiplexing with our CRISPR-Cas9/sgRNA system is nearly 100% using one sgRNA , we next tested its ability for trans-multiplexed CRISPR-Cas9 genome editing , that is introducing double-stranded DNA breaks at multiple genomic loci simultaneously , using tRNAs to drive the expression of the HDV-sgRNAs . We simultaneously targeted two and three unlinked genomic loci in diploid cells for loss-of-function mutagenesis by introducing a single plasmid expressing Cas9 and two or three sgRNAs with two or three gene-specific barcoded DNA molecules , respectively . The efficiency of duplex and triplex targeting ( requiring four and six chromosome cuts , respectively ) with two sgRNAs ( URA3 and LYP1 ) or three ( URA3 , LYP1 and COX10 ) was 43% and 19% , respectively ( Figure 3D; Supplementary file 1A ) . This multiplexed efficiency was dependent on the presence of the HDV ribozyme 5′ of the guide sequence as duplex targeting dropped to 3 . 5% in cells with sgRNA lacking the 5′ HDV ribozyme ( Figure 3D; Supplementary file 1A ) . We measured the relative cellular abundance of sgRNAs expressed by PSNR52 , with and without the 5′ HDV ribozyme , by reverse transcription quantitative polymerase chain reaction ( RT-qPCR ) and found that the presence of the ribozyme increases the intracellular abundance of the sgRNAs by sixfold ( Figure 3E ) , consistent with the model that sgRNA abundance is rate limiting for CRISPR/Cas targeting ( Hsu et al . , 2013 ) . Haploid trans-multiplexing showed a minimal loss in activity scaling from one to three loci ( up to three chromosome cuts ) ( Figure 3F; Supplementary file 1A ) . These experiments demonstrate that our multiplexed CRISPR/Cas9 system , which we term CRISPRm , is powerful enough to generate multiple marker-free targeted mutations in the yeast genome in a single experiment and that the 5′ tRNA-HDV ribozyme sequence is required for higher order multiplex targeting . Having established CRISPRm for cis- and trans-multiplex targeting , we next tested its capabilities in engineering genes and pathways integrated in the yeast genome . We first tested whether CRISPRm could be used to assemble foreign genes as chromosomal integrations . We tested the efficiency of in vivo assembly of a functional nourseothricin-resistance ( NatR ) gene from three overlapping PCR products encoding a transcription promoter , protein open reading frame ( ORF ) and transcription terminator at a single locus using cis-multiplexing in the diploid S288C strain . The efficiency of Cas9-mediated integration and assembly of these three DNA fragments to the correct ( URA3 ) locus was measured by a combination of 5-fluoroorotic acid resistance ( 5-FOAR ) and NatR . We found 85% efficiency of targeting and assembly in both copies of the URA3 locus in diploid yeast S288C cells , and 70% in polyploid ATCC4124 by using the tRNAPro sequence as the sgRNA promoter ( Figure 3C; Supplementary file 1A ) , whereas PSNR52-dependent targeting was weaker in S288C cells and non-existent in ATCC4124 cells ( Figure 3C; Supplementary file 1A ) . Thus , CRISPRm enables the one-step markerless assembly of functional genes from multiple fragments into the S . cerevisiae genome , resulting in homozygous insertions into both diploid and polyploid yeast strains . Finally , we tested whether CRISPRm could be used for in vivo selections of improved protein function . We targeted a cellodextrin utilization pathway from the cellulolytic fungus Neurospora crassa , comprised of the cellodextrin transporter ( cdt-1 ) and an intracellular β-glucosidase ( gh1-1 ) , to assemble in vivo in diploid S . cerevisiae . In combination , these two genes result in yeast cells capable of consuming the disaccharide cellobiose , which could be used to improve biofuel production ( Galazka et al . , 2010 ) . To select for improved cellobiose utilizing strains we used error-prone PCR to amplify the cdt-1 gene . We co-transformed the library of mutated cdt-1 alleles , along with overlapping dsDNA fragments for the promoter and terminator , for assembly and integration into the URA3 locus ( ura3::PPGK1-cdt-1-TADH1 library ) of a diploid strain with a previously integrated β-glucosidase gh1-1 gene ( lyp1::PTDH3-gh1-1-TCYC1 ) using CRISPRm . We then grew the transformants in medium containing cellobiose as the sole carbon source to select for functional cdt-1 alleles and identified one strain with 2 . 6-fold ( homozygous cdt-1 N209S/I354N/S360P/T406S/W531L allele ) and one strain with 1 . 7-fold ( homozygous cdt-1 Q45H/F262Y/F533Y allele ) increased cellobiose utilization capacity over wild type cdt-1 ( Figure 4A ) . The mutations were mapped in CDT-1 onto structures of the homologous major facilitator superfamily ( MFS ) hexose-like transporters XylE from Escherichia coli ( Sun et al . , 2012 ) and GlcP from Staphylococcus epidermidis ( Iancu et al . , 2013 ) ( Figure 4B ) . Notably , one site maps to a site that likely interacts directly with the sugar substrate ( N209S ) ( Figure 4B ) . Whereas two mutations in the second allele ( Q45H , F533Y ) map to non-conserved regions of the transporters , one ( F262Y ) maps to a site immediately adjacent to the nearly universally-conserved PESPR motif in hexose transporters that is involved in transporter dynamics ( Sun et al . , 2012 ) . Furthermore , the mutation from phenylalanine to tyrosine results in an expanded motif ( PESPRY ) that is present in all of the major hexose transporters in S . cerevisiae ( Figure 4B ) . We then used CRISPRm to introduce the N209S mutation ( nucleotide G626A ) in the cdt-1Q45H/F262Y/F533Y diploid yeast background . The quadruple mutant cdt-1 allele resulted in synergistic cellobiose utilization with 3 . 9-fold increased growth over strains expressing wild-type cdt-1 ( Figure 4A ) . We next tested the selected cdt-1 alleles in anaerobic fermentations using cellobiose as the sole carbon source . We constructed a cdt-1N209S/F262Y double-mutant in diploid yeast using CRISPRm , and compared it to wild-type cdt-1 and the cdt-1Q45H/N209S/F262Y/F533Y quadruple mutant . We found that wild-type cdt-1 integrated into the chromosome was barely able to support cellobiose fermentation ( Figure 4C ) . By contrast , the cdt-1Q45H/N209S/F262Y/F533Y quadruple mutant or the cdt-1N209S/F262Y double-mutant expressed from the chromosome resulted in complete fermentation of cellobiose , comparable to wild-type cdt-1 expressed from a high-copy 2µ plasmid ( Figure 4C; Ha et al . , 2011 ) . 10 . 7554/eLife . 03703 . 006Figure 4 . CRISPRm mediated insertion and evolution of chromosomal DNA libraries for in vivo protein engineering . ( A ) . Utilization of cellobiose in CRISPRm-engineered diploid yeast strains . S288C wild type cdt-1− ( black ) , cdt-1 ( blue ) , cdt-1Q45H/F262Y/F533Y ( indicated as cdt-1F262Y* on the figure , gray ) , cdt-1N209S/I354N/S360P/T406S/W531L ( indicated as cdt-1N209S* on the figure , green ) and cdt-1N209S+Q45H/F262Y/F533Y ( indicated as cdt-1N209S+F262Y* on the figure , red ) . ( B ) Location of mutations at conserved sites in the evolved CDT-1 transporter structure . The mutated residues are colored magenta in the top figure , and are mapped onto the E . coli XylE transporter in the outward-facing configuration bound to glucose ( upper panels and lower left , PDB entry 4GBZ ) , and onto the S . epidermidis GlcP transporter in the inward-facing configuration ( lower right , PDB entry 4LDS ) . Amino acids in parentheses are the sequences in XylE or GlcP . ( C ) Fermentation of cellobiose by wild-type cdt-1 expressed from a chromosomally-integrated copy in diploid S288C yeast . Cellobiose is indicated as G2 , ethanol as EtOH . The rate of cellobiose consumption was 0 . 13 g L−1 hr−1 . ( D ) Fermentation of cellobiose by mutant versions of cdt-1 expressed from chromosomally-integrated copies . Glucose consumption ( gray ) was identical in both strains . The rate of cellobiose consumption for both strains was over 2 . 0 g L−1 hr−1 . In ( C and D ) , values and error bars represent the means and standard deviations of three independent biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 03703 . 006
We have demonstrated that CRISPRm can be used to quickly generate markerless loss-of-function alleles , heterologous insertion alleles , allele swaps and engineer proteins in yeast by in vivo selections for greatly improved metabolic activity . Although we selected for improved protein function , as suggested by the location of the mutations in CDT-1 ( Figure 4B ) , it is also possible that selection enriched for sequences that confer improved mRNA stability or translation efficiency , due to the fact that the selection was carried out in vivo . Future experiments will be needed to distinguish between these possibilities . CRISPRm could be used to integrate DNA libraries to interrogate cis-regulatory elements such as transcription promoters , 5′- and 3′-untranslated regions ( 5′-UTRs and 3′-UTRs ) in mRNAs , or riboswitches . CRISPRm enables these types of experiments because it allows for near quantitative integration of DNA libraries , which can be a single fragment or complex assemblies of multiple fragments . This could enable integration and selection of multi-gene pathways , or specific mutation and selection of subsets of a protein or multiple DNA-encoded functions . The ability to use chromosomally-integrated DNA libraries removes a key limitation in in vivo selection experiments by eliminating biomolecule expression variation due to variable plasmid copy numbers . Although we used a relatively small library size in the present experiments , even in this small library we observed impressive gains in transporter function . We observed 2–3 mutations per kilobase in the selected transporter genes , resulting in 3–5 amino acid substitutions per transporter . We were then able to combine the mutations in a single transporter with superior fermentation performance . Library sizes introduced by CRISPRm should be scalable to the limits of transformation efficiency , to 107 or larger ( Kawai et al . , 2010 ) . CRISPRm should also enable multiple rounds of selection for directed evolution experiments . We envision that each cycle would involve error-prone PCR of the gene of interest , either from a starting sequence or an enriched and improved cell population , followed by CRISPRm-mediated integration of the amplified gene and selection . This cycle would compare favorably with that required for plasmid-based selections , yet would avoid the problem of plasmid copy number variation . CRISPRm should also be adaptable to screens that rely on sorting large populations of cells , that is by fluorescence-activated cell sorting . We envision that CRISPRm could improve the discovery of human therapeutics by yeast display ( Krugel et al . , 1988 ) , and could be used to map the importance of protein–protein interactions ( Ovchinnikov et al . , 2014 ) via multiplexed library insertion . To achieve high efficiency multiplexing with CRISPRm , we found it was necessary to modify the sgRNA design by including a 5′ HDV ribozyme , and to assay multiple RNA polymerase III promoters for a given strain background . The ease of discovering tRNAs in any genome and the universality of the sgRNA construct means that , with few modifications , CRISPRm can be used directly in non-model fungal hosts , such as pathogens or organisms used in the biotechnology industry . The genetics of many of these organisms have not been studied in any depth due to the technological limitations of available genetic manipulation techniques . For example , industrial S . cerevisiae strains are more stress tolerant and produce much higher yields of desired biofuel or renewable chemical end products than laboratory strains ( Kerr and Service , 2005; Farrell et al . , 2006; Rubin , 2008 ) . However , linking genotypes of industrial yeasts to their phenotypes remains difficult . CRISPRm should serve as a rapid and high throughput means for connecting the genotypes of these organisms to their phenotypes , that is for generating marker-free barcoded alleles for large-scale pooled fitness studies of loss-of-function mutants in these organisms ( Giaever et al . , 2002 ) . In the future , CRISPRm may also be applicable to interrogating mammalian cells , if the levels of HDR are sufficiently high . Finally , the ability to use CRISPRm for multiplexed targeting paves the way for applying directed evolution to cellular pathways and genetic circuits for higher order synthetic biology applications in any host strain .
Gibson Assembly Mastermix ( E2611L ) ( New England Biolabs , Ipswich , MA ) ( Gibson et al . , 2009 ) was used to fuse the KANMX ( Yeastdeletionpages . com ) cassette to the pUC bacterial origin of replication from pESC-URA ( Agilent Technologies , Santa Clara , CA ) . Restriction—free ( RF ) ( van den Ent and Löwe , 2006 ) cloning was used to add a yeast 2µ origin of replication from pESC-URA to the pCAS backbone . The resulting pCAS backbone plasmid was propagated in yeast to confirm functionality . The Cas9 gene from S . pyogenes was amplified from clone MJ824 ( Jinek et al . , 2012 ) and cloned into the pCAS backbone plasmid by RF cloning . A yeast nuclear localization signal ( NLS ) sequence , codon optimized using IDT software ( Integrated DNA Technologies , Coralville , IO ) , was then cloned into the plasmid by RF cloning . Additional elements fused by RF cloning to the Cas9-NLS sequence included the GFP gene , the CYC1 terminator from S . cerevisiae strain S288c ( Yeastgenome . org ) and the promoters ( 800 base pairs upstream ) from the genes TDH3 , TEF1 , RNR2 and REV1 , also taken from strain S288c ( Lee et al . , 2013 ) . For genome editing experiments , the GFP sequence was removed from the Cas9 gene and replaced with a C-terminal His8 affinity tag , by RF cloning . Synthetic DNA ( Integrated DNA Technologies , Coralville , IO ) for the sgRNA and for a catalytically active form of the Hepatitis Delta Virus ( HDV ) Ribozyme was sequentially cloned by RF cloning into the Cas9 containing vector ( pCAS ) . The terminator ( 200 bp ) of SNR52 ( Yeastgenome . org ) was cloned 3′ of the ribozyme-sgRNA sequence by RF cloning . A series of RNA polymerase III ( Pol III ) promoters were PCR amplified from S288c genomic DNA and cloned 5′ of the ribozyme-sgRNA sequence by RF cloning . The tRNA promoters included the full-length tRNA plus 100 base pairs upstream of the tRNA gene ( Supplementary file 1C ) . The sgRNAs used for multiplex targeting were PCR amplified using primers containing 5′ and 3′ restriction sequences and sub-cloned into pCAS by ligation dependent cloning into SalI , SpeI and SacII unique restriction sites . S . cerevisiae strain S288c ( 204508; ATCC ) ( American Type Culture Collection , Manassas , VA ) was used as the haploid , and then mated to form the homozygous diploid for the diploid experiments . Yeast strain ATCC4124 is an industrial polyploid yeast isolated from a molasses distillery , and was obtained from ATCC ( American Type Culture Collection , Manassas , VA ) . Expression and localization of Cas9-GFP was verified by imaging haploid S . cerevisiae S288c cells transformed with pCas9-GFP::KAN using fluorescence microscopy ( Leica Epifluorescence , Leica Microsystems , Buffalo Grove , IL ) . Cells were grown overnight and nuclear localization visualized at 100× magnification . Yeast cells containing pCAS ( Cas9-His8 variant ) were grown in a Bioscreen C Growth Curve Analyzer ( Growth Curve USA , Piscataway , NJ ) in 200 µl of YPD + G418 liquid medium ( 20 g/l Peptone [211667; Bacto] , 10 g/l Yeast Extract [212750; Bacto] , 0 . 15 g/l Adenine hemisulfate [A9126; Sigma] and 20 g/l Glucose [G8270; Sigma] + 200 mg/l G418 [29065A; Santa Cruz Biotechnology] ) . Cells were grown in three biological replicates each with five technical replicates for 48 hr at 30°C under constant shaking . The wild-type control containing an empty vector ( pOR1 . 1 ) was also grown in five technical replicates . Mean and standard deviations of the optical density at 600 nm were calculated for each time point measured by the Bioscreen . The Cas9 transformation mix consisted of 90 µl yeast competent cell mix ( OD600 = 1 . 0 ) , 10 . 0 µl × 10 mg/ml ssDNA ( D9156; Sigma , St . Louis , MO ) , 1 . 0 µg pCAS plasmid , 5 . 0 µg of linear repair DNA and 900 µl Polyethyleneglycol2000 ( 295906; Sigma ) , 0 . 1 M Lithium acetate ( 517992; Sigma ) 0 . 05 M Tris–HCl ( 155568; Invitrogen ) EDTA ( 10618973; Fisher Scientific ) . To measure Cas9 independent integration , the linear DNA was co-transformed with a plasmid lacking the Cas9 protein and sgRNA ( pOR1 . 1 ) ( Supplementary file 2B ) . Cells were incubated 30 min at 30°C , and then subjected to heat shock at 42°C for 17 min . Following heat shock , cells were re-suspended in 250 µl YPD at 30°C for 2 hr and then the entire contents were plated onto YPD+G418 plates ( 20 g/l Peptone , 10 g/l Yeast Extract , 20 g/l Agar , 0 . 15 g/l Adenine hemisulfate , 20 g/l Glucose and G418 at 200 mg/l ) . Cells were grown for 48 hr at 37°C , and colonies imaged using the Biorad ChemiDoc Imager ( Biorad , Hercules , CA ) before replica plating onto phenotype-selective media . The guide sequences in the sgRNAs used for targeting the various loci are shown in Supplementary file 1D . URA3 mutants were selected on 2 . 0 g/l Yeast nitrogen base without amino acids or ammonium sulfate ( Y1251; Sigma ) , 5 . 0 g/l Ammonium sulfate ( A4418; Sigma ) , 1 . 0 g/l CSM ( 4500-012; MP Biosciences ) , 20 g/l Glucose , 20 g/l Agar + 5-fluoroorotic acid ( 1 g/l ) ( F-230-25; Goldbio ) ; LYP1 mutants were selected on 2 . 0 g/l Yeast nitrogen base without amino acids or ammonium sulfate , 5 . 0 g/l Ammonium sulfate , 1 . 0 g/l CSM-lysine ( 4510-612; MP Biosciences ) , 20 g/l Glucose , 20 g/l Agar + thialysine ( 100 mg/l ) ( A2636; Sigma ) ; CAN1 mutants were selected on 2 . 0 g/l Yeast nitrogen base without amino acids or ammonium sulfate , 5 . 0 g/l Ammonium sulfate , 1 . 0 g/l CSM-arginine ( 4510-112; MP Biosciences ) , 20 g/l Glucose , 20 g/l Agar + canavanine sulfate ( 50 mg/l ) ( C9758; Sigma ) ; the remaining auxotrophic mutants were selected on 2 . 0 g/l Yeast nitrogen base without amino acids or ammonium sulfate , 5 . 0 g/l Ammonium sulfate , 1 . 0 g/l CSM , 20 g/l Glucose , 20 g/l Agar; and aerobic respiration deficient mutants ( petites ) were selected on 20 g/l Peptone , 10 g/l Yeast Extract , 20 g/l Agar , 0 . 15 g/l Adenine hemisulfate , 20 g/l Glycerol ( G5516; Sigma ) . Colonies from the YPD+G418 plates were picked and grown overnight in 0 . 8 ml of YPD . Genomic DNA was extracted from these cultures using the MasterPure Yeast DNA Extraction Kit ( MPY80200; Epicentre ) . PCR confirmation of the 60-mer integration allele was performed using primers flanking the target site . PCR products were purified by Exo-SAP-IT ( 78201; Affymetrix ) and Sanger sequenced at the UC Berkeley , Sequencing Facility ( Berkeley , CA ) to confirm barcode sequence in the amplicon . Cells containing the pCAS plasmid with sgRNA inserts were grown in 900 µl of YPD+G418 medium for 24 hr at 30°C and 750 rpm . Total RNA was extracted from exponentially growing yeast cells using Ambion RNA RiboPure Yeast Kit ( AM1926 ) ( Life Technologies , Carlsbad , CA ) . RT-qPCR was performed on the Applied Biosciences StepOne Real-Time PCR System ( Applied Biosystems , Foster City , CA ) using the Invitrogen EXPRESS One-Step SYBR GreenER Kit ( Life Technologies , Carlsbad , CA ) . The RT-qPCR expression level data was quantified using the Comparative CTT ( ΔΔCT ) method and relative abundance of the sgRNA was normalized to the mRNA transcript UBC6 , which was used as the endogenous control ( Teste et al . , 2009 ) . The primer sequence used for the RT reaction was 5′-AAAAGCACCGACTCGGT-3′ and the additional q-PCR primer used was 5′-GTTTTAGAGCTAGAAATAGCAAG-3′ . The primers used for the UBC6 endogenous control were ( RT ) 5′-CATTTCATAAAAAGGCCAACC-3′ and ( qPCR ) 5′-CCTAATGATAGTTCTTCAATGG-3′ . DNA sequences for the sgRNAs used to test HDV ribozyme function are shown in Supplementary file 2C and Supplementary file 2D . Multiplex targeting was performed as described above using pCAS plasmids containing more than one sgRNA expression construct cloned into one of the restriction sites by ligation dependent cloning . Single vs double mutant efficiency was scored relative to the number of colonies present on the YPD+G418 plate . Genomic DNA isolation and PCR of the integration site was performed as described above . Drug resistance cassettes were assembled in vivo from three linear double-stranded DNA fragments PCR amplified from the Ashbya gossypii TEF1 promoter ( AgPTEF1 ) , the nourseothricin open reading frame ( NatR ) and A . gossypii TEF1 ( AgTTEF1 ) terminator in separate reactions . The primers used to amplify the promoter and terminators contained 50 bp of homology to the nourseothricin ORF and 50 bp of homology to the genomic target . Colonies exhibiting drug resistance ( nourseothricin 100 mg/l ) ( N-500-1; Goldbio ) following replica plating were compared to the number of colonies on the YPD+G418 to determine efficiency of multiplex assembly . To generate CDT1 mutant allele libraries , the GeneMorph II Random Mutagenesis Kit ( 200550; Aglient ) ( Agilent Technologies , Santa Clara , CA ) was used to amplify the N . crassa cdt-1 open reading frame . The library of cdt-1 mutant alleles was co-transformed with PCR-amplified linear dsDNA fragments encoding the ScPPGK1 promoter and ScTADH1 terminator into a diploid S288c yeast strain containing a previously-integrated gh1-1 gene . The gh1-1 gene included the ScPTDH3 promoter , the N . crassa gh1-1 open reading frame and ScTCYC1 terminator . The primers used to amplify the promoters and terminators contained 50 bp of homology to either the cdt-1 or gh1-1 ORFs and 50 bp of homology to the respective the genomic targets . 5 µg of each of the three PCR products ( promoter , open reading frame , terminator ) were co-transformed with the pCAS plasmid containing the URA3 guide sequences and screened for G418 resistance as described . Approximately 1600 G418-resistant colonies were pooled and resuspended in minimal cellobiose medium ( SCel ) ( 2 . 0 g/l Yeast nitrogen base without amino acids or ammonium sulfate , 5 . 0 g/l Ammonium sulfate , 1 . 0 g/l CSM , 20 g/l Cellobiose ) . Resuspended cells were immediately spread evenly on SCel plates for initial analysis prior to cellobiose selection ( t = 0 samples ) . Ten microliters of the resuspended cells were inoculated in 50 ml of SC medium in biological triplicate . Cells were harvested after 3 days and spread onto SCel plates . Cells were grown at 30°C for 4 days , and the largest colonies were chosen for further analysis . Cells were grown overnight in 0 . 5 ml of Synthetic Dextrose ( 2% ) ( SD ) in 96 well plates . Cultures were diluted 1:500 in SCel media ( 4% cellobiose ) and 150 µl were grown using the Tecan Sunrise ( Tecan Systems Inc . , San Jose , CA ) in biological triplicate for 3 days at 30°C . Average and standard deviation was calculated for each biological sample . Relative fitness was calculated by measuring area between the curves ( ABC ) which normalizes growth to the area under the curve ( AUC ) for diploid cells lacking cdt-1 ( ABC = AUC cdt-1+−AUC cdt-1− ) . Fold cellobiose utilization capacity = ( ABC cdt-1S209/ABC cdt-1 ) . Integration of the repair oligonucleotide into the chromosomal wild-type cdt-1 gene or cdt-1Q45H/F262Y/F533Y was performed as described for integrating barcoded DNA . The sgRNA sequence used to target the wild type allele of cdt-1 was cloned into the pCAS vector with the sequence 5′-TGCACTGGCTTCTACAACTG-3′ . The repair oligonucleotide used to make the A626G ( i . e . , N209S ) mutation was 5′-CGGCCGCTGCACTGGCTTCTACAGCTGCGGTTGGTTCGGAGGTTCATTCCTGCCGCCTG-3′ with 50 more base pairs of homology to cdt-1 , on both sides of the above oligonucleotide . Genomic DNA was extracted as described and Sanger sequencing confirmed the incorporation of the A626G mutation . Similarly , the T785A mutation ( F262Y ) was incorporated into the cdt-1N209S allele using an sgRNA with guide sequence 5′-CCTCGCTTCCTATTTGCCAA-3′ , and a repair oligonucleotide with sequence 5′-AGAATCCCCTCGCTACCTATTTGCCAACGGCCGCGACGCTGAGGCTGTTGCCTTTCTTGT-3′ . Genomic DNA was extracted as described and Sanger sequencing confirmed the incorporation of the T785A mutation . The sequences of N . crassa CDT-1 ( accession XP_963801 ) , E . coli XlyE ( accession YP_492174 . 1 ) , and S . epidermidis GlcP glucose transporter ( accession ZP_04818045 . 1 ) were aligned with the MUSCLE algorithm ( Edgar , 2004 ) . Figures were prepared with the PyMol molecular graphics system ( http://www . pymol . org/ ) . Strain yML068 was generated by integrating a short DNA ‘spacer’ sequence and Leu2 into the LEU2 locus of BY4741 . Strain yML069 was generated by integrating PTDH3-mRuby2-TTDH1 and Ura3 into the URA3 locus , and integrating PTDH3-Venus-TTDH1 and Leu2 into the LEU2 locus of BY4741 . Strain yML104 was generated by integrating PTDH3-mRuby2-TTDH1-PTDH3-Venus-TTDH1 and Ura3 into the URA3 locus of yML068 . Plasmid pML1350 is a Ura3 CEN6/ARS4 plasmid with PTDH3-mRuby2-TTDH1 . pML1362 is a Leu2 CEN6/ARS4 plasmid with PTDH3-Venus-TTDH1 . pML1177 is a Ura3 CEN6/ARS4 plasmid with PTDH3-mRuby2-TTDH1-PTDH3-Venus-TTDH1 . Synthetic defined media lacking leucine and uracil ( SD-LU ) was made by adding 6 . 7 g/l Difco Yeast Nitrogen Base without amino acids; 2 g/l Drop-out Mix Synthetic Minus Leucine , Uracil without Yeast Nitrogen Base ( US Biological ) ; and 20 g/l Dextrose to distilled water . Colonies were picked into SD-LU media and grown to exponential phase at 30°C . Cultures were concentrated by centrifugation , spotted onto plain glass slides , and examined on a Zeiss Observer D1 microscope using a 100× DIC objective . Images were captured using a Hamamatsu Orca-flash 4 . 0 ( C11440 ) camera using auto-exposure . Fluorescence images were taken using an X-Cite Series 120 lamp , Zeiss filter set 45 ( excitation at 560/40 nm and emission at 630/75 nm ) for mRuby2 , and Zeiss filter set 46 ( excitation at 500/20 nm and emission at 535/30 nm ) for Venus . Images were analyzed and composites were created using Fiji ( http://fiji . sc ) . To rule out that our CRISPR-Cas9/sgRNA system resulted in non-specific genome targeting , we performed whole genome sequencing for URA3 and LYP1 targeted strains and searched for insertion/deletions ( INDELs ) , single nucleotide polymorphisms ( SNPs ) and multi-nucleotide polymorphisms ( MNPs ) . We identified 21 sequence variants across the nine URA3 and LYP1 targeted strains ( Supplementary file 1B ) . Whole genome sequencing was performed by the UC Davis Genome Center ( Davis , CA ) using the Illumina MiSeq platform ( Illumina , Hayward , CA ) to produce 150 bp paired-end reads . The software package versions used for sequencing data analysis were as follows: BWA ( v . 0 . 7 . 5ar405 ) , Picard ( v . 1 . 92[1464] ) , SAMtools ( v . 0 . 1 . 19-44428cd ) and the GATK ( 2 . 7-2-g6bda569 ) . The S288C reference genome ( v . R64-1-1 , release date 3 Feb 2011 ) was obtained from the Saccharomyces Genome Database ( yeastgenome . org ) and prepared for use in sequencing data analysis with bwa index , CreateSequenceDictionary from Picard , and samtools faidx . Sequencing reads were processed with Scythe ( v . 0 . 991 ) to remove adapter contamination and Sickle ( v . 1 . 210 ) to trim low quality bases . Processed reads were mapped to the S288C reference genome using bwa mem with the—M option for picard and GATK compatibility . The mapped reads were sorted with SortSam and duplicate reads were marked with MarkDuplicates from Picard . Read alignments were refined by performing local realignment with the RealignerTargetCreator and IndelRealigner walkers from the GATK on all samples collectively . Variant detection for both SNPs and INDELs was performed with GATK's UnifiedGenotyper , with parameters adjusted for haploid genomes and no downsampling of coverage , for each sample independently . The resulting SNP and INDEL calls were filtered with the VariantFiltration walker from GATK ( see header of the VCF file , supplemental VCF file , for details ) . A custom Perl script ( Supplementary file 3 ) was written to identify all GG dinucleotide sequences in the S288C reference genome , extract every Cas9 target sequence ( i . e . , 20 nt sequence corresponding to the 20 nucleotides immediately 5′ of the ‘NGG’ PAM site ) , and obtain the genome coordinates ranging from 30 nucleotides upstream and downstream of the PAM site . Cas9 target sequences were added to VCF files as custom annotations using snpEff ( v3 . 3h ) , and SnpSift ( v3 . 3h ) was used to extract desired fields into tables for analysis with a custom R script ( Supplementary file 4 ) . Needleman–Wunsch global alignments between our guide sequences and Cas9 target sequences were performed using the pairwiseAlignment function ( Biostrings package , Bioconductor ) in R , with a substitution matrix of −1 for mismatches and 2 for matches , produced with the nucleotideSubstitutionMatrix function ( Biostrings package , Bioconductor ) . The probability of there being a better match for the guide sequence to a given Cas9 target sequence was calculated as the frequency of Cas9 target sequences with better alignments to the same guide sequence , amongst 10 , 000 randomly selected Cas9 target sequences . To compile counts of all variants and various subclasses , a GATKReport was generated from the VCF files with GATK's VariantEval walker , read into R using the ‘gsalib’ library , and the desired categories were extracted with a custom R script ( Supplementary file 4 ) . Without the repair DNA template required for HDR , the majority of mutations caused by Cas9 are expected to be INDELs , SNPs or MNPs that initiate within the protospacer sequence , which is the 20 nucleotide ( nt ) sequence 5′ of the PAM ( Fu et al . , 2013 ) . Therefore , for completeness , we searched for all PAM sites within 30 nt upstream and downstream for each of the 21 variants . We then compared the 20 nt URA3 and LYP1 guide sequences to the putative protospacer sequences within the regions flanking each variant . An end-to-end alignment identified 10 or fewer nucleotide matches between the URA3 or LYP1 guide sequences and the variant sequences ( Supplementary file 1B ) . Cas9 requires at least 12 perfect base pair matches within the guide-target sequence ( Hsu et al . , 2013 ) so it is highly unlikely that the URA3 and LYP1 guide sequences directed Cas9 to any of these potential protospacers that lie within variant sites . These mutations likely arose as natural variants during the course of the experiment . As a second method to evaluate the likelihood of off-target mutations , we performed local alignments of our guide sequences to all Cas9 target sequences whose PAM site was within 30 nucleotides upstream and downstream of a detected variant , as well as to 10 , 000 randomly selected Cas9 target sequences from the genome . Since our guide sequences are expected to have a better match to 13% or more of all Cas9 target sequences ( ∼126 , 000 or more sites ) than to the best matching Cas9 target sequence with a nearby variant , and the number of nucleotide matches in end-to-end alignments is at most 10 , we argue that the variants identified in the genomes of URA3- and LYP1-targeted strains are unlikely the result of off-target Cas9 modifications . Yeast strain colonies were inoculated in 20 ml of oMM ( optimized minimal media ) ( Yuping Lin , personal communication ) with 2% glucose in 50 ml Falcon tubes and grown aerobically at 30°C to saturation for 24 hr . The oMM contained 1 . 7 g/l YNB ( Y1251; Sigma , Saint Louis , MO , USA ) , 2× CSM , 10 g/l ( NH4 ) 2SO4 , 1 g/l MgSO4 . 7H2O , 6 g/l KH2PO4 , 100 mg/l adenine hemisulfate , 10 mg/l inositol , 100 mg/l glutamic acid , 20 mg/l lysine , 375 mg/l serine and 0 . 1 M 2- ( N-morpholino ) ethanesulfonic acid ( MES ) pH 6 . 0 . The saturated cultures were then inoculated to starting OD600 of 0 . 2 in 500 ml of oMM with 2% glucose in 1 l Erlenmeyer flasks and grown aerobically at 30°C to mid-log phase at a final OD600 of 2 . 5 . Cells were harvested and washed twice with sterile ddH2O . The washed cells were then inoculated at OD600 of 20 in 50 ml oMM with 1% glucose and 8% cellobiose in 125 ml serum flasks . After inoculation , the flasks were sealed with rubber stoppers and clamped with an aluminum seal . To achieve anaerobic conditions , the headspaces of the sealed flasks were purged with nitrogen gas for 30 min . These were then cultivated at 30°C and 220 rpm . Using sterile needles and syringes , 1 ml samples were removed through the rubber stoppers at the indicated time points . The cells were pelleted and 5 μl of the supernatants were analyzed for cellobiose , glucose , glycerol , and ethanol content by high performance liquid chromatography on a Prominence HPLC ( Shimadzu , Kyoto , Japan ) equipped with Rezex RFQ-FastAcid H 10 × 7 . 8 mm column . The column was eluted with 0 . 01 N of H2SO4 at a flow rate of 1 ml/min , 55°C . | Over the course of billions of years , natural evolution has produced new proteins and adapted existing ones so that they work better . Scientists have learned how to use the principles that underlie evolution to similarly engineer proteins in the laboratory . This process , known as directed evolution , is a powerful tool for improving how proteins function . Directed evolution normally involves mutating the gene that encodes the protein of interest , selecting the genes that produce the most promising proteins for another round of mutation , and repeating the process until the desired protein function is achieved . In the first step of directed evolution , a gene is usually mutated randomly in order to create a large ‘library’ of different forms of the gene . These are joined to circular pieces of DNA known as plasmids that can replicate themselves inside cells . However , the number of plasmids than can be taken up differs from cell to cell . This complicates experiments , and the ideal directed evolution experiment would have the same number of plasmids , or target genes , being delivered into each cell . Ryan et al . have developed a new method for performing directed evolution experiments that uses a recently developed technique called the CRISPR-Cas9 system . This can make direct changes to a DNA strand such as inserting or deleting specific sequences that code for proteins . Ryan et al . used the CRISPR-Cas9 system to create multiple DNA breaks simultaneously across the genome of yeast cells , and joined ‘barcoded’ DNA or DNA for intact genes to these breaks . This avoids the need to use plasmids to introduce foreign DNA into cells . Ryan et al . have named this method the Multiplex CRISPR ( or CRISPRm ) system . Having established CRISPRm , Ryan et al . tested whether it could be used to engineer improved proteins by attempting to modify a transporter protein called CDT-1 . This protein transports the sugar cellobiose into yeast cells , where it can be converted into alcohol by fermentation . This is important for making biofuel from plants . After just one round of directed evolution using CRISPRm , Ryan et al . successfully isolated a form of the CDT-1 protein that increased the rate of fermentation over 10-fold; hence this CDT-1 variant could be used to increase biofuel production . In the future , it will be important to implement multiple selection rounds with CRISPRm , and to test how large the DNA libraries can be for directed evolution . In time , CRISPRm could find use in evolving and engineering different combinations of genes , metabolic pathways , and possibly entire genomes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Selection of chromosomal DNA libraries using a multiplex CRISPR system |
The Rho GTPase Rac1 activates the WAVE regulatory complex ( WRC ) to drive Arp2/3 complex-mediated actin polymerization , which underpins diverse cellular processes . Here we report the structure of a WRC-Rac1 complex determined by cryo-electron microscopy . Surprisingly , Rac1 is not located at the binding site on the Sra1 subunit of the WRC previously identified by mutagenesis and biochemical data . Rather , it binds to a distinct , conserved site on the opposite end of Sra1 . Biophysical and biochemical data on WRC mutants confirm that Rac1 binds to both sites , with the newly identified site having higher affinity and both sites required for WRC activation . Our data reveal that the WRC is activated by simultaneous engagement of two Rac1 molecules , suggesting a mechanism by which cells may sense the density of active Rac1 at membranes to precisely control actin assembly .
Dynamic rearrangements of the actin cytoskeleton play central roles in cellular processes , ranging from cell migration and adhesion to endocytosis and intracellular vesicle trafficking ( Skau and Waterman , 2015 ) . In many of these processes , actin dynamics are spatially and temporally controlled by members of the Wiskott-Aldrich Syndrome Protein ( WASP ) family . These proteins integrate a diverse array of upstream signals and transmit them through their conserved VCA sequence to the Arp2/3 complex , which , in turn , nucleates actin filaments to create branched actin networks at membranes ( Campellone and Welch , 2010; Padrick and Rosen , 2010 ) . In most WASP family proteins , the VCA element is inhibited in the resting state . Inhibition is mediated either by autoinhibition within a single polypeptide chain , as in WASP and N-WASP , or by trans-inhibition within large multi-protein assemblies , as in WAVE and WASH ( Rotty et al . , 2013 ) . Multiple signals , including binding to ligands ( e . g . Rho family GTPases , phosphoinositide lipids and membrane receptors ) and covalent modifications ( e . g . phosphorylation and ubiquitination ) , often act cooperatively to relieve inhibition and concomitantly recruit WASP proteins to their sites of action at membranes ( Chen et al . , 2014a , 2010; Hao et al . , 2013; Lebensohn and Kirschner , 2009; Padrick and Rosen , 2010; Prehoda et al . , 2000; Torres and Rosen , 2003; 2006 ) . The WASP family members WAVE1 , WAVE2 and WAVE3 are essential to actin dynamics needed for normal development and function of most eukaryotic organisms , and are also implicated in many cancers ( Bisi et al . , 2013; Lane et al . , 2014; Takenawa and Suetsugu , 2007; Yanagisawa et al . , 2013 ) . In cells , each WAVE protein functions exclusively within a 400 kDa , heteropentameric assembly , termed the WAVE Regulatory Complex ( WRC ) , which also contains the proteins Sra1 , Nap1 , Abi2 , and HSPC300 ( or their homologs ) ( Stovold et al . , 2005 ) . Previously , we reported crystal structures of an inhibited WRC assembly containing all five subunits , but lacking the disordered C terminus and SH3 domain of Abi2 and the proline-rich region of WAVE1 ( mini-WRC ) ( Chen et al . , 2010 ) , and of this assembly bound to a WRC Interacting Receptor Sequence ( WIRS ) motif peptide derived from the adhesion receptor , protocadherin 10 ( Chen et al . , 2014a ) . The structure revealed that Sra1 and Nap1 form an elongated dimer of about 100 × 100 ×200 Å . WAVE1 , Abi2 and HSPC300 form a trimer as an elongated 4-helix bundle , which aligns along the long axis of the Sra1/Nap1 dimer ( Figure 1A ) . Immediately following the helix bundle , an extended sequence of ~90 a . a . from WAVE1 ‘meanders’ across the surface of Sra1 as a loose collection of loops and helices ( the meander region ) . The VCA element of WAVE1 is sequestered at the VCA-binding site by multiple interactions between its V and C helices and a group of helices from both Sra1 and the meander region of WAVE1 itself , explaining how WAVE1 is inhibited within the WRC ( Chen et al . , 2010 ) . The structures also suggested how the WRC might be activated and recruited to membranes by the combined actions of the Rho GTPase Rac1 , the acidic phospholipid PIP3 , phosphorylation on both Ser and Tyr residues , and membrane receptors containing WIRS motifs . Among these , the GTPase Rac1 functions as a canonical activator , whose direct interaction with the WRC is necessary , and in some reports , sufficient to drive activation ( Chen et al . , 2014a; Chen et al . , 2010; Eden et al . , 2002; Ismail et al . , 2009 ) . Based on mutagenesis and biochemical data , we previously proposed that Rac1 binds to a conserved surface at one end of Sra1 adjacent to the VCA-binding site ( referred to as the ‘A Site’ in Figure 1B ) . We hypothesized that binding of Rac1 could cause structural changes that propagate through the meander region to the VCA to promote VCA release and consequent activation toward the Arp2/3 complex ( Chen et al . , 2010 ) . However , in the absence of a WRC-Rac1 complex structure , it remained uncertain exactly how Rac1 binds to and activates the WRC . Here we report the structure of a WRC-Rac1 complex at 7 Å resolution determined by single-particle cryo-electron microscopy ( cryo-EM ) , together with complementary mutagenesis and biochemical analyses of the complex . Surprisingly , Rac1 is not found on the binding site we predicted previously . Rather , it binds to a conserved site on the opposite end of Sra1 . Our biochemical data , including quantitative pull-down , membrane recruitment , and analytical ultracentrifugation ( AUC ) assays with WRC mutants , confirm that both the original site and the new site can bind Rac1 , the latter with approximately 40-fold higher affinity . Actin assembly experiments further reveal that both sites contribute to activation of the WRC toward the Arp2/3 complex . These data lead to a model in which maximal WRC activation requires simultaneous engagement of two Rac1 molecules on the membrane . This bivalent activation model suggests a mechanism by which cells may precisely control actin dynamics at membranes by sensing the density of active Rac1 produced by upstream signaling pathways .
Initially we sought to determine the crystal structure of an intermolecular complex of the WRC bound to Rac1 . Despite extensive efforts , however , we were unable to obtain diffracting crystals of such an assembly . These investigations led us to an engineered WRC-Rac1 complex with improved stability , which proved suitable for structure determination by cryo-EM ( Figure 1A , B ) . Compared to the mini-WRC construct previously used for crystallography ( Chen et al . , 2010 ) , we incorporated the following modifications to stabilize Rac1 binding ( Figure 1A ) . First , we removed the VCA sequence from the WRC ( herein named ΔWRC ) , a modification that we previously showed increases the affinity for Rac1 ~10 fold ( Chen et al . , 2010 ) . Second , we genetically fused Rac1 to the C terminus of WAVE1 ( a . a . 1–230 ) through a flexible linker ( GGS ) 6 . The previous crystal structures ( and the cryo-EM structure reported here ) showed that WAVE1 is only ordered to residue L184 within the WRC , affording the system here an effective linker of 65 amino acids . This linker , ~230 Å if fully extended , would provide ample length to connect Rac1 to the previously demonstrated binding site , which is approximately 70 Å from residue L184 of WAVE1 . Based on experiments measuring binding to immobilized GST-Rac1 , the tethered Rac1 protected the WRC from binding to GST-Rac1 much more efficiently than the un-tethered Rac1 protected the wild type WRC , suggesting the tethered Rac1 binds the WRC in a physiologically relevant manner ( Figure 1—figure supplement 1A ) . Third , we introduced two point mutations , Q61L and P29S , into the fused Rac1 . The Q61L mutation abolishes the GTP hydrolysis activity of Rac1 ( Der et al . , 1986 ) , maintaining the protein in the GTP-bound state . The P29S mutation was identified as one of the major somatic mutations in melanoma ( Hodis et al . , 2012; Krauthammer et al . , 2012 ) , and was shown to increase the affinity of Rac1 for the effectors PAK1 ( p21 protein activated kinase 1 ) and MLK3 ( mixed-lineage kinase 3 ) ( Hodis et al . , 2012; Krauthammer et al . , 2012 ) . Similarly , we found that the P29S mutation also increases the affinity of Rac1 for the WRC ( Figure 1—figure supplement 1B and C ) . Together , these optimizations facilitated stable association of Rac1 with the WRC . To distinguish this construct from the mini-WRC used for crystallography , we name it the ΔWRC230-Rac1 complex . We first examined the ΔWRC230-Rac1 complex by electron microscopy of negatively stained specimens , which showed images of monodisperse particles of homogeneous size and shape ( Figure 1—figure supplement 2A ) . This prompted us to collect cryo-EM images of vitrified samples for structure determination ( Figure 1—figure supplement 2B ) . We first manually picked ~12 , 000 particles from 489 images and calculated class averages using the iterative stable alignment and clustering ( ISAC ) procedure ( Yang et al . , 2012 ) . Some of the resulting averages showed extra density in addition to the WRC , indicating the presence of Rac1 ( Figure 1—figure supplement 2C , arrows ) . We further collected 1684 images , from which a total of 160 , 591 particles were automatically picked , and then processed using Relion ( Scheres , 2012 ) . The pooled particle images were aligned to the low-pass filtered crystal structure of the mini-WRC ( PDB 3P8C ) ( Chen et al . , 2010 ) and subjected to 3D classification ( Figure 1—figure supplement 2D ) . After further refinement , we obtained a final density map from 29 , 784 particles at a resolution of 7 Å ( Figure 1—figure supplement 3 ) . The overall EM density map clearly revealed a structure containing both the WRC and Rac1 ( Figure 1—figure supplement 3A ) , with Rac1 bound to Sra1 at one end of the WRC ( detailed below ) . The mini-WRC crystal structure ( PDB 3P8C ) could be unambiguously docked into the EM density map , with the majority of the secondary structural elements fitting into the density without need for adjustment ( Figure 1B ) . Several surface loops in Sra1 and Nap1 that were not observed in the crystal structure also have defined EM densities ( Figure 2A ) . The EM map shows no density for the V and C helices of the VCA region of WAVE1 ( Figure 2B , right panel , elements colored black ) , which was present in the mini-WRC crystal structure but was removed in the ΔWRC230-Rac1 complex used here . No density was observed for the first helix ( α1 ) of Sra1 ( Figure 2B , left panel ) . In crystals of mini-WRC this helix did not bind in intra-complex fashion , but rather to an adjacent WRC in the crystal lattice . Density was also not observed for residues 161–230 of WAVE1 or the ( GGS ) 6 linker ( Figure 2B , right panel ) , which leads to the N terminus of the fused Rac1 , suggesting that this sequence of 88 residues does not form a stable structure in the complex . No major conformational changes of the WRC were observed within the limits of our 7 Å resolution , except for the meander region of WAVE1 ( a . a . 90–184 ) ( Figure 2C ) , which formed a series of helices ( α2-α6 ) arrayed across the surface of Sra1 in the mini-WRC crystal structure ( Chen et al . , 2010 ) . In the EM map , the meander region generally has weak density . Several regions either adopt an orientation different from the crystal structure ( a . a . 90–97 in α2 , a . a . 113–118 prior to α3 , and a . a . 148–160 between α5 and α6 ) , or have missing density ( a . a . 130–147 between α3 and α5 and a . a . 161–182 of the long α6 helix ) ( Figure 2C ) . In the crystal structure of the inhibited mini-WRC , helices α2 and α6 directly bind the V and C regions of the VCA , respectively . Mutagenesis studies suggest that these interactions act cooperatively to sequester the VCA , inhibiting its activity toward the Arp2/3 complex ( Chen et al . , 2010 ) . The observations here support this idea , since removing the VCA ( and possibly Rac1 binding as well ) reciprocally appears to destabilize the meander region . The only major extra density not accounted for by the WRC was located on a surface of Sra1 ~120 Å from the previously proposed Rac1 binding site , on the opposite end of the long axis of the WRC ( Figure 1B ) . This density unambiguously accommodated the crystal structure of Rac1 P29S , including all secondary structure elements and most loops ( PDB 3SBD ) ( Krauthammer et al . , 2012 ) ( Figure 1C ) . In this position , the N terminus of Rac1 is located ~100 Å from the last ordered residue of WAVE1 to which it is linked , a distance that could readily be spanned by the 88 disordered residues of the tether . Note that none of our EM maps resulting from 3D classification indicated that Rac1 was bound to other regions of the WRC . Even supervised classification that included as reference a model of the WRC with Rac1 bound to the previously predicted binding site failed to produce a map that revealed additional density at this site ( data not shown ) . Rac1 and other small GTPases use two conserved structural elements , Switch I ( a . a . 25–39 ) and Switch II ( a . a . 57–75 ) , to bind and activate downstream signaling proteins in a nucleotide-dependent manner ( Mott and Owen , 2015 ) . In our model of the complex , both Switch I and Switch II of Rac1 make direct contact with the WRC , through a conserved surface on the Sra1 subunit ( a . a . 957–979 , arrow heads in Figure 1C ) . This observation further validates the EM density map , which was derived without using crystal structures of Rac1 as a reference . In our structure Rac1 is not located at the site predicted by our previous biochemical/mutagenesis analyses . That site is a conserved surface patch on Sra1 near the α4-α6 helices of the WAVE1 meander region , adjacent to the VCA binding site ( Figure 1B , Figure 3A ) ( Chen et al . , 2010 ) . Instead , in the EM reconstruction , Rac1 binds to a different conserved surface at the opposite end of the rod-shaped Sra1 subunit , which is distant from the meander-VCA region ( Figure 1B , Figure 3A ) . For clarity , here we name the previously described site the ‘A Site’ ( for Adjacent Site ) , and the site observed in the EM structure the ‘D Site’ ( for Distant Site ) . How can the previous biochemical analyses be reconciled with the location of Rac1 found in the present structure ? In our previous work , we found that the A Site was one of the few highly conserved surface patches on the WRC . We showed that mutating any of three surface residues ( C179R , R190D , or M632D ) , or one pair of residues ( E434K/F626A ) , at the A Site disrupted Rac1 binding in qualitative GST pull-down experiments , and that the two mutations analyzed in a quantitative equilibrium dialysis experiment ( R190D and E434K/F626A ) both decreased affinity ( Chen et al . , 2010 ) . In comparison , the D Site is also conserved , although less so than the A Site ( Supplementary file 1 and Figure 3—figure supplement 1 ) . To test if Rac1 also binds to the D Site , we introduced point mutations to its surface residues and examined the mutants in a battery of assays . We divided the surface into six distinct patches and mutated them separately ( color coded in Figure 3B ) : P957A/K958D/I959A , R961D/P963A/R964D , Y967A , G971W , E974A/F975A/H978A/Q979A , and D1122A/E1123A . We also examined a point mutation , S969F , which is buried immediately beneath the D Site ( labeled in red in Figure 3C and D ) . The equivalent mutation , S968F , in Cyfip2 ( the mouse ortholog of Sra1 ) was shown to cause an altered cocaine response in mice through an unknown mechanism ( Kumar et al . , 2013 ) . Finally , we also reanalyzed a previously tested mutation at the A Site , C179R ( labeled in black in Figure 3A ) . As described in the Protein Purification section of the Materials and Methods , all of the WRC mutants appear to be properly folded and assembled . We first examined the ability of immobilized GST-Rac1 to retain the WRC mutants . GST-Rac1 loaded with GMPPNP ( a poorly hydrolyzable GTP analog ) could efficiently retain wild type ( WT ) WRC as well as one of the six D Site mutants , D1122A/E1123A ( Figure 3E ) . All other mutations at the D Site substantially diminished binding , suggesting that this region mediates high-affinity interactions with Rac1 . The D1122A/E1123A mutation is located at the periphery of the D Site , likely explaining its lack of effect on binding ( Figure 3B , grey patch ) . Consistent with our previous report ( Chen et al . , 2010 ) , the A Site mutation , C179R , also decreased binding under the same experimental conditions , albeit more moderately ( Figure 3E ) . The observed interactions are specific to active Rac1 , since Rac1-GDP did not bind to any of the WRCs ( Figure 3—figure supplement 2A ) . The S969F mutation did not appreciably affect binding ( Figure 3E ) . We obtained similar results under different reaction conditions , employing the high-affinity mutant Rac1 ( Q61L/P29S ) at a higher salt concentration ( 100 mM NaCl , vs . 50 mM above ) ( Figure 3—figure supplement 2B ) . Under these conditions , which favor binding to the WRC , the A Site mutation C179R , and the D Site mutations , P957A/K958D/I959A , G971W and E974A/F975A/H978A/Q979A , all weakened binding to a moderate degree , whereas the R961D/P963A/R964D and Y967A mutations at the D Site still eliminated observable binding . The D1122A/E1123A and S969F mutations had no effect . These data suggest that the A Site makes a modest contribution to the total observed binding in pull-down experiments , while the D Site , especially the surface patch formed by R961/P963/R964/Y967 , has a major contribution . It is notable that P963 and Y967 are highly conserved in eukaryotic organisms including animals , fungi , amoebas and many protozoans , except plants ( Supplementary file 1 ) . In the experiments below , we focused on the Y967A mutation to disrupt the D Site . Because Rac1 activates the WRC at membranes in vivo , we further developed an assay to examine if mutating the A or D Site could similarly disrupt the binding when Rac1 is attached to membranes ( Figure 3F ) . We fused an N-terminal mCherry tag and a C-terminal His8 tag to Rac1 ( either Q61L/P29S or wild type , as indicated in Figure 3F ) . The latter allowed the protein to be anchored on supported bilayers containing 2% Ni2+-NTA-DGS lipid ( see schematic presentation in Figure 3F ) ( Banjade and Rosen , 2014 ) . We then quantified the amount of EGFP-tagged WRC recruited to the Rac1-containing membrane using Total Internal Reflection Fluorescence ( TIRF ) microscopy . Paralleling the above GST pull-down experiments , mutating either the A Site ( C179R ) or the D Site ( Y967A ) significantly decreased membrane recruitment of the WRC , with the D Site mutation having a larger effect . In contrast , the S969F mutation did not decrease , but rather slightly increased membrane recruitment ( Figure 3F ) . These data suggest that Rac1 can bind to both the A and D sites . Nevertheless , it remained possible that only the D Site directly contacts the GTPase ( as observed in the cryo-EM structure ) and the A Site mutations disrupted binding through allosteric effects on the D site . To examine this possibility , we used Multi-Signal Sedimentation Velocity Analytical Ultracentrifugation ( MSSV AUC ) to directly measure the stoichiometry of the interaction between Rac1 ( Q61L/P29S ) and ΔWRC230 ( Figure 4 ) ( Balbo et al . , 2005; Padrick et al . , 2010 ) . To separately track sedimentation of the two components , we fused EGFP to the N-terminus of Rac1 . This labeling allowed us to use absorbance at 490 nm to specifically monitor EGFP-Rac1 and use interference signals to record the sedimentation of all proteins . Thus , in our assay EGFP-Rac1 was tracked by both signals , and the WRC by interference only . We note that MSSV AUC enables direct quantification of the EGFP-Rac1/WRC stoichiometry for any given species , as well as its sedimentation coefficient . Individually , both EGFP-Rac1 and the WRC sedimented as high-purity , monodisperse species , with sedimentation coefficients of 3 . 5 s and 10 . 5 s , respectively ( Figure 4 , and see Figure 4—figure supplement 1 for SDS PAGE gels and gel filtration chromatography profiles of the protein samples ) . When the two proteins were mixed , keeping WRC at approximately 1 µM concentration and increasing EGFP-Rac1 from 1 . 2 µM to 15 µM , a fraction of EGFP-Rac1 began co-sedimenting with the WRC ( Figure 4A , left panel , green curves ) . Addition of EGFP-Rac1 also shifted the WRC peak toward a higher sedimentation coefficient in a concentration dependent manner ( Figure 4A , left panel , black curves ) . At an intermediate concentration of EGFP-Rac1 ( 3 . 7 μM ) , the WRC peak shifted to 11 . 7S with a Rac1:WRC stoichiometry of 1 . 3:1 , suggesting more than one binding site . Increasing the concentration of EGFP-Rac1 up to 15 . 1 μM shifted the sedimentation coefficient of the WRC peak to 12 . 1S and stoichiometry to 1 . 9:1 , confirming that two Rac1 molecules can simultaneously bind to a single WRC ( Figure 4B ) . The MSSV AUC data for the mutants were quite different . For the A Site mutant ( C179R ) , at a nearly saturating concentration of EGFP-Rac1 ( 8 . 4 μM vs . dissociation constant KD of 0 . 27 μM as determined below ) , the binding stoichiometry only reached 1:1 ( Figure 4A , blue curves , 4B ) . For the D Site mutant ( Y967A ) , the binding stoichiometry only reached to 0 . 6:1 with 8 . 5 μM Rac1 , consistent with the lower affinity ( KD ~11 . 5 μM ) of the A Site as determined below ( Figure 4A , red curves , 4B ) and implicated by the qualitative GST-pull down data above . Together , the AUC data confirmed that two molecules of Rac1 can engage the WRC at the A and D sites simultaneously , and that our mutations could selectively weaken binding to the individual sites . The above data all suggest that the A Site has a weaker affinity than the D Site . To quantify this difference , we used an equilibrium pull-down ( EPD ) assay to determine the binding affinity of GST-Rac1 ( Q61L/P29S ) for each site ( Figure 5 ) ( Lee et al . , 2000; Pollard , 2010 ) . In this assay , we mixed a constant amount of ΔWRC230 ( 0 . 1 μM ) with varying concentrations of GST-Rac1 ( 0 . 01 μM to 140 μM ) and a fixed amount of Glutathione Sepharose beads . The beads were sufficient , given their quantity and the extremely high affinity of GST for glutathione ( KD ~7 pM , [Waterboer et al . , 2005] ) to retain virtually all GST-Rac1 and its bound WRC , even at the highest concentrations ( Figure 5—figure supplement 1 ) . After a brief centrifugation to pellet the beads , uncomplexed WRC remained in the supernatant and could be quantified by SDS PAGE ( example gels shown in Figure 5A ) , allowing us to determine the binding isotherms , which we could numerically fit to different binding models to obtain dissociation constants ( Figure 5B and Figure 5—figure supplement 2 ) . Note that the binding isotherms determined by the EPD assays do not distinguish the WRC species with one or two bound Rac1 molecules , or the WRC species with Rac1 bound to the D-Site or the A-Site , because all species can be equally pelleted with the glutathione beads . Therefore , fitting the binding isotherm of the WT WRC to a cyclic two-site binding model ( Figure 5—figure supplement 2A ) to extract the dissociation constants for both the D Site and the A Site could not converge , with large uncertainty for the derived KD ( not shown ) . Fitting to a simplified sequential two-site binding model or to a single-site model resulted in a similar dissociation constant for the D Site ( KD , D of 0 . 169 ± 0 . 008 μM in Figure 5C , black solid curve compared to 0 . 106 ± 0 . 002 μM in Figure 5C , grey dotted curve ) . Fitting to the sequential two-site binding model was statistically better than the single-site model ( p=5 . 45×10−9 ) , and yielded a dissociation constant for the A Site for the D Site-bound WRC ( KD , A ( D ) ) of 0 . 262 ± 0 . 044 μM , suggesting positive cooperativity between the A and D Sites compared to the dissociation constant for the A Site alone ( KD , A of 11 . 52 ± 0 . 11 µM as determined below ) . Such cooperativity likely results from avidity , as two nearby Rac1 molecules , either partners in a GST fusion or adjacent on the solid matrix , bind the same WRC . In contrast to the WT WRC , the isotherm for the A Site mutant ( C179R ) , fit well to a single-site binding model , yielding a KD , D of 0 . 265 ± 0 . 004 μM for the D Site ( Figure 5C , light blue dotted curve ) , whereas fitting to a two-site binding model resulted in an identical KD , D with no statistical improvement ( Figure 5C , blue solid curve ) , supporting the idea that the C179R mutation effectively impaired the A Site . For the D Site mutant ( Y967A ) , the strong shift of the data to higher Rac1 concentrations clearly shows that the interaction that recruits the WRC to the beads is appreciably weaker than that for the wild type or A site mutant WRCs . But the isotherm did not fit well to either a single- or a two-site binding model , although both yielded a similar KD , A for the A Site ( 10 . 85 ± 0 . 4 vs 9 . 6 ± 0 . 38 μM , Figure 5C red dotted curve , and see Figure 5—figure supplement 2 ) . More complicated models involving more than two dissociation constants did not converge ( not shown ) . However , when we modeled a dimeric GST-Rac1 binding two WRCs cooperatively ( Figure 5B , GST-dimer model ) , the fitting was significantly improved ( p=1 . 39×10−11 , Figure 5C , red solid curve ) . This model yielded a KD , A of 11 . 52 ± 0 . 11 μM for binding of the first WRC , and a KD , A’ of 0 . 033 ± 0 . 003 μM for binding of the second WRC to this initial ( GST-Rac1 ) 2-WRC assembly . These data would be consistent with strong positive cooperativity between WRC binding events when Rac1 only binds to the A Site . The better fit afforded by this model does not authenticate the molecular mechanism it describes . We note , however , that the potential for cooperative interactions between closely-positioned WRCs has been suggested by the crystal structure of miniWRC . There , the N-terminal helix of Sra1 , which is not observed in the EM density of the dilute WRC studied here , bound between adjacent WRC assemblies in the crystal lattice ( Chen et al . , 2010 ) . Future studies to examine such an interaction and its functional significance may be revealing . The GST-dimer model was not statistically better than the best fits for the WT or C179R WRC ( Figure 5—figure supplement 2B; Figure 5—figure supplement 2C , thin solid curves ) . In summary , our data demonstrate that both the previously identified A Site and the newly identified D Site are bona fide binding sites for Rac1 , the latter with ~40 fold higher affinity than the former . The two sites can act together to recruit the WRC to membranes containing active Rac1 . In retrospect , in our previous equilibrium dialysis experiments , the A Site mutants did not reach saturation at the highest concentrations of Rac1 used ( 8 μM ) , behavior we interpreted as indicating reduced affinity at a single site ( see Figure 3B of [Chen et al . , 2010] ) . Our new data provide a new interpretation for this observation , suggesting that the A Site mutants might have approached an asymptote at ~50% of the level reached by the WT WRC ( i . e . with one site effectively eliminated by the mutation and a second site intact ) . We next asked whether both Rac1-binding sites are important for activating the WRC toward the Arp2/3 complex in pyrene-actin assembly assays ( Figure 6 ) . In these assays , actin assembly ( monitored by an increase in fluorescence of pyrene-labeled actin ) occurs slowly and with a long initial lag in the presence of WRC230VCA and Arp2/3 complex alone ( orange curve in Figure 6A ) . Addition of Rac1 ( Q61L ) -GMPPNP activates the WRC , which , in turn , stimulates the actin-nucleating function of the Arp2/3 complex , decreasing the lag and increasing the maximum rate of actin assembly ( dashed curves in Figure 6A ) . For the WT WRC , a saturating concentration of Rac1 ( 10 μM ) could produce actin-assembly rates equivalent to that of the constitutively active WAVE1 VCA peptide ( dashed cyan curve in Figure 6A ) . Mutating the D Site ( Y967A , Figure 6B , E ) or the A Site ( C179A , Figure 6D , E ) completely abrogated activation , suggesting that activation of the WRC requires binding of Rac1 to both sites . Interestingly , the S969F mutant , which had slightly higher affinity for Rac1 ( Figure 3F ) , was more sensitive to the GTPase than was WT WRC ( Figure 6A , C , E ) . For example , the S969F mutant WRC was fully activated by only 3 μM Rac1 , instead of 10 μM ( green dashed curve in Figure 6C vs . cyan dashed curve in Figure 6A ) . Similarly , 1 μM Rac1 plus the S969F mutant WRC produced actin-assembly rates equivalent to 3 μM Rac1 plus the WT WRC ( magenta dashed curve in Figure 6C vs . green dashed curve in Figure 6A ) . These data suggest that both Rac1-binding sites are important for activation of the WRC . To further examine this notion , we re-analyzed the actin assembly data for Rac1 Q61L/P29S + WRC shown in Figure 1—figure supplement 1C in light of the dissociation constants derived from the EPD assays . As shown by the black data in Figure 6F , actin assembly rate is strongly non-linear with the total concentration of Rac1-bound WRC species , i . e . the sum of both WRC- ( Rac1 ) 1 complexes ( A or D site engaged ) and WRC- ( Rac1 ) 2 . The lack of response to WRC-Rac1 concentrations up to 80 nM suggests suggests that engagement of a single site ( primarily the D Site , due to its high affinity ) is not sufficient to drive WRC activation . Further , plotting actin assembly rate versus the concentration of WRC- ( Rac1 ) 2 yields a linear relationship ( red data in Figure 6F ) . The slope of this line gives a specific activity of ~0 . 22 nM actin/s/nM WRC- ( Rac1 ) 2 . Together , the qualitative and quantitative actin assembly data show that engagement of both the A and D sites is necessary for WRC activation .
The Rho family GTPase , Rac1 , plays important roles in directing signals from upstream pathways to the WRC to promote actin cytoskeletal assembly . We have determined the structure of a WRC-Rac1 complex using cryo-EM , revealing a GTPase-binding site distinct , and physically distant , from the site indicated previously by mutagenesis . Nevertheless , several lines of biochemical data demonstrate that both sites do , in fact , bind Rac1 . Moreover , both are required for activation of the WRC toward the Arp2/3 complex . These findings have several important implications . First , Rac1 activates the WRC through two binding sites that are distinct from the VCA-binding surface of Sra1 . We previously posited that binding to the A Site might trigger conformational changes in the adjacent meander region . These , in turn , could destabilize the meander-VCA contacts , leading to release of the VCA from the body of the WRC ( Chen et al . , 2010 ) . Such a mechanism , involving destabilization of a VCA-containing structural element , would be analogous to the activation of the WAVE1 relatives WASP and N-WASP by the Rho family GTPase Cdc42 ( Kim et al . , 2000; Prehoda et al . , 2000; Torres and Rosen , 2003 ) . Our current data suggest that the A Site and the D Site likely cooperate to trigger conformational changes that release the VCA . These changes may be focused in the meander region , as in our initial hypothesis , or may involve more substantial rearrangements of the WRC body , as suggested by the large distance between the D Site and the VCA-meander element of the assembly . In the latter scenario , the requirement for engagement of Rac1 at both sites to activate the WRC may explain why we did not observe major conformational changes on the WRC in our current structure , in which only the D Site is occupied . It remains to be understood how Rac1 binds the A Site and how the A and D Sites cooperate to drive allosteric activation of the WRC . Second , our data inform on the recent observation that a point mutation , S968F , in Cyfip2 ( corresponding to S969F in Sra1 ) caused phenotypes in mice including decreased response to cocaine and a reduced number of dendritic spines in neurons of the nucleus accumbens ( Kumar et al . , 2013 ) . We found that S969 in Sra1 is buried immediately underneath the D Site . Unlike the previous hypothesis that this mutation may act by destabilizing the WRC , our results show that S969F in Sra1 facilitates Rac1 binding and WRC activation . It is possible that S969 may be part of the pathway for the conformational propagation from the D Site to the VCA-binding pocket upon Rac1 binding . Assuming the S968F mutation in Cyfip2 has a similar biochemical effect , our data suggest that excessive activation of the WRC by Rac1 might be responsible for the observed changes in cocaine response and neuronal morphology . Third , our data depict how Rac1 could work together with other signaling molecules to regulate the WRC on membranes ( Figure 7 ) . One face of the WRC is highly basic , and the opposite face is highly acidic ( Chen et al . , 2010 ) . This charge distribution suggests that when the WRC is bound to the negatively charged inner leaflet of the plasma membrane , and/or specific phospholipids such as PIP3 , its basic surface will lie against the membrane and its acidic surface will face the cytoplasm . In this orientation , both Rac1 binding sites can be positioned to enable a bound GTPase to bury its prenylated tail in the bilayer , consistent with an activation model involving simultaneous engagement of both sites . This orientation is also compatible with other potential interactions . For example , VCA can be readily released from above the membrane to interact with the Arp2/3 complex further in the cytoplasm ( Chazeau et al . , 2014 ) . In addition , WIRS-containing transmembrane receptors could reach the WIRS-binding site on the acidic face of the WRC through < 23 residues ( or ~80 Å in distance; Figure 7A ) ( Chen et al . , 2014a ) . Indeed , with few exceptions , the WIRS elements in previously predicted WIRS-containing receptors in humans are generally located > 25 residues from their transmembrane sequence ( not shown ) ( Chen et al . , 2014a ) . These restraints together suggest a plausible model showing how the WRC may be oriented at membranes to interact with various regulatory molecules . Finally , our bivalent interaction model provides an appealing mechanism to explain how cells could sense the level of active Rac1 at membranes to spatially and temporally control WRC activation and actin polymerization . Our data suggest that WRC activation requires simultaneous engagement of two Rac1 molecules at two distinct binding sites with ~40 fold different affinity . When the density of Rac1 at membranes is low , Rac1 should interact with the WRC preferentially through the high-affinity site , which on its own is not sufficient to trigger activation but primes the complex . The WRC only becomes activated when Rac1 density is sufficient to enable binding to both sites . Consequently , activation of the WRC at membranes is likely non-linear with respect to Rac1 density ( note that technical limitations in working with low Rac1 density on membranes prevented us from examining this experimentally ) . This behavior could be harnessed by cells to determine precisely where and when to turn on actin assembly . Analogous ideas have been suggested previously , based on observations that Rac1 and the Arf GTPase can cooperatively recruit and activate the WRC at membranes ( Koronakis et al . , 2011 ) . Given the low sequence homology between Rac1 and Arf ( 22% and 38% identity in Switch I and Switch II , respectively ) , we feel it is unlikely that either the A or D site that we have characterized is also the binding site for Arf . Further , these cooperative effects could modulate the actions of other WRC ligands . For example , we previously discovered a large array of WIRS-containing receptors , which could potentially recruit the WRC to membranes ( Chen et al . , 2014a ) . For a given receptor , whether the recruitment would lead to productive activation of the WRC and actin polymerization would depend on the local membrane density of active Rac . Other regulators of the WRC , including kinases , acidic phospholipids and scaffolding proteins , might also cooperate with Rac1 activation in similar fashion ( Chen et al . , 2010; Lebensohn and Kirschner , 2009; Miki et al . , 2000; Padrick et al . , 2011 ) . Moreover , the unstructured regions of WAVE and Abi ( including the C terminal SH3 domain of Abi ) , which are not included in our current construct , may make additional contributions to Rac1 binding and activation . These regions have been shown to be important elements for regulating WRC activity , primarily by recruiting other regulatory molecules through interactions mediated by poly-proline sequences and/or SH3 domains ( Chen et al . , 2014c; Dai and Pendergast , 1995; Miki et al . , 2000; Soderling et al . , 2002 ) or by other mechanisms yet to be understood ( Miyamoto et al . , 2008 ) . Together , these effects would lead to precise control of the timing and degree of actin assembly depending on the specific collection of upstream signals and their spatial organization at the membrane .
Similar to mini-WRC , all WRC constructs used in this work contain human Sra1 , Nap1 , HSPC300 ( or EGFP-HSPC300 ) , and Abi2 ( 1-158 ) , in addition to different WAVE1 constructs , including WAVE1 ( 1–230 ) - ( GGS ) 6-Rac1 ( 1-177 ) ( Q61L/P29S ) in ΔWRC230-Rac1 , WAVE1 ( 1–217 ) in ΔWRC217 , WAVE1 ( 1–230 ) in ΔWRC230 , WAVE1 ( 1–217 ) - ( GGS ) 6- ( 485-559 ) in WRC217VCA , and WAVE1 ( 1–230 ) - ( GGS ) 6- ( 485-559 ) in WRC230VCA . The WRCs were expressed and purified essentially as previously described for the mini-WRC ( Chen et al . , 2010; Ismail et al . , 2009 ) , except that the MBP-tagged WAVE1 proteins were expressed in ArcticExpressTM ( DE3 ) RIL cells ( Stratagene ) at 10°C ( Chen et al . , 2014b ) . Based on the following arguments the mutant WRCs all appear to be properly assembled and stable . ( 1 ) All mutations were made to surface residues , except the S967F mutation , which is buried immediately underneath the D Site ( and nevertheless did not negatively affect Rac1 binding/activation ) . Such mutations likely only disrupt the local binding surface , rather than the overall folding of the WRC , especially since the WRC is ~400 kDa in size and is held together through multiple , extensive inter-subunit interactions . ( 2 ) Reconstitution of the recombinant WRC is a multi-step process , involving purification of individual proteins from different host cells and assembly/purification of subcomplexes and ultimately the WRC by a variety of affinity , ion exchange and gel filtration chromatography steps . All the mutant WRCs behaved identically to the wild type WRC during each step of the reconstitution ( see example in Figure 4—figure supplement 1 top panel , and Figure 6—figure supplement 1 , comparing gel filtration chromatography profiles of mutant WRCs to the wild type ) . Furthermore , the C179R and Y967A mutants , each at a distinct Rac1 binding site , behaved identically to the wild type WRC in the MSSV AUC experiments . A mutation that disrupts the overall folding would very likely cause aberrant behaviors during certain steps in reconstitution , leading to lower purity and yield , or complete failure of reconstitution . MBP-tagged mCherry- ( GGS ) 6-Rac1 ( 1-188 ) -His8 ( mCherry-Rac1-His8 for short ) and MBP-tagged mEGFP- ( GGS ) 6-Rac1 ( 1-188 ) ( EGFP-Rac1 for short ) were expressed in BL21 ( DE3 ) T1R cells at 18°C overnight and purified using amylose beads ( New England Biolabs ) . After the MBP tag was removed by TEV-protease cleavage , the protein was further purified by Ni-NTA agarose beads ( Qiagen , Germany ) , followed by cation-exchange chromatography through a Source SP15 column ( GE Healthcare ) . All constructs were verified by DNA sequencing . Other proteins , including Arp2/3 complex , actin , WAVE1 VCA , GST-Rac1 , Rac1 and TEV protease , were purified as previously described ( Ismail et al . , 2009 ) . Prior to the EM experiments , an aliquot of the purified ΔWRC230-Rac1 complex was thawed and passed through a 2 . 4 mL Superdex 200 gel-filtration column ( GE Healthcare ) to remove glycerol from previous purification steps and exchange buffer to 10 mM HEPES pH 7 . 0 , 100 mM NaCl , 1 mM MgCl2 and 2 mM TCEP . The protein sample ( 3 . 5 μl at 0 . 1 mg/mL ) was applied to glow-discharged Quantifoil holey carbon grids ( 400 copper mesh , R1 . 2/1 . 3 ) , which were then flash frozen in liquid ethane using a Gatan CryoPlunge 3 . Grids were imaged with an FEI Tecnai F20 operated at an acceleration voltage of 200 kV and a calibrated magnification of 40 , 410x ( nominal magnification of 29 , 000x ) , yielding a pixel size of 0 . 62 Å on the specimen level . A Gatan K2 Summit direct detector device ( DDD ) camera was used to collect dose-fractionated image stacks in super-resolution counting mode using the UCSF Figure 4 data acquisition software ( Li et al . , 2013 ) . The dose rate used was 6 . 4 e–/Å2/s . Of a total of 2173 image stacks , 30 frames were recorded with 200 ms per frame ( total exposure time of 6 s ) for 1366 stacks , 34 frames were recorded with 300 ms per frame ( total exposure time of 10 . 2 s ) for 407 stacks , and 51 frames were recorded with 200 ms per frame ( total exposure time of 10 . 2 s ) for 400 stacks . The frames were binned over 2 × 2 pixels ( yielding a pixel size of 1 . 24 Å ) , aligned to each other using motioncorr ( Li et al . , 2013 ) , and summed . The defocus parameters of each micrograph were determined with CTFFIND3 ( Mindell and Grigorieff , 2003 ) . Particles from the first 489 images were manually picked using e2boxer ( Tang et al . , 2007 ) and windowed into 220 × 220 pixel images , which were then reduced to 64 × 64 pixel images . The 11 , 874 particle images were subjected to the iterative stable alignment and clustering ( ISAC ) procedure ( Yang et al . , 2012 ) implemented in SPARX ( Hohn et al . , 2007 ) . Three ISAC generations , specifying 200 particles per group and a pixel error threshold of 0 . 7 , resulted in classification of 2607 particles ( ~22% of the entire data set ) into 56 classes . All images were then automatically picked with Relion1 . 3 ( Scheres , 2012 ) , yielding 160 , 591 particles that were windowed into 250 × 250 pixel images . The particle images were subjected to 2D classification specifying 300 classes , and classes producing poor averages were discarded . This step was repeated specifying 200 classes . The remaining 78 , 406 particles were subjected to 3D classification specifying 12 classes and using as initial model the crystal structure of the mini-WRC ( PDB 3P8C ) ( Chen et al . , 2010 ) low-pass filtered to 60 Å . Six classes showing extra density for Rac1 ( 39 , 145 particles ) were combined and subjected to a second round of 3D classification specifying 10 classes . Of the resulting maps , 8 classes showed extra density for Rac1 ( Figure 1—figure supplement 2D ) and were combined ( 29 , 784 particles ) for refinement , yielding a map at a resolution of 7 . 4 Å . For particle polishing in Relion ( Scheres , 2012 ) , image stacks were created for which each frame represented 600 ms of exposure . For instance , for image stacks recorded with 200 ms frames , 3 frames were averaged; for image stacks recorded with 300 ms frames , 2 frames were averaged . Only the first 10 frames , corresponding to an exposure time of 6 s , were used for particle polishing . The final map had a resolution of 7 . 0 Å , as estimated by Fourier shell correlation of independently refined 3D reconstructions from half data sets using the 0 . 143 cut-off criterion ( Figure 1—figure supplement 3B ) . The local resolution was assessed using the program ResMap ( Kucukelbir et al . , 2014 ) ( Figure 1—figure supplement 3C ) . Atomic models of the mini-WRC ( PDB 3P8C ) and Rac1 ( PDB 3SBD ) were docked into the EM map using UCSF Chimera ( Pettersen et al . , 2004 ) . Figures were prepared with UCSF Chimera and Pymol ( Schrodinger , 2015 ) . The EM density map has been deposited in EMDB with accession number EMD-6642 . Non-equilibrium GST pull-down experiments in Figure 3E were performed as previously described ( Chen et al . , 2014a ) . Briefly , 130 pmol of GST-Rac1 ( 1-177 ) -GMPPNP and 260 pmol of WRC ( composed of Sra1 , Nap1 , MBP-EGFP-HSPC300 , MBP-Abi2 ( 1-158 ) and MBP-WAVE1 ( 1–230 ) ) were mixed with 20 μL of Glutathione Sepharose beads ( GE Healthcare ) in 1 mL of binding buffer ( 20 mM HEPES pH 7 , 50 or 100 mM NaCl , 5% ( w/v ) glycerol , 2 mM MgCl2 and 5 mM β-mercaptoethanol ) at 4°C for 30 min , followed by three washes using 1 mL of the binding buffer . Bound proteins were eluted with GST elution buffer ( 100 mM Tris-HCl pH 8 . 5 , 50 mM NaCl , 5% ( w/v ) glycerol , 2 mM MgCl2 , 5 mM β-mercaptoethanol and 30 mM reduced glutathione ) and examined by SDS-PAGE . Figure 3—figure supplement 2A and B used GST-Rac1 ( 1-177 ) -GDP and GST-Rac1 ( 1-177 ) ( Q61L/P29S ) , respectively and the same WRC assemblies as Figure 3E . Equilibrium GST pull-down ( EPD ) experiments were performed essentially as previously described ( Lee et al . , 2000; Pollard , 2010 ) . Glutathione Sepharose beads ( GE Healthcare ) were first equilibrated in EPD buffer ( 20 mM HEPES pH 7 , 100 mM NaCl , 5% ( w/v ) glycerol , 2 mM MgCl2 , and 5 mM β-mercaptoethanol ) and stored as a 50% ( v/v ) slurry . All protein samples were dialyzed against EPD buffer overnight at 4°C to maximize buffer match . After dialysis , GST-Rac1 ( 1-188 ) ( Q61L/P29S ) was concentrated to ~500 μM using an Amicon Ultra centrifuge concentrator ( 3 kDa MWCO , Millipore ) . All proteins were centrifuged at ~21 , 000 g at 4°C for 10 min to remove denatured proteins before use . Each reaction was assembled in 100 μL total volume of EPD buffer in a 200 μL PCR tube ( Axygen ) , which contained 0 . 1 μM ΔWRC230 ( consisting of Sra1 , Nap1 , HSPC300 , MBP-Abi2 ( 1-158 ) and MBP-WAVE1 ( 1–230 ) ) varying concentrations of GST-Rac1 ( 1-188 ) ( Q61L/P29S ) , 30 μL of the Glutathione Sepharose beads ( by aliquoting 60 μL of the 50% ( v/v ) slurry using a wide-bore pipette tip ) , and 0 . 05% Triton X100 to facilitate mixing . The reactions were gently mixed at 4°C on a rotary mixer for 30 min . After a brief centrifugation ( ~10 , 000 g for 30 s ) to pellet the beads , 40 µL of the supernatant was immediately transferred to 8 µL of 6X loading buffer ( 360 mM Tris-HCl pH 6 . 8 , 12% ( w/v ) SDS , 60% ( w/v ) glycerol , 0 . 00012% ( w/v ) bromophenol blue , and 140 mM freshly added 2-mercaptoethanol ) , and analyzed by Coomassie blue-stained SDS-PAGE gels . The gels were imaged by a ChemiDocTMXRS + system ( BioRAD ) using its standard protocols . Total intensity of the Sra1 and Nap1 bands was quantified by ImageJ ( FIJI ) to determine the unbound WRC . The derived fractional occupancy from 2 to 5 independent experiments was directly merged to obtain the binding isotherms . The program DynaFit ( BioKin [Kuzmic , 1996] ) was used to numerically fit the binding isotherms to different equilibrium models to obtain dissociation constants KD ( see Supplementary file 2 for scripts ) . The uncertainty of the derived KDs was further evaluated by Monte Carlo simulations with a ‘shuffle’ algorithm implemented in DynaFit . The final KD values reported in Figure 5 together with the standard errors of the fit were determined by the histograms generated by 5 , 000 Monte Carlo simulations ( Kuzmic , 1996 ) . The fitting results were compared by the F-test using Matlab ( Mathworks ) . Supported lipid bilayers ( SLBs ) containing 98% POPC and 2% Ni2+-NTA DGS ( Avanti Polar lipids ) were prepared as previously described ( Banjade and Rosen , 2014; Su et al . , 2016 ) , with the following modifications . After washing with 5% Hellmanex III ( Hëlma Analytics ) , glass bottom 96-well plates ( Matrical ) were washed with 6 M NaOH for 30 min at 45°C twice , and thoroughly rinsed with MilliQ H2O . Small unilamellar vesicles ( SUVs ) were added to cleaned wells equilibrated with 50 mM HEPES pH 7 . 5 , 150 mM NaCl and 1 mM TCEP , and incubated at 40°C for 1 hr . The membrane recruitment assay was performed at room temperature . The SLBs were first washed twice with binding buffer containing 10 mM HEPES pH 7 . 0 , 100 mM NaCl , 2 mM MgCl2 , 1 mM TCEP and 1 mg/mL BSA . mCherry-Rac1-His8 protein ( either Q61L/P29S or wild type , as indicated in Figure 3F ) was added at 20 nM concentration , and allowed to bind the lipid bilayers for 30 min . Unbound proteins were then removed by three washes , which together afforded a 125-fold dilution of the initial solution . EGFP-tagged WRCs ( composed of Sra1 , Nap1 , MBP-EGFP-HSPC300 , MBP-Abi2 ( 1-158 ) and MBP-WAVE1 ( 1–230 ) ) were then added at 80 nM and incubated for 30 min , followed by three washes affording a 125-fold dilution of the initial solution to reduce non-specific binding of the WRC as well as background fluorescence . TIRF images were taken using a TIRF/iLas2 module ( Biovision ) mounted on a Leica DMI6000 microscope with a 100 × 1 . 49 NA objective ( EM-CCD camera , ImagEMX2 , Hamamatsu ) . Images were acquired at 488 nm for EGFP-WRC and 561 nm for mCherry-Rac1 . Images were processed using ImageJ ( FIJI ) . Signal from the 488 nm channel from samples with only mCherry Rac1 and binding buffer was used as background for EGFP-WRC measurements . Signal from the 561 nm channel from samples with only SLBs and EGFP-WRC was used as background for mCherry-Rac1 measurements . EGFP/mCherry signals were calculated from 10 images from two independent experiments . Data are reported as mean ± SEM . The autofocus module on our microscope was used to find the focal plane for all measurements . The Multi-Signal Sedimentation Velocity ( MSSV ) Analytical Ultracentrifugation ( AUC ) assay was performed as previously described ( Balbo et al . , 2005; Padrick et al . , 2010 ) , to measure the stoichiometry of binding between EGFP-Rac1 ( 1-188 ) ( Q61L/S29S ) and ΔWRC230 ( consisting of Sra1 , Nap1 , HSPC300 , MBP-Abi2 ( 1-158 ) and MBP-WAVE1 ( 1–230 ) ) . Extra care was given when preparing the protein samples to ensure good data quality since the WRC is observed to be sensitive to denaturation , likely caused by liquid surface tension from air bubbles and mechanical shaking , especially in the absence of glycerol in the buffer . The sample preparation required gentle pipetting , avoiding air bubbles or foaming , and removing sticky , denatured WRC floating on the surface after centrifugation steps prior to AUC analyses ( same cautions were applied to all other experiments reported here ) . Both the WRC and EGFP-Rac1 samples were passed through a Hiload Superdex 200 gel filtration column ( GE Healthcare ) equilibrated with 20 mM HEPES pH 7 , 150 mM NaCl , 10 or 20% ( w/v ) glycerol and 1 mM DTT ( and 2 mM MgCl2 for EGFP-Rac1 ) to remove potential aggregates . The proteins were aliquoted , flash frozen in liquid nitrogen , and stored at −80°C . Before use , proteins were thawed in a water bath at room temperature and subjected to extensive dialysis ( for 3 continuous days with multiple buffer exchanges at 4°C ) in the same beaker containing AUC buffer ( 10 mM HEPES pH 7 , 100 mM NaCl , 2 mM MgCl2 , and 1 mM DTT ) to remove glycerol ( which could affect the AUC interference signal ) and to maximize buffer match . After dialysis , the samples were centrifuged at ~21 , 000 g for 10 min at 4°C to remove denatured proteins . Approximately 400 µL of the samples were introduced into the sample sectors of dual-sector Epon centerpieces that had been placed between sapphire windows in a standard AUC cell . The reference sectors were filled with the same volumes of AUC buffer . The cells were inserted into an An50Ti rotor and put under vacuum for temperature equilibration for a minimum of 2 hr . Subsequently , centrifugation was commenced at 50 , 000 rpm . Data were acquired using both absorbance at 490 nm and interference optics . All experiments were performed overnight at 20°C . Interferometric molar signal increments for the two protein species were calculated based on their respective amino-acid compositions ( Zhao et al . , 2011 ) , resulting in 141 , 801 fringes M−1cm−1 for EGFP-Rac1 and 904 , 251 fringes M−1cm−1 for the WRC . Using the former value as a standard , the data sets for EGFP-Rac1 alone were globally analyzed to establish its sedimentation coefficient and the extinction coefficient ( Padrick et al . , 2010 ) at 490 nm ( 32 , 310 . 8 AU M−1cm−1 . This value is lower than the reported value of 56 , 000 AU M−1cm−1 ( Cranfill et al . , 2016 ) , likely due to incomplete maturation of the EGFP tag during overnight expression ( Heim et al . , 1995 ) . A data set for WRC alone was analyzed using only interference data to determine its sedimentation coefficient . The actual concentrations of each protein species in the assembled cells were calculated by integrating the ck ( s ) distributions as described below . The MSSV data were analyzed using SEDPHAT ( Balbo et al . , 2005 ) . Using the extinction coefficients obtained above , the interferometric and absorbance ( at 490 nm ) data from mixtures of the WRC and EGFP-Rac1 were globally analyzed to yield component ck ( s ) distributions that reported on the concentrations of WRC and EGFP-Rac1 as a function of sedimentation coefficient . By integrating these distributions over sedimentation coefficient ranges where co-sedimentation was evident , the concentrations of the individual proteins in the complex could be calculated , and thus the molar ratios of the complexed WRC and EGFP-Rac1 were derived . Sedimentation coefficient values for the individual proteins and complexes were determined by a weighted integration scheme ( Schuck , 2003 ) , yielding signal-average sedimentation coefficients . In combination with the calculated molar ratios and refined frictional ratios ( Padrick et al . , 2010 ) , these values allow the determination of the stoichiometries of the complexes . Actin polymerization assays were performed at 22°C using a PTI Fluorometer ( Photon Technology International ) as previously described ( Chen et al . , 2014a ) . Reactions contained 4 μM actin ( 5% pyrene labeled ) , 10 nM Arp2/3 complex , 100 nM WRC217VCA ( Figure 1—figure supplement 1c; consisting of Sra1 , Nap1 , HSPC300 , Abi2 ( 1-158 ) and WAVE1 ( 1–217 ) - ( GGS ) 6- ( 485-559 ) ) or WRC230VCA ( Figure 6; consisting of Sra1 , Nap1 , HSPC300 , Abi2 ( 1-158 ) and WAVE1 ( 1–230 ) - ( GGS ) 6- ( 485-559 ) ) or VCA , and various amounts of Rac1 Q61L loaded with GMPPNP . | Our cells contain a network of filaments made up of a protein called actin . Just like the skeleton that supports our body , the actin ‘cytoskeleton’ gives a cell its shape and strength . Actin filaments are also critical for many other processes including enabling cells to move and divide . The assembly of actin filaments must be properly controlled so that they are formed at the right time and place within the cell . A complex of proteins known as the WAVE Regulatory Complex ( WRC ) promotes the assembly of actin filaments . The complex contains a region called the VCA , which is able to bind to and activate another protein to make the new actin filaments . The WRC regulates filament assembly by controlling the availability of the VCA in a way that is similar to opening and closing a safe box . When new actin filaments are not needed , the safe box is closed and the VCA is not available . However , when cells need to make new actin filaments , the WRC is opened to release the VCA region so that it is able to bind to the filament-producing protein . Previous studies have shown that a protein called Rac1 acts as a key to open the WRC and trigger actin filament assembly . But it remains unclear how this works . A major obstacle to studying this process is that Rac1 and the WRC only weakly interact with each other , which makes it difficult to capture the interaction under experimental conditions . To overcome this obstacle , Chen et al . tethered a Rac1 molecule to the WRC in order to make the interaction more stable . A technique called cryo-electron microscopy was used to study the three-dimensional shape of this Rac1-WRC complex . Unexpectedly , Rac1 was attached to a different part of the WRC than the site predicted by previous studies . Further experiments showed that Rac1 needs to bind to both of these sites at the same time in order to open the WRC and release VCA , similar to using two keys to open one safe box for increased security . Some cancers , neurological disorders and other diseases can be caused by defects in WRC and Rac1 activity . Therefore , these findings could lead to new ways to treat these conditions in human patients . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | Rac1 GTPase activates the WAVE regulatory complex through two distinct binding sites |
Assessing the imminence of threatening events using environmental cues enables proactive engagement of appropriate avoidance responses . The neural processes employed to anticipate event occurrence depend upon which cue properties are used to formulate predictions . In serial compound stimulus ( SCS ) conditioning in mice , repeated presentations of sequential tone ( CS1 ) and white noise ( CS2 ) auditory stimuli immediately prior to an aversive event ( US ) produces freezing and flight responses to CS1 and CS2 , respectively ( Fadok et al . , 2017 ) . Recent work reported that these responses reflect learned temporal relationships of CS1 and CS2 to the US ( Dong et al . , 2019 ) . However , we find that frequency and sound pressure levels , not temporal proximity to the US , are the key factors underlying SCS-driven conditioned responses . Moreover , white noise elicits greater physiological and behavioral responses than tones even prior to conditioning . Thus , stimulus salience is the primary determinant of behavior in the SCS paradigm , and represents a potential confound in experiments utilizing multiple sensory stimuli .
Learned temporal relationships between cue stimuli and aversive events allow individuals to avoid danger . For example , progressive darkening of clouds often precedes lightning storms , and dark skies prompt evacuation from exposed spaces . Other forms of threat prediction derive not from cue timing or sequencing but rather from the intensity or salience of a stimulus , such as an entrenched soldier who uses relative volume of auditory threat stimuli ( e . g . foreign vehicles or voices ) to gauge proximity of an advancing enemy and determine when to retreat . Human studies indicate that the specific neural circuits engaged during prediction of event occurrence depend on which cognitive strategy is used to solve a particular task ( Breska and Ivry , 2018 ) . Thus , determination of the neural mechanisms that regulate different forms of threat prediction , and the consequences when such mechanisms are dysfunctional , requires behavioral paradigms in which the cognitive processes engaged are clearly defined . In SCS conditioning ( Fadok et al . , 2017 ) , sequential presentation of two different auditory stimuli ( pure tones followed by white noise , in that order ) precedes delivery of an aversive unconditioned stimulus ( US , footshock ) . Following repeated SCS-US presentations , mice exhibit distinct defensive behaviors to each SCS component: tones elicit freezing whereas white noise elicits flight . The paradigm thus appears to model natural behavioral shifts that occur as the perceived probability of directly encountering threat increases . As posited by ‘predatory imminence theory’ , prey animals initially freeze ( to avoid detection ) when predators are present at a distance , but then switch to flight ( escape ) to avoid entrapment if a predator becomes close enough that avoiding detection is no longer possible ( Blanchard and Blanchard , 1989; Blanchard et al . , 1995; Bouton and Bolles , 1980; Fanselow , 1994; Fanselow and Lester , 1988; Perusini and Fanselow , 2015 ) . Given the presence of a specific , repeating sequence of auditory stimuli preceding shock during conditioning , defensive behaviors elicited by individual components of the SCS could in principle be driven by learned CS-US temporal relationships . However , the form of both appetitive and aversive conditioned responses is known to vary substantially according to the particular properties of a given conditioned stimulus , even when the same underlying construct has been learned ( Holland , 1977; Holland , 1979; Holland , 1980 ) . Therefore , differences in the intrinsic properties of tone and white noise stimuli themselves , rather than their temporal relationship to the US , could underlie the distinct behaviors these stimuli evoke during SCS conditioning . To define the key factors responsible for the topography of behavioral responding in this paradigm , we systematically varied CS-US temporal relationships and properties of SCS component stimuli during or following conditioning . We found that when presented at equal sound pressure levels ( SPL ) , white noise elicits greater active defensive behavior than tones , irrespective of stimulus order during conditioning . Following standard tone-white noise SCS conditioning , each stimulus was also capable on its own of evoking either conditioned freezing or flight , according to SPL . Furthermore , when presented at equivalent SPL , white noise promoted greater arousal and simple locomotor responding than pure tone stimuli , even in the absence of any prior conditioning . Together , these data argue that stimulus salience is the major factor determining the form of conditioned responses during SCS conditioning .
We first tested whether reversing the order of 7 . 5 kHz tone ( TN ) and white noise ( WN ) presentation during SCS conditioning reverses the behaviors these stimuli elicit . To distinguish responses due to learned CS-US associations from those due to sensitization or generalization , a control group was included with a 60 s ‘gap’ between the SCS and the US ( Figure 1A–C ) . As evident from the motion traces ( Figure 1D–F ) , all groups exhibited significantly greater motion during WN than TN , irrespective of the order that these stimuli were presented during training ( Figure 1J–L ) . As conditioning progressed , mice in all groups began to exhibit active responses to the WN , including darting and jumping , behaviors quantified using an ‘escape score’ ( Figure 1M–O , see Materials and methods ) . Evidence that Pavlovian conditioning occurred to individual components of the SCS is as follows . First , freezing to the TN differed between paired Group 1 ( G1 ) and gap Group 3 ( G3 ) during conditioning . Freezing to TN was higher in G1 than G3 ( 3-Way ANOVA on Day 2 , G1 vs . G3 , Main Effect of Stimulus ( F ( 1 , 23 ) = 429 . 5 , p<0 . 0001 ) , Main Effect of Trial ( F ( 4 , 92 ) = 5 . 083 , p=0 . 001 ) , Group X Stimulus Interaction , ( F ( 1 , 23 ) = 27 . 51 , p<0 . 0001 ) ; Follow-Up Two-Way RM ANOVA for freezing just to the tone stimulus , Main Effect of Trial ( F ( 3 . 2 , 75 . 9 ) =3 . 79 , p<0 . 05 ) , Main Effect of Group ( F ( 1 , 23 ) = 6 . 41 , p<0 . 05 ) ) . Second , a separate cohort of mice trained on the same protocol were tested for TN-elicited freezing in a novel context ( Figure 1—figure supplement 1 ) . Whereas mice in the gap group ( G3 protocol ) did not show significantly increased freezing between baseline and tone onset ( p>0 . 05 ) , mice in the paired group ( G1 protocol ) exhibited robust acute freezing upon tone onset ( p<0 . 001 ) ( Two-Way RM ANOVA , Main Effect of Stimulus ( F ( 1 , 13 ) = 19 . 98 , p<0 . 001 ) , Stimulus X Group Interaction ( F ( 1 , 13 ) = 5 . 492 , p<0 . 05 ) , Sidak’s comparisons to determine which group drives the Main Effect of Stimulus ) . Third , motion and escape score during WN presentations differed between G1 and G3 during conditioning . Mice in G1 had higher motion to WN than G3 mice ( 3-Way ANOVA on Day 2 , G1 vs . G3 Activity , Main Effect of Stimulus ( F ( 1 , 23 ) = 69 . 89 , p<0 . 0001 ) , Main Effect of Group ( F ( 1 , 23 ) = 11 . 75 , p<0 . 01 ) , Group X Stimulus Interaction ( F ( 1 , 23 ) = 19 . 77 , p<0 . 001 ) ; Follow-up Two-Way RM ANOVA for motion just to WN stimulus , Main Effect of Group ( F ( 1 , 23 ) = 15 . 79 , p<0 . 001 ) ) , and also had higher escape scores to the WN stimulus ( 3-Way ANOVA on Day 2 , G1 vs . G3 Escape Score , Main Effect of Stimulus ( F ( 1 , 23 ) = 67 . 85 , p<0 . 0001 ) , Main Effect of Group ( F ( 1 , 23 ) = 15 . 98 , p<0 . 001 ) , Group X Stimulus Interaction ( F ( 1 , 23 ) = 20 . 41 , p<0 . 001 ) ; Follow-up Two-Way RM ANOVA for escape score just to WN , Main Effect of Group ( F ( 1 , 23 ) = 18 . 25 , p<0 . 001 ) ) . Although Group 2 ( G2 ) did not show significantly different freezing , motion , or escape score compared to G3 ( 3-Way ANOVA on Day 2 , G2 vs G3 , no group differences or interactions for freezing , motion , or escape score ) , G2 did display differential behavior to the two CS stimuli across these same metrics and in the same direction as G1 ( Figure 1E–N; 2-Way ANOVA with Trial and Stimulus as factors; details in Source Data ) . Notably , G1 motion responses to WN on day 2 ( Figure 1D ) were largest immediately following stimulus onset and decreased thereafter until US exposure ( paired t-test , average motion first two vs . last two seconds of CS2 , trials 6 , 7 , p<0 . 01; trials 8 , 9 , p<0 . 05 ) . Thus , imminence in the SCS paradigm does not appear to be determined by a cognitive process that uses cue order or hazard rate , and reversing stimulus order does not reverse behavior . Similar results were observed when these same experiments were performed with C57Bl/6J mice ( Figure 1—figure supplement 2 ) , the strain most comparable to that used in previous studies ( Dong et al . , 2019; Fadok et al . , 2017 ) . Together , these results suggest that threat prediction in the SCS paradigm may be related to intrinsic properties of the auditory stimuli themselves . Mice can hear sounds from 1 kHz to 100 kHz , but sensitivity to specific frequencies varies dramatically over this range . For example , the minimal sound pressure levels ( SPL ) that mice can reliably detect for 16 kHz tones is ~10 x lower ( 10 dB ) than for 7 . 5 kHz tones ( 20 dB ) ( Koay et al . , 2002 ) . Given that the WN stimulus used here and previously ( Dong et al . , 2019; Fadok et al . , 2017 ) is composed of frequencies between 1–20 kHz , one explanation for the above results is that WN stimuli are more efficiently detected and so of higher salience to mice than 7 . 5 kHz tones . To test this idea , we measured physiological and behavioral responses to unconditioned TN and WN stimuli from naive , head-fixed mice on running wheels that had not undergone conditioning of any kind or previously been exposed to these stimuli ( Figure 2A–C ) . Surprisingly , we found that pupil dilation and simple locomotor responses on the running wheel were significantly greater to WN than TN ( Figure 2D–F ) . In addition , comparison of the first three versus last three trials revealed that whereas TN responses habituate with repeated presentations , WN responses do not ( Figure 2G–N ) . Thus , even in the absence of any association with an aversive US , TN and WN differ significantly in the magnitude of the physiological and behavioral responses they elicit . This suggests that TN and WN are differentially salient to mice , which perceive the two stimuli as reflecting distinct points along the threat imminence continuum . A prediction of this model is that a 7 . 5 kHz CS presented at high SPL should be perceived as more imminent and elicit more escape than the exact same CS presented at low SPL . To test this , we performed a ‘SPL step test’ in which mice were presented with a SCS composed of two 7 . 5 kHz tones: CS1 is held constant at 75 dB while CS2 SPL magnitude begins at 55 dB and is stepped up by 5 dB each trial , finishing at 105 dB ( Figure 3A–C ) . While predominantly freezing was observed at ≤85 dB , 7 . 5 kHz tones began to elicit escape behaviors in the paired group when CS2 ≥90 dB ( Figure 3D , F , H ) . Further , escape scores for trials where CS2 ≥90 dB were significantly higher in group 1 ( paired ) than group 3 ( gap; Figure 3H , I ) : 2-Way Repeated Measures ANOVA , Main Effect of Trial ( F ( 4 , 92 ) =3 . 208 , p<0 . 05 ) , Main Effect of Group ( F ( 1 , 23 ) =4 . 613 , p<0 . 05 ) . This argues that group one responses are at least in part influenced by perceived threat levels which are a function of conditioned fear , and are not simply a reflexive reaction to loud sounds . Moreover , escape at later trials was observed in response to CS2 but not CS1 , demonstrating that these behavioral changes were not due solely to enhanced responsivity to any stimulus following repeated US exposure . To determine whether behavioral responses to WN also scale with SPL , we performed a SPL step test using a simple WN CS presented in a novel context ( Figure 3J–L ) . At low SPL ( 40–45 dB ) , WN elicited robust freezing and little to no escape behavior . In contrast , at higher SPL ( ≥60 dB ) , escape responses were common and freezing was minimal during WN presentations ( Figure 3M–O ) . Thus , SCS fear conditioned TN and WN stimuli elicit freezing or flight behavior according to the SPL magnitude at which they are presented . Elicitation of robust escape by SCS conditioned 7 . 5 kHz tones required presentation at ≥90 dB , whereas both paired and unpaired mice began responding actively to WN stimuli at SPL as low as 50 dB . Although these stimuli differ in terms of frequency , they also differ with regards to signal regularity: whereas the 7 . 5 kHz tone is sinusoidal and periodic , WN is random and aperiodic . Therefore , although the above results could reflect differential sensitivity of mice to stimuli of different frequencies , they might alternatively be due to distinct defensive responses triggered by periodic versus aperiodic signals . To test if frequency alone can influence defensive behaviors , we performed fear conditioning using a SCS composed of 3 and 12 kHz pure tones ( Figure 4 ) . These frequencies were chosen as: a ) the threshold SPL in mice is ~100 x lower for 12 kHz than 3 kHz pure tones ( Koay et al . , 2002 ) ; perceived loudness of these two stimuli should thus differ when presented at standard SPL used during conditioning , similar to a 7 . 5 kHz/WN SCS; and b ) 12 kHz is well separated from 17 to 20 kHz , a range that may be innately aversive in mice ( Beckett et al . , 1996; Blanchard et al . , 1992; Cuomo et al . , 1992; Evans et al . , 2018; Mongeau et al . , 2003 ) . As conditioning progressed , paired groups exhibited higher motion , less freezing , and more escape to the 12 kHz than 3 kHz CS , regardless of the order in which the stimuli were presented during training ( Figure 4E–M ) . Thus , despite having no apparent intrinsic aversive valence , 12 kHz tones can elicit greater active threat responses than 3 kHz tones presented at equivalent SPL during SCS conditioning . Freezing and escape behavior in this ‘two-tone’ SCS protocol resulted from Pavlovian Conditioning . Though groups did not differ in freezing behavior to the 3 kHz tone during conditioning , this difference was revealed in a novel context tone test ( Figure 4N–P ) . Elevated motion and escape behaviors to the 12 kHz tone occurred only in the paired groups , indicating that these behaviors are conditioned responses ( 2-Way RM ANOVA , G4 vs . G6 , Motion to 12 kHz Tone: Main Effect of Trial ( F ( 2 . 5 , 45 . 7 ) =3 . 31 , p<0 . 05 ) , Main Effect of Group ( F ( 1 , 18 ) = 8 . 64 , p<0 . 01 ) ; 2-Way RM ANOVA , G4 vs . G6 , Escape score to higher Tone: Main Effect of Trial ( F ( 2 . 7 , 48 . 0 ) =4 . 36 , p<0 . 05 ) , Main Effect of Group ( F ( 1 , 18 ) = 10 . 1 , p<0 . 01 ) ; 2-Way RM ANOVA , G5 vs G6 , Escape score to 12 kHz Tone: Main Effect of Group ( F ( 1 , 18 ) = 4 . 49 , p<0 . 05 ) . As observed for the TN and WN stimuli ( Figure 1 ) , reversing stimulus order reduced the magnitude of the elevated activity and escape to the high-salience stimulus , but did not reverse the behaviors elicited by the two stimuli .
In conclusion , we found that audio frequency properties strongly influence the defensive behaviors elicited by SCS fear conditioned auditory stimuli . Escape behaviors were most potently triggered by stimuli that contain frequencies to which mouse hearing is most sensitive , an effect that was independent of the order in which auditory stimuli were presented during learning . In addition , pure tones that elicit freezing at typical experimental sound pressure levels can promote conditioned escape when presented at higher levels . These data argue that stimulus salience , not temporal proximity to the US , is the primary means by which mice assess imminence and engage appropriate defensive strategies in the SCS paradigm . This would appear to be similar mechanistically to how mice respond to innately threatening visual stimuli , where the probability and intensity of escape behaviors scale with visual stimulus salience ( Evans et al . , 2018 ) . An implication of this work is the critical need to consider the behavioral sensitivity of experimental subjects to auditory stimuli of different frequencies . Psychophysical studies have demonstrated that all species have a particular range of frequencies that they hear well ( i . e . which are audible at 10 dB ) ; stimuli outside of this range may need to presented at substantially higher SPL in order to be efficiently detected . In addition , although most laboratory animals exhibit overlap in their hearing ranges , there can be significant differences in their sensitivity to particular frequencies , even among closely related species . For example , whereas the 10 dB threshold includes frequencies ranging from ~5–40 kHz in rats , this range is very narrow in mice and limited to frequencies close to 16 kHz ( Heffner and Heffner , 2007 ) . Differences can also exist across mouse strains and between different ages of the same strain . For example , C57BL/6J mice undergo hearing-loss induced plasticity that by 5 months of age results in loss of responsivity to high frequency tones ( >20 kHz ) with concomitantly enhanced behavioral sensitivity to middle ( 12–16 kHz ) but not low ( 4–8 kHz ) frequency stimuli ( Carlson and Willott , 1996; Willott et al . , 1994 ) . Moreover , certain frequencies may be innately aversive in rodents: rats emit and respond defensively to alarm vocalizations near 20 kHz ( Beckett et al . , 1996; Blanchard et al . , 1992; Cuomo et al . , 1992 ) , and 17–20 kHz ultrasonic sweeps can elicit robust freezing and flight behaviors in mice ( Evans et al . , 2018; Mongeau et al . , 2003 ) . White noise stimuli , which are both aperiodic and include 17–20 kHz frequencies , may thus be uniquely salient to mice under conditions of impending potential threats due to recruitment of dedicated defensive circuits tuned to innately threatening auditory stimuli . Indeed , in conventional fear conditioning to a simple CS composed of a single auditory stimulus , significantly more flight behavior was evoked by a white noise CS than a tone CS ( Fadok et al . , 2017 ) . Importantly , discrimination studies that employ multiple auditory cues could be complicated both by variations in the ability of subjects to perceive different frequencies as well as potential innate valence associated with certain stimuli . For example , aversively conditioning a high intensity US with a 5 kHz CS+ followed by a generalization test using a higher salience CS- such as white noise could yield misleading conclusions if subjects exhibit escape behaviors to the CS- and , as is common , freezing is the only metric used to assess cue responsivity . Such confounds may be best avoided by assaying discrimination using tasks which measure behavioral responses to distinct patterns of a single , constant intensity sensory stimulus ( e . g . drifting visual gratings of different orientation [Burgess et al . , 2016] ) . Interpretation of discrimination studies that employ auditory stimuli would benefit from counterbalancing assignment of CS+ and CS- stimuli ( Sanford et al . , 2017 ) , and also from use of stimuli at frequencies and SPL that are detectable but do not trigger active fear behaviors . Severe stress can result in persistent generalization or sensitization of threat responding , such that stimuli which normally elicit little to no response come to evoke robust defensive behaviors . For example , in the stress-enhanced fear learning ( SEFL ) model , exposure to inescapable shocks in the absence of auditory stimuli results in nonassociative freezing to white noise in a novel context on the following day ( Perusini et al . , 2016 ) . Although we observed white noise-elicited escape behavior in group 3 ( ‘gap’ , Figure 1F and O ) , these responses cannot be attributed directly to generalization as it remains possible that some CS-US association formed despite the 60 s gap ( i . e . via trace conditioning ) . Thus , the extent to which white noise can nonassociatively elicit active defensive behavior will need to be determined in future experiments where training is performed with a US presented in the complete absence of CS stimuli , as done in the SEFL model . Previous work provided behavioral and neurophysiological evidence that SCS fear conditioned tone and white noise stimuli acutely elicit distinct defensive states indicative of different points along the threat imminence continuum ( Fadok et al . , 2017 ) . We have found that these defensive states track with the frequency and intensity of the conditioned stimuli , not order of CS presentation during learning . This argues that threat imminence in this model is determined primarily via the salience of threat-predictive auditory stimuli which , together with recent experience ( Mongeau et al . , 2003 ) and current fear levels , determines the threshold for switching from freezing to flight . Our results contrast with those of another study ( Dong et al . , 2019 ) , which reported that training with a ‘reversed SCS’ ( white noise-tone-US ) reverses the behaviors elicited ( i . e . mice freeze to the WN but exhibit flight to the tones ) . As the experimental procedures used in both studies were essentially the same , the explanation for the discrepant results is presently unclear . However , while the B6J mice used here were obtained from JAX , the mice used in Dong et al . were of undefined substrain ( ‘C57Bl/6’ ) and obtained from a different vendor . Therefore , it is possible that different mouse strains utilize distinct neural processes to assess threat imminence . Future work will be required to determine if this is indeed the case and if so , the mechanistic underpinnings of such differences . Although reversing the order of the white noise and tone stimuli during training did not qualitatively alter the type of behaviors elicited by the CSs , this switch did have a quantitative effect . Specifically , white noise elicited significantly less escape behavior when it preceded rather than followed the tone during training ( Figure 1 ) . One potential explanation for this result is that compound stimuli which increase in salience from CS1 to CS2 are more naturalistic and produce higher arousal levels and greater learning than the reverse order . Indeed , tonal stimuli which sweep from low up to high frequencies are rated by human observers as more alarming than high to low sweeps ( Catchpole et al . , 2004 ) . Similarly , frequency upsweeps are associated with elevation of attention and arousal , whereas downsweeps are thought to have a calming effect ( Owren and Rendall , 2001 ) . Use of compound stimuli that either increase or decrease in salience from CS1 to CS2 might thus have opposing influences on arousal , resulting in either optimal or suboptimal states for sensory signal processing and learning ( Aston-Jones and Cohen , 2005; McGinley et al . , 2015; Yerkes and Dodson , 1908 ) . Finally , we note that conditioned responses exhibited at the onset of a CS can differ qualitatively from those displayed near CS offset ( Holland , 1980 ) . It thus remains possible that temporal factors make some contribution to defensive responding in SCS conditioning . Given the potent influence of stimulus salience , resolution of this issue will likely require the use of a SCS comprised of distinct component stimuli that can be clearly discriminated and yet are also matched for salience .
Male FVBB6 F1 hybrid mice ( 3–5 months of age , 25–30 g ) were used for all experiments except those in Figure 1—figure supplement 1 , which used C57Bl/6J mice ( JAX ) . All mice were singly housed beginning one week prior to and throughout training and testing , and maintained on a 12 hr reverse light/dark cycle with access to food and water ad libitum . All behavioral tests were conducted during the dark phase , beginning not before one hour of lights OFF and ending not later than one hour before lights ON . Animals were randomly assigned to the experimental groups . The behavioral procedures used in this study were approved by the Institutional Animal Care and Use Committee at Boston Children’s Hospital . Behavioral training used fear conditioning chambers ( 30 × 25×25 cm , Med-Associates , Inc St . Albans , VT ) , equipped with a Med-Associates VideoFreeze system . The boxes were enclosed in larger sound-attenuating chambers . Aspects of the boxes were varied to create two distinct contexts . The pre-exposure and testing context were composed of a white Plexiglas floor insert and a curved white Plexiglas wall insert with a hole over the wall speaker , making the rear walls of the chamber into a semi-circle . The ceiling and front door were composed of clear Plexiglas . The overhead light was off and the box was cleaned with 1% acetic acid . The conditioning context was comprised of a rectangular chamber with aluminum sidewalls and a white Plexiglas rear wall . The grid floor consisted of 16 stainless steel rods ( 4 . 8 mm thick ) spaced 1 . 6 cm apart ( center to center ) . Pans underlying each box were sprayed and cleaned between mice . Fans mounted above each chamber provided background noise ( 65 dB ) . The experimental room was brightly lit with an overhead white light . Animals were kept in a holding room and individually transported to the experimental room in their home cage . Chambers were cleaned with soap and water following each day of behavioral testing . For tone-white noise SCS , three groups of mice were conditioned with compound stimuli consisting of ten pure tone pips ( 7 . 5 KHz , 75 dB , 0 . 5 s duration at 1 Hz ) , ten white noise pips ( WN , 75 dB , 0 . 5 s duration at 1 Hz ) , and a foot shock ( 0 . 9mA , 1 s duration ) . The order and pairing differed between groups: Group one received Tone-WN paired with shock , Group two received WN-Tone paired with shock , and Group three received Tone-WN not directly paired with shock ( i . e . 60 s gap in between CS2 and US ) . All groups had a 3 min baseline period prior to the first CS and 30 s after the final shock . Groups 1 and 2 had a 60 s average pseudorandom ITI ( range 50–90 s ) , while Group 3 had a 180 s average pseudorandom ITI ( range 150–200 ) . For pure tone SCS conditioning , the protocols were the same except that the tone and white noise stimuli were replaced with two pure tone stimuli: 3 KHz ( 75 dB , 10 × 0 . 5 s duration pips at 1 Hz ) and 12 KHz ( 75 dB , 10 × 0 . 5 s duration pips at 1 Hz ) . On the day 0 of both experiments , mice were placed into the pre-exposure context and received four CS-alone trials . On Days 1 and 2 , mice were placed into the conditioning context , where they received five CS trials that included shock . SPL step tests were run as indicated in the figures . Freezing behavior , average motion , and maximum motion were calculated using motion indices determined using automated near infrared ( NIR ) video tracking equipment and computer software ( VideoFreeze , Med-Associates Inc ) , as previously described ( Zelikowsky et al . , 2013 ) . Escape behaviors were scored manually from video files to count the number of darts and jumps . Darts were defined as rapid crossings preceded by immobility; jumps were defined as rapid movements in which all four paws left the floor . These behaviors were summed to determine the number of escape behaviors per mouse per trial , and used to quantify the vigor of responses to particular auditory stimuli via an ‘escape score’ . As most mice were freezing throughout baseline ( BL ) periods on conditioning day 2 ( resulting in a motion index = 0 ) , computation of a ‘flight score’ which compares motion during CS presentation versus BL as a CS/BL ratio ( similar to what was done previously using velocity [Fadok et al . , 2017] ) was problematic due to most ratios having 0 in the denominator . We therefore calculated an ‘escape score’ by taking the difference in average motion index ( MI ) during CS versus the baseline for each trial ( i . e . the 10 s period preceding delivery of a CS ) , dividing this by 100 , and then adding one point for each dart or two points for each jump observed during that particular stimulus and trial: escape score = ( MICS – MIBL ) /100 + 1 ( for each dart ) + 2 ( for each jump ) . Mice with stainless steel head posts were head-fixed on a running wheel , and pupils illuminated with an infrared LED and imaging with a FLIR Flea3 USB 3 . 0 camera at 30fps . Importantly , mice used for these experiments had not previously received any type of conditioning nor been exposed to either tone or white noise stimuli . To extract pupil diameter traces , the pupil was thresholded and binarized in Bonsai 2 . 3 using a custom workflow ( OpenCV ) . The resulting image was dilated and eroded to remove noise from the pupil edge , and the largest radius of the oval is extracted as pupil diameter . Blinks were removed in MATLAB . Following habituation to head-fixation on the wheel for three days ( 10 min per day ) , mice were exposed to ten trials of the Tone-WN stimuli alone; the following day they received ten trials of the WN-Tone stimuli alone . To minimize the influence of ‘ceiling effects’ , trials were excluded when pupil diameter exceeded the mouse’s own 50th percentile in the 5 s prior to stimulus onset . All velocity traces were included . Data were analyzed with t-tests or two-way repeated-measures ANOVAs , with Sidak post hoc analysis correcting for multiple comparisons where appropriate . Sample size was pre-determined from previously published research and from pilot experiments performed in the laboratory . Experiments in Figure 1 were replicated two ( groups 2 and 3 ) or three ( group 1 ) times using separate cohorts of animals . Experiments in Figure 2 were replicated twice using separate groups of animals . Experiments in Figures 3 and 4 were performed once . Experiments in Figure 1—figure supplement 1 were performed once . Experiments in Figure 1—figure supplement 2 were replicated twice with separate cohorts of animals . In all instances , these were ‘biological replicates’ ( i . e . different mice for each experiment ) . Lab personnel were blind to experimental group during scoring . Statistical significance is labeled as *p<0 . 05 , **p<0 . 01 , and ***p<0 . 001 . | If you notice the skies above you becoming darker , your first thought might be to seek shelter . Experience will have taught you that darkening skies are often a sign of an approaching storm . Learning to recognise changes that occur prior to an unpleasant event can help us avoid danger . But this is not the only strategy people can use to predict when something bad is about to happen . Another option is to use the intensity , or salience , of sensory information . Soldiers fighting on the front line , for example , might rely on the loudness of enemy voices or vehicles to judge how close an advancing enemy is . This information will help them decide when to retreat . Different brain processes are active when individuals use each of these two strategies to predict when an upcoming event will occur . One approach to study these processes is to use a technique called “SCS conditioning” . This involves exposing mice to two sounds , followed by a mild electric shock administered to the feet . The first sound is a pure tone; the second is a burst of white noise . After repeated trials , mice begin to show distinct responses to the two sounds . They freeze in response to the tone but run away upon hearing the white noise . These responses parallel behaviors seen in the wild . When mice detect a distant predator , they freeze to avoid detection . But if the predator comes too close for the mice to avoid being spotted , they instead try to flee . Some have argued that in the SCS task , mice learn that the white noise predicts an imminent shock . The mice therefore flee as soon as they hear it . By contrast , they learn that the tone predicts a delayed shock and therefore choose to freeze instead . However , by tweaking the SCS procedure , Hersman et al . now show that even if the white noise occurs before the tone , it is still more likely than the tone to trigger an escape response . In fact , mice are more reactive to white noise than tones even if the sounds are never paired with shocks . This suggests that mice find white noise naturally more noticeable than tones . Moreover , Hersman et al . show that tones can also trigger escape responses if they are sufficiently intense . Together these results suggest that mice use the intensity of the stimuli – rather than the length of time between each stimulus and the shock – to decide whether to freeze or flee . People with anxiety disorders often show exaggerated responses to things that do not pose a genuine threat . At present the pathways in the brain that are responsible for these excessive reactions are unclear . The results of Hersman et al . will aid research into the brain circuits that detect , assess and respond to threats . Understanding these circuits could in the future lead to better treatments for anxiety disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Stimulus salience determines defensive behaviors elicited by aversively conditioned serial compound auditory stimuli |
The sense of smell enables animals to react to long-distance cues according to learned and innate valences . Here , we have mapped with electron microscopy the complete wiring diagram of the Drosophila larval antennal lobe , an olfactory neuropil similar to the vertebrate olfactory bulb . We found a canonical circuit with uniglomerular projection neurons ( uPNs ) relaying gain-controlled ORN activity to the mushroom body and the lateral horn . A second , parallel circuit with multiglomerular projection neurons ( mPNs ) and hierarchically connected local neurons ( LNs ) selectively integrates multiple ORN signals already at the first synapse . LN-LN synaptic connections putatively implement a bistable gain control mechanism that either computes odor saliency through panglomerular inhibition , or allows some glomeruli to respond to faint aversive odors in the presence of strong appetitive odors . This complete wiring diagram will support experimental and theoretical studies towards bridging the gap between circuits and behavior .
An animal uses its sense of smell to navigate odor gradients , and to detect the threat or reward associated with an odor . In the nervous system , odorants are detected by olfactory receptor neurons ( ORNs ) whose axons organize centrally into glomeruli by olfactory receptor type ( Wang et al . , 1998; Vosshall et al . , 2000 ) . Uni- and multi-glomerular projection neurons ( PNs ) relay olfactory inputs to higher-order brain areas ( Stocker et al . , 1990; Liang et al . , 2013 ) . Common between mammals and insects ( Vosshall and Stocker , 2007; Su et al . , 2009 ) , PNs target two major brain centers , one associated with learning and memory ( such as the mushroom bodies ( MB ) in insects ) , and another that mediates some innate behaviors ( such as the lateral horn ( LH ) in insects ) ( Fischbach and Heisenberg , 1984; Heisenberg et al . , 1985; Stocker et al . , 1990; Sosulski et al . , 2011 ) . Local neurons ( LNs ) mediate communication between glomeruli , implementing computations such as gain control ( Olsen and Wilson , 2008 ) . While the connectivity of a few glomeruli has been recently partially reconstructed in the adult fly ( Rybak et al . , 2016 ) , the complete number and morphology of cell types and the circuit structure with synaptic resolution is not known for any glomerular olfactory system . In the Drosophila larva , we find a similarly organized glomerular olfactory system of minimal numerical complexity ( Figure 1a ) . In this tractable system , each glomerulus is defined by a single , uniquely identifiable ORN ( Fishilevich et al . , 2005; Masuda-Nakagawa et al . , 2009 ) , and almost all neurons throughout the nervous system are expected to be uniquely identifiable and stereotyped ( Manning et al . , 2012; Vogelstein et al . , 2014; Li et al . , 2014; Ohyama et al . , 2015 ) . Some of the olfactory LNs and PNs have already been identified ( Masuda-Nakagawa et al . , 2009; Thum et al . , 2011; Das et al . , 2013 ) . This minimal glomerular olfactory system exhibits the general capabilities of the more numerically complex systems . For example , as in other organisms ( Friedrich and Korsching , 1997; Nagayama et al . , 2004 ) and in the adult fly ( Bhandawat et al . , 2007; Nagel and Wilson , 2011; Kim et al . , 2015 ) , the output of the uniglomerular PNs tracks the ORN response ( Asahina et al . , 2009 ) , which represents both the first derivative of the odorant concentration and the time course of the odorant concentration itself ( Schulze et al . , 2015 ) . Like in the adult fly ( Olsen and Wilson , 2008 ) and zebrafish ( Zhu et al . , 2013 ) , gain control permits the larval olfactory system to operate over a wide range of odorant concentrations ( Asahina et al . , 2009 ) . The olfactory behaviors exhibited by the larva have been well studied ( mostly in 2nd and 3rd instar larvae ) , in particular chemotaxis ( Cobb , 1999; Bellmann et al . , 2010; Gomez-Marin et al . , 2011; Gershow et al . , 2012; Schulze et al . , 2015; Gepner et al . , 2015; Hernandez-Nunez et al . , 2015 ) , as well as the odor tuning and physiological responses of ORNs ( Fishilevich et al . , 2005; Louis et al . , 2008; Asahina et al . , 2009; Kreher et al . , 2008; Montague et al . , 2011; Mathew et al . , 2013 ) . Additionally the larva presents odor associative learning ( Gerber and Stocker , 2007 ) . Obtaining the wiring diagram of all neurons synaptically connected to the ORNs would enable the formulation of system-level hypotheses of olfactory circuit function to explain the observed behavioral and functional properties . The reduced numerical complexity and dimensions of the larval olfactory system , the similarity of its organization and capabilities to other organisms , and the tractability of the larva as a transparent genetic model organism , make it an ideal model system in which to study the complete circuit architecture of a glomerularly organized olfactory processing center . 10 . 7554/eLife . 14859 . 003Figure 1 . Overview of the wiring diagram of the glomerular olfactory system of the larval Drosophila . ( a ) Schematic of the olfactory system of the larval Drosophila with EM-reconstructed skeletons overlaid . The ORN cell bodies are housed in the dorsal organ ganglion , extend dendrites into the dome of the dorsal organ , and emit axons to the brain via the antennal nerve . Like in all insects , neuron cell bodies ( circles ) reside in the outer layer of the nervous system ( grey ) , and project their arbors into the neuropil ( white ) where they form synapses . Also shown are the major classes of local neurons ( Broad LNs , Picky LNs and Keystone ) and the 2 classes of projection neurons ( uPNs and mPNs ) . The arbors of the Broad LNs ( black ) specifically innervate the AL . LNs and mPN dendrites can extend into the subesophageal zone ( SEZ ) , innervated by sensory neurons of other modalities . uPNs project to specific brain areas ( mushroom body calyx and lateral horn; LH ) , and mPNs mostly project to other nearby brain areas . ( b ) The larva presents 21 unique olfactory glomeruli , each defined by a single ORN expressing a single or a unique pair of olfactory receptors . We reconstructed each ORN with a skeleton and annotated its synapses , here colored like the skeleton to better illustrate each glomerulus . See Figure 1—figure supplement 1 for individual renderings that aided in the identification of each unique ORN . ( c ) Summary connectivity table for the right antennal lobe with all major neuron classes ( 4 neuromodulatory neurons and the descending neuron from the brain were omitted ) , indicating the percent of postsynaptic sites of a column neuron contributed by a row neuron . For most neurons , the vast majority of their inputs originates in other neurons within the antennal lobe . In parentheses , the number of neurons that belong to each cell type . We show only connections with at least two synapses , consistently found among homologous identified neurons in both the left and right antennal lobes . Percentages between 0 and 0 . 5 are rounded down to 0 . ( d ) Schematic of the innervation patterns of the main classes of LNs and PNs in the antennal lobe . White ovals represent the glomeruli . Solid circles are cell bodies . Shaded areas with dotted outlines represent the extent of the PN dendritic arbors , with each uPN ( green ) innervating one glomerulus and each mPN ( blue ) innervating multiple glomeruli . Their axons ( arrows ) project to the brain . Broad LNs ( black ) are axonless and present panglomerular arbors . Picky LN ( orange ) dendrites span multiple glomeruli and their axons ( arrow; not shown ) target a different yet overlapping set of glomeruli as well as regions outside the olfactory system . Choosy LNs are similar to the Picky LNs but their axons remain within the antennal lobe . ( e ) A simplified wiring diagram of the larval olfactory system with only the main connections . ORNs are excitatory . All shown LNs are inhibitory . Broad LNs reciprocally connect to all glomeruli and each other and thus engage in presynaptic inhibition ( on ORNs ) and postsynaptic inhibition ( on uPNs ) . Picky LNs form a hierarchical circuit and selectively synapse onto mPNs . Another LN , Keystone , receives inputs from ORNs , one Picky LN and non-ORN sensory neurons , and can potentially alter the operational mode of the entire olfactory system by altering the pattern of inhibition ( see text ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 00310 . 7554/eLife . 14859 . 004Figure 1—figure supplement 1 . Electron microscopy view of the antennal lobe of Drosophila larva . ( A ) Cross-section of the antennal lobe , with dorsal up and lateral to the left . Notice most of the antennal lobe is wrapped by glial cells ( darker profiles ) , but these do not fully enclose it ( not shown ) , opening towards the SEZ ( bottom right ) . Individual glomeruli are not wrapped in glia like in other insects ( Oland et al . , 2008 ) , but glia do separate the antennal lobe from the neuronal somas and from nearby neuropils . ALT , antennal lobe tract . Corresponds to serial section 648 in the online volume , at bottom left ( right antennal lobe from anterior view ) . ( B ) Magnification of the box in ( A ) , showing an axon terminal of the 24a ORN synapsing onto the 24a uPN and multiple LNs . The dendrite of the 24a PN contains mitochondria , vesicles and presynaptic sites ( magenta arrows ) , synapsing in this section onto e . g . one of the two Keystone LNs among others . The ORN axon bouton hosts multiple synapses ( red arrows ) with prominent ribbons or T-bars . The ORN boutons are packed with vesicles , giving them a darker appearance than surrounding PN and LN neurites; also contain numerous mitochondria ( not shown in this image ) . Only some LNs are labeled for clarity; all neurites in this image were reconstructed . Blue arrows point to synapses of labeled LNs . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 00410 . 7554/eLife . 14859 . 005Figure 1—figure supplement 2 . A single , identified ORN for each glomerulus in the antennal lobe of the first instar larva . Each panel shows an EM-reconstructed arbor of an ORN ( colored ) over the background of a Broad LN Duet ( grey ) . ORN synapses are rendered in the same color as the skeleton . To the left , all ORNs of each half of the antennal lobe are rendered together . The orientation ( lateral to the left , dorsal up ) and relative position of each ORN has been chosen to exactly match the arrangement in the supplementary Figure 1 of Masuda-Nakagawa et al . ( 2009 ) , where each individual ORN was identified and labeled with GFP using genetic driver lines . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 00510 . 7554/eLife . 14859 . 006Figure 1—figure supplement 3 . EM-reconstructed arbors of all LNs . Left dorsal and posterior views of all LNs . The bundling of their primary axon tracts suggests that all LNs derive from 5 neuronal lineages ( 5 on the left and 5 on the right ) , shown in 5 different colors . Right , renderings of the left antennal lobe , posterior view . These identified neurons present similar morphology and connectivity in the right antennal lobe . Broad LNs are shown for reference . The morphology , cell body position and number of Choosy LNs matches that of the pair of GABAergic LNs described in Figure 2 L–O of Thum et al . , 2011 . Scale bar: a cell body measures about 4 micrometers in diameter . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 006 We reconstructed from electron microscopy all synaptic partners of the 21 ORNs for both the left and right antennal lobes of a first instar larva ( Figure 1b; Figure 1—figure supplement 1 and 2 ) . Per side , we found 21 uniglomerular PNs ( uPNs; one per glomerulus ) , 14 LNs , 14 multiglomerular PNs ( mPNs ) , 4 neuromodulatory neurons , 6 subesophageal zone ( SEZ ) interneurons and 1 descending neuron ( Figure 1c , d ) . These identified neurons present stereotyped connectivity when comparing the left and right antennal lobes . The lack of undifferentiated neurons in the 1st instar antennal lobe , and comparisons with light-microscopy images of other instars suggests that the 1st instar antennal lobe contains all the neurons present throughout larval life . Here , we analyze this complete wiring diagram on the basis of the known function of circuit motifs in the adult fly and other organisms and known physiological properties and behavioral roles of identified larval neurons . We found two distinct circuit architectures structured around the two types of PNs: a uniglomerular system where each glomerulus participates in a repeated , canonical circuit , centered on its uPN ( Python and Stocker , 2002 ) ; and a multiglomerular system where all glomeruli are embedded in structured , hetereogeneous circuits read out by mPNs ( Figure 1e; [Das et al . , 2013] ) . We also found that the inhibitory LNs structure a circuit that putatively implements a bistable inhibitory system . One state could compute odor saliency through panglomerular lateral inhibition , that is , by suppressing the less active glomeruli in favor of the more active ones . The other may enable select glomeruli , specialized for aversive odors , to respond to faint stimuli in a background of high , appetitive odor stimuli . We discuss the role of these two possible operational states and how neuromodulatory neurons and brain feedback neurons participate in the interglomerular circuits .
We mapped the wiring diagram of the first olfactory neuropil of the larva by reconstructing the left and right ORNs and all their synaptic partners . We used a complete volume of the central nervous system ( CNS ) of a first instar larva , imaged with serial section electron microscopy ( [Ohyama et al . , 2015]; see Materials and methods for online image data availability; Figure 1—figure supplement 1 ) . We reconstructed 160 neuronal arbors using the software CATMAID ( Saalfeld et al . , 2009; Schneider-Mizell et al . , 2016 ) . All together , the 160 neurons add up to a total of 38 , 684 postsynaptic sites and 55 millimeters of cable , requiring about 600 , 000 mouse clicks over 736 hr of reconstruction and 431 hr of proofreading . Only 136 of 14 , 346 ( 0 . 9% ) postsynaptic sites of ORNs remained as small arbor fragments ( comprising a total of 0 . 25 millimeters of cable , or 0 . 5% of the total reconstructed ) that could not be assigned to any neuron . We sorted the 160 reconstructed neurons into 78 pairs of bilaterally homologous neurons and 4 ventral unpaired medial ( VUM ) neurons ( 2 are mPNs and 2 are octopaminergic 'tdc' neurons; [Selcho et al . , 2014] ) . These 78 pairs we further subdivided into 21 pairs of ORNs , 21 pairs of uPNs , 13 pairs of mPNs ( plus 2 additional VUM mPNs ) , 14 pairs of LNs , 6 pairs of neurons projecting to the SEZ ( 'SEZ neurons' ) , 1 pair of descending neurons from the brain , 1 pair of serotonergic neurons ( CSD; [Roy et al . , 2007] ) , and 1 pair of octopaminergic non-VUM neurons ( 'lAL-1'; [Selcho et al . , 2014] ) . The 14 pairs of LNs originate in 5 different lineages ( Figure 1—figure supplement 3 ) . We assigned the same name to neurons of the same lineage , and numbered each when there is more than one per lineage . LNs connect to other neuron classes stereotypically in the two antennal lobes ( Figure 2 ) . We selected names reminiscent of either their circuit role or anatomical feature , including 'Broad' to refer to panglomerular arbors; 'Picky' and 'Choosy' for LNs of two different lineages ( and different neurotransmitter; see below ) with arbors innervating select subsets of glomeruli; 'Keystone' for a single pair that mediate interactions between LNs of different circuits; and 'Ventral LN' for a single pair of LNs with ventral cell bodies . We also determined the neurotransmitters of LNs that were previously unknown ( Figure 2—figure supplement 1 ) . We introduce the properties of each LN type below with the olfactory circuits that they participate in . 10 . 7554/eLife . 14859 . 007Figure 2 . Percentage of synapses of LNs from/onto specific cell types . The entry for each neuron presents two bars , for the left and right homologs . Top row , Broad LNs and Keystone . T: Trio , D: Duet . Left , differences between the Trio and Duet subtypes are evident in the fraction of inputs that originates in ORNs , uPNs and Keystone . The Duet subtype presents a far larger fraction of its inputs from ORNs , and barely receives any inputs from Keystone . By its pattern of inputs , Keystone resembles a Broad LN Trio neuron , except for the large fraction of non-ORN inputs and the inputs from Picky LNs ( specifically from Picky LN 0 ) . Right , note how the Trio subtype devote about 25% of their synapses to each other , whereas the Duet subtype preferrentially targets uPNs , providing postsynaptic inhibition to the glomeruli ( both lateral and feedforward inhibition ) . Keystone differs from the Broad LNs in that it targets uPNs much more weakly , preferring instead the Broad LN Trio and a variety of other neurons . Bottom row , Picky LNs , Choosy LNs and Ventral LN . Left , the fraction of inputs from ORNs stands out as a large difference among Picky LNs , with Picky LN 3 and 4 receiving substantially fewer , similarly to Choosy LNs . The fraction of inputs received from other Picky LNs ( green ) is among the most distinguishing feature of Picky LN 0 , which receives close to none . Right , in contrast to the similar patterns of inputs onto all Picky LNs , Picky LN 0 stands out as very different from other Picky LNs in its choice of downstream synaptic partners , spreading approximately evenly between ORNs , uPNs , mPNs , other Picky LNs and Keystone . Choosy LNs strongly prefer uPNs , being therefore strong providers of postsynaptic inhibition to glomeruli . Notice that Picky LNs , Choosy LNs and Ventral LN have a larger fraction of synapses to/from 'others' , with their arbors spreading towards adjacent sensory neuropils in the SEZ . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 00710 . 7554/eLife . 14859 . 008Figure 2—figure supplement 1 . Neurotransmitters of Keystone LN and Picky LNs . Genetic driver lines specific for Keystone LN ( GAL4 line GMR27F08 ) and Picky LNs ( split-GAL4 lines JRC_SS04499 , JRC_SS04500 , JRC_SS04260 ) driving GFP expression specifically in these neurons were labeled with anti-GABA and anti-vGlut ( A–U ) , and also anti-Chat ( all negative; not shown ) . Keystone presents immunoreactivity to anti-GABA ( textbf A–C ) , and at least 4 of the 5 Picky LNs are positive to anti-vGlut and negative to anti-GABA ( D–U ) . These neurons derive from the BAla2 lineage ( Das et al . , 2013 ) . JRC_SS04260 drives expression specifically and uniquely a Picky LN , likely Picky LN 4 , which presents anti-vGLut immunoreactivity ( P–R ) . Left unlettered panels show the homologous identified EM-reconstructed neurons , with Broad LNs in grey for reference . Asterisks mark the location of cell bodies when there is not labeling , such as in panels I , O and U . Broad LNs and Choosy LNs are GABAergic ( see Thum et al . , 2011 at Figure 2 D–G for Broad LNs and L–O for Choosy LNs ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 008 In vertebrates and most arthropods , olfactory glomeruli are defined by a group of same-receptor ORNs converging onto a set of glomerular-specific PNs ( mitral and tufted cells in the mouse and zebrafish olfactory bulb ) ( Stocker et al . , 1990; Satou , 1992; Ressler et al . , 1994; Wang et al . , 1998; Distler and Boeckh , 1996 ) . In Drosophila larva , this system is reduced to a single ORN and a single uPN per glomerulus ( Python and Stocker , 2002; Ramaekers et al . , 2005; Masuda-Nakagawa et al . , 2009 ) . With one exception ( 35a , which has 2 bilateral uPNs ) , our EM-reconstructed wiring diagram is in complete agreement with these findings ( Figures 1b , 3a ) . Most of the larval uPNs project to both the MB and the LH ( Figure 3a ) , like in the adult fly ( Stocker et al . , 1990; Luo et al . , 2010 ) . 10 . 7554/eLife . 14859 . 009Figure 3 . The uniglomerular circuit consists of 21 glomerular-specific projection neurons , which interact primarily with their corresponding ORN and with the 5 panglomerular LNs ( Broad LNs ) , each an identified neuron . ( a ) Posterior view of the EM-reconstructed uPNs of the right antennal lobe . The dendrites of each uPN delineate the glomerular boundaries , and the axons project to both the mushroom body ( required for learning and memory; [Michels et al . , 2011] ) and the lateral horn ( implicated in innate behaviors; [de Belle and Heisenberg , 1994] ) . 19 uPNs are likely generated by the same neuroblast lineage BAmv3 ( Das et al . , 2013 ) ( although the uPNs for 42a , 74a , and 67b are slightly separated from the rest ) , and the other two ( the uPNs for 33a and 35a ) clearly derive from two other neuroblasts . Notice that the 35a uPN is bilateral , ascends through a different tract , and receives additional inputs outside of the antennal lobe . The 33a uPN does not synapse within the calyx and the 82a uPN does not continue to the lateral horn . The left antennal lobe ( not shown ) is a mirror image of the right one . ( b ) Dorsal view of the EM-reconstructed , axonless Broad LNs ( Duet in orange; Trio in blue ) shown together and individually . All neurons are on the same lineage: BAlc ( Das et al . , 2013 ) . The pre- ( red ) and post- ( cyan ) synaptic sites on these panglomerular neurons are fairly uniformly distributed . ORNs in grey for reference . These neurons extend posteriorly out of the olfactory glomeruli to receive synapses from 2 non-ORN sensory neurons that enter the brain via the antennal nerve . ( c ) Percentage of the total number of postsynaptic sites on the dendrite of an uPN , Broad LN or Choosy LN ( columns ) that originate in a given ORN or LN ( rows ) for the right antennal lobe . Since the 35a uPN is bilateral , we include inputs to it from both antennal lobes . We show only connections with at least two synapses , consistently found among homologous identified neurons in both the left and right antennal lobes . Percentages between 0 and 1 are rounded to 1 , but totals are computed from raw numbers . The uniglomerular nature of uPNs ( notice the green diagonal ) and panglomerular nature of Broad LNs is evident . The Broad LN Duet generally contributes more synapses onto uPNs than the Broad LN Trio does . While the number of synapses that an ORN makes onto its uPN varies widely ( 24–120 synapses; see Supplementary file 1 and 2 ) , this number is tailored to the size of the target uPN dendrite given that percentage of inputs the ORN contributes to the uPN is much less varied ( mostly 45–65% ) . For an extended version of this table that includes all LNs , see Figure 3—figure supplement 1 . ( d ) Both Broad LN types ( Trio and Duet ) mediate presynaptic inhibition ( synapses onto ORN axons ) similarly , but the Duet shows far stronger postsynaptic inhibition ( synapses onto uPN dendrites ) while the Trio receives far more dendro-dendritic synapses from uPNs . Connections among Broad LNs are not shown for simplicity . Each arrow is weighted linearly by the number of synapses for an average single Broad LN of each type . ( e ) The 5 Broad LNs that govern this circuit synapse reciprocally , with the Trio type synapsing more strongly onto each other . Shown here for the right antennal lobe with arrow thickness weighted by the square root of the number of synapses . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 00910 . 7554/eLife . 14859 . 010Figure 3—figure supplement 1 . Extended version of table in Figure 3c , including all other olfactory-related neurons . Tables show percent of postsynaptic sites of a column neuron contributed by a row neuron . We show only connections with at least two synapses , consistently found among homologous identified neurons in both the left and right antennal lobes . Percentages between 0 and 0 . 5 are removed . For bilateral neurons , inputs from both sides are included . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 01010 . 7554/eLife . 14859 . 011Figure 3—figure supplement 2 . Distribution of postsynaptic sites on the uPN dendrites . We show 5 examples , plotting the distance ( along the cable ) of individual postsynaptic sites ( colored dots ) to the axon initial segment of each uPN . The same type of presynaptic neuron presents the same color across all plots . Notice how Choosy LN inputs ( red , framed in a red box ) onto uPNs are generally more proximal to the axon initial segment than other inhibitory inputs such as from Broad LNs; particularly noticeable for 42a PN ( top row ) and 94 & 94b PN ( bottom row ) . No noticeable difference exists between Broad LN Duet and Trio . Notice that the left 49a PN presents an arbor with two main dendrites , with one being further than the other from the axon initial segment , explaining the split in the distribution of distances of postsynaptic sites . While 67b PN ( third row ) does not receive inputs from Choosy LNs , the Picky LN 3 ( light green ) , which specifically targets 67b PN and no other uPN , provides proximal inputs . Presynaptic neurons are ordered with the largest contributor at the bottom of each plot . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 011 In insects ( adult fly , bee , locust ) and in vertebrates , the excitation of glomeruli is under control of inhibitory LNs that mediate functions such as gain control , which define an expanded dynamic range of uPN responses to odors ( Sachse and Galizia , 2002; Lei et al . , 2002; Olsen and Wilson , 2008; Olsen et al . , 2010; Zhu et al . , 2013 ) . We found that most non-sensory inputs to the larval uPNs ( Figure 1c ) are from a set of 5 panglomerular , axonless , and GABAergic ( Thum et al . , 2011 ) neurons that we named Broad LNs ( Figure 3b , c; Figure 3—figure supplement 1 ) . These 5 Broad LNs also account for most inputs onto the ORN axons ( Figure 1c ) , therefore being prime candidates for mediating both intra- and interglomerular presynaptic inhibition ( onto ORNs ) as observed in the adult fly with morphologically equivalent cells ( Wilson and Laurent , 2005; Olsen and Wilson , 2008; Olsen et al . , 2010; Chou et al . , 2010 ) , and in the larva ( Asahina et al . , 2009 ) . We divided the 5 Broad LNs into two classes , Trio and Duet , based on the number of neurons of each type ( Figure 3b ) . While both types provide panglomerular presynaptic inhibition ( onto ORN axons ) , the Duet makes far more synapses onto the dendrites of the uPNs . This may indicate a far stronger role for the Broad LN Duet in postsynaptic inhibition ( onto uPN dendrites; Figure 3d ) . In the adult fly , presynaptic inhibition implements gain control ( Olsen and Wilson , 2008 ) , and postsynaptic inhibition plays a role in uPNs responding to the change in ORN activity ( Nagel et al . , 2015; Kim et al . , 2015 ) . The two types of glomerular inhibition are provided by two separate cell types , and may therefore be modulated independently . For example , the uPNs emit dendritic outputs that primarily target the Broad LN Trio ( Figure 3d ) , indicating that the output of the glomerulus contributes more to presynaptic than to postsynaptic inhibition . Similar excitatory synapses from uPNs to inhibitory LNs have been shown in vertebrates ( Rall et al . , 1966 ) , and synapses from PNs to LNs and vice versa have been described in the adult fly ( Rybak et al . , 2016 ) . Beyond their role in pre- and postsynaptic inhibition of ORNs and uPNs respectively , the Broad LNs synapse onto all neurons of the system , including other LNs and mPNs ( Figure 1c , Figure 2 ) . Therefore Broad LNs may be defining a specific dynamic range for the entire antennal lobe , enabling the system to remain responsive to changes in odorant intensities within a wide range . Importantly , Broad LNs also synapse onto each other ( Figure 3c , e ) like in the adult ( Okada et al . , 2009; Rybak et al . , 2016 ) . Furthermore , the two types of Broad LNs have a different ratio of excitation and inhibition , originating in the preference of Trio to synapse far more often onto each other than onto Duet ( Figure 3e , Figure 2 ) . This suggests that the two types not only have different circuit roles but also have different properties . Another GABAergic cell type , that we call the Choosy LNs ( two neurons; Figures 1c , 3c , Figure 1—figure supplement 3 ) , contributes exclusively to postsynaptic inhibition for most glomeruli . Unlike the Broad LNs , Choosy LNs have a clear axon innervating most glomeruli , while their dendrites collect inputs from only a small subset of glomeruli ( Figure 3c; Figure 1—figure supplement 3; Figure 3—figure supplement 1 ) . Therefore some glomeruli can drive postsynaptic inhibition of most glomeruli . Additionally , the inputs from Choosy LNs tend to be more proximal to the axon initial segment of the uPNs ( Gouwens and Wilson , 2009 ) unlike those of Broad LNs which are more uniformly distributed throughout the uPN dendritic arbor ( Figure 3—figure supplement 2 ) . In the adult , ORNs tend to synapse at the most distal PN dendritic terminals , allowing for some LN inhibition to occur via synapses more proximal to the axon initial segment ( Rybak et al . , 2016 ) . This pattern of spatially structured inputs suggests that different inhibitory LN types may exert different effects on uPN dendritic integration . Parallel to the uniglomerular readout by the 21 uPNs , we found 14 multiglomerular PNs ( mPNs; Figure 4a ) . Each mPN receives unique and stereotyped inputs from multiple ORNs ( Figure 4c ) or at least from one ORN and multiple unidentified non-ORN sensory neurons in the SEZ ( Figure 4a ) . The mPNs originate in multiple neuronal lineages and project to multiple brain regions; most commonly the lateral horn ( LH ) but also regions surrounding the MB calyx . Of the 14 mPNs , three project to the calyx itself ( mPNs b-upper , b-lower and C2 ) and another ( mPN cobra ) to the MB vertical lobe ( Figure 4a ) . In addition to the 14 mPNs that project to the brain , we identified an extra 6 oligoglomerular neurons that project to the SEZ ( SEZ neurons; Figure 1c; Figure 4—figure supplement 1 ) . A class of mPNs has been described in the adult fly ( Liang et al . , 2013 ) but their projection pattern does not match any of the larval mPNs . In strong contrast to uPNs , mPNs are very diverse in their lineage of origin , their pattern of inputs , and the brain areas they target . A small subset of mPNs has been identified via light microscopy before ( Thum et al . , 2011; Das et al . , 2013 ) . 10 . 7554/eLife . 14859 . 012Figure 4 . The multiglomerular circuit consists of 14 mPNs that project to the brain and 5 Picky LNs , each an identified neuron . ( a ) Posterior view of EM-reconstructed mPNs that innervate the right antennal lobe ( in color; uPNs in grey for reference ) , each receiving inputs from a subset of olfactory glomeruli but many also from non-ORN sensory neurons in the subesophageal zone ( SEZ ) . Most mPNs ( green ) project via the same tract as the uPNs ( mALT ) . They can project via other tracts ( other colors ) , but never via the mlALT used by the iPNs of the adult Drosophila . The mPNs project to many regions including a pre-calyx area , a post-calyx area , the lateral horn ( LH ) and the mushroom body vertical lobe ( MB vl ) . mPNs are generated by diverse neuroblast lineages including BAlp4 , BAla1 , and others ( Das et al . , 2013 ) . ( b ) Dorsal view of the EM-reconstructed Picky LNs shown together and individually . When shown individually , the Picky LNs are in 2 colors: blue for the dendrites and soma , and green for the axon . Zoom in to observe that presynaptic sites ( red ) are predominantly on the axon , whereas postsynaptic sites ( cyan ) are mostly on dendrites . Collectively , the dendritic arbors of the 5 Picky LNs tile the olfactory glomeruli . The dendrites of the Picky LN 3 and 4 extend significantly into the SEZ . They all originate from the same neuroblast lineage: BAla2 ( Das et al . , 2013 ) . ( c ) Percentage of the total number of postsynaptic sites on the dendrite of a mPN or Picky LN ( column neuron ) that originate from a given glomerulus or Picky LN ( row neurons ) . Here we define the glomerulus as connections from the ORN or via dendro-dendritic synapses from a given ORN’s uPN . This is most relevant for mPN A1 , which can receive more synapses from an ORN’s uPN than the ORN itself ( see suppl . Adjacency Matrix ) . We show the inputs to the mPNs and Picky LNs for the right antennal lobe , but for all bilateral mPNs ( bil . -lower , bil . -upper , and VUM ) we include inputs from both sides . We show only connections with at least two synapses , consistently found among homologous identified neurons in both the left and right antennal lobes . Percentages between 0 and 1 are rounded to 1 , but totals are computed from raw numbers . Connections in this table are stereotyped ( when comparing the left and right antennal lobes ) and selective . Note that mPNs that receive many inputs from non-ORN sensory neurons in the SEZ have a low total of ORN+uPN input . For an extended version of this table that includes all LNs see Figure 4—figure supplement 2 . ( d ) The direct upstream connectivity for two mPNs , with ORNs colored by the groups emerging from the PCA analysis of odor tuning . Connections from ORNs and Picky LNs to mPNs create 3 different types of motifs: direct excitatory connections from ORNs , lateral inhibitory connections from ORNs only via Picky LNs , and feedforward loops where an ORN connects both directly to the mPN and laterally through a Picky LN . Note that the activity of Picky LN 0 could alter the integration function for mPN A3 and indirectly for B2 , as well as many other mPNs ( not shown ) . Arrow thicknesses are weighted by the square root of the number of synapses between neurons . ( e ) The Picky LN hierarchy , dominated by Picky LN 0 , here showing connections with 2 or more consistent synapses between bilaterally homologous neurons . Some of these connections are axo-axonic ( see Figure 4—figure supplement 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 01210 . 7554/eLife . 14859 . 013Figure 4—figure supplement 1 . Six SEZ neurons receive specific inputs from some ORNs and from some antennal lobe LNs . Left , EM-reconstruction of the 6 SEZ neurons ( vine , cypress , clamp , spruce and ginkgo 1 and 2 ) , with their axons labeled green and their dendrites blue . Presynaptic sites in red and postsynaptic sites in cyan . Middle , 3 of these SEZ neurons project to the same unidentified region of the SEZ . Spruce projects to a more posterior area . Lateral view , anterior to the left . Right , table of percent of postsynaptic sites of a column neuron contributed by a row neuron , illustrating how some ORNs and LNs specifically target these SEZ neurons . We show only connections with at least two synapses , consistently found among homologous identified neurons in both the left and right antennal lobes . Percentages between 0 and 0 . 5 are removed . Notice how Picky LNs 2 , 3 and 4 synapse strongly onto SEZ neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 01310 . 7554/eLife . 14859 . 014Figure 4—figure supplement 2 . Extended version of table in Figure 4c , including all other olfactory-related neurons . Tables show percent of postsynaptic sites of a column neuron contributed by a row neuron . We show only connections with at least two synapses , consistently found among homologous identified neurons in both the left and right antennal lobes . Percentages between 0 and 0 . 5 are removed . For bilateral neurons , inputs from both sides are included . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 01410 . 7554/eLife . 14859 . 015Figure 4—figure supplement 3 . principal component analysis of odors leading to a principled clustering of orns . ( a ) Clustering of odors by odorant-descriptor . Results of K-means clustering of odors in the 32 dimensional odor-descriptor space proposed in Haddad et al . ( 2008 ) . Odors cluster into five groups that are well correlated with odor chemical type ( alcohols , aromatics , esters , pyrazines , and others ) . ( b–e ) Clustering of odors by ORN response . ( b ) The variance explained for the odors in ORN response space as a function of the number of principal components ( dimensions ) . The 'elbow' of this curve is composed of the principal components used for the clustering analysis of the odors by ORN-response . ( c ) How the odors span the space of the first 3 principal components of ORN response space . The odors are individual points colored by which of the five clusters , calculated via an affinity propagation clustering algorithm , they belong to . ( d ) How each of the odors fit into the clusters in ORN response space . Each cluster tends to group odors of similar chemical type . ( e ) The ORNs that represent the centroid of each cluster , calculated using a threshold obtained via Otsu’s method . See Materials and methods for further details . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 01510 . 7554/eLife . 14859 . 016Figure 4—figure supplement 4 . Pattern of ORN inputs onto Picky LNs . ( a ) The connections of ORNs onto the hierarchy of Picky LNs . ORNs are colored by the groups emerging from the PCA analysis of odor tuning . Inhibitory connections from Picky LNs are shown in black ( only connections with 2 or more synapses among bilaterally homologous neuron pairs are show ) . Excitatory connections from ORNs are shown in grey ( only connections with 4 or more synapses among bilaterally homologous neuron pairs are shown ) . See Supplementary file 1 and 2 ( containing the adjacency matrices ) for the complete set of connections . The thickness of the arrows is proportional to the square root of the number of synapses . Some of these connections are axo-axonic ( see c ) . ( c ) ORNs can synapse onto the Picky neurons at either their dendrites or their axons . This table shows values from -1 to 1 based on the written formula . Values between -1 and 0 correspond to the ORN synapsing more to the axon of the Picky LN than the dendrite , and values between 0 and 1 correspond to the ORN synapsing more to the dendrite of the Picky LN than the axon . Only consistent connections between ORNs and Picky dendrites or ORNs and Picky axons with a threshold of at least 2 consistent synapses per side are used to calculate these ratios . For values that are not 1 or -1 , the value can differ from side to side . Because the threshold is lowered from that of A , more connections appear , but since we only consider connections consistent in how they connect to the Picky LNs ( to dendrite or axon ) , some of the weakest connections also drop out compared to a . ( c ) The Picky LN hierarchy shown with the Picky LNs split into axon and dendrite , showing that not all connections are from the axon of a Picky LN to the dendrites of another . We are only showing connections that are consistent both in their motif ( axo-axonic , dendro-dendritic , etc ) and with a consistent threshold of 2 synapses on both sides . Because these criteria are more stringent than those used in a , some connections drop out ( such as Picky LN 4 to Picky LN 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 016 In addition to inputs from Broad LNs ( Figure 4—figure supplement 2 ) , mPNs also receive up to 26% of inputs from 5 stereotypically connected , oligoglomerular LNs that we call Picky LNs ( Figure 4b , c ) . While both Choosy LNs and Picky LNs are oligoglomerular and present distinct axons , the Choosy LNs are GABAergic whereas at least 4 of the 5 Picky LNs are instead glutamatergic ( Figure 2—figure supplement 1 for Picky LNs and Figure 2 panels L-O in ( Thum et al . , 2011 ) for Choosy LNs; see also Supp . Fig . 2 of Das et al . [2013] ) . The difference in neurotransmitter is consistent with Picky LNs deriving from a different lineage than Choosy LNs ( Figure 1—figure supplement 3 ) . In addition , the two Choosy LNs present indistinguishable connectivity , whereas each Picky LN has its own preferred synaptic partners ( see Supplementary file 1 and 2 ) . Additionally , unlike the Choosy LNs , Picky LNs rarely target uPNs ( Figure 3—figure supplement 1 ) . Glutamate has been shown to act as a postsynaptic inhibitory neurotransmitter in the adult fly antennal lobe for both PNs and LNs ( Liu and Wilson , 2013 ) , and therefore in larva , Picky LNs may provide inhibition onto both mPNs and other LNs . Unlike the Broad LNs , which are panglomerular and axonless , the Picky LNs present separated dendrites and axons ( Figure 4b ) . Collectively , Picky LN dendrites roughly tile the antennal lobe ( Figure 4b ) . While some Picky LN axons target select uPNs , about 40% of Picky LN outputs are dedicated to mPNs or each other ( Figure 4c; Figure 2 ) . Similarly to the mPNs , Picky LNs 2 , 3 , and 4 receive inputs from unidentified non-ORN sensory neurons in the SEZ ( Figure 4b , Figure 2 ) . Given that ORNs present overlapping odor tuning profiles ( Kreher et al . , 2008 ) , we applied dimensionality-reduction techniques and discovered that ORNs cluster into 5 groups by odorant preference ( Figure 4—figure supplement 3; see the Materials and methods section for more detail ) . This helped interpret the pattern of ORNs onto Picky LNs and mPNs . We found that some Picky LNs aggregate similarly responding ORNs ( Figure 4d; Figure 4—figure supplement 4 ) . For example , Picky LN 2 receives inputs preferentially from ORNs that respond to aromatic compounds , and Picky LN 3 and 4 similarly for aliphatic compounds ( esters and alcohols; Figure 4—figure supplement 4 ) . On the other hand , Picky LN 0 and 1 aggregate inputs from ORNs from different clusters , suggesting that these Picky LNs may select for ORNs that are similar in a dimension other than odorant binding profile . The stereotyped and unique convergence of different sets of ORNs onto both mPNs and Picky LNs , and the selective connections from Picky LNs to mPNs , suggest that each mPN responds to specific features in odor space , defined by the combinations of ORN and Picky LN inputs . These features are implemented through direct excitatory connections from ORNs or indirect inhibitory connections via Picky LNs ( lateral inhibition; Figure 4d ) . Some ORNs affect the activity of the same mPN through both direct excitatory and lateral inhibitory connections through Picky LNs ( incoherent feedforward loop , ( Alon , 2007 ) ; Figure 4d ) . The combination of these motifs may enable an mPN to respond more narrowly to odor stimuli than the ORNs themselves , many of which are broadly tuned ( Kreher et al . , 2008 ) , or to respond to a combinatorial function of multiple ORNs that describe an evolutionarily learned feature meaningful for the larva . For example , one mPN ( A1 ) reads out the total output of the uniglomerular system by integrating inputs across most ORNs and uPNs ( Figure 4c ) . Another mPN ( B2 ) could respond to the linear combination of ORNs sensitive to aromatic compounds ( direct connections ) , but its response could change in the presence of alcohols and esters due to feedforward loops ( Figure 4d ) . And mPNs A3 and B3 both collect inputs from ORNs ( Figure 4c , d ) known to respond to aversive compounds ( 22c , 45b , 49a , 59a , and 82a; [Kreher et al . , 2008; Ebrahim et al . , 2015] ) or whose ORN drives negative chemotaxis ( 45a; [Bellmann et al . , 2010; Hernandez-Nunez et al . , 2015] ) . Additionally , mPN B3 receives inputs from 33a , an ORN whose receptor lacks a known binding compound , and therefore is likely narrowly tuned to other ecologically relevant odorants as was shown for 49a and the pheromone of a parasitic wasp ( Ebrahim et al . , 2015 ) . The connectivity patterns of the 14 types of mPNs are vastly diverse from each other and likely each one extracts different features from odor space , often integrating inputs from non-ORN sensory neurons as well . In contrast to the all-to-all connectivity of the Broad LNs , the Picky LNs synapse onto each other in a selective , hierarchical fashion ( Figure 4e ) . The structure of the Picky LN hierarchy suggests that Picky LNs 0 and 3 can operate in parallel , while the activity of the other Picky LNs is dependent on Picky LN 0 ( Figure 4e ) . These connections among Picky LNs include axo-axonic connections , and some Picky LNs receive stereotypic ORN inputs onto their axons ( Figure 4—figure supplement 4 ) . The stereotyped hierarchy among Picky LNs defines yet another layer of computations in the integration function of each mPN . Picky LN 0 not only dominates the Picky LN hierarchy , and with it the multiglomerular system , but also may dramatically alter the inhibition of the entire olfactory system . This is because the main synaptic target of Picky LN 0 ( Figure 2 ) is a bilateral , axonless , GABAergic LN called Keystone ( Figure 5a; Figure 2—figure supplement 1 ) , which in turn strongly synapses onto the Broad LN Trio–a major provider of presynaptic inhibition ( Figure 5b ) . Interestingly , Keystone is also a major provider of presynaptic inhibition , but selectively avoids some glomeruli ( Figure 5c; Figure 3—figure supplement 1 ) . Therefore the wiring diagram predicts that these two parallel systems for presynaptic inhibition can directly and strongly inhibit each other ( Figure 5b ) : homogeneous across all glomeruli when provided by the Broad LN Trio , and heterogeneous when provided by Keystone ( Figure 5c ) . In conclusion , the circuit structure and the known synaptic signs predict that Picky LN 0 can promote a state of homogeneous presynaptic inhibition by disinhibiting the Broad LN Trio ( Figure 5d ) . 10 . 7554/eLife . 14859 . 017Figure 5 . The wiring diagram suggests two operational states: homogeneous or heterogeneous presynaptic inhibition . ( a ) Posterior view of the EM-reconstructed neurons innervating the left antennal lobe that could govern the switch ( uPNs in grey and right ORNs in dark grey for reference ) . The Keystone LN ( blue ) has a symmetric bilateral arbor and additionally innervates the SEZ , receiving inputs from non-ORN sensory neurons ( in black ) . Neuromodulatory neurons that make direct morphological synapses onto LNs are serotonergic ( CSD in pink; projects contralaterally after collecting inputs from near the MB calyx ) and octopaminergic ( lAL-1 and two tdc , in dark and light green ) , and all arborize well beyond the antennal lobe . Also included is Picky LN 0 ( red ) . ( b ) A wiring diagram outlining the strong LN-LN connections , showing the core reciprocal inhibition between Broad LN Trio and Keystone that could mediate the switch between homogeneous ( panglomerular ) presynaptic inhibition and heterogenous ( selective ) presynaptic inhibition . For simplicity , neurons are grouped together if they belong to the same neuron type , with the number of neurons belonging to each group indicated in parentheses . Connections are weighted by the square root of the number of synapses between groups of neurons . The self-arrow for the Broad LN Trio represents the average number of synapses that one of the Trio neurons receives from the other two . Picky LN 0 inhibits Keystone , thereby disinhibiting the Broad LN Trio and promoting homogeneous presynaptic inhibition . The maxillary nerve sensory neurons are the top input providers of Keystone and may drive the system towards heterogeneous presynaptic inhibition ( see C ) . The effect of direct inputs from neuromodulatory neurons is unknown , but at least it has been suggested that octopaminergic neurons may have an excitatory effect on inhibitory LNs ( Linster and Smith , 1997 ) . ( c ) Cartoon of glomeruli colored by the percentage of inputs onto ORN axon terminals provided by the Broad LN Trio and from Keystone , indicating the amount of presynaptic inhibition ( onto ORNs ) in either state . The inhibition provided by Broad LN Trio is much more uniform than the inhibition provided by Keystone . Dotted lines indicate glomeruli that receive Picky LN 0 input on either the ORN or uPN . ( d ) The LNs putatively active in each state . ( e ) Unlike other Picky LNs , Picky LN 0 makes synapses onto ORN axon terminals and many uPNs . Here connections with 2 or more synapses consistent between bilaterally homologous neuron pairs are shown . Arrow thicknesses are weighted by the square root of the number of synapses between neurons . With the exception of 45a , all shown ORNs and uPNs belong to glomeruli that synapse onto Picky LN 0 as well . Thus Picky LN 0 provides both pre- and postsynaptic inhibition to a small set of glomeruli . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 01710 . 7554/eLife . 14859 . 018Figure 5—figure supplement 1 . EM-reconstructed arbor of the descending neuron . Renderings of the left antennal lobe , posterior view . This identified neuron exists in the left and right antennal lobes , presenting similar morphology and connectivity in the right antennal lobe . Broad LNs and uPNs are shown for reference . Scale bar: a cell body measures about 4 micrometers in diameter . DOI: http://dx . doi . org/10 . 7554/eLife . 14859 . 018 The alternative state of heterogeneous presynaptic inhibition implemented by Keystone could be triggered by select non-ORN sensory neurons that synapse onto Keystone in the SEZ ( Figure 5a , b ) . These non-ORN sensory neurons are the top inputs of Keystone and do not synapse onto any other olfactory LN . In contrast , ORNs that synapse onto Keystone also synapse onto the Broad LN Trio ( Figure 3—figure supplement 1 ) , suggesting a role for non-ORN sensory inputs in tilting the balance towards Keystone and therefore the heterogeneous state . However , the subset of ORNs that also synapse onto Picky LN 0 ( Figure 4c ) could oppose the effect of the non-ORN sensory neurons by inhibiting keystone and therefore disinhibiting the Broad LN Trio . Neuromodulatory neurons could also affect the balance between Keystone and Broad LN Trio . Beyond the possible effect of volume release of serotonin ( Dacks et al . , 2009 ) and octopamine ( Linster and Smith , 1997; Selcho et al . , 2012 ) within the olfactory system , we found that these neuromodulatory neurons synapse directly and specifically onto Keystone or Broad LN Trio , respectively ( Figure 5b ) . Beyond non-ORN inputs , ORNs synapse selectively onto these neuromodulatory neurons . Two ORNs ( 74a and 82a ) synapse onto the serotonergic neuron CSD ( Roy et al . , 2007 ) , and five ORNs ( 42b , 74a , 42a , 35a and 1a ) onto an octopaminergic neuron ( lAL-1; see Figure 4k in Selcho et al . , 2014 ) , suggesting that specific ORNs may contribute to tilting the balance between homogeneous and heterogeneous presynaptic inhibition via neuromodulation . The only other provider of panglomerular presynaptic inhibition is the Broad Duet , which is the main provider of panglomerular postsynaptic inhibition . These neurons may operate similarly in both states given that they are inhibited by both Keystone and Broad LN Trio ( Figure 1c ) . The higher fraction of inputs from Broad LN Trio onto Duet might be compensated by the fact that the Trio LNs inhibit each other ( Figures 3e , 5b ) , whereas the two Keystone LNs do not ( Figure 5b ) . Therefore , potentially the Broad LN Duet could be similarly active in either state ( Figure 5d ) . The possibility of heterogeneous presynaptic inhibition promoted by Keystone suggests that some ORNs can escape divisive normalization of their outputs relative to the rest . Not surprisingly , one such ORN is 49a ( Figure 5c ) , which is extremely specific for the sexual pheromone of a parasitic wasp that predates upon larvae ( Ebrahim et al . , 2015 ) . The other two ORNs that escape fully are 1a and 45b . 1a activation drives negative chemotaxis ( Hernandez-Nuñez et al . , unpublished ) . 45b senses compounds that elicit negative chemotaxis in larvae ( Kreher et al . , 2008 ) . These three ORNs , and in particular 49a , are under strong postsynaptic inhibition by both Broad LN Duet and Choosy LNs ( Figure 3c ) . In summary , reducing presynaptic inhibition in these 3 ORNs may enable larvae to perceive odors evolutionarily associated with life-threatening situations less dependently of the response intensity of other ORNs ( i . e . overall odorant concentration ) . This is consistent with the finding that responses to aversive odors may rely on specific activity patterns in individual ORNs ( Gao et al . , 2015 ) . The strong postsynaptic inhibition might be instrumental for their corresponding uPN to respond to the derivative of the ORN activity ( with an incoherent feedforward loop; [Alon , 2007] ) , as shown in the adult fly ( Kim et al . , 2015 ) , facilitating detection of concentration changes . A key neuron in tilting the balance between homogeneous and heterogeneous presynaptic inhibition in the Broad-Keystone circuit is Picky LN 0 ( Figure 5b ) . Remarkably , one of the two top ORN partners of Picky LN 0 is ORN 42a ( Figure 4c ) , the strongest driver of appetitive chemotaxis in larvae ( Fishilevich et al . , 2005; Asahina et al . , 2009; Schulze et al . , 2015; Hernandez-Nunez et al . , 2015 ) . The connections of Picky LN 0 extend beyond that of other oligoglomerular LNs , and include both pre- and postsynaptic inhibition of a small subset of glomeruli , including 42a ( Figure 5e ) . The wiring diagram therefore indicates that Picky LN 0 , a likely glutamatergic LN , engages in seemingly contradictory circuit motifs: simultaneously inhibiting specific ORNs and their uPNs , while also disinhibiting them by inhibiting Keystone . The suppression of Keystone disinhibits the Broad LN Trio and therefore promotes homogeneous inhibition . However this is further nuanced by reciprocal connections between Picky LN 0 and Broad LN Trio ( Figure 5b ) . This push-pull effect of glutamatergic LNs on PNs has been described for the olfactory system of the adult fly as conducive to more robust gain control and rapid transitions between network states ( Liu and Wilson , 2013 ) . This refined control could endow Picky LN 0-innervated glomeruli like 42a ( Figure 5e ) with the ability to better detect odor gradients , consistent with 42a being a strong and reliable driver of appetitive chemotaxis ( Fishilevich et al . , 2005; Asahina et al . , 2009; Schulze et al . , 2015 ) . Picky LN 0 and its push-pull effect on PNs not only can have an effect on positive chemotaxis but also on negative . A clear example is the 82a glomerulus ( known to respond to an aversive odor that drives negative chemotaxis [Kreher et al . , 2008] ) which lacks a well-developed uPN but engages in strong connections with mPNs such as A3 ( Figure 4c , d ) . We found that , like for the appetitive case of 42a , Picky LN 0 engages in both presynaptic inhibition onto 82a ORN and also postsynaptic inhibition onto mPN A3 , one of the top PNs of 82a ORN . And like other ORNs mediating aversive responses ( e . g 49a ) , the 82a uPN is also under strong postsynaptic inhibition ( Figure 3c ) . Finally , we found evidence that an individual glomerulus can have a global effect on the olfactory system . All LNs ( except Picky LN 3 ) receive inputs from Ventral LN ( Figure 1c , Figure 3—figure supplement 1 ) , an interneuron of unknown neurotransmitter , which is primarily driven by the 13a glomerulus . This suggest that 13a , an ORN sensitive to alcohols ( Kreher et al . , 2008 ) , could potentially alter the overall olfactory processing . In the mammalian olfactory bulb , descending inputs from the brain target granule cells ( the multiglomerular inhibitory LNs ) , shaping the level of inhibition ( Balu et al . , 2007 ) . In addition to descending neuromodulatory neurons ( CSD; Figure 5a ) , in the larva we found a descending neuron ( Figure 5—figure supplement 1 ) that targets specific mPNs and LNs ( Figure 4—figure supplement 2 ) . In addition to other mPNs , this descending neuron targets the two mPNs that we postulate are aversive ( mPNs A3 and B3 ) . Together with the axo-axonic inputs it receives from 45a ORN ( an aversive ORN , [Bellmann et al . , 2010; Hernandez-Nunez et al . , 2015] ) , this descending neuron is associated with the processing of aversive stimuli . Additional descending neurons affecting PNs and LNs might exist but were beyond the scope of this study , where we focused on neurons directly synapsing with ORNs .
The glomerular olfactory system of the larva develops in a similar fashion to the vertebrate olfactory bulb where the afferents ( i . e . ORNs ) organize the central neurons , unlike in the adult fly ( Prieto-Godino et al . , 2012 ) . In zebrafish , GABAergic LNs provide depolarizing currents to PNs ( mitral cells ) via gap junctions at low stimulus intensities , enhancing low signals , and inhibit the same PNs at high stimulus intensity via GABA release , implementing a form of gain control ( Zhu et al . , 2013 ) . This role is played by a class of panglomerular excitatory LNs in the adult fly that make gap junctions onto PNs and excite inhibitory LNs ( Yaksi and Wilson , 2010 ) . In the larva , all panglomerular neurons are GABAergic; if any were to present gap junctions with uPNs , a cell type for gain control in larva would be equivalent to the one in zebrafish . Particularly good candidates are the Broad LN Duet , which provide the bulk of feedforward inhibitory synapses onto uPNs in larva . Interestingly , postsynaptic inhibition might not be mediated by GABA in the adult fly ( supp . fig . 5 in Olsen and Wilson [2008]] ) , rendering olfactory circuits in larva more similar to vertebrates . Presynaptic inhibition exists both in the adult fly and , as suggested by the present work , in larva , and is mediated by the same kind of panglomerular GABAergic neurons ( the Broad Trio LNs in larva; and see [Olsen and Wilson , 2008] ) . The uniglomerular circuit is the most studied in all species both anatomically and physiologically . We found that each uPN receives an unusually large number of inputs from an individual ORN compared to other sensory systems in the larva ( Ohyama et al . , 2015 ) . This large number of morphological synapses could be interpreted as a strong connection , which would support faster or more reliable signal transmission . In the adult fly , the convergence of multiple ORNs onto an individual PN enables both a fast and reliable PN response to odors ( Bhandawat et al . , 2007 ) . The temporal dynamics of crawling are far slower than that of flying , and therefore we speculate that the integration over time of the output of a single ORN might suffice for reliability , demanding only numerous synapses to avoid saturation . Positive , appetitive chemotaxis involves odor gradient navigation , leading to a goal area where food is abundant which may overwhelm olfaction . We postulate that navigation and feeding correspond to the homogeneous and heterogeneous states of presynaptic inhibition that we described . During navigation , homogeneous presynaptic inhibition ( via Broad LN Trio ) could best enhance salient stimuli and therefore chemotaxis , enabling the olfactory system to operate over a wide range of odorant intensities ( Asahina et al . , 2009 ) . During feeding , strongly stimulated ORNs could scale down the inputs provided by other , less stimulated , ORNs . In other words , if homogeneous presynaptic inhibition persisted during feeding , the larvae would lose the ability to detect important odorants that are likely to be faint , for example the scent of a predator such as a parasitic wasp via 49a ( Ebrahim et al . , 2015 ) . The larva can selectively release presynaptic inhibition via Keystone , which provides presynaptic inhibition to appetitive glomeruli while also inhibiting the Broad LN Trio–the major providers of panglomerular presynaptic inhibition . So the larva could feed and remain vigilant to evolutionarily important cues at the same time . Not surprisingly , the switch might be triggered by neuromodulatory neurons and non-ORN sensory neurons , potentially gustatory , that synapse onto Keystone . In addition to the uniglomerular system that is present across multiple vertebrate and invertebrate species ( Satou , 1992; Wang et al . , 1998; Vosshall et al . , 2000 ) , we found , in the Drosophila larva , a multiglomerular system that presumably performs diverse processing tasks already at the first synapse . One such task could be the detection of concentration gradients for some odorant mixtures , suggesting an explanation for the observation that some ORNs can only drive chemotaxis when co-activated with other ORNs ( Fishilevich et al . , 2005 ) . Similar glomerular-mixing circuits have been described in higher brain areas ( lateral horn ) of the fly ( Wong et al . , 2002; Fişek and Wilson , 2014 ) and of mammals ( Sosulski et al . , 2011 ) . We hypothesize that in the larva , the morphological adaptations to a life of burrowing might have led to specific adaptations , relevant to an animal that eats with its head , and therefore the dorsal organ housing the ORNs , immersed in food . It is perhaps not surprising that we found multisensory integration across ORNs and non-ORNs ( likely gustatory ) already at the first synapse . And we hypothesize that the pooling of chemosensors ( ORNs and non-ORNs ) onto mPNs and Picky LNs may be related to the reduction in the number of ORNs relative to insects with airborne antennae . With our complete wiring diagram of this tractable , transparent model system and genetic tools for manipulating and monitoring the activity of single identified neurons , we have now the opportunity to bridge the gap between neural circuits and behavior ( Carandini , 2012 ) .
We reconstructed neurons and annotated synapses in a single , complete central nervous system from a 6-h-old [iso] Canton S G1 x w1118 larva imaged at 4 . 4 x 4 . 4 x 50 nm resolution , as described in Ohyama et al . ( 2015 ) . The volume is available at http://openconnecto . me/catmaid/ , titled ”'acardona_0111_8' . To map the wiring diagram we used the web-based software CATMAID ( Saalfeld et al . , 2009 ) , updated with the novel suite of neuron skeletonization and analysis tools ( Schneider-Mizell et al . , 2016 ) , and applied the iterative reconstruction method described in Ohyama et al . ( 2015 ) ; Schneider-Mizell et al . ( 2016 ) . CNS was dissected from 3rd instar larvae . 4% formaldehyde was used as fixative for all antibodies except anti-dVGlut that required bouin fixation ( Drobysheva et al . , 2008 ) . After fixation , brain samples were stained with rat anti-flag ( 1:600 , Novus Biologics ) and chicken anti-HA ( 1:500 , Abcam , ab9111 ) for labeling individual neurons in the multi-color flip-out system ( Nern et al . , 2015 ) , while mouse anti-Chat ( 1:150 , Developmental Studies Hybridoma Bank , ChaT4B1 ) and rabbit anti-GABA ( 1:500 , Sigma A2052 , Lot# 103M4793 ) or anti-dVGlut ( Daniels et al . , 2004 ) were used for identifying neurotransmitters . Antibodies were incubated at 4°C for 24 hr . Preparations were then washed 3 times for 30 min . each . with 1% PBT and then incubated with secondary antibodies ( including: goat anti-Mouse Alexa Fluor 488 , goat anti-rabbit Alexa Fluor 568 , donkey anti-rat Alexa Fluor 647 , Thermofisher; and goat anti-chicken Alexa Fluor 405 , Abcam ) at 1/500 dilution for 2 hr at room temperature , followed by further washes . Nervous systems were mounted in Vectashield ( Vector Labs , Burlingame , CA ) and imaged with a laser-scanning confocal microscope ( Zeiss LSM 710 ) . Extensive screens have been conducted to identify which odorants activate each ORN ( Kreher et al . , 2008; Montague et al . , 2011; Mathew et al . , 2013 ) . These data can be used as a starting point to determine whether the multiglomerular circuit extracts relevant components of odorant physical descriptor space , that is , the chemical structure of the odorant as sampled by ORNs . Using the data from ( Kreher et al . , 2008 ) and ( Montague et al . , 2011 ) , we conducted a dimensionality reduction via PCA followed by a clustering analysis , and then used the data from ( Mathew et al . , 2013 ) to verify our findings . To determine how ORNs encode odors we followed the PCA analysis in ORN space performed in ( Haddad et al . , 2010 ) adding the data of pyrazines from ( Montague et al . , 2011 ) . Then , using the Scree test we selected the first 3 components of the PCA as the relevant ones to use for clustering , and we ran the clustering minimization using the affinity propagation algorithm ( which doesn’t require the number of clusters as an input ) ( Figure 4—figure supplement 3b , c ) . Four of the obtained clusters correspond very well with odorant type ( alcohols , aromatics , esters , and pyrazines; Figure 4—figure supplement 3d ) . The fifth cluster is mixed and mainly includes odorants with very low or no ORN response . Consistent with that , k-means clustering in the 32-dimensional odorant physical descriptor space described in ( Haddad et al . , 2008 ) results in 5 clusters , 4 of them matching the non-mixed clusters obtained in ORN space ( Figure 4—figure supplement 3a ) . To determine which ORNs encode the regions of each cluster , we projected back the centroid of each cluster onto ORN space using the inverse transformation ( Figure 4—figure supplement 3e ) . Different subsets of ORNs were more likely to encode each cluster . The projections of the cluster centroids in ORN space are not discrete numbers; in order to make these results easier to interpret a threshold can be established to determine which ORNs encode a cluster centroid and which ones don’t . A suitable approach is to use Otsu’s method , which can be considered a one-dimensional discrete analog of Fisher’s discriminant analysis ( Otsu , 1975 ) . We obtained a threshold of 0 . 4725 , which we used to determine the ORNs that encode each cluster ( dashed red line in Figure 4—figure supplement 3e ) . In ( Mathew et al . , 2013 ) a set of odorants that specifically activate single ORNs at low concentrations were identified . These data can easily be used to cross-validate the predicted receptive field of the different ORNs in our analysis . Four of the odorants tested were alcohols and activated Or13a , 35a , 67b and 85c all of which are in our alcohols group . Other three were aromatics and activated Or22c , 24a and 30a , which are all in our aromatics group . One was a pyrazine and activated Or33b which is in our pyrazine group . Finally pentyl acetate ( an ester ) activated 47a which is in our esters group . The other odorants in ( Mathew et al . , 2013 ) were in regions of odor space ( mostly ketones and aldehydes ) that were not sampled in ( Kreher et al . , 2008 ) or ( Montague et al . , 2011 ) and therefore their responses cannot be predicted with our analysis . As more datasets are collected , approaches like the one we present here can be used to better establish the receptive field of each ORN . | Our sense of smell can tell us about bread being baked faraway in the kitchen , or whether a leftover piece finally went bad . Similarly to the eyes , the nose enables us to make up a mental image of what lies at a distance . In mammals , the surface of the nose hosts a huge number of olfactory sensory cells , each of which is tuned to respond to a small set of scent molecules . The olfactory sensory cells communicate with a region of the brain called the olfactory bulb . Olfactory sensory cells of the same type converge onto the same small pocket of the olfactory bulb , forming a structure called a glomerulus . Similarly to how the retina generates an image , the combined activity of multiple glomeruli defines an odor . A particular smell is the combination of many volatile compounds , the odorants . Therefore the interactions between different olfactory glomeruli are important for defining the nature of the perceived odor . Although the types of neurons involved in these interactions were known in insects , fish and mice , a precise wiring diagram of a complete set of glomeruli had not been described . In particular , the points of contact through which neurons communicate with each other – known as synapses – among all the neurons participating in an olfactory system were not known . Berck , Khandelwal et al . have now taken advantage of the small size of the olfactory system of the larvae of Drosophila fruit flies to fully describe , using high-resolution imaging , all its neurons and their synapses . The results define the complete wiring diagram of the neural circuit that processes the signals sent by olfactory sensory neurons in the larva’s olfactory circuits . In addition to the neurons that read out the activity of a single glomerulus and send it to higher areas of the brain for further processing , there are also numerous neurons that read out activity from multiple glomeruli . These neurons represent a system , encoded in the genome , for quickly extracting valuable olfactory information and then relaying it to other areas of the brain . An essential aspect of sensation is the ability to stop noticing a stimulus if it doesn't change . This allows an animal to , for example , find food by moving in a direction that increases the intensity of an odor . Inhibition mediates some aspects of this capability . The discovery of structure in the inhibitory connections among glomeruli , together with prior findings on the inner workings of the olfactory system , enabled Berck , Khandelwal et al . to hypothesize how the olfactory circuits enable odor gradients to be navigated . Further investigation revealed more about how the circuits could detect slight changes in odor concentration regardless of whether the overall odor intensity is strong or faint . And , crucially , it revealed how the worst odors – which can signal danger – can still be perceived in the presence of very strong pleasant odors . With the wiring diagram , theories about the sense of smell can now be tested using the genetic tools available for Drosophila , leading to an understanding of how neural circuits work . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | The wiring diagram of a glomerular olfactory system |
The mechanism that leads to liquid-liquid phase separation ( LLPS ) of the tau protein , whose pathological aggregation is implicated in neurodegenerative disorders , is not well understood . Establishing a phase diagram that delineates the boundaries of phase co-existence is key to understanding whether LLPS is an equilibrium or intermediate state . We demonstrate that tau and RNA reversibly form complex coacervates . While the equilibrium phase diagram can be fit to an analytical theory , a more advanced model is investigated through field theoretic simulations ( FTS ) that provided direct insight into the thermodynamic driving forces of tau LLPS . Together , experiment and simulation reveal that tau-RNA LLPS is stable within a narrow equilibrium window near physiological conditions over experimentally tunable parameters including temperature , salt and tau concentrations , and is entropy-driven . Guided by our phase diagram , we show that tau can be driven toward LLPS under live cell coculturing conditions with rationally chosen experimental parameters .
Protein liquid-liquid phase separation ( LLPS ) is a process in which proteins assemble and partition into a protein-dense phase and a protein-dilute phase . The proteins in the dense phase form droplets , and retain liquid-like mobility , as shown by NMR measurements ( Burke et al . , 2015; Brady et al . , 2017 ) . The process of LLPS in vitro has been observed for decades ( Anderson and Kedersha , 2006; Brangwynne et al . , 2009; Wippich et al . , 2013; Veis and Aranyi , 1960; Arneodo et al . , 1988; Water et al . , 2014 ) , but the field has recently been invigorated by the realization that LLPS also occurs in vivo , suggesting a possible physiological role for these assemblies ( Brangwynne et al . , 2009; Molliex et al . , 2015; Hyman et al . , 2014 ) . The overwhelming majority of proteins observed to undergo LLPS are intrinsically disordered proteins ( IDPs ) ( Uversky et al . , 2015 ) , and much of the research thus far has focused on ALS-related IDPs , including FUS ( Molliex et al . , 2015; Li et al . , 2013; Murakami et al . , 2015; Patel et al . , 2015 ) , hnRNPA2B1 and hnRNPA1 ( Kato et al . , 2012 ) , TDP-43 ( Kato et al . , 2012; Li et al . , 2018b ) , C9ORF72 ( Kwon et al . , 2014; Lee et al . , 2016; Boeynaems et al . , 2017 ) and Ddx4 ( Nott et al . , 2015 ) . Recently , we and others discovered that another amyloid forming IDP , the microtubule binding protein tau , also undergoes LLPS ( Ambadipudi et al . , 2017; Zhang et al . , 2017; Hernández-Vega et al . , 2017; Ferreon et al . , 2018; Wegmann et al . , 2018 ) . Interestingly , many of the LLPS forming IDPs have been observed to form amyloid fibrils in cell-free systems ( Murakami et al . , 2015; Kato et al . , 2012 ) , leading to a number of hypotheses regarding the physiological role of LLPS in regulating aggregation . In particular , a compelling idea is that protein LLPS may be an intermediate regulatory state , which could redissolve into a soluble state or transition to irreversible aggregation/amyloid fibrils ( Murakami et al . , 2015; Patel et al . , 2015; Kato et al . , 2012; Ambadipudi et al . , 2017; Zhang et al . , 2017 ) . In a healthy neuron , tau is bound to microtubules . When tau falls off the microtubule under adverse conditions to the cell , tau is solubilized in the intracellular space as an IDP . Under certain conditions , tau forms intracellular fibrillary tangles , a process linked to neurodegenerative tauopathies that include Alzheimer’s disease . In recent work , we showed that tau in neurons strongly ( nanomolar dissociation constant ) and selectively associates with smaller RNA species , most notably tRNA ( Zhang et al . , 2017 ) . We also found tau and RNA , under charge matching conditions , to undergo LLPS ( Zhang et al . , 2017 ) in a process determined to be complex coacervation ( CC ) ( Bungenberg de Jong , 1949 ) . We found that tau-RNA LLPS is reversible , and persisted for >15 hr without subsequent fibrilization of tau , and hypothesized that LLPS is potentially a physiological and regulatory state of tau . In this work , we characterize the phase diagram of tau-RNA LLPS using a combination of experiment and simulation , and thereby specify the conditions that drive the system toward a homogeneous phase or an LLPS state . We study a N-terminus truncated version of the longest isoform of human 4R tau in vitro , and first demonstrate that tau-RNA complexation is reversible , and that tau remains dynamic and without a persistent structure within the dense phase . The phase coexistence curve separating a supernatant phase from a condensate phase is determined by the system's free energy , which in turn is state dependent , that is dependent on concentration , temperature , salt , and the nature of the interaction strength between the various solution constituents , including the solvent . We construct the phase diagram from cloud-point measurements of the onset of complex coacervation under varying conditions of temperature , salt , and polymer concentrations . These experiments establish the features and phase coexistence boundaries of the phase diagram , which we then model using theory and simulation to rationalize and understand the physical mechanisms that drive and stabilize LLPS . A number of theoretical models can be used to model LLPS , each with their own advantages and disadvantages . Ideally , one would turn to simulations at atomic resolution in explicit solvent; however , such models are computationally prohibitive given the multiple orders of magnitude in time and length scales involved in LLPS . Turning to the polymer physics literature , theoretical treatments of simplified coarse-grained models are much more computationally tractable , and offer useful insight . Although approximate , analytical theories can be formulated , providing an extremely efficient platform for describing the thermodynamics of polyelectrolyte mixtures ( Sing , 2017 ) . These include the Flory-Huggins model ( Flory , 1953 ) , the Voorn-Overbeek model ( Veis and Aranyi , 1960; Spruijt et al . , 2010; Overbeek and Voorn , 1957; Tainaka , 1979; Veis , 1963; Tainaka , 1980; Nakajima and Sato , 1972 ) , the random-phase approximation ( Borue and Erukhimovich , 1988; Borue and Erukhimovich , 1990; Castelnovo and Joanny , 2001 ) , the Poisson-Boltzmann cell model ( Biesheuvel and Cohen Stuart , 2004a; Biesheuvel and Stuart , 2004b ) , as well as other more sophisticated approaches ( Lytle and Sing , 2017; Shen and Wang , 2018; Shen and Wang , 2017 ) , which have been applied to synthetic polymers with low sequence heterogeneity ( Spruijt et al . , 2010; Chollakup et al . , 2010; Zalusky et al . , 2002; Li et al . , 2018a; de la Cruz et al . , 1995 ) , and to proteins with single composition ( Brady et al . , 2017; Nott et al . , 2015; Banjade and Rosen , 2014; Banjade et al . , 2015 ) . While such models have been successful in describing simpler polyelectrolytes , it is less apparent that these models are suitable to describe the complex coacervation of the more complicated tau-RNA system . The simplest approach that one can use is the Flory-Huggins ( FH ) model , augmented by the Voorn and Overbeek ( VO ) correction to describe electrostatic correlations . This model is widely used to model LLPS; however , while experimental data can be fit to the model ( Brady et al . , 2017; Nott et al . , 2015 ) , ultimately the FH-VO model has serious inadequacies . The original Flory-Huggins model is a mean-field theory , which means that fluctuations in polymer densities away from their average value in each phase are neglected . Augmenting the FH model with a VO treatment of electrostatics approximately accounts for charge correlations , but it entirely neglects chain-connectivity ( Qin and de Pablo , 2016 ) . Thus , the FH-VO model is unable to model the spatially varying charge distribution along the polymer backbone . Ideally , one would like to introduce chain connectivity , charge correlation , and uneven charge distribution into a more realistic polymer physics model; however , a full treatment of polymer density fluctuations is analytically intractable . One possible approach is to pursue a Gaussian approximation to field fluctuations , also known as the random phase approximation ( RPA ) ( Kudlay and Olvera de la Cruz , 2004; Kudlay et al . , 2004; Castelnovo and Joanny , 2000 ) . The RPA model can be viewed as a lowest-order correction to the mean field approximation and was recently introduced to describe the charge pattern and sequence-dependent LLPS of IDPs ( Lin et al . , 2017; Lin et al . , 2016 ) . The advantage of the RPA model , over the mean-field FH-VO model , is that charge correlations are introduced in a formally consistent manner . Nonetheless , it has been recently demonstrated that the RPA model fails to quantitatively predict polymer concentrations in the dilute phase , given that higher order fluctuations are important in this regime ( Delaney and Fredrickson , 2017; Das et al . , 2018 ) . Of all the models described above , fitting experimental data with the FH or FH-VO theory is currently the preferred methodology in the LLPS community to describe and analyze phase diagrams . We demonstrate that this model can be fit to describe our experimental data , but the learning outcome from this modeling is limited . Thus , we take a different approach by computing the exact phase diagram of an off-lattice coarse-grained polyelectrolyte model using field theoretic simulations ( FTS ) . FTS is a numerical approach that allows one to fully account for fluctuations , and thus to compute equilibrium properties from a suitably chosen coarse-grained representation of the true system without the need for analytical approximation . The ability to perform field theoretic simulations enables us to include the important physics of polymer sequence-specificity that cannot be captured by FH-VO , including charge distribution and chain connectivity . Results from FTS are compared to those obtained from the FH-VO model . The model substantiates the experimental phase diagram that the equilibrium window for the complex coacervation of tau and RNA under cellular conditions is narrow . Guided by the phase diagram , empirically obtained from in vitro experiments and validated by simulation , we finally show that LLPS of tau-RNA can be established and rationalized under cellular co-culturing conditions in the presence of live cells .
Truncated versions of the longest isoform of human 4R tau , residues 255–441 ( Peterson et al . , 2008 ) and residues 255–368 were used to study tau-RNA complex coacervation ( CC ) . A C291S mutation was introduced to either tau variant , resulting in single-cysteine constructs . Thioflavin T assays and TEM imaging were performed showing these variants retain the capability to form fibrils with morphology similar to full length tau . Unless otherwise specified , we refer to these two single-cysteine tau constructs as tau187 and tau114 ( tau114 is close to K18 , 244–372 [Gustke et al . , 1994] ) , respectively , while tau refers collectively to any of these variants ( see Materials and methods for experimental details ) . Importantly , experiments were performed with freshly eluted tau within 30 min upon purification to minimize the effects of possible disulfide bond formation . This minimizes the influence of the cysteine mutations on the LLPS behavior of tau-RNA CC . The single-cysteine containing tau187 can be singly spin labeled at site 322 , referred to as tau187-SL ( see Materials and methods ) . Full length tau , tau187 and tau114 are overall positively charged with an estimated +3 , +11 and+11 charge per molecule at neutral pH , respectively , based on their primary sequences . The charged residues of tau are more concentrated in the four repeat domains ( Figure 1A ) . PolyU RNA ( 800 ~ 1000 kDa ) , which is a polyanion carrying one negative charge per uracil nucleotide , was used in this study and henceforth referred to as RNA ( Figure 1A ) . Under ambient conditions , both tau and RNA are soluble and stable in solution . By mixing tau and RNA under certain conditions , a turbid and milky suspension was obtained within seconds , where tau and RNA formed polymer-rich droplets ( dense phase ) separated from polymer-depleted supernatants ( dilute phase ) ( Figure 1B ) . These polymer-rich droplets are tau-RNA CCs . We began by determining the concentration of the dense and dilute phases . After mixing and centrifuging 60 μL tau187-RNA droplet suspension , we separated a polymer-rich phase of volume <1 μL with a clear boundary against the dilute supernatant phase . Applying UV-Vis spectroscopy ( see Materials and methods ) , we determined the concentration of tau and RNA inside the droplets as >76 mg/mL and >17 mg/mL with partitioning factors of >15 and >700 , respectively . This is consistent with our previously findings that tau is virtually exclusively partitioned within the dense phase ( Zhang et al . , 2017 ) . High-protein concentrations are typically correlated with higher propensity for irreversible protein aggregations . In order to verify that there was indeed no fibril formation , tau187-RNA CCs were prepared by mixing tau187-SL and RNA ( see Materials and methods ) and monitored by continuous wave electron paramagnetic resonance spectroscopy ( For details of cw-EPR experiments see Materials and methods ) . The cw-EPR spectra shows no broadening ( Figure 1C ) , and the cw-EPR spectra analysis reveals an unchanged rotational correlation time for the spin label of tau187-SL , τ , of 437 ± 37 ps as a function of time after >96 hr of incubation at room temperature ( Figure 1D , turquoise ) ( see Materials and methods ) . For comparison , cw-EPR spectra and τ were recorded of tau187-SL alone in buffer , and of tau187-SL in the presence of heparin under fibril forming conditions . Tau187-SL alone in buffer showed cw-EPR spectra overlapping with those of tau187-RNA CC , and rotational correlation time τ , 425 ± 16 ps , nearly identical to the τ of tau187-SL CCs ( Figure 1D , red ) . In contrast , tau187-SL with heparin shows a significantly broadened cw-EPR spectrum and an increasing τ to 2 . 3 ± 0 . 7 ns ( Figure 1C , D , green ) . Note that a hundreds of ps range of τ corresponds to rapid tumbling of the spin label , whose rotational degree of freedom is minimally hindered by molecular associations , while a several ns range of τ corresponds to slow tumbling and molecular hindering by association or confinement . The Thioflavin T ( ThT ) fluorescence curves of the same sample system as a function of time confirms the absence of amyloid aggregate formation in tau187-RNA CCs ( Figure 1—figure supplement 1 ) . These results together suggest that tau187-RNA CCs are in an equilibrium state , in which tau retains its solution-like dynamics . Next , we investigated the reversibility of tau187-RNA complex coacervation . Tau187-RNA CCs were prepared again and incubated by cyclically ramping the temperatures ( 1°C/min ) upwards and downwards , while the absorbance at λ = 500 nm was monitored , referred to as turbidity hereafter . Ramping rates of 0 . 5°C/min and 1°C/min were tested , but the results shown to be indistinguishable . Microscopy images were concurrently acquired at low and high turbidity , confirming the appearance and abundance of CC droplets correlating with turbidity increase , and vice versa ( Figure 1E ) . The turbidity-temperature curves show that at high temperature , samples became turbid with Abs500 ~1 . 5 and abundance of CCs , while at low temperature , samples became transparent with Abs500 ~0 and absence of CCs . This demonstrates tau187-RNA CC formation is favored at higher temperature , following clearly a lower critical solution temperature behavior ( LCST ) ( Figure 1E ) ( Siow et al . , 1972 ) . By cycling the temperature , we robustly and reversibly changed the tau187-RNA mixture between a turbid state to a completely transparent state ( Figure 1E ) . The transition temperatures at which the turbidity emerged during heating and vanished during cooling stay invariant with repeated heating-cooling cycles . The method of extracting a cloud point for the LCST transition temperature from such data will be described in detail in the next section . Importantly , the history of temperature change does not affect the resulting state . Hence the formation and dissolution of tau187-RNA CCs are reversible and consistent with a path-independent equilibrium process . We point out that the maximum turbidity value successively decreases with each heating cycle ( Figure 1E ) , even though the transition temperatures remain invariant . This can be attributed to slow degradation of RNA with time , ( as demonstrated in Figure 1—figure supplement 2 ) by verifying an altered turbidity change in the presence of RNase or RNase inhibitor . It is understood that upon gradual heating of the solution phase , the mechanism of LLPS proceeds via a nucleation process ( Berry et al . , 2015 ) , and hence there is a kinetic barrier evidenced by the observed hysteresis in Figure 1E . Nonetheless , we conclude that the final tau-RNA CC state reached upon heating is a true thermodynamic state , and thus can be modeled by an equilibrium theory of phase separation . To understand the principles and governing interactions driving tau-RNA CC formation , we constructed a phase diagram for tau187-RNA CC by measuring the transition temperature – to be described in greater detail below – as a function of protein concentration and salt concentration . We first recorded tau187-RNA turbidity at various [tau] , [RNA] and [NaCl] values , ranging from 2 to 240 μM , 6–720 μg/mL and 30–120 mM , respectively . Titrating RNA to tau187 , the turbidity was found to be peaked when [RNA]:[tau] reached charge matching condition at which the charge ratio between net positive and negative charges was 1:1 ( which for tau187 and RNA used in this study corresponded to [tau187]:[RNA]=1 μM: 3 μg/mL ) , validating once more that LLPS is driven by complex coacervation ( CC ) ( Figure 1—figure supplement 3 ) . Henceforth , all phase diagram data are acquired at a charge matching condition between RNA and tau . Titrating NaCl to tau187-RNA , CC formation showed a steady decrease of turbidity ( Figure 1—figure supplement 3 ) . Combined , these demonstrate that tau187-RNA CC favors the condition of charge balance and low ionic strength , which is consistent with known properties of CC and previous findings ( Zhang et al . , 2017 ) . We next investigated the phase separation temperatures under various sample compositions . Tau187-RNA CCs were prepared with a fixed [tau]:[RNA] ratio corresponding to the condition of net charge balance . Therefore , the composition of tau187-RNA CC can be determined by [tau] and [NaCl] . Samples were heated at 1°C/min between T = 15–25°C , while the turbidity was monitored . The turbidity-temperature data of the heating curves were then fit to a sigmoidal function , so that the cloud point temperature , Tcp , could be extracted as shown in Figure 2A ( Tcp was determined from heating curves out of practical utility; Tcp from cooling curves is possibly closer to thermodynamic transitions ) . The experimental cloud-point temperature Tcp for CC formation as a function of [tau] and [NaCl] are shown ( as points ) in Figure 2B and Figure 2C . The experimental data points show that increasing [tau] lowers Tcp , favoring CC formation , while increasing [NaCl] raises Tcp , disfavoring CC formation . Such trends were observed at two [NaCl] and two [tau] values , respectively ( Figure 2B and C ) . Experimentally , Tcp was determined for a range of [tau] and [NaCl] conditions ( see Figure 2—figure supplement 1 ) . We point out that there is certain level of variability in the observed Tcp , which can result from pH fluctuation of the ammonium acetate buffer upon tau-RNA addition , as well as RNA degradation as demonstrated in Figure 1—figure supplement 2 . The features of the Tau-RNA CC phase diagram were also investigated by comparing tau187 and tau114 . Tau187-RNA CC and tau114-RNA CC were prepared with 20 μM tau187 and 28 μM tau114 , so that the total concentration of polymer , that is tau and RNA , reaches 0 . 5 mg/mL . Turbidity was recorded at varying [NaCl] . Similar to the observation with tau187-RNA CC , tau114-RNA CC showed decreasing turbidity at increasing [NaCl] ( Figure 2—figure supplement 2 ) . The [NaCl] values where turbidity reaches 0 were estimated as 131 mM and 150 mM for tau187 and tau114 , respectively , implying CC formation is more favorable with tau114 that hence can sustain higher [NaCl] . Based on this , 20 μM of tau187 , 131 mM of NaCl and room temperature , 20°C , were used as the phase separation conditions ( [tau] , [NaCl] and Tcp ) for tau187 , and 28 μM , 150 mM and 20°C for tau114 . These two experimental conditions were used in the next section for comparing the two constructs of tau . We next used the FH-VO model to fit the experimental data for the tau187-RNA CC system , as is commonly done in LLPS studies . Despite its theoretical deficiencies , the FH-VO model is commonly used for its simplicity and ease of implementation . Our system consists of five species: tau187 , RNA , monovalent cation ( Na+ ) , anion ( Cl- ) and water . For simplicity , we explicitly consider only the effect of excess salt , and do not include polymer counterions . The FH-VO model maps these five species onto a three-dimensional lattice ( Figure 2D ) . Each polymer is treated as a uniform chain with degree of polymerization N and average charge per monomer σ . N was taken as the average chain length of the species ( one for monovalent ions ) . The charge density σ of RNA , monovalent ions and water were set to 1 , 1 and 0 , respectively . The values for σ of tau187 or tau114 were calculated from the net charge at neutral pH divided by the chain length . The composition of the species is expressed in terms of the volume fraction ϕ of the occupied lattice sites , which are proportional to the molar concentrations ( see Materials and methods for details ) . As in experiments , tau187-RNA CCs were prepared at fixed [tau]:[RNA] and [Na+]:[Cl−] ratios . Under these two constraints , the volume fraction of all five species in tau187-RNA CC listed above can be determined with two variables , [tau] and [NaCl] , which are experimentally measurable . Given N , σ , [tau] , [NaCl] and Tcp , the task is to find ϕtauI and ϕtauII , the volume fractions of tau in the dilute and dense , coacervate , phase at equilibrium , that is the binodal coexistence points . The model and procedure is described in detail in the Materials and methods . For each experimental observation of Tcp determined for a given [tau] and [NaCl] ( Figure 2—figure supplement 1 ) , the FH-VO expression has one unknown parameter , the Flory-Huggins χ term . The Flory-Huggins χ parameter is introduced as an energetic cost to having an adjacent lattice site to a polymer segment occupied by a solvent molecule ( Brangwynne et al . , 2015 ) . Here , we take χ to be an adjustable parameter , such that given a suitable expression for χ , the complete binodal curve can be modeled with the FH-VO theory . Consequently , we first solved for χ at each given experimental condition , so that the theoretical binodal curve intersects the experimental data point . Figure 2E shows two representative examples of a theoretical binodal curve ( solid line ) intersecting a single experimental data point at the given [NaCl] and [tau] . This procedure gives an empirical χ parameter for each experimental data point , as collated in Figure 2F as a function of 1/Tcp . We then performed to this set of experimental data a least-squares fit of the empirical χ parameter to the form A + B/T ( Figure 2F ) , yielding an expression of the temperature dependence of χ ofχ ( T ) =1 . 8-390T , R2=0 . 67 A temperature dependence of χ in the form of Equation 1 ( consistent with the observed LCST ) , can originate from hydrophobic interactions between non-polar groups , whose interaction strength tends to increase with temperature ( Lin et al . , 2018; Dias and Chan , 2014 ) . This explanation has also been used to describe cold denaturation of proteins ( Dill et al . , 1989 ) . Finally , from this expression for χ ( T ) , we computed the binodal curves that establishes the phase coexistence as a function of Tcp , [tau] and [NaCl] , shown as solid lines , only for the dilute phase coexistence for Tcp vs [tau] ( Figure 2B ) and Tcp vs [NaCl] ( Figure 2C ) . For the full phase diagram showing both dilute and dense binodal curves see Figure 2—figure supplement 3 . The experimental data ( shown as points ) and computed binodal curves both exhibited a decreasing Tcp with increasing [tau] and an increasing Tcp with increasing [NaCl] . This simply establishes that tau-RNA CC favors higher tau concentrations in the 1–240 μM range and lower ionic strength in the 30–120 mM range tested here . Binodal curves for tau114-RNA CC were also computed and are compared with tau187-RNA CC , along with experimental data ( Figure 2—figure supplement 2 ) . Comparison of the two constructs shows that tau114-RNA CC has a lower Tcp than tau187-RNA CC , suggesting it is more favorable to phase separation . This qualitatively agrees with experimental observations . Notice that the shorter tau114 has a slightly higher propensity to form CC as compared to the longer tau187 fragment , an observation that is opposite of what one would expect from purely entropic considerations based on the mixing of homopolymers or simple coacervation . One possible explanation could be the increased charge density of tau114 with respect to tau187 , indicating the importance of both charge sequence and charge density for the phase diagram . Additional short-ranged sequence-specific interactions between tau114 and RNA that are not present in tau187 is another possibility that is not considered in the present model . Although the FH-VO model can be brought into agreement with experiment through a judicious choice of χ , it is fundamentally unsound from a theoretical perspective , noticeably because it neglects connectivity between charges on the same chain . This is a severe limitation because it is expected that subtle difference in primary amino acid sequences may have a profound effect on the phase diagram . A particularly appealing alternative to gain insights into the thermodynamics of LLPS is to perform field theoretic simulations ( FTS ) on a physically motivated polyelectrolyte model ( Figure 3 ) , in which each amino acid is represented by a single monomeric unit of length b in a coarse-grained bead-spring polymer model . The charge of each segment is unambiguously assigned from the particular amino acid charge at pH 7 . 0 . In addition to harmonic bonds between nearest neighbors , which enforces chain connectivity , all segment pairs interact via two types of non-bonded potentials: a short-ranged excluded volume repulsion and a long-range electrostatic interaction between charged monomers ( see Figure 3 ) . We take the polymers to be in a slightly good solvent , meaning that favorable interactions between monomers and solvent cause chain swelling . In such cases , the excluded volume interaction is modeled as a repulsive Gaussian function between all monomer pairs with a strength that increases with solvent quality ( Doi and Edwards , 1988 ) . Conversely , as solvent quality decreases , the excluded volume repulsion decreases , approaching zero at the so-called theta condition . In the present case , we limit ourselves to the case where the excluded volume is positive and small , that is a good solvent near the theta condition . Simulations are performed using a single excluded volume strength , v , identical for all monomers , which is an input parameter in the model and can be adjusted to parameterize the favorable monomer-solvent interactions . Additionally , the long-range electrostatic interactions are described by a Coulomb potential in a screened , uniform , dielectric background . The length scale of the electrostatic interactions is parameterized by the Bjerrum length lB , which is the distance at which the electrostatic interactions become comparable to the thermal energy kBT and is defined as ( 2 ) lB=e24πϵ0ϵrkBT where e is the unit of electronic charge , ϵr the dielectric constant ( ϵr=80 for water ) , and ϵ0 the vacuum permittivity . The main features of the model used for FTS here are the inclusion of chain connectivity , charge sequence-dependence for the electrostatic interactions based on the primary amino acid sequence of tau , solvation effects which are parameterized by the single excluded volume parameter , v , and an electrostatic strength parameterized by the Bjerrum length , lB . FTS is performed in implicit solvent with a uniform dielectric background . We assume that the polymer chains are in a fully dissociated state , and we do not explicitly represent counter ions . The effect of excess salt is included in our model by introducing point charges explicitly , which engage in Coulomb interactions with all other charged species and repel other ions and polymer segments at short distances by the same Gaussian excluded volume repulsion . By introducing explicit small ions in this manner , we are neglecting strong correlations such as counter ion condensation; however , we are allowing for weak correlations of the Debye-Hückel type . The explicit addition of salt will serve to screen the electrostatic interactions and inhibit the driving force for CC , in agreement with the experiments . Details of the FTS protocol are described in the Materials and methods . By performing FTS at various state points and computing equilibrium properties , we first set out to fully explore the parameter space relevant for LLPS in this model . This involves running simulations at different conditions analogous to experiments . For each simulation , the thermodynamic state of the system is determined by specifying a particular value for the dimensionless excluded volume parameter v/b3 , the dimensionless Bjerrum length lB/b , and the dimensionless monomer number density ρb3 . Figure 4 shows the final polymer density configuration for two representative simulations at a monomer density of ρb3=0 . 22 at different thermodynamic conditions ( see caption for Figure 4 for details ) . Although the bulk density is fixed and identical for the two cases , the local polymer density is free to fluctuate . The left simulation box ( Figure 4 ) shows a case where a single phase is favored , indicated by a nearly homogenous polymer density throughout the simulation box ( white/blue ) . This is contrasted by the right simulation box ( Figure 4 ) depicting the case where the system phase separates into a dilute polymer-deplete region ( white ) and a dense polymer-rich droplet region ( red ) —the coacervate phase with the color signifying the polymer density . Figure 4 shows that given suitable parameterization , FTS can be used to study complex coacervation of a coarse-grained tau-RNA model . Given this observation , we next map out the full phase diagram in the parameter space of the model while fixing the physical parameters of charge sequence , chain length , and chain volume fractions that are consistent with the experimental conditions . The parameters to be explored in connection with phase behavior are the strength of the interactions in the polyelectrolyte model: the excluded volume strength v and the Bjerrum length lB . A direct comparison between FTS and the experimental phase diagram will be deferred until the following section . The phase coexistence points ( binodal conditions ) for a given value of the excluded volume v and Bjerrum length lB can be obtained by running many simulations over a range of concentrations , and finding the concentration values at which the chemical potential and the osmotic pressure are equal in both phases ( see Figure 5—figure supplement 1 ) . The procedure is described in the SI and is repeated for many different v and lB combinations . The resulting phase diagram will be a three-dimensional surface which is a function of ρ , v , and lB . In Figure 5A , we show a slice of this surface along the lB-ρ plane with a fixed value of v=0 . 0068 b3 , and in Figure 5B we show a slice along the v-ρ plane with a fixed lB=1 . 79 b ( at T = 293 K , Equation 3 ) . It should be noted that Figure 5 presents the first complete phase diagrams from FTS presented in the literature of a theoretical model describing a biological complex coacervate system . From Figure 5A , one can see that lB and v have counteracting effects , namely increasing v that is caused by increased solvent quality destabilizes the coacervate phase and favors the single phase , whereas increasing lB that is caused by reduced electrostatic screening favors coacervation , and destabilizes the single phase . The physical interpretation of the trends in Figure 5 is that the actual binodal for the experimental system will depend on two competing features: the solvent quality proportional to v , which inhibits coacervation , and the electrostatic strength of the media proportional to lB which promotes coacervation . The FTS-derived phase diagram shown in Figure 5 provides a guide how to experimentally tune the window for complex coacervation by changing the relative contribution of the solvent quality or the dielectric strength . Experimentally , the solvent quality can be decreased by adding crowding agents or by changing the hydrophilic/hydrophobic amino acid composition , while the electrostatic strength can be controlled by the salt concentration . Increasing salt concentration tends to decrease the bare electrostatic strength by screening the charges , and this is predicted to stabilize the single phase solution mixture against coacervation , in agreement with experimental observation . We explore these ideas further below in the context of tau coacervation in vivo . Despite the simplicity of the coarse-grained description , the model predicts that these two competing parameters , excluded volume vs . electrostatic interactions , are nearly balanced around physiological salt concentration , temperature , and protein concentration . Assuming that the relative dielectric constant for water is ϵr=80 , and that the segment size b is approximately equivalent to the distance between Cα carbons , that is b~4Å , it follows that lB=1 . 75b at 300 K . ( lB=0 . 7 nm at 300 K ) . In the lB-ρ plane ( shown in Figure 5A ) , at the cross-section of lB=1 . 75 , three points for ρb3 are indicated that correspond to 1 , 5 , and 10 μM for tau concentrations at 300 K . Here , we have implicitly assumed that at physiological temperature and in a crowded cellular environment tau is near the theta condition , and thus v is small . This analysis suggests that small modulation in the experimental conditions , such as changes in the temperature or salt concentration , local pH or crowding effects ( via the excluded volume parameter v ) can readily and reversibly induce complex coacervation in vivo under physiological conditions . In the preceding section , we presented the phase diagram from FTS explicitly in terms of the model parameters of the excluded volume v and Bjerrum length lB . We now seek to compare our simulation results directly with the experimental phase diagram . This requires knowing precisely how the model parameters depend on temperature . We again take the monomer size b to be approximately the distance between the Cα carbons b~4 Å , and assume a dielectric constant of ϵw=80 for pure water . Although ϵr will depend on temperature , for simplicity , we treat this parameter as a constant such that the Bjerrum length lB~1/T . Thus , lB can be estimated at the experimental cloud point temperature directly from Equation 2 ( Figure 6A ) , which leaves only one unknown parameter v . The excluded volume parameter v can be related to the residue-residue non-Coulombic interaction potential as ( 3 ) ν=∫1-e- UrkBTd3r ( Equation 3 ) and is typically taken to be proportional to ( 1-θ/T ) where θ is the theta temperature , the temperature at which the chain follows ideal chain statistics ( Doi and Edwards , 1988; Stockmayer , 1955; Rubinstein and Colby , 2003 ) . For LCST behavior , it is customary to introduce the form v=-v0 ( 1-θ/T ) where v0 controls the magnitude of the excluded volume interactions ( Suzuki et al . , 1982 ) . This form of the excluded volume implies that at temperatures lower than the theta temperature , the excluded volume is repulsive ( v > 0 , meaning a good solvent ) and for temperatures above the theta point , the excluded volume becomes attractive ( v < 0 , poor solvent conditions ) . By adjusting the excluded volume in FTS to fit a subset of the experimental data , ( shown in Figure 6B ) , we then perform a linear fit to obtain a value of v0=0 . 25b3 and θ=309 K . Note that in the range of temperatures considered , the excluded volume remains positive . However , for temperatures higher than θ , when the excluded volume becomes negative , the polymer chain will collapse , consistent with the observation in the literature ( Bianconi et al . , 2012 ) , which showed that tau undergoes a thermal compaction at high temperatures due to entropic factors ( Stockmayer , 1955 ) In such high-temperature regimes , a more sophisticated treatment is needed; however , all our experimental conditions remain below this threshold . Having mapped the two model parameters v and lB to the experimental temperature , we can compare directly the FTS with the experimental results ( Figure 6C ) . The calculated FTS data points under the condition of low-salt concentration are shown as filled green squares in Figure 6C . Next , explicit salt ions were introduced as point charges to simulate an excess salt concentration of 120 mM . We make the assumption that the salt is equally partitioned in both phases , and thus the concentration of salt is a constant , allowing us to sweep the polymer concentration at fixed salt concentration to find the phase coexistence points . A more detailed FTS study of salt partitioning performed using a Gibbs ensemble method found that under conditions of nearly charge-balanced polymers , as is the case in the system of this study , the salts are nearly equipartitioned and counterion condensation is not a dominant factor ( private communication with Danielsen SPO , McCarty J , Shea J-E , Delaney KT , Fredrickson GH on the "Small ion effects on Self-Coacervation phenomena in block polyampholytes" ) . Simulations performed in this manner with explicit salt are shown as open green squares in Figure 6C . The FTS data clearly demonstrate that the effect of added salt is to stabilize the single solution phase , and to raise the binodal closer to physiological temperature 37°C , in agreement with experiments ( filled red and blue circles Figure 6C ) . The complementary dense branch of the binodal curve is also predicted from FTS and is shown in Figure 6—figure supplement 1 . To demonstrate that FTS of tau-RNA LLPS can be applied to consider the effect of post translational charge modification of tau , we tested FTS of phosphorylation as an example . We first phosphorylated tau187 in vitro using mouse brain extract , and confirmed the occurrence of phosphorylation using SDS-PAGE and western blot analysis ( see Materials and methods and Figure 6—figure supplement 2 for details ) . Next , we titrated phosphorylated tau187 ( P-tau ) and non-phosphorylated tau187 ( Tau ) with RNA , while recording turbidity values of the sample . Results showed that after phosphorylation the optimal droplet amount emerged at a lower RNA concentration , as supported by both turbidity and microscopy ( Figure 6—figure supplement 3A , B ) . Assuming P-tau-RNA LLPS also follows complex coacervation , we can estimate from the results > 5 additional negative charges are added to P-tau , which is consistent with the fact that phosphorylation adds additional negative charges to protein . At optimal droplet conditions , we further prepared P-tau-RNA and Tau-RNA samples using 20 μM P-tau with 50 μg/mL RNA and 20 μM Tau with 100 μg/mL RNA , respectively . We titrated the two samples with NaCl , while monitoring turbidity . Results showed the droplet amount in P-tau-RNA sample vanishes at [NaCl]~80 mM , while in Tau-RNA sample vanishes at [NaCl]~260 mM ( Figure 6—figure supplement 3C , D ) . These results demonstrated phosphorylation reduces the propensity of tau-RNA LLPS . For qualitative comparison , we performed FTS on our parameterized tau187 model at equivalent temperature and concentration assuming complete phosphorylation of serine residues S262 , S396 , S404 , S416 , and S422 , by giving them a charge of −2 in the FTS model . Phosphorylated serines were identified according to the literature ( Mair et al . , 2016 ) . For simplicity we assumed a fully phosphorylated tau , which serves as a limiting case , recognizing that in reality , tau will be phosphorylated with some level of variability . To maintain charge neutrality in the simulation box , we add excess salt concentration , so that the total concentration of small ions is estimated to be ~292 mM . Figure 6—figure supplement 4 shows that added phosphates increases the concentration of tau in the solution phase and decreases tau concentration in the coacervate phase ( narrowing the two phase window ) , which is consistent to the narrowed two phase window observed in Figure 6—figure supplement 3C . This is likely due to electrostatic repulsion between the phosphorylated serines and the negatively charged RNA . In contrast , hyper-phosphorylated full-length tau has been shown to favor simple coacervation ( Wegmann et al . , 2018 ) . In full-length tau , most of the phosphorylated sites are not in the positively charged repeat domain region , indicating that its LLPS might follow the same principles as the self-coacervation seen in polyampholytes ( Delaney and Fredrickson , 2017 ) . Looking at the experimental and calculated phase diagrams ( Figure 2B and C ) , it is seen that under physiological conditions ( Tcp ~37°C , [NaCl]~100 mM ) it is principally feasible for cells to tune the formation of tau-RNA CCs . This has important implications for studying the physiological roles of tau-RNA CCs , and thus we asked if tau-RNA CCs could indeed exist in a biologically relevant media in the presence of living cells . Both the FH-VO theory and FTS predict that the conditions of high-protein concentration , low ionic strength , high temperature and high crowding reagents ( leading to solution conditions with a lower effective excluded volume parameter to model the poorer solvent environment in an implicit solvent model [Jeon et al . , 2016] ) would independently favor tau-RNA CC formation . Using these tuning parameters as a guide , we designed several experiments to test the ability for tau-RNA CCs to form in a co-culture with H4 neuroglioma cells . We incubated H4 cells with tau187/tau114-RNA under CC conditions at varying temperatures , polymer concentrations and crowding reagent concentrations . At low polymer concentrations ( 10 μM tau , 30 μg/ml RNA ) no LLPS was observed in the cellular media ( Figure 7 , first column ) , where increasing the temperature to 37°C did not apparently influence the solution phase ( Figure 7 , first column , first and third row ) . However , when tau and RNA concentrations were increased ( 100 μM tau , 300 μg/ml RNA ) LLPS could be observed ( Figure 7 , second column ) . Further , LLPS could also be achieved by adding an additional crowding reagent ( here PEG ) to low concentration samples of tau and RNA ( Figure 7 , third column ) . As predicted , LLPS of tau-RNA CC was modulated by ( i ) temperature , ( ii ) tau and RNA concentration and/or ( iii ) the presence of crowding reagent PEG ( Figure 7 ) . Lowering the temperature to 18°C significantly reduced the number and size of fluorescent droplets , demonstrating that tau-RNA LLPS is indeed tunable by temperature , and demonstrate the biological consequence of the LCST behavior ( Figure 7 , first and third row ) . These results were consistently found for both tau187 and tau114 systems . The successful application of FTS for tuning and predicting tau-RNA CCs in cellular media is a first step toward understanding the physiological condition under which tau-RNA LLPS , which follows the CC mechanism , can occur . Notice that our truncated tau construct has been demonstrated to undergo LLPS at similar conditions ( [tau] , [RNA] , [NaCl] and temperature ) compared with full length tau , 2N4R , in vitro ( Zhang et al . , 2017 ) . The conditions described for LLPS here suggests that conditions exist in vivo under which LLPS by complex coacervation may be achieved by biological regulation mechanisms , and under conditions where tau and the LLPS forming constituents are available in the cytoplasm .
We report here the first detailed picture of the thermodynamics of tau-RNA complex coacervation . The observation of an LCST phase diagram implies that although electrostatic interactions are key to CC formation , factors that contribute to solvation entropy gain are key to driving liquid-liquid phase separation . We have computed the first approximation-free theoretical phase diagram for tau-RNA complex coacervation from FTS , where we introduced a temperature-dependent excluded volume term . Simulations show a competition between electrostatic strength ( parameterized by the salt concentration ) and excluded volume ( parameterized by the solvent quality ) . This knowledge can be used to design experiments that perturb this parameter space in vivo , as well as predict or understand biological mechanisms that may be favorable towards liquid-liquid phase separation . As a proof of this concept we have shown that by deliberately changing salt concentration , temperature , and solvent quality ( by the addition of PEG ) , we can make tau-RNA LLPS appear or disappear in cellular medium with live cells . Interestingly , we find that without any adjustable parameters our simulations predict that tau-RNA is positioned near the binodal phase boundary around physiological conditions . This suggests that small and subtle changes within the cellular environment may be sufficient to induce LLPS in otherwise healthy neurons . Even if the conditions that induce LLPS in the cell is transient , the LLPS state can facilitate irreversible protein aggregation if aggregation-promoting factors are already available , giving credence to the idea that LLPS may play a role in neurodegenerative diseases . However , we speculate that LLPS is reversible in the majority of biological events that drive LLPS , making it hard to observe this state within the cellular context .
Unless stated , a 20 mM ammonium acetate buffer at pH 7 . 0 was used and referred to here as final buffer . Tau , RNA , NaCl , PEG and other stocks were prepared using final buffer . Measurements were taken in final buffer at room temperature unless stated . N-terminal truncated , microtubule binding domain containing tau187 ( residues 255–441 with a His-tag at the N-terminus ) were used for in vitro studies . The cloning , expression , and purification have been previously described ( Peterson et al . , 2008; Pavlova et al . , 2009 ) . The single cysteine variant of tau187 ( tau187C291S ) were generated via site-direct mutagenesis . E . coli BL21 ( DE3 ) cells previously transfected were cultured from frozen glycerol stock overnight in 10 mL luria broth ( LB ) which was used to inoculate 1 L of fresh LB . Culturing and inoculation were performed at 37°C with shaking of 200 rpm . At OD600 of 0 . 6–0 . 8 , tau187 variant expression was induced by incubation with 1 mM isopropylß-D-thiogalactoside ( Sigma Aldrich ) for 2–3 hr . Cells were harvested by centrifugation for 30 min at 5000 × g ( Beckman J-10; Beckman Instruments , Inc ) , and the pellets were stored at −20°C until further use . Cell pellets were resuspended in lysis buffer ( Tris-HCl pH 7 . 4 , 100 mM NaCl , 0 . 5 mM DTT , 0 . 1 mM EDTA , 1 mM PMSF ) with 1 Pierce protease inhibitor tablet ( Thermo Fisher ) . Lysis was initiated by the addition of lysozyme ( 2 mg/ml ) , DNase ( 20 µg/ml ) , and MgCl2 ( 10 mM ) and incubated for 30 min on ice . Lysate was then heated to 65°C for 13 min , cooled on ice for 20 min and then centrifuged to remove the precipitant . The supernatant was loaded onto a Ni-NTA agarose column pre-equilibrated with wash buffer A ( 20 mM sodium phosphate pH 7 . 0 , 500 mM NaCl , 10 mM imidazole , 100 µM EDTA ) . The column was then washed with 20 ml of buffer A , 15 ml buffer B ( 20 mM sodium phosphate pH 7 . 0 , 1 M NaCl , 20 mM imidazole , 0 . 5 mM DTT , 100 µM EDTA ) . Purified tau187 was eluted with buffer C ( 20 mM sodium phosphate pH 7 . 0 , 0 . 5 mM DTT , 100 mM NaCl ) supplemented with varying amounts of imidazole increasing from 100 mM to 300 mM . The protein was then concentrated via centrifugal filters ( MWCO 10 kDa; Millipore Sigma ) and the buffer was exchanged into final buffer by PD-10 desalting column ( GE Healthcare ) . The final protein concentration was determined by UV-Vis absorption at 274 nm using an extinction coefficient of 2 . 8 cm−1mM−1 , calculated from absorption of Tyrosine [3] . Freshly eluted tau187C291S ( with one cysteine at site 322 ) was replaced in final buffer using a PD-10 desalting column ( GE Healthcare ) . Protein after PD-10 was labeled overnight at 4°C by immediately mixing with a 10-fold molar excess of the spin label ( 1-oxyl-2 , 2 , 5 , 5-tetramethylpyrroline-3-methyl ) methanethiosulfonate ( MTSL; Toronto Research Chemicals ) , resulting in spin labeled tau ( tau187C291S-SL ) . Excess label was removed using PD-10 . The protein was concentrated using centrifugal filter ( MWCO 10 kDa; Amicon ) and the final protein concentration was determined by UV-Vis absorption at 274 nm as mentioned above . Non-labeled tau187C291S was used in order to achieve spin dilution . Cw EPR measurements were carried out using a X-band spectrometer operating at 9 . 8 GHz ( EMX; Bruker Biospin , Billerica , MA ) and a dielectric cavity ( ER 4123D; Bruker Biospin , Billerica , MA ) . 100 μM tau187C291S-SL was mixed with 400 μM tau187C291S to reach 20% spin labeling . Samples under droplet forming condition were prepared by adding 1 . 5 mg/mL RNA , and tau samples under aggregation-inducing conditions prepared by adding 125 μM heparin ( 15 kDa average MW; Sigma-Aldrich ) . A sample of 4 . 0 μL volume was loaded into a quartz capillary ( CV6084; VitroCom ) and sealed at both ends with critoseal , and then placed in the dielectric cavity for measurements . Cw EPR spectra were acquired by using 6 mW of microwave power , 0 . 5 gauss modulation amplitude , 100 gauss sweep width , and 8–64 scans for signal averaging . The recorded cw EPR spectra were subjected to single- or double-component simulation . EPR simulation and fitting were performed using MultiComponent , a program developed by Christian Altenbach ( University of California , Los Angeles ) . For all spectra fitting , the magnetic tensors A and g were fixed and used as constraints as previously reported ( Pavlova et al . , 2016 ) . These values are Axx = 6 . 2 G , Ayy = 5 . 9 G , Azz = 37 . 0 G , and gxx = 2 . 0078 , gyy = 2 . 0058 , and gzz = 2 . 0022 . For soluble tau , the cw EPR spectra were best fitted with a single-component simulation and the rotational diffusion constant ( R ) can be extracted . The rotation correlation time τR was calculated using τR = 1/ ( 6R ) . For tau-heparin aggregates , the cw EPR were subjected to double-component simulation , where the parameters of the fitted single-component were used as a mobile-component . The immobile-component were set to be identical to the mobile-component , except the diffusion tensor tilt angle βD = 36° and the order parameter S . The fitting parameters were limited at a minimum , which includes the population , p , rotational diffusion constants of mobile- and immobile-component , R1 and R2 , and the order paramter , S of the immobile-component . The fitted immobile-component were used to represent the rotational correlation time for tau-heparin fibrils . For tau-RNA CC . the cw EPR spectra were subjected to both single- and double-component fitting . Comparing the two fitting schemes showed that singl-component fitting has almost overlapped the cw EPR spectra , while double-component fitting results in a immobile-component population of ~10% ( data not shown ) . This showed that tau-RNA CC cw EPR spectra can be sufficiently fit with single-component . The fitted rotational correlation time was calculated and plotted against tau-heparin samples . Turbidity of samples at room temperature were represented by optical density at a 500 nm wavelength ( OD500 ) , using a Shimadzu UV-1601 spectrophotometer ( Shimadzu Inc ) . The amount of coacervates in a sample were approximated to be propotional to its OD500 . Tubidity of samples at ramping temperatures were represented by OD500 measured using Jasco J-1500 CD Spectrometer ( JASCO Inc ) equipped with temperature controller and spectrophotometer . 120 μL of 20 μM tau187C291S , 60 μg/mL polyU RNA and 30 mM NaCl in working buffer were prepared in a 100 μL cuvette ( Starna Scientific Ltd ) and kept at 4°C for 5 min before cycling . Heating and cooling temperatures were ramped at 1°C/min while OD500 was monitored . Bright field images were examed to confirm the presence of tau-RNA CC . 100 μM tau187C291S and 300 μg/mL polyU RNA was mixed in presence of 20 mM ammonium acetate and 30 mM NaCl . 10 μL of the mixture was pipetted onto a microscope slide with a cover slide gapped by two layers of double-sided sticky tape . Temperatures were controlled using an incubator . Bright field images were acquired using a spectral confocal microscope ( Olympus Fluoview 1000; Olympus , Center Valley , PA ) . It was shown by fluorescence microscopy in protein-RNA LLPS that protein is concentrated inside the droplet ( Patel et al . , 2015; Elbaum-Garfinkle et al . , 2015 ) . For representing tau inside the droplets with measurement taken from droplet suspension , we quantified the percentage of tau present as droplets . After mixing and centrifuging 60 μL droplet suspension of 400 µM tau187/322C and 1500 µg/mL polyU , ~1 μL dense phase was generated with clear boundary against dilute phase . Dissolving dense phase in high concentration of NaCl resulted in transparent solution thus UV absorption can be measured . Due to the difficulty of preparing large volume of pure dense phase , we can only underestimate the tau and polyU concentration in dense phase . Since tau and RNA have different UV absorbance spectra , fitting spectra of the tau-RNA mixed sample with those of pure tau and polyU generated the concentration of both . Fitting results showed that over 99% of the tau and over 99 . 9% of polyU were condensed inside the dense phase . This partitioning guaranteed that the property of tau in the droplet suspension represents those in the droplets . Protein ( tau187 or K18 ) was labeled with Alexa Fluor 488 or 555 5-SDP ester ( Life Technologies ) according to the suppliers instructions . After labeling , 100 mM glycine was added to quench the reaction and the proteins were subjected to Zeba desalting columns ( Thermo Scientific ) to remove any unreacted label . Average label incorporation was between 1 and 1 . 5 moles/mole of protein , as determined by measuring fluorescence and protein concentration ( Amax × MW of protein / [protein]×εdye ) . H4 neuroglioma cells ( ATCC HTB-148 ) were cultured in DMEM supplemented with 10% FBS , 100 μg/mL penicillin/streptomycin . Cultures were maintained in a humidified atmosphere of 5% CO2 at 37°C . Tau protein ( 1:20 labeled K18:unlabeled tau114 ) , RNA , PEG , and media ( DMEM , 10% FBS , 1% Pen/Strep ) were mixed at the indicated concentrations and added to cells at varying temperatures . Images were obtained on a Leica SP8 Resonant Scanning Confocal . Phosphorylation of tau was performed as previously described ( Despres et al . , 2017 ) . In brief , tau protein ( 40 μL at 6 mM ) was mixed with 200 μL mouse brain extract and incubated overnight at 37°C in phosphorylation buffer ( 40 mM HEPES pH 7 . 3 , 2 mM MgCl2 , 5 mM EGTA , 2 mM DTT , 2 mM ATP , 1 μM okadaic acid , protease inhibitors ) . After incubation , samples were centrifuged and the supernatant was buffer exchanged using zeba desalting columns ( Thermo Fisher ) into buffer ( 20 mM ammonium acetate , pH 7 ) . Concentration was determined by BCA assay . Phosphorylation was confirmed using a western blot assay to look at phospho-epitopes 396/404 using PHF-1 Antibody ( Peter Davies ) . FH-VO is based on a Flory-Huggins ( FH ) treatment , where the polymer system is mapped onto a lattice . Voorn and Overbeek extended the FH formalism to polyelectrolytes by including long-ranged electrostatic interactions with a Debye–Hückel term . The resulting expression for the free energy of mixing ( ΔGmix ) per lattice site is ( S1 ) ΔGmMkBT=∑ϕiNilnϕi−α[Σσiϕi]32+Σχijϕiϕjwhere M=V/lw3 is the total number of lattice sites . In Equation S1 , the index i refers to one of the five species . Ni is the degree of polymerization for species i . For tau187 and tau114 , Ni equals to the length of the polypeptides ( Figure 1—source data 1 ) ; while for RNA , Ni is estimated by the average MW 900 kDa for polyU RNA and the MW of condensated uridine monophosphate , 306 Da . For monovalent ions and water , Ni = 1 . σi is the average charge per monomer , which is determined by ( net charge ) /Ni . The net charges of tau at experimental pH conditions ( pH = 7 ) were estimated based on primary sequences in Figure 1—source data 1 , using pepcalc . com . σi for other species were listed in Figure 2—source data 1 . In FH-VO model , σi is fixed . We also consider a modified version , a FH-VO-CR model , where σi of RNA is set to a function of temperature as discussed further below . In Equation S1 , ϕi is the volume fraction of species ( tau , RNA , Na+ , Cl- , H2O ) . ϕi was computed by ϕi=ci×Ni×1cw where ci is the molar concentration and cw the molar concentration of pure water computed from water volume molarity: cw = 55 . 56 mol/L . In experiments , ctau and cRNA were designed to reach a 1:1 charge ratio , therefore , we have NRNA ×cRNA = 11 × ctau = 11×[tau] . In addition to NaCl , there is 20 mM ammonium acetate in the buffer . The total monovalent salt concentration is csalt=cNaCl+20 mM=NaCl+20 mM . Therefore , ϕi were calculated from experimental [tau] and [NaCl] as , ( S2 ) ϕtau=[tau]×207×1cw ϕRNA=[tau]×11×1cwϕsalt= ( [NaCl]+20mM ) ×1cw ϕpolymer=ϕtau+ϕRNA=[tau]×218×1cw ϕwater=1−ϕpolymer−ϕsalt α is the strength of the electrostatic interactions defined as ( S3 ) α=23πlw3 ( e24πϵrϵ0kBT ) 3/2where lw is the length of a lattice , computed from cw , lw=1×10−3m3cwNA3 , ϵrϵ0 the water permitivity , ϵrϵ0=80×8 . 85×10-12 F/m , kB the Boltzmann constant and T the absolute temperature . χij is the Flory-Huggins interaction parameter between species i and j , which will be defined and discussed below . The three terms on the right-hand side of Equation S1 are respectively: ( 1 ) the ideal Flory-Huggins mixing entropy , ( 2 ) the mixing enthalpy due to Coulombic interactions based on Debye–Hückel approximation ( Hückel and Debye , 1923 ) and ( 3 ) the excess free energy to account for the non-Coulombic interactions , which can include contributions from water perturbation ( Fu and Schlenoff , 2016 ) , cation-π interaction ( Kim et al . , 2016 ) and dipole-dipole interactions ( Holehouse et al . , 2015 ) . Equation S1 has been successfully applied in PDMAEMA-PAA complex coacervate ( Spruijt et al . , 2010 ) . In this work , we refer to Equation S1 as FH-VO model , which is a minimal model for complex coacervation . A phase separation temperature , Tcp , was assigned to the cloud point of the sample . Tcp was determined by fitting normalized turbidity-temperature curves to a sigmoid function as followsnormalizedturbidity=11+exp ( −k× ( T−Tcp ) ) , to findT=Tcp ϕi and T can be converted from/to experimental conditions as described , where tau and RNA are added at a fixed charge neutrality ratio . Therefore , ΔGmix depends on four variables: total polymer volume fraction ϕpolymer=ϕp+ϕq , total salt volume fraction ϕsalt=ϕs++ϕs- , temperature T and Χ , a matrix of χpp , χpq , χps+ , … , χqp , χqq , … . A two-phase equilibrium exists where the sum of mixing free energy of two coexisting phases are lower than that of the homogeneous mixture . For simplicity , we adopt the assumption that the salt concentration in both two phases are identical ( Spruijt et al . , 2010 ) , leaving the system a binary mixture of polymer and buffer . Binodal compositions are defined by pairs of points on the curve of ΔGmixing vs . ϕpolymer that have common tangents , corresponding to compositions of equal chemical potentials of both buffer and polymer in dense and dilute phases . A binodal composition curve ( binodal curve ) was computed by finding the bi-tangent points of ΔGmixing vs . ϕpolymer at a series of ϕsalt at given temperature T and given parameters . Given ϕsalt , T and Χ , the mixing free energy is solely dependent on ϕ=ϕpolymer:fϕ=ΔGmixingϕpolymer A bi-tangent pair ϕ1 , f1 , ( ϕ2 , f2 ) was calculated by solving the set of nonlinear equations ( Rubinstein and Colby , 2003; Kwon et al . , 2015 ) , {∂∂ϕ|Φ=Φ1−∂∂ϕ|Φ=Φ2=0 ( f−Φ∂∂ϕ ) |Φ=Φ1− ( f−Φ∂∂ϕ ) |Φ=Φ2=0which was solved by R function nleqslv using Newton-Ralphson algorithm at given initial guess . Finally , the ϕpolymer and ϕsalt were converted into [tau] and [NaCl] as described . Our system consists of n total polymers made up of nτ tau molecules of length Nτ and np RNA molecules of length Np . Each amino acid is treated as a single Kuhn segment of length b . The solvent is treated implicitly with a uniform dielectric background ϵ . For simplicity , we only consider the symmetric case of Np=Nτ . Chain connectivity is enforced by a harmonic bond potential of the form βUbond=32b2∑α=1n∑j=1N ( |rα , j-rα , j-1| ) 2 where rα , j is the coordinates of bead j on chain α . In addition to chain connectivity , all monomers interact with a short-ranged excluded volume potential ( Doi and Edwards , 1988 ) . We take the well-known Edward's delta function model for the excluded volume interaction βUex=vδr where v is the excluded volume parameter ( Doi and Edwards , 1988 ) . The charge of each bead j for the tau molecule zτ , j is determined from the primary amino acid sequence with aspartic ( D ) and glutamic ( E ) acid being zτ , j=-1 , arginine ( R ) and lysine ( K ) being zτ , j=+1 and all other amino acids being neutral zτ , j=0 . The RNA chain is treated as a fully-charged chain with zp , j=-1 for all monomers . Charged segments interact via a long-ranged Coulomb potential βUel=lB zizjr with lB=e24πϵ0ϵrkBT being the Bjerrum length , e is the unit of electronic charge , ϵ0 is the vacuum permittivity , and ϵr is the dielectric constant . For a schematic depiction of the polymer physics model see Figure 3 in the main text . The model is ‘regularized’ by smearing all statistical segments over a finite volume instead of treating them as point particles ( Delaney and Fredrickson , 2016 ) . This is accomplished by endowing each bead with a Gaussian profile with a width on the order of the statistical segment length Γr= ( 3/πb2 ) 2 exp ( -3r2/b2 ) where r is a radial distance from the monomer center . As a consequence of this density smearing , the interactions between monomers ‘softens’ ( Villet , 2012 ) . The advantage of the coarse-grained polyelectrolyte model employed in this work is that it can be exactly converted to a statistical field theory by utilizing a Hubbard-Stratonovich transformation as described in Fredrickson ( 2006 ) . Invoking this transformation , the canonical partition function is expressed in terms of two fluctuating auxiliary fields w and φ which serve to decouple the excluded volume and Coulombic interactions , respectively ( Patel et al . , 2015; Li et al . , 2018b; Kwon et al . , 2014; Lee et al . , 2016; Boeynaems et al . , 2017 ) . In the statistical field representations the canonical partition function is ( S4 ) Z=Z0∫Dw∫Dφ exp ( −H[w , φ] ) where Z0 contains the ideal gas partition function and self-interaction terms . The field-theoretic Hamiltonian for this model is ( S5 ) Hw , φ=12v∫dr w ( r ) 2+18πlB∫dr |∇φ|2-nτlnQτw , φ-nplnQp[w , φ]where Qτ[w , φ] and Qpw , φ are the partition functions for a single tau and a single RNA molecule in the conjugate fields . These single chain partition functions can be computed using a Gaussian chain propagator such that ( S6 ) Ql[ψ]=1V∫drql ( r , Nl;ψ ) where l indexes the chain type ( tau/RNA ) and ψj=i Γ⋆w+zjφ with i=-1 and ⋆ a spatial convolution . The chain propagator qlr , j;ψ is constructed from a Chapman-Kolmogorov-type equation ( S7 ) ql ( r , j+1;ψ ) = ( 32πb2 ) 3/2exp[−ψ ( r , j+1 ) ]∫dr′ql ( r′ , j;ψ ) exp ( −3|r−r′|22b2 ) with initial condition qlr , 0;ψ=exp[-ψr , 0] . From the field theoretic Hamiltonian any thermodynamic observable may be computed as an ensemble average of a corresponding operator expressed in terms of the field configurations G~w , φ ( S8 ) ⟨G⟩=Z0Z∫Dw ∫DφG∼[w , φ]exp ( −H[w , φ] ) We stress that no additional approximations are made in moving from a particle-based model to a statistical field theory . The advantage of such a transformation is that the pairwise interactions between monomers are decoupled in favor of interactions between monomers and a complex-valued field . This transformation is particularly suited to our purposes here as conventional particle simulations can only study the earliest stages of protein aggregation . Field theoretic simulation ( FTS ) has been widely used to model synthetic polymers ( Delaney and Fredrickson , 2016 ) , including LLPS in polyelectrolytes and polyampholytes ( Delaney and Fredrickson , 2017; Lee et al . , 2008; Popov et al . , 2007 ) . The interested reader is directed to this literature for further detail of the method . Here we apply this powerful numerical method in a new context to model LLPS of a biological system ( tau-RNA under cellular conditions ) . The main advantage of this approach is that it does not rely on analytical approximations , and thus can be useful when comparing directly to experiment at conditions where such approximate theories break down . FTS allows one to numerically compute ensemble averages of the form of Equation S8 by sampling along a stationary stochastic trajectory in the space of the field variables . The method has been presented in detail elsewhere ( Delaney and Fredrickson , 2016; Fredrickson , 2006; Alexander-Katz et al . , 2005 ) . We use complex Langevin ( CL ) sampling ( Klauder , 1983; Parisi , 1983 ) to stochastically sample the auxiliary fields . The method involves promoting the fields to be complex-valued and numerically propagating the CL equations of motion ( S9 ) ∂w ( r , t ) ∂t=−λwδH[w , φ]δw ( r , t ) +ηw ( r , t ) ∂φ ( r , t ) ∂t=−λφδH[w , φ]δφ ( r , t ) +ηφ ( r , t ) where ηwr , t and ηφ ( r , t ) are real-valued Gaussian white-noise random variables with zero mean and variance proportional to the dissipative coefficients λw and λφ . A single FTS step involves computing the single chain partition functions for a given field configurations w , φ given by Equation S6 along with any operators G~w , φ followed by updating the field configurations according to Equation S9 . Under the condition that the system is ergotic , ensemble averages are computed as time averages over the CL trajectory . All FTS-CL simulations were performed in reduced units by scaling spatial lengths by a reference distance R0=b/6 corresponding to the prefactor in the scaling relation of an ideal homopolymer radius of gyration with respect to the chain length Rg=R0N1/2 ( Flory and Volkenstein , 1969 ) . Simulations were performed in a cubic box of length L=34 . 0 R0 using periodic boundary conditions . Fields were sampled with a spatial collocation mesh of 32 ( Anderson and Kedersha , 2006 ) sites . An exponential time difference ( ETD ) algorithm ( Villet and Fredrickson , 2014; Düchs et al . , 2014 ) with Δt=0 . 01 was used to numerically propagate the CL equations of motion Equation S13 . All simulations were performed on NVIDIA Tesla M2070 or K80 graphics processing units ( GPUs ) . The thermodynamic state of the system is fully determined by specifying a dimensionless excluded volume parameter B=v/R03 , a dimensionless Bjerrum length E=4πlB/R0 , and a dimensionless polymer chain number density C=ρR03 with ρ=∑lnlNl/V where l indexes the chain type ( tau or RNA ) . Additionally , we require the fraction of chains of each type ϕl=nlNl/∑lnlNl . In this work , we consider only a 1:1 charge ratio , which for the model shown in Figure 1 corresponds to ϕτ=0 . 954 and ϕp=0 . 046 . In order to compute the phase coexistence points from FTS , we need to compute the chemical potential μ and the osmotic pressure Π . The chemical potential operator for chain type l is μl∼=lnρR03ϕl−Nl−lnQl . For the pressure operator , we use the form given in Appendix B of Delaney and Fredrickson ( 2017 ) . The conditions for the stable coexistence of two phases is given by the chemical equilibrium condition ∑lνlμlI=∑lνlμlII and the mechanical equilibrium condition ΠI=ΠII . The stoichiometric coefficient νl ensures charge neutrality . The procedure we employ in this work is that of Delaney and Fredrickson ( 2017 ) and involves computing the chemical potential and pressure for a range of polymer concentrations . Figure 5-figure supplement 1 ( Left ) shows a plot of the osmotic pressure vs . the chemical potential for different polymer conentrations . The simulation data represent three branches: a dilute branch ( red ) , a conentrated branch ( blue ) , and an unstable branch ( orange ) . The equilibrium condition of equal chemical potential and equal osmotic pressure can be directly gleaned from the intersection of the dilute and concentrated branch . This gives the critical conditions for phase coexistence . Figure 5—figure supplement 1 ( Right ) show a plot of the chemical potential vs . polymer concentration . The critical chemical potential value is shown by the dashed horizontal line . The intersection of this line with the polymer concentration data points from FTS gives the dilute supernatant polymer concetration ρI and the coacervate concentration of the coexisting phase ρII . By repeating this procedure for many different thermodynamic conditions , we can construct the phase diagrams shown in Figure 5 of the main text . As discussed in the main text , the interaction parameter is decomposed into an entropic and enthalpic contribution χ=ϵs+ϵH/T . According to the Flory-Huggins treatment , the non-combinatoric contribution to the Gibbs free energy of mixing isΔGmix=RTnpχϕwwhere R is the ideal gas constant , T is the temperature , np is the total number of moles of monomer units , and ϕw is the volume fraction of water . From the relation ΔSmix=-∂ΔGmix∂T , the non-ideal entropy of mixing isΔSmix=-Rnpϕwϵs . From the relation ΔHmix=ΔGmix+TΔSmix , the enthalpy of mixing arising from non-ionic interactions isΔHmix=RnpϕwϵH . Values in the main text are computed using a water volume fraction of ϕw=0 . 722 . See Figure 2—source data 2 for further details . | Proteins make up much of the machinery of cells and perform many roles that are essential for life . Some important proteins – known as intrinsically disordered proteins – lack any stable three-dimensional structure . One such protein , called tau , is best known for its ability to form tangles in the brain , and a buildup of these tangles is a hallmark of Alzheimer’s disease and many other dementias . Tau is also one of a number of proteins that can undergo a process called liquid-liquid phase separation: essentially , a solution of tau separates into a very dilute solution interspersed with droplets of a concentrated tau solution , similar to an oil-water mixture separating into a very watery solution with drops of oil . Understanding the conditions that lead to spontaneous liquid-liquid phase separation might give insight into how the tau tangles form . However , it was not known whether it is possible in principle for liquid-liquid phase separation of tau to occur in a living brain . Lin , McCarty et al . have now used an advanced computer simulation method together with experiments to map the conditions under which a solution containing tau undergoes liquid-liquid phase separation . Temperature as well as the concentrations of salt and the tau protein all influenced how easily tau droplets formed or dissolved , and the narrow range of conditions that encouraged droplet formation fell within the normal conditions found in the body , also known as “physiological conditions” . This suggested that tau droplets might form and dissolve easily in living systems , and possibly in the brain , depending on the precise physiological conditions . To explore this possibility further , tau protein was added to a dish containing living cells . As the map suggested , slightly adjusting temperature or protein concentrations caused tau droplets to form and dissolve , all while the cells remained alive . The map provided by this study may offer guides to researchers looking for liquid-liquid phase separation in the brain . If liquid-liquid phase separation of tau occurs in living brains , it may be important for determining whether and when damaging tau tangles emerge . For example , the high concentration of tau in droplets might speed up tangle formation . Ultimately , a better understanding of the conditions and mechanism for liquid-liquid phase separation of tau can help researchers understand the role of protein droplet formation in living systems . This may be a process that promotes , or possibly a regulatory mechanism that prevents , the formation of tau tangles associated with dementia . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2019 | Narrow equilibrium window for complex coacervation of tau and RNA under cellular conditions |
In presynaptic boutons , calcium ( Ca2+ ) triggers both neurotransmitter release and short-term synaptic plasticity . Whereas synaptotagmins are known to mediate vesicle fusion through binding of high local Ca2+ to their C2 domains , the proteins that sense smaller global Ca2+ increases to produce short-term plasticity have remained elusive . Here , we identify a Ca2+ sensor for post-tetanic potentiation ( PTP ) , a form of plasticity thought to underlie short-term memory . We find that at the functionally mature calyx of Held synapse the Ca2+-dependent protein kinase C isoforms α and β are necessary for PTP , and the expression of PKCβ in PKCαβ double knockout mice rescues PTP . Disruption of Ca2+ binding to the PKCβ C2 domain specifically prevents PTP without impairing other PKCβ-dependent forms of synaptic enhancement . We conclude that different C2-domain-containing presynaptic proteins are engaged by different Ca2+ signals , and that Ca2+ increases evoked by tetanic stimulation are sensed by PKCβ to produce PTP .
The complex manner in which patterns of action potentials ( AP ) are transformed into neurotransmitter release suggests the existence of multiple presynaptic calcium ( Ca2+ ) sensors ( Kaeser and Regehr , 2013 ) . Synaptotagmin-1 , synaptotagmin-2 , and synaptotagmin-9 have been identified as Ca2+ sensors for synchronous release ( Sudhof , 2013 ) , but the Ca2+ sensors that regulate short-term use-dependent plasticity remain elusive . For a widespread form of short-term plasticity termed post-tetanic potentiation ( PTP ) , a high-frequency burst of presynaptic APs enhances subsequent AP-evoked release for tens of seconds . PTP requires sustained elevation of presynaptic Ca2+ , and in most cases synaptic enhancement outlives Ca2+ increases ( Regehr et al . , 1994; Brager et al . , 2003; Korogod et al . , 2005; Habets and Borst , 2007; Fioravante et al . , 2011 , 2012 ) . Although PTP is thought to contribute to short-term memory ( Silva et al . , 1996; Abbott and Regehr , 2004 ) , the Ca2+ sensor that mediates this plasticity has not been identified . Three Ca2+-dependent isoforms of protein kinase C ( PKCCa; PKCα , PKCβ , and PKCγ ) play crucial roles in PTP ( Fioravante et al . , 2011 , 2012; Chu et al . , 2014 ) . Because these isoforms contain Ca2+-binding C2 domains ( Shao et al . , 1996; Sutton and Sprang , 1998 ) , we hypothesize that they function as Ca2+ sensors for PTP . However , it is unclear whether the Ca2+-binding properties of the C2 domain of PKCCa ( Kohout et al . , 2002 ) are well-suited to mediate PTP , which is thought to rely , at least in part , on the waning residual Ca2+ after the AP burst ( Fioravante and Regehr , 2011 ) . Moreover , diacylglycerol ( DAG ) binding to the C1 domain of PKCCa ( Figure 1A ) can regulate the activity of PKCCa ( Newton , 2010 ) , and it has been proposed that PKC could play a permissive role in PTP rather than function as the Ca2+ sensor ( Saitoh et al . , 2001 ) . Indeed , presynaptic C2-domain proteins do not necessarily function as Ca2+ sensors; rather , they can regulate release independent of their Ca2+-binding properties ( for an example , see Groffen et al . , 2010; Pang et al . , 2011 ) . Furthermore , additional Ca2+-binding proteins have been implicated in short-term plasticity ( Sakaba and Neher , 2001; Junge et al . , 2004; Mochida et al . , 2008; He et al . , 2009; Shin et al . , 2010 ) , but it has not been established that Ca2+ binding to these proteins is required for short-term plasticity . Thus , in order to determine whether PKCCa isoforms are Ca2+ sensors that mediate PTP , it must be determined if PTP relies on Ca2+ binding to the PKCCa C2 domain . 10 . 7554/eLife . 03011 . 003Figure 1 . Expression of PKCβ rescues synaptic potentiation in animals lacking calcium-dependent PKCs . Synaptic plasticity was examined at the calyx of Held following tetanic stimulation ( B and F ) or bath application of the phorbol ester PDBu ( C and G ) for wild-type ( wt , black ) , PKCαβ dko animals ( purple ) , and PKCαβ dko animals expressing PKCβWT-YFP ( green ) . ( A ) Domain arrangement of PKCCa . DAG and PDBu bind to the C1 domain and Ca2+ binds to the C2 domain . ( B , C , F , G ) Left , example EPSCs recorded prior to ( bold traces ) and after ( light traces ) synaptic enhancement for each experimental condition . Right , EPSCs are plotted as a function of time ( mean ± SEM ) . For ( B ) , wild-type: 62 ± 12%; αβ dko: 2 . 4 ± 1 . 8% . Also see Figure 1—figure supplement 1 and accompanying legend for PTP induced under elevated-temperature conditions . Similar to PTP induced at room temperature , PTP at near-physiological temperature requires PKCCa ( Figure 1—figure supplement 2 ) . For ( C ) , at steady state: wild-type: 97 ± 12%; αβ dko: 3 . 2 ± 3 . 4%; for ( F ) , PKCβWT-YFP: 61 ± 7%; for ( G ) , 84 ± 11% . In F and G , the αβ dko group data from B and C respectively are re-plotted for comparison . Also see Figure 1—figure supplement 3 and Figure 1—figure supplement 4 . ( D ) In this schematic of the auditory brainstem , the ventral cochlear nucleus ( VCn ) and medial nuclei of the trapezoid body ( MNTB ) are labeled . An AAV expressing PKCβWT-YFP was injected in the VCn at postnatal day 4 . ( E ) Confocal images of a brain section labeled with an antibody against vGlut1 ( red ) are shown for a calyx of Held expressing PKCβWT-YFP ( green ) in a PKCαβ dko animal at postnatal day 18 . Scale bar: 10 µm . ( H and I ) The synaptic mechanism through which PKCβ rescues PTP was examined under conditions that relieve AMPA receptor desensitization and saturation . ( H ) Left , overlay of EPSCs ( 10 ms inter-stimulus interval ) delivered prior to ( bold traces ) and 10 s after ( light traces ) PTP-inducing tetanus . Middle , traces are normalized to the first EPSC to allow comparison of PPR . Right , PPRPOST ( after tetanus ) over PPRPRE ( before tetanus ) ( mean ± SEM , see Figure 1—source data 1 and 2 ) . Wild-type: p=0 . 49; αβ dko expressing PKCβWT-YFP: p=0 . 68 . ( I ) Summary of the readily releasable pool ( RRP ) and release probability ( p ) contributions to PTP ( mean ± SEM , also see Figure 1—figure supplements 5 and 6 and Figure 1—source data 1 and 2 ) . RRPWT: 37 ± 9%; RRPPKCβWT-YFP: 39 ± 12%; p=0 . 88 . Scale bars in B , C , F , and G: 2 nA , 1 ms . Scale bars in H: 2 nA , 5 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00310 . 7554/eLife . 03011 . 004Figure 1—source data 1 . Summary and statistical analyses of synaptic properties during PTP and PDBu-induced potentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00410 . 7554/eLife . 03011 . 005Figure 1—source data 2 . Summary and statistical analyses of basal synaptic properties . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00510 . 7554/eLife . 03011 . 006Figure 1—figure supplement 1 . PTP can be induced under near-physiological conditions . Most studies of PTP are performed at room temperature in 2 mM external Ca2+ , 1 mM external Mg2+ . We tested whether PTP could also be induced under more physiological conditions ( 34°C , 1 . 5 mM external Ca2+ , 1 . 2 mM external Mg2+ ) . In this recording from a representative cell , we first induced PTP with our standard stimulation protocol ( 100 Hz , 4 s ) at 24°C ( top ) . We then increased the temperature to 34°C , repeated the induction protocol , but failed to obtain PTP ( middle ) . However , we successfully induced PTP at 34°C in the same cell with higher frequency stimulation ( 400 Hz , 2 s ) . We found that at elevated-temperature , PTP could not be reliably induced with the 100 Hz , 4 s stimulation protocol ( 2 . 9% ± 0 . 6% , n = 8 cells ) but it could be readily obtained with the 400 Hz , 2 s protocol ( 83 ± 14% , n = 5 cells ) . These results indicate that PTP can be obtained under near-physiological conditions . The higher stimulation frequency needed to induce the plasticity at 34°C is within the physiological firing frequency range observed at matured calyces of Held in vivo ( Sonntag et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00610 . 7554/eLife . 03011 . 007Figure 1—figure supplement 2 . Under near-physiological conditions PTP is mediated by PKCCa . We used a selective PKCαβ inhibitor ( Calbiochem 539654; 250 nM ) ( Tanaka et al . , 2004 ) to determine whether PKCCa isoforms are essential for PTP obtained under near-physiological conditions ( 34°C , 1 . 5 mM external Ca2+ , 1 . 2 mM external Mg2+ ) . We have previously established the specificity of this inhibitor ( Chu et al . , 2014 ) , which potently blocks PKCβ ( Km ∼ 5–20 nM; Km for PKCα ∼ 300 nM ) . In wild-type ( wt ) mice , we found that the PTP induced by a 400 Hz , 2 s tetanus ( 83 ± 14%; n = 5 ) was dramatically reduced in the presence of the PKCαβ inhibitor ( 17 ± 6 . 6%; n = 7; p<0 . 001 ) . This result establishes that PKCCa isoforms play a crucial role in PTP under near-physiological conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00710 . 7554/eLife . 03011 . 008Figure 1—figure supplement 3 . At the functionally mature calyx of Held , PKCα does not contribute to PTP but plays a small role in phorbol ester-induced potentiation . PTP ( top left ) and PDBu-induced potentiation ( top right ) are plotted as a function of time ( mean ± SEM ) , for wild-type ( black ) and PKCα ko ( red ) groups . Bottom , summary plots ( mean ± SEM ) of peak PTP ( 10 s after tetanus ) and steady-state PDBu-induced potentiation ( 430–600 s in PDBu ) for indicated groups . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00810 . 7554/eLife . 03011 . 009Figure 1—figure supplement 4 . PKCCa isoforms do not regulate basal synaptic properties . Box-plots of basal synaptic properties for wild-type ( black ) , PKCαβ dko ( purple ) , and PKCαβ dko groups expressing wild-type PKCβ ( βWT-YFP; green ) . Medians and interquartile ranges ( Q3-Q1 ) are shown . Whiskers extend to max and min values for each group . Box-plots were used to illustrate the full data range of each group; additionally , the data sets for mEPSC amplitude were not normally distributed . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 00910 . 7554/eLife . 03011 . 010Figure 1—figure supplement 5 . Determining the contributions of RRP and p in wild-type and rescued PTP at the functionally mature calyx of Held . Synaptic mechanisms of PTP were examined using stimulus trains in the presence of kynurenate and CTZ . Left , example synaptic currents evoked by the first 40 stimuli of a 4 s , 100 Hz train ( dark traces ) and by a 40-pulse 100-Hz train ( light traces ) at the peak of PTP are shown for wild-type group ( top ) and PKCαβ dko group expressing wild-type PKCβ ( βWT-YFP; bottom ) . Right , example plots of cumulative EPSC as a function of stimuli . The linear fits to the last 15 points , back-extrapolated to the y-axis , are shown and were used to determine the change in RRP size and p for the examples shown to the left according to the train method ( see ‘Materials and methods’ for methodology ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 01010 . 7554/eLife . 03011 . 011Figure 1—figure supplement 6 . Determining the contributions of RRP and p in wild-type and rescued PTP at the functionally mature calyx of Held . Synaptic mechanisms of PTP were examined using stimulus trains in the presence of kynurenate and CTZ . Left , box-plots of basal RRP size ( RRP1; left ) and basal p ( p1; right ) , estimated from the 1st AP train for wild-type group ( black ) and αβ dko group expressing βWT-YFP ( green ) . Medians and interquartile ranges ( Q3-Q1 ) are shown . Whiskers extend to max and min values for each group . Box-plots were used to illustrate the full data range of each group . Right , cumulative histograms of changes in RRP ( RRP2/RRP1 ) and p ( p2/p1 ) with PTP . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 011
To investigate the function of PKCCa isoforms in PTP , we first examined their role at the functionally mature calyx of Held synapse ( postnatal day 17–22 ) ( Fedchyshyn and Wang , 2005; Yang et al . , 2010 ) using double knockout mice for PKCα and β ( αβ dko ) . We recorded excitatory postsynaptic currents ( EPSCs ) from principal neurons in the medial nucleus of the trapezoid body ( MNTB ) in response to extracellular stimulation . Tetanic stimulation induced PTP in wild-type animals ( Figure 1B , black ) but not in PKCαβ dko animals ( Figure 1B , purple ) . Thus , in contrast to the immature calyx of Held where a substantial component of PTP ( ∼20% ) is independent of PKCCa ( Fioravante et al . , 2011 ) , PTP at the functionally mature calyx of Held relies entirely on PKCCa isoforms . We further tested whether the contribution of PKCCa to a related form of potentiation that occludes PTP also increases with development . Phorbol 12 , 13-dibutyrate ( PDBu ) , a DAG analog , can enhance transmission by activating not only PKCCa ( Figure 1A ) but also Ca2+-insensitive PKC isoforms and other presynaptic proteins ( Brose and Rosenmund , 2002; Newton , 2010 ) . At immature calyces , ∼35% of PDBu-mediated enhancement is independent of PKCCa ( Fioravante et al . , 2011 ) . We found that PDBu enhances release at functionally mature wild-type calyces ( Figure 1C , black ) but not at age-matched αβ dko calyces ( Figure 1C , purple ) . Thus , at the functionally mature calyx of Held , both PTP and PDBu-mediated enhancement rely entirely on PKCCa , suggesting that the contributions of parallel mechanisms to these forms of plasticity ( e . g . , Wierda et al . , 2007; Shin et al . , 2010 ) diminish with development . Although most of the studies presented here were performed at room temperature , we also examined PTP at near-physiological temperatures ( 34°C ) . Higher stimulus frequencies were required to induce PTP ( Figure 1—figure supplement 1 ) , but PTP was still dependent on PKCCa ( Figure 1—figure supplement 2 ) . We next assessed whether presynaptic expression of PKCβ in αβ dko animals rescues PTP . PKCβ was chosen because genetic deletion of PKCα had little effect on PTP ( Figure 1—figure supplement 3 ) , suggesting that PTP is mediated primarily by PKCβ at the functionally mature calyx . We generated an adeno-associated virus ( AAV ) , which we used to express wild-type PKCβ fused to yellow fluorescent protein ( PKCβWT-YFP ) in αβ dko animals ( Figure 1D ) . Two weeks after injection , virally expressed PKCβWT-YFP localized to glutamatergic terminals positive for the marker vGlut1 ( Figure 1E ) . In contrast to non-injected αβ dko animals , we observed reliable PTP at synapses expressing PKCβWT-YFP ( Figure 1F , green; compare to non-injected age-matched αβ dko animals , purple ) . Expression of PKCβWT-YFP did not alter basal synaptic properties ( Figure 1—figure supplement 4 ) . Moreover , PKCβWT-YFP expression supported PDBu-induced potentiation in αβ dko mice ( Figure 1G , green ) , which was very similar in amplitude to that observed in wild-type animals ( Figure 1C , black; Figure 1—source data 1 ) . Thus , expression of PKCβ is sufficient to rescue PTP and PDBu-mediated enhancement in PKC αβ dko animals . To determine whether rescued PTP and PTP in wild-type animals are mediated by the same synaptic mechanism , we calculated the paired-pulse ratios ( PPR = EPSC2/EPSC1 ) before and at the peak of PTP . If PTP reflects an increase in vesicular release probability ( p ) , which is inversely related to PPR , then PPRPOST/PPRPRE should decrease . However , PPRPOST/PPRPRE was unchanged in both wild-type ( Figure 1H , black ) and αβ dko animals expressing PKCβWT-YFP ( Figure 1H , green ) . This suggests that in both groups PTP is not mediated by an increase in p . We next examined the contribution of the readily releasable pool ( RRP ) of vesicles to PTP by evoking EPSCs with AP trains before and at the peak of PTP ( Figure 1—figure supplement 5 ) and found that PTP was mediated by equivalent increases in the RRP in wild-type and rescued groups ( Figure 1I , Figure 1—figure supplement 6 ) . Thus , at the functionally mature calyx , the same mechanism mediates PTP in wild-type animals and at synapses in PKCαβ dko animals that express PKCβWT-YFP . To determine if PKCβ is the Ca2+ sensor for PTP , it is necessary to abolish Ca2+ binding to PKCβ . To this end , we mutated five C2-domain aspartates ( D ) to alanines ( A ) and examined the effect on Ca2+ binding using purified recombinant wild-type C2 ( C2WT ) and mutant C2D/A domains ( Figure 2A , B , Figure 2—figure supplement 1 ) . These aspartates are predicted to mediate Ca2+ binding based on structural similarity ( Nalefski and Falke , 1996; Ubach et al . , 1998 ) . We assessed Ca2+ binding through changes in intrinsic fluorescence of tryptophan residues adjacent to the predicted Ca2+-binding sites ( Figure 2—figure supplement 1; Nalefski and Newton , 2001 ) . In the absence of Ca2+ , C2WT displayed a characteristic intrinsic fluorescence emission spectrum that peaked around 340 nm ( Figure 2C , dark green ) . A similar basal emission spectrum was observed for C2D/A ( Figure 2C , dark blue ) , suggesting that D-to-A mutations did not affect domain folding ( Nalefski and Newton , 2001 ) . Addition of 1 mM Ca2+ increased the fluorescence intensity of C2WT ( Figure 2C , light green ) but not C2D/A ( light blue , Figure 2C ) , indicating that D-to-A mutations prevent Ca2+-dependent rearrangements . 10 . 7554/eLife . 03011 . 012Figure 2 . C2-domain mutations of PKCβ abolish Ca2+ binding and Ca2+-induced translocation without impairing phorbol ester-induced translocation . ( A ) A partial sequence of the PKCβ C2 domain is shown with Ca2+-coordinating aspartates in green . These aspartates were mutated to alanines ( blue ) in the C2D/A construct . Also see Figure 2—figure supplement 1 . ( B ) Coomassie blue-stained gel of recombinant wild-type ( C2WT ) and mutant ( C2D/A ) PKCβ C2 domains . ( C ) Averaged intrinsic tryptophan fluorescence is shown for C2WT and C2D/A . Fluorescence emission spectra were recorded in 0 mM Ca2+ ( bold traces ) and 1 mM Ca2+ ( light traces ) . Peak fluorescence intensity change: C2WT:17 ± 1 . 3%; C2D/A: −1 . 3 ± 2 . 0% . ( D ) Translocation of PKCβWT-YFP ( left ) and PKCβD/A-YFP ( right ) in HEK293T cells was monitored in response to the Ca2+ ionophore ionomycin and in response to PDBu . Ca2+ increases caused PKCβWT-YFP to translocate , but not PKCβD/A-YFP . Both PKCβWT-YFP and PKCβD/A-YFP translocated in response to PDBu . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 01210 . 7554/eLife . 03011 . 013Figure 2—figure supplement 1 . Protein sequence alignment for PKCβ C2WT and PKCβ C2D/A . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 013 When activated by either phorbol esters or Ca2+ , PKC translocates from the cytoplasm to the plasma membrane ( Newton , 2010 ) . We utilized this property to test the effects of the D-to-A mutations on the response of PKCβ to Ca2+ increases . We expressed PKCβWT-YFP or PKCβD/A-YFP in HEK293T cells and monitored the subcellular distribution of the kinase . The Ca2+ ionophore ionomycin induced translocation of PKCβWT-YFP ( Figure 2D , top left ) , but did not alter the intracellular distribution of PKCβD/A-YFP ( Figure 2D , top right ) . In contrast , PDBu , which binds to the C1 domain , caused both PKCβWT-YFP and PKCβD/A-YFP to translocate . This result indicates that Ca2+ binding to the PKC C2 domain is necessary for Ca2+-induced , but not PDBu-induced , translocation of PKCβ . Moreover , it suggests that D-to-A mutations in the C2 domain prevent Ca2+ activation of PKCβ without interfering with C1-domain-mediated membrane recruitment of PKCβ . We next tested whether Ca2+ binding to the PKCβ C2 domain is required for PTP . Using AAV to express PKCβD/A-YFP , we found that PKCβD/A-YFP localized to vGlut1-positive areas and distributed similarly to PKCβWT-YFP ( Figure 3A , compare to Figure 1E ) . Expression of PKCβD/A-YFP in αβ dko calyces , similar to wild-type PKCβ , did not affect basal synaptic properties ( Figure 3—figure supplement 1 , Figure 3—source data 2 ) . However , in stark contrast to wild-type PKCβ , PKCβD/A failed to rescue PTP ( Figure 3B , blue; also see Figure 3D , left ) . The inability of PKCβD/A-YFP to support PTP could be due to a loss of Ca2+ binding to PKCβ; alternatively , the D-to-A mutations may have induced more profound impairments of PKCβ , rendering it unable to enhance neurotransmitter release . To distinguish between these possibilities , we tested PDBu-induced potentiation in PKCβD/A-YFP-expressing calyces . Compellingly , PDBu-induced potentiation in PKCβD/A-YFP-expressing calyces was rescued to wild-type levels ( Figure 3C , blue; also see Figure 3D , right ) . This indicates that PKCβD/A-YFP retained its ability to enhance synaptic transmission . We conclude that PKCβD/A-YFP is unable to mediate PTP because it is unable to bind Ca2+ , and that Ca2+ binding to the C2 domain of PKCβ is required for PTP . Therefore , PKCβ is a Ca2+ sensor for PTP . 10 . 7554/eLife . 03011 . 014Figure 3 . PTP requires Ca2+ binding to PKCβ but phorbol ester-induced potentiation does not . ( A ) Confocal images of a brain section labeled with an antibody against vGlut1 ( red ) are shown for a calyx of Held expressing PKCβD/A-YFP ( green ) in a PKCαβ dko animal . ( B and C ) Synaptic plasticity was examined in PKCαβ dko animals at calyces of Held expressing Ca2+-insensitive PKCβ ( PKCβD/A-YFP , blue traces ) . Representative traces and time-courses ( mean ± SEM ) are shown following tetanic stimulation ( B ) and during bath application of PDBu ( C ) . For ( B ) , PKCβD/A: 3 . 6 ± 2 . 2%; for ( C ) , PKCβD/A: 98 ± 23% . Scale bars: 1 nA , 1 ms . In ( B and C ) , the αβ dko group data from Figure 1B , C respectively are re-plotted for comparison . For basal synaptic properties of the PKCβD/A-YFP-expressing group , see Figure 3—figure supplement 1 and Figure 3—source data 2 . ( D ) Summary plots ( mean ± SEM ) of the magnitude of synaptic enhancement produced by tetanic stimulation ( left ) and by PDBu ( right ) . Source data are provided in Figure 3—source data 1 and 2 . See also Figure 1—source data 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 01410 . 7554/eLife . 03011 . 015Figure 3—source data 1 . Summary and statistical analyses of synaptic properties during PTP and PDBu-induced potentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 01510 . 7554/eLife . 03011 . 016Figure 3—source data 2 . Summary and statistical analyses of basal synaptic properties . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 01610 . 7554/eLife . 03011 . 017Figure 3—figure supplement 1 . Basal synaptic properties are not altered by PKCβD/A-YFP expression . Box-plots of basal synaptic properties for wild-type ( black ) and PKCαβ dko groups expressing D-to-A mutant PKCβ ( βD/A-YFP; blue ) . Medians and interquartile ranges ( Q3–Q1 ) are shown . Whiskers extend to max and min values for each group . DOI: http://dx . doi . org/10 . 7554/eLife . 03011 . 017
To the best of our knowledge , PKCβ is the first Ca2+ sensor to be identified specifically for short-term synaptic plasticity . Similar to synaptotagmin-1 , synaptotagmin-2 , and synaptotagmin-9 , PKCβ requires binding of Ca2+ to its C2 domain for its Ca2+-sensing function ( Figure 3 ) . However , PKCβ acts upstream of vesicle fusion ( de Jong and Verhage , 2009 ) , does not regulate basal transmission or paired-pulse plasticity ( Figure 1—figure supplement 4 ) , and is not expected to be activated by single stimuli . How do PKCβ and synaptotagmins respond to such different activity patterns and , consequently , such different Ca2+ signals ? It is likely a combination of differences in Ca2+-binding properties and subcellular localization that underlie these contrasting responses ( Nalefski et al . , 2001 ) . Synaptotagmin-1 binds Ca2+ cooperatively with low affinity and fast kinetics , and localizes close to release sites on synaptic vesicles; therefore , it is poised to detect large , transient Ca2+ signals near open voltage-gated Ca2+ channels ( Sudhof , 2013 ) . In contrast , PKCβ is activated by lower Ca2+ levels with lower cooperativity and is cytosolic ( Nalefski and Newton , 2001; Kohout et al . , 2002 ) . Prolonged stimulation is necessary to produce a sufficient buildup of Ca2+ to activate PKCβ , which is consistent with the prolonged activity requirement for PTP ( Habets and Borst , 2005; Korogod et al . , 2005 ) . It is unlikely that PKCβ is the sole Ca2+ sensor that triggers PTP . At the granule cell to Purkinje cell synapse , PTP can be dependent on either PKCα or PKCβ ( Fioravante et al . , 2012 ) , and at the calyx of Held synapse prior to the onset of hearing PTP depends on PKCγ ( Chu et al . , 2014 ) . These findings suggest that Ca2+-sensitive PKC isoforms , including PKCα and PKCγ , may constitute a class of proteins that acts as Ca2+ sensors for PTP , much as multiple isoforms of synaptotagmin act as Ca2+ sensors for fast synaptic transmission ( Sudhof , 2012 ) . Further studies are required to determine whether PKCα , PKCγ , and other Ca2+-sensitive proteins implicated in PTP , such as Munc13 ( Wierda et al . , 2007 ) , calmodulin ( Junge et al . , 2004 ) , and synaptotagmin-2 ( He et al . , 2009; Xue and Wu , 2010 ) , can also act as Ca2+ sensors to produce PTP . PKCβ mediates PTP by phosphorylating downstream targets ( Genç et al . , 2014 ) , which could explain how PTP outlives elevations of presynaptic Ca2+ ( Fioravante and Regehr , 2011 ) . We suggest that PKCβ is a founding member of a new class of Ca2+ sensors that function upstream of vesicle fusion to regulate short-term plasticity .
Cloning was performed by Genscript . Viruses were generated by the University of Pennsylvania Vector Core . All constructs were verified by sequencing . Wild-type PKCβ-YFP ( βWT-YFP ) was obtained through PCR from Addgene plasmid #14866 ( Violin et al . , 2003 ) , using the following primers: 5′-GACACAACAGTCTCGAACTTAATCGAACCCGCGGCACGAGCCTCGACG-3′; 3′-GGGAAAAAGATCGGATCCTCAGGCGTCGACGGGCCCTCTAGATTACTTG-5′ . To generate an adeno-associated viral vector , PKCβWT-YFP was inserted into a pENN . AAV . CMV . TurboRFP . RBG cis-plasmid ( courtesy of the University of Pennsylvania Vector Core ) using SacII and SalI , after removal of the TurboRFP sequence with SpeI and XhoI . Mutant PKCβ-YFP ( βD/A-YFP ) was generated by replacing the 5 aspartates that coordinate calcium binding ( Sutton and Sprang , 1998 ) with alanines through PCR . To generate bacterial expression plasmids of the PKCβ C2 domain , the sequences for C2WT or C2D/A ( a . a . 157–294 ) ( Torrecillas et al . , 2003 ) ( see also Figure 2—figure supplement 1 ) were inserted into a pGEX-KG vector ( Addgene database #2890 ) using XbaI and NcoI and the following primers: 5′-TCCGGTGGTGGTGGTGGAATTCTAGAAGAACGCCGTGGCCGCATC-3′ and 3’-AAGCTTGAGCTCGAGTCGACCCATGGTCATCCTTCCGGCGGCACCG-5′ . All animal experiments were conducted at Harvard Medical School and were completed in accordance with guidelines by the Harvard Medical Area Standing Committee on Animals . PKCαβ double knockout ( dko ) mice were obtained through breeding of PKCα and PKCβ single knockout ( ko ) animals generated by M Leitges ( Leitges et al . , 1996 , 2002 ) . The probability of obtaining an αβ double knockout animal from heterologous ( het ) crosses is very low ( 1:16 ) , making viral injection experiments unfeasible . We therefore bred het-ko animals together to increase the probability of getting desired animals . Similarly , to increase the probability of obtaining wild-type mice , we crossed PKC het–het or het-wild-type mice to use as wild-type controls . Wild-type mice were derived from the same genetic line as αβ dko animals . To prevent genetic drift in the inbred ko lines , we backcrossed them every second generation to C57BL/6J or 129S2 . For experiments , animals of both sexes were used and age-matched wild-type , PKCα ko and PKCαβ dko mice from our colony were interleaved . P4 pups were stereotactically and unilaterally injected under isofluorane anesthesia with AAVs into the VCn ( from lambda: 1 . 3 mm lateral , 0 . 9 mm caudal , 3 mm ventral ) , where globular bushy cells that give rise to calyx of Held synapses in the contralateral MNTB reside . Injections ( 600 nl at a rate of 1 nl/s ) were performed with an UltraMicroPump ( UMP3 , WPI , Sarasota FL ) and Wiretrol II capillary micropipettes ( Drummond Scientific , Broomall PA ) pulled to a fine tip ( 10–20 µm diameter ) . At this age , the skull is sufficiently soft so it can be penetrated with a 281/2–gage needle without the need for drilling . After the injection , the skin was closed with Gluture ( Abbott Laboratories , Irving TX ) , and pups were allowed to recover on a heating pad prior to returning to the home cage . 14–18 days were allowed for expression prior to slice preparation . The calyx of Held synapse in the auditory brainstem was chosen for this study because of its monosynaptic innervation and its amenability to viral manipulations . Indeed , our study would not be possible at more ‘conventional’ polysynaptic preparations such as the CA3-CA1 hippocampal synapse because it is not possible to infect 100% of synapses with viruses . Transverse 190-µm to 200-µm-thick brainstem slices containing the MNTB were made with a vibratome slicer ( VT1000S , Leica , Buffalo Grove IL ) from juvenile ( postnatal day 17–22 ) mice deeply anesthetized with isoflurane . Brains were dissected and sliced at 4°C in cutting solution consisting of the following ( in mM ) : 125 NaCl , 25 NaHCO3 , 1 . 25 NaH2PO4 , 2 . 5 KCl , 0 . 1 CaCl2 , 3 MgCl2 , 25 glucose , 3 myo-inositol , 2 Na-pyruvate , 0 . 4 ascorbic acid , continuously bubbled with 95% O2/5% CO2 ( pH 7 . 4 ) . Slices were incubated at 32°C for 30 min in a bicarbonate-buffered solution composed of the following ( in mM ) : 125 NaCl , 25 NaHCO3 , 1 . 25 NaH2PO4 , 2 . 5 KCl , 2 CaCl2 , 1 MgCl2 , 25 glucose , 3 myo-inositol , 2 Na-pyruvate , 0 . 4 ascorbic acid , continuously bubbled with 95% O2/5% CO2 ( pH 7 . 4 ) . Slices were transferred to a recording chamber at room temperature ( 21–24°C ) under an upright microscope ( Olympus , Center Valley PA ) equipped with a 60× objective . During recordings , the standard perfusion solution consisted of the bicarbonate-buffered solution ( see above ) with 1 µM strychnine and 25 µM bicuculline ( R&D Systems , Minneapolis MN ) to block inhibition . Slices were superfused at 1–3 ml/min with this external solution . Whole-cell postsynaptic patch-clamp recordings were made from visually identified cells in the MNTB region using glass pipettes of 2–3 MΩ resistance , filled with an internal recording solution of the following ( in mM ) : 20 CsCl , 140 Cs-gluconate , 20 TEA-Cl , 10 HEPES , 5 EGTA , 5 Na2-phosphocreatine , 4 ATP-Mg , 0 . 3 GTP-Na . Series resistance ( Rs ) was compensated by up to 70% , and the membrane potential was held at −70 mV . EPSCs were evoked by stimulating presynaptic axons with a custom-made bipolar stimulating electrode midway between the medial border of the MNTB and the midline of the brainstem . For slice recordings from injected animals , principal neurons in the MNTB contralateral to the injection site were selected based on the presence of YFP-expressing presynaptic terminals . A Multiclamp 700B ( Axon Instruments/Molecular Devices , Sunnyvale CA ) amplifier was used . Recordings were digitized at 20 KHz with an ITC-18 A/D converter ( Instrutech Corp . /HEKA Elektronik , Bellmore NY ) using custom macros ( written by MA Xu-Friedman ) in Igor Pro ( Wavemetrics , Portland OR ) and filtered at 8 kHz . Macros can be found on Dr Xu-Friedman's website http://biology . buffalo . edu/Faculty/Xu_Friedman/mafPC/sign_in . html . The protocol for inducing PTP was as follows: an estimate of baseline synaptic strength was obtained through low-frequency stimulation at 0 . 2 Hz for 25 s . PTP was induced with a 4-s stimulus train at 100 Hz , followed by low-frequency stimulation to test for PTP . For phorbol ester experiments , PDBu ( 1 µM; Tocris , UK ) was washed in for 10 min once a stable baseline of at least 3 min was established . Synaptic strength was evaluated by afferent fiber stimuli , repeated every 20 s . During the inter-trial intervals , 5 s stretches of postsynaptic current were recorded to assess the frequency and amplitude of mEPSCs . To assess PPR , pulses were delivered at an inter-stimulus interval of 10 ms . For all recordings , the access resistance and leak current were monitored , and experiments were rejected if either of these parameters changed significantly . Data analysis was performed using routines written in IgorPro ( WaveMetrics ) . PTP magnitude was calculated as the ratio of EPSC amplitude 10 s after the 4-s , 100 Hz train over the average baseline . The magnitude of PDBu-induced potentiation was estimated by averaging the steady-state responses , 430–600 s from wash-in onset . To analyze spontaneous events , mEPSCs were detected using a threshold ( average peak to peak noise in the baseline ) of the first derivative of the raw current trace and confirmed visually . Statistical analyses were done using one-way ANOVA tests for multiple group comparisons followed by Tukey post-hoc analysis , or Kruskal–Wallis non-parametric ANOVA for data sets that were not normally distributed . Pairwise comparisons were performed with Student's t tests . Level of significance was set at p<0 . 05 . To determine the contributions of RRP and p to wild-type and rescued PTP , stimulus trains were used in the presence of kynurenate ( 1 mM ) and CTZ ( 0 . 1 mM ) to prevent postsynaptic receptor saturation and desensitization . Briefly , the amplitude of the first 40 responses to the stimulus train used to induce PTP and to a stimulus train ( 400 ms , 100 Hz ) 10 s later ( at the peak of PTP ) were measured , and a plot of the cumulative EPSC for each train vs the stimulus number was made . The key to this approach is that the EPSC amplitude eventually reaches a steady-state level , and under these conditions the RRP is depleted and the remaining release is due to replenishment from a recycling/reserve pool ( Schneggenburger et al . , 1999 ) . The size of the RRP can then be determined by a linear fit to the steady-state responses ( last 15 EPSCs ) , which is extrapolated back to the y-axis ( Moulder and Mennerick , 2005; Thanawala and Regehr , 2013 ) . p is then calculated from EPSC1/RRP . 150-µm thick transverse brainstem slices were prepared as described above from P18–P22 animals injected with AAVs and fixed with 4% paraformaldehyde for 2 hr at 4°C . At the end of fixation , slices were transferred to phosphate buffered saline ( Sigma-Aldrich , St . Louis MO ) and stored at 4°C until further processing . Slices were then incubated in blocking solution ( phosphate buffered saline +0 . 25% Triton X-100 [PBST] +10% normal goat serum ) for 1 hr at room temperature . Slices were incubated with primary antibody ( anti-vGlut1 guinea pig polyclonal [Synaptic Systems , Germany] ) in PBST overnight at 4°C , followed by incubation with secondary antibody ( goat anti-guinea pig Alexa 568-conjugated [Life Technologies , Carlsbad CA] ) in PBST for 2 hr . Slices were mounted to Superfrost glass slides ( VWR , Visalia CA ) and air-dried for 30 min . Following application of Prolong anti-fade medium ( Invitrogen ) , slices were covered with a top glass coverslip ( VWR ) and allowed to dry for 24 hr prior to imaging . Antibodies were used at 1:500 dilution . Images were acquired with a Zeiss 510 Meta confocal microscope using a Plan-apochromat 1 . 4 NA 63x oil lens . Emission filters were BP570-670 nm for the red channel ( vGlut1 ) and BP500-550 for YFP ( PKCβ ) . Single optical sections at 1024 × 1024 ( average of three scans ) were obtained sequentially for the different channels . Color channels were split and merged in ImageJ to obtain the composite images in RGB . N-terminal GST fusion proteins of PKCβ C2WT and C2D/A were expressed in Escherichia coli BL21 cells . Pelleted bacteria were resuspended in ice-cold PBS supplemented with 500 µM EDTA , 0 . 5 mg/ml lysozyme ( Amresco , Solon OH ) , and protease inhibitor cocktail ( Easypack; Roche , South San Francisco CA ) , and the bacteria were lysed by sonication . After centrifugation at 11 , 200 RPM for 30 min , the soluble fraction was collected and incubated with glutathione sepharose 4B beads ( GE healthcare , Pittsburgh PA ) for 1 hr at 4°C . Samples were cleared from nucleic acid contaminants with benzonase ( 40 U/ml , Sigma ) for 3 hr at RT , and subsequently eluted from the beads with solution containing 100 mM Tris , 10 mM CaCl2 , 5 mM Glutathione ( pH 7 . 4 ) for 1 hr at 4°C . GST was cleaved with thrombin-agarose ( 100 µl resin/mg protein , Sigma ) for 24 hr at 4°C , and samples were dialyzed to solution containing 40 mM Tris–HCl pH 7 . 4 , 100 mM NaCl , and 0 . 5 mM Na-EGTA . GST was removed from the samples using glutathione sepharose 4B beads . 10 µl of purified protein was run on a 12% SDS gel and Coomassie blue-stained to check for purity ( Figure 2B ) . Intrinsic tryptophan fluorescence of purified recombinant C2WT and C2D/A was monitored in dialysis buffer ( see above ) . Emission spectra were recorded from 325 to 425 nm on a Spectramax M5 microplate reader ( Molecular Devices ) . Excitation was set at 295 nm and peak intrinsic fluorescence change ( ΔF ) upon addition of 1 mM free Ca2+ was estimated at 341 nm . To correct for the effect of volume increase on fluorescence readings upon addition of Ca2+-containing buffer , ΔF in buffer-alone controls was subtracted from fluorescence values in buffer+Ca2+ groups . Experiments were repeated with two independently purified batches of protein , for a total of seven times . Similar results were obtained every time . HEK293T cells plated on glass coverslips were transfected with PKCβWT-YFP or βD/A-YFP expression vectors using Lipofectamine 2000 ( Life Technologies ) . 24 hr after transfection , the coverslips were transferred to the imaging chamber of a custom-built 2-photon laser scanning microscope system and superfused with buffer ( 138 mM NaCl , 1 . 5 mM KCl , 10 mM HEPES , 1 mM MgCl2 , 2 mM CaCl2 , 10 mM glucose , pH 7 . 4 ) at 2 ml/min . YFP was excited at 840 nm with a Ti-Sapphire laser through a 60× , 1 . 1 NA water-immersion Olympus lens . A 500–550 BP emission filter was used . 512 × 512 frame scans were acquired at a rate of 1 line/4 ms , every 30 s . To stimulate translocation , the superfusion solution was switched to one containing 1 µM PDBu or 10 µM ionomycin ( R&D Systems ) for 15 min . The experiment was repeated three times for PKCβWT-YFP and twice for PKCβD/A-YFP , with similar results . Acquired images were exported to ImageJ and brightness/contrast was adjusted equally for all images within an experiment for display purposes . | Brain function is dependent upon the rapid transfer of information from one brain cell to the next at junctions known as synapses . When an electrical signal called an action potential is generated by the cell before the synapse , the presynaptic cell , it triggers an influx of calcium ions into that cell . These ions activate specific calcium sensors , triggering release of molecules called neurotransmitters from the presynaptic cell through exocytosis of synaptic vesicles . These neurotransmitters bind to receptors on the membrane of the postsynaptic cell , and produce an electrical signal whose size is a measure of synaptic strength . The strength of a synapse can change over time—a property that is called plasticity . Synapses can undergo both long-term and short-term increases in strength . Post-tetanic potentiation is a short-term increase in strength that lasts for tens of seconds: it is triggered by a calcium increase in the presynaptic cell and involves an increase in the amount of neurotransmitter released in response to each presynaptic action potential . Post-tetanic potentiation is thought to underlie short-term memory . However , the identity of the sensor that detects the build-up of calcium in post-tetanic potentiation was not known . Now , Fioravante , Chu et al . have provided the first direct evidence that an enzyme called protein kinase C is responsible . Electrophysiological recordings in brain slices from genetically modified mice revealed that animals that lack protein kinase C do not show post-tetanic potentiation . However , potentiation can be restored by re-introducing the enzyme into presynaptic cells . Importantly , a mutated version of protein kinase C that lacks the ability to bind calcium is unable to trigger post-tetanic potentiation . Protein kinase C represents a new class of presynaptic calcium sensors that supports short-term plasticity . It is likely that future studies will identify additional members of this class of sensors that allow different synapses to have different forms of short-term plasticity . Further research is also needed to clarify the mechanisms underlying short-term plasticity and to understand how different forms of short-term plasticity are associated with different functions and behaviors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2014 | Protein kinase C is a calcium sensor for presynaptic short-term plasticity |
The oncoprotein transcription factor MYC is a major driver of malignancy and a highly validated but challenging target for the development of anticancer therapies . Novel strategies to inhibit MYC may come from understanding the co-factors it uses to drive pro-tumorigenic gene expression programs , providing their role in MYC activity is understood . Here we interrogate how one MYC co-factor , host cell factor ( HCF ) –1 , contributes to MYC activity in a human Burkitt lymphoma setting . We identify genes connected to mitochondrial function and ribosome biogenesis as direct MYC/HCF-1 targets and demonstrate how modulation of the MYC–HCF-1 interaction influences cell growth , metabolite profiles , global gene expression patterns , and tumor growth in vivo . This work defines HCF-1 as a critical MYC co-factor , places the MYC–HCF-1 interaction in biological context , and highlights HCF-1 as a focal point for development of novel anti-MYC therapies .
MYC oncogenes ( c- , L- , and N- ) encode a family of related transcription factors that are overexpressed in a majority of cancers and responsible for ~100 , 000 cancer-related deaths in the United States each year ( Schaub et al . , 2018 ) . Capable of acting as both transcriptional activators and repressors , MYC proteins ( hereafter 'MYC' ) dimerize with their obligate partner MAX ( Blackwood and Eisenman , 1991 ) to bind and regulate the expression of thousands of genes connected to the cell cycle , protein synthesis , metabolism , genome stability , apoptosis , and angiogenesis ( Tansey , 2014 ) . Fueled by reports that experimental inactivation of MYC promotes tumor regression in mice ( Alimova et al . , 2019; Beaulieu et al . , 2019; Giuriato et al . , 2006; Jain , 2002; Soucek et al . , 2013 ) , there is considerable interest in the idea that MYC inhibitors could form the basis of broadly effective anticancer therapies . MYC itself , however , is widely viewed as undruggable ( Dang et al . , 2017 ) , meaning that effective strategies to pharmacologically inhibit MYC will most likely come from targeting the co-factors with which it interacts to drive and sustain the malignant state ( Brockmann et al . , 2013; Bryan et al . , 2020 ) . The interactome of MYC has been extensively interrogated ( reviewed by Baluapuri et al . , 2020 ) . One effective strategy for prioritizing which of these interaction partners to study has been to focus on those that interact with conserved segments of the MYC protein , which are referred to as ‘MYC boxes’ ( Mb ) ( Meyer and Penn , 2008 ) . In addition to the highly conserved basic helix-loop-helix domain that interacts with MAX , six evolutionarily conserved MYC boxes have been described ( Baluapuri et al . , 2020 ) . On average , MYC boxes are around 15 amino acid residues in length , and although it is clear that they each mediate multiple protein–protein interactions ( Kalkat et al . , 2018 ) , a number of predominant interactors have been described for most of these segments: Mb0 , for example , interacts with the general transcription factor TFIIF to stimulate transcription ( Kalkat et al . , 2018 ) , MbI interacts with the ubiquitin ligase SCFFBW7 to control MYC protein stability ( Welcker et al . , 2004 ) , MbII interacts with the STAGA component TRRAP ( McMahon et al . , 1998 ) to regulate histone acetylation ( Kalkat et al . , 2018 ) , and MbIIIb interacts with the chromatin-associated protein WDR5 ( Thomas et al . , 2015 ) to facilitate its recruitment to ribosomal protein genes ( Thomas et al . , 2019 ) . The two remaining MYC boxes are less well understood , but MbIIIa is important for tumorigenesis ( Herbst et al . , 2005 ) and recruitment of HDAC3 to chromatin ( Kurland and Tansey , 2008 ) , and MbIV interacts with the ubiquitous chromatin-associated protein host cell factor ( HCF ) –1 ( Thomas et al . , 2016 ) . HCF-1 is an essential nuclear protein ( Goto et al . , 1997 ) that is synthesized as a 2035 amino acid precursor and proteolytically cleaved by O-GlcNAc transferase ( OGT ) ( Capotosti et al . , 2011 ) into a collection of amino- ( HCF-1N ) and carboxy- ( HCF-1C ) terminal fragments that remain associated . HCF-1 was first identified through its ability to assemble into a multiprotein–DNA complex with the herpes simplex virus transactivator VP16 , but was later shown to function in uninfected cells as a co-factor for cellular transcription factors , and as part of the Sin3 and MLL/SET histone modifying complexes ( Wysocka and Herr , 2003 ) . The interaction of HCF-1 with VP16 is likely direct and is mediated by a tetrapeptide ‘EHAY’ motif in VP16 that binds to a region within the HCF-1N fragment known as the VP16-induced complex ( VIC ) domain ( Freiman and Herr , 1997 ) . This four residue HCF-1-binding motif ( HBM ) —consensus ( D/E ) -H-x-Y—is present in other viral and cellular transcription factors that interact directly with the HCF-1 VIC domain , including key cell cycle regulators such as the E2F family of proteins ( Tyagi et al . , 2007 ) . We identified HCF-1 as a MYC-associated protein through proteomic approaches , and demonstrated that the interaction occurs through the VIC domain of HCF-1 and an atypical HBM within MbIV that carries the sequence 'QHNY' ( Thomas et al . , 2016 ) . Mutation of these four HBM residues to alanine disrupts the interaction of MYC with HCF-1 in vitro and reduces the ability of MYC to promote murine fibroblast tumor growth in nude mice ( Thomas et al . , 2016 ) . The small and well-defined interaction point between MYC and HCF-1 , and the importance of this interaction to tumorigenesis , raise the possibility that the MYC–HCF-1 nexus could be a viable venue for discovery of novel anti-MYC therapies . If that venue is to be pursued , however , we need to place this interaction in biological context , identify the gene networks that are under its control , and determine whether the MYC–HCF-1 interaction is required for tumor initiation , maintenance , or both . Here we use a combination of loss- and gain-of function approaches to interrogate the role of the MYC–HCF-1 interaction in the context of a canonically MYC-driven cancer—Burkitt lymphoma . We demonstrate that the interaction between MYC and HCF-1 is directly involved in controlling the expression of genes linked to ribosome biogenesis , translation , and mitochondrial function . We define the impact of modulation of this interaction on cell growth , metabolism , and global gene expression patterns . And we show that disrupting the MYC–HCF-1 interaction promotes rapid and persistent tumor regression in vivo . This work reveals how MYC executes a core arm of its pro-tumorigenic gene expression changes , defines HCF-1 as a tumor-critical MYC co-factor , and provides proof-of-concept for a new way to inhibit MYC in the clinic .
To understand the role of HCF-1 in MYC function , we sought to use separation-of-function mutations in MYC that modulate interaction with HCF-1 in a predictable way . We therefore introduced a number of mutations in the atypical HBM of MYC ( QHNY ) that we expected to decrease—or increase—interaction with HCF-1 , based on properties of prototypical HBM sequences ( Freiman and Herr , 1997; Figure 1A ) . We substituted the MYC HBM for the canonical HBM from VP16 ( VP16 HBM ) ; we also mutated the invariant histidine of the HBM to glycine in the MYC ( H307G ) and VP16 ( VP16 HBM:H307G ) contexts , or we changed all four HBM residues to alanine ( 4A ) . A mutation in the separate WDR5-binding motif ( WBM; Thomas et al . , 2015 ) was our specificity control . We transiently expressed these full-length FLAG-tagged MYC proteins in 293T cells and measured their ability to interact with endogenous HCF-1 in a co-immunoprecipitation ( co-IP ) assay ( Figure 1B ) . As expected , the 4A mutation disrupts the MYC–HCF-1 interaction , as do both histidine to glycine substitutions—confirming the essentiality of this core HBM residue to the MYC–HCF-1 association . In contrast , replacing the MYC HBM with the canonical VP16 sequence increases the amount of HCF-1 recovered in the co-IP . The enhanced binding of the VP16 HBM MYC mutant to HCF-1 is also observed in vitro using purified recombinant MYC ( Figure 1—figure supplement 1A ) and in vitro translated HCF-1VIC ( Figure 1—figure supplement 1B ) . Based on these data , we conclude that the MYC HBM is an authentic HBM , and that its variation from the canonical HBM sequence leads to a tempered interaction with HCF-1 . We also conclude that we can use the 4A and VP16 HBM mutations to probe the significance of the MYC–HCF-1 interaction through both loss- and gain-of-function approaches . To understand the cellular consequences of modulating the MYC–HCF-1 interaction , we engineered a system that allows us to express the 4A or VP16 HBM mutant MYC proteins as the sole form of MYC in a cell . We chose Ramos cells , a Burkitt lymphoma ( BL ) -derived line in which a t ( 8;14 ) translocation places one MYC allele under regulatory control of the immunoglobulin heavy chain enhancer ( Figure 1—figure supplement 1C; Wiman et al . , 1984 ) . The untranslocated MYC allele is not expressed in these cells ( Bemark and Neuberger , 2000 ) . Because sequences encoding the MYC HBM are contained within exon 3 , we used CRISPR/Cas9-triggered homologous recombination of the translocated MYC allele to integrate an exon 3 switchable cassette for wild-type ( WT ) MYC , 4A , or VP16 HBM mutants , and confirmed appropriate integration by Southern blotting ( Figure 1—figure supplement 1D and E ) . In cells expressing an inducible CRE-ERT2 recombinase , treatment with 4-hydroxytamoxifen ( 4-OHT ) results in the excision of exon 3 of MYC , bringing in place a modified exon 3 that carries an HA-epitope tag and drives expression of P2A-linked green fluorescent protein ( GFP ) . Twenty-four hours after 4-OHT treatment , at least 85% of cells in each population are switched—as monitored by GFP expression ( Figure 1—figure supplement 1F ) , and we observe the expected appearance of HA-tagged MYC proteins , which migrate more slowly due to the presence of the epitope tag ( Figure 1C ) . Importantly , the exchanged MYC proteins are expressed at levels comparable to endogenous MYC ( Figure 1C ) , are predominantly nuclear ( Figure 1—figure supplement 2A and B ) , and behave as expected , with the 4A mutant showing reduced ( Figure 1D ) , and the VP16 HBM mutant enhanced ( Figure 1E ) , interaction with endogenous HCF-1 . Also as expected , these mutations have minimal impact on the interaction of MYC with WDR5 . Thus , we successfully generated a system for inducible , selective , and bidirectional modulation of the MYC−HCF-1 interaction in the context of an archetypal MYC-driven cancer cell line . To monitor the contribution of the MYC–HCF-1 interaction to cell proliferation , we pulsed each of our engineered Ramos lines with 4-OHT for 2 hr to generate approximately equally mixed populations of switched and unswitched cells . Based on the ability of MYC to drive glutamine addiction ( Jeong et al . , 2014 ) and cell cycle progression ( Pajic et al . , 2000 ) , we monitored how the GFP-positive switched cells in the population compared to their unswitched counterparts in terms of glutamine-dependency ( Figure 1F ) , cell cycle profiles ( Figure 1G ) , and proliferation ( Figure 1H and Figure 1—source data 1 ) . We see that 4A switched cells have a selective advantage over the WT switch in their ability to grow without exogenous glutamine ( Figure 1F ) . This advantage is likely due to loss of the MYC–HCF-1 interaction , as the VP16 HBM mutant cells have a corresponding deficit in growth under glutamine-starvation conditions ( Figure 1F ) . When assayed in media replete with glutamine , cell cycle profiles for the two mutants are modestly altered compared to their WT counterparts , including small but statistically significant changes in the proportion of cells in G2/M ( Figure 1G ) , which again trend in opposite directions for the two MYC mutants—decreasing for the 4A-expressing cells and increasing for those that express the VP16 HBM mutant ( Figure 1G ) . Finally , in long-term growth assays in complete media , we observe that 4A mutant cells are gradually lost from the culture over time , whereas there is a significant enrichment of VP16 HBM cells , compared to the WT control ( Figure 1H and Figure 1—source data 1 ) . The differences in representation of the two MYC mutants in these populations is unlikely due to apoptosis—we observe no differences in the proportion of sub-G1 cells between the different switches ( Figure 1G ) —but tracks with changes in cell doubling time ( Figure 1—figure supplement 2C and D and Figure 1—source data 1 ) , which are increased for the 4A , and decreased for the VP16 HBM mutant cells . The altered and opposing impact of the 4A and VP16 HBM mutations in these assays leads us to conclude that the MYC–HCF-1 interaction promotes the glutamine-dependency—and rapid proliferative status—of these BL cells in culture . As part of our survey of the impact of the MYC–HCF-1 interaction on Ramos cell processes , and because of its influence on glutamine dependency , we determined whether metabolite levels are altered in response to expression of the 4A or VP16 HBM MYC mutants . We performed global , untargeted , mass spectrometry-based metabolomics on switched cells using reverse-phase liquid chromatography ( RPLC ) and hydrophilic interaction liquid chromatography ( HILIC ) separation methods . We detected ~2000 metabolites with each approach ( Figure 2A–F ) , and there is strong consistency among biological replicates ( Figure 2—figure supplement 1A and B ) . In general , more metabolites are significantly changed , and with a greater magnitude , for the 4A than the VP16 HBM MYC mutant ( Figure 2A–B and D–E and Figure 2—source data 1 and 2 ) . For both mutants , significantly changed metabolites group into a variety of categories , with a particular enrichment for those related to amino acid and lipids ( Figure 2C and F ) . Comparing the direction of individual metabolite changes for the 4A and VP16 HBM mutants ( Figure 2—figure supplement 1C ) , we note that a significant portion of the metabolite changes detected by both the RPLC and HILIC methods are in the same direction for the two MYC mutants . In general , these shared metabolite changes fail to cluster strongly into biological pathways; the only significantly enrichment being glycerophospholipid metabolism ( Figure 2—figure supplement 1D ) . Focusing on metabolite changes that occur in opposite directions for the 4A and VP16 HBM mutants , however , we observe significant enrichment in pathways linked to nitrogen and amino acid metabolism ( Figure 2G ) . There is a clear anti-correlation between the impact of the 4A and VP16 HBM mutations on metabolites connected to aspartic acid ( Figure 2H ) , and we observe that intracellular levels of glutamine ( and associated metabolites ) are increased in the 4A and decreased in the VP16 HBM mutant cells ( Figure 2—figure supplement 1E and F ) . Notably , these changes in intracellular amino acid levels are not confined to aspartic acid and glutamine , but there is a general tendency for amino acid levels to be increased in 4A and decreased in VP16 HBM mutant cells , compared to the WT switch ( Figure 2—figure supplement 1C and Table 1 ) . Based on these data , we conclude that the MYC–HCF-1 interaction , directly or indirectly , plays a global role in influencing intracellular amino acid levels in this setting . Next , we used RNA-sequencing ( RNA-Seq ) to monitor transcriptomic changes associated with modulating the MYC–HCF-1 interaction . Twenty-four hours after switching , we observed changes in the levels of ~4000 transcripts in the 4A , and ~3600 transcripts in the VP16 HBM , cells compared to the WT switch ( Figure 3—figure supplement 1A and Figure 3—source data 1 and 2 ) . These changes are highly consistent among biological replicates ( Figure 3—figure supplement 1B ) and modest in magnitude ( Figure 3A ) , congruous with what is typically reported for MYC ( Levens , 2002; Nie et al . , 2012 ) . We confirmed for a representative set of genes that these changes are dependent on switching ( Figure 3—figure supplement 1C ) . Gene ontology ( GO ) enrichment analysis revealed that transcripts decreased by introduction of the 4A mutant are strongly linked to ribosome biogenesis , tRNA metabolism , and the mitochondrial matrix ( Figure 3B ) , while those that are induced have links to transcription , cholesterol biosynthesis , and chromatin . For the VP16 HBM MYC mutant , decreased transcripts cluster in categories mostly related to the centrosome and the cell cycle ( Figure 3C ) . What is particularly striking , however , is that transcripts that are induced by the VP16 HBM protein have a pattern of clustering that is almost the exact opposite of those suppressed by the 4A mutant—including ribosome biogenesis , tRNA metabolism , and the mitochondrial matrix . If anti-correlations between these gain- and loss-of-function mutants can be used to reveal MYC–HCF-1 co-regulated processes , the above data highlight protein synthesis and mitochondrial function as key points of convergence for the interaction of MYC with HCF-1 . To explore this on a gene-by-gene basis , we compared individual gene expression changes that were either the same , or opposite , in direction for the 4A and VP16 HBM mutants ( Figure 3D , Figure 3—figure supplement 1D and Figure 3—source data 3 ) . Transcripts decreased by both mutations show modest enrichment in categories connected to immune signaling and cell adhesion ( Figure 3E , left ) , whereas increased transcripts are primarily enriched in those encoding histones ( Figure 3E , right ) . Turning to transcripts that change in opposite directions with each mutant , those that are induced by the 4A mutant are moderately enriched in categories relating to kinase function and the cell cycle ( Figure 3F , left ) , while those that are reduced by the 4A mutant are strongly enriched in categories connected to ribosome biogenesis and the mitochondrial matrix ( Figure 3F , right ) . The genes represented in each of these categories are shown in Figure 3—figure supplement 2A and B . This analysis confirms that reciprocal changes we observed for the GO categories in Figure 3B and C results from reciprocal changes in the expression of a common set of genes . From our data , we conclude that the MYC–HCF-1 interaction plays an important role in influencing the expression of genes that promote ribosome biogenesis and maintain mitochondrial function . Finally , we interrogated our RNA-Seq data set for transcript changes that correlate with the widespread changes in amino acid levels that occur upon modulation of the MYC–HCF-1 interaction . Here we discovered that the accumulation of amino acids we observe with the 4A mutant is generally matched with a decrease in transcripts of cognate aminoacyl-tRNA synthetases ( Figure 3—figure supplement 3A ) —and vice versa for the decreased amino acid levels in the gain-of-function VP16 HBM mutant ( Figure 3—figure supplement 3B ) . The reciprocal relationship of amino acid levels and tRNA ligase expression changes in response to the 4A and VP16 HBM mutants is consistent with the notion that defects in tRNA charging lead to compensatory changes in amino acid uptake ( Guan et al . , 2014; Harding et al . , 2000 ) , further reinforcing the concept that a key biological context in which MYC and HCF-1 function together is protein synthesis . As a challenge to the concept that ribosome biogenesis and mitochondrial matrix genes are controlled via the MYC–HCF-1 interaction , we asked whether expression of these genes is impacted by acute depletion of HCF-1 , mediated via the dTAG method ( Nabet et al . , 2018 ) . We used CRISPR/Cas9-triggered homologous recombination to integrate an mCherry-P2A-FLAG-FKBP12F36V cassette into the HCFC1 locus in Ramos cells; the effect of which is to amino-terminally tag HCF-1N with the FLAG epitope and FKBP12F36V tags , and to mark the population of modified cells by mCherry expression ( Figure 4—figure supplement 1A ) . Because the HCFC1 locus resides on the X-chromosome , and because Ramos cells are derived from an XY patient , only a single integration event is needed . Tagged cells sorted by fluorescence-activated cell sorting display the expected shift in apparent molecular weight of HCF-1N and the appearance of an appropriately-sized FLAG-tagged species ( Figure 4A ) . Addition of the dTAG-47 degrader results in the rapid and selective disappearance of the HCF-1N fragment; the HCF-1C fragment is largely unaffected by up to 24 hr of dTAG-47 treatment ( Figure 4B ) . Consistent with the known functions of HCF-1 ( Julien and Herr , 2003 ) , treated cells display altered cell cycle profiles ( Figure 4—figure supplement 1B ) , but appear to be able to complete at least one round of cell division , as notable deficits in proliferation are only evident 48 hr after dTAG-47 addition ( Figure 4C and Figure 4—source data 1 ) . These data reveal that the HCF-1N fragment is essential in Ramos cells , and that early time point analyses should be resistant to complicating effects of HCF-1N degradation on cell proliferation . We performed RNA-Seq analysis 3 hr after addition of dTAG-47—a time point at which the majority of HFC-1N is degraded ( Figure 4B ) . Despite the early time point , we identified ~4500 significant transcript changes associated with dTAG-47 treatment of sorted cells ( Figure 4—figure supplement 1C and Figure 4—source data 2 ) . These changes are equally divided between increased and decreased , although decreased transcripts are generally more impacted ( larger median fold-change [FC] ) than those that are induced ( Figure 4D ) . Seventy-five of these differentially expressed genes are also altered in response to dTAG-47 treatment of unmodified Ramos cells ( Figure 4—figure supplement 1D and E and Figure 4—source data 3 ) and were excluded from further analyses . GO enrichment analysis showed that transcripts reduced by HCF-1N degradation are similar in kind to those reduced by the 4A mutation in MYC—including ribosome biogenesis and tRNA metabolic processes ( Figure 4E ) —while those induced by HCF-1N degradation tend to be cell cycle-connected ( Figure 4E and Figure 4—figure supplement 1F ) . We validated representative transcript changes by reverse transcription and quantitative PCR ( RT-qPCR; Figure 4—figure supplement 1G ) . Importantly , many of the genes that are differentially expressed upon HCF-1N degradation are also differentially expressed in the presence of either the 4A or VP16 HBM mutants ( Figure 4—figure supplement 1H ) , and we identified a union set of ~450 genes—oppositely regulated by the 4A and VP16 HBM mutants—the expression of which also changes when HCF-1N is destroyed ( Figure 4F ) . Within this set , loss of HCF-1N tends to mimic the loss of interaction 4A mutant ( Figure 4G ) . Moreover , within the cohort of transcripts that are reduced by both HCF-1N destruction and the 4A mutation , we see clear representation of genes connected to ribosome biogenesis and the mitochondrial matrix ( Figure 4—figure supplement 1I and J ) . Although performing RNA-Seq at this early time likely underestimates the impact of loss of HCF-1N on the transcriptome , the presence of these ribosome biogenesis and mitochondrial matrix genes at the point of coalescence of all our RNA-Seq experiments strongly suggests that they are directly controlled by the MYC–HCF-1 interaction . To help identify direct transcriptional targets of the MYC–HCF-1 interaction , we next compared the genomic locations bound by MYC and HCF-1N in Ramos cells . We performed ChIP-Seq using an antibody against the amino-terminus of HCF-1 ( Machida et al . , 2009 ) , and identified ~1900 peaks for HCF-1N ( Figure 5—source data 1 ) , the majority of which are promoter proximal ( Figure 5A ) . These peaks occur at genes enriched in functions connected to HCF-1 ( Minocha et al . , 2019 ) , including the mitochondrial envelope , the cell cycle , as well as ribonucleoprotein complex biogenesis ( Figure 5B ) . Known ( Figure 5C ) and de novo ( Figure 5—figure supplement 1A ) motif analysis revealed that HCF-1N peaks are enriched in DNA sequences linked to nuclear respiratory factor ( NRF ) –1 , as well as the Sp1/Sp2 family of transcription factors . Although linked to NRF-1 , the ‘CATGCG’ motif is also a non-canonical E-box that MYC binds in vitro and in vivo ( Blackwell et al . , 1993; Haggerty et al . , 2003; Morrish et al . , 2003; Shi et al . , 2014 ) . Overlaying these data with our previous ChIP-Seq analysis of MYC in Ramos cells ( Thomas et al . , 2019 ) , we see that 85% of these HCF-1N peaks are also bound by MYC ( Figure 5D and Figure 5—source data 2 ) . The relationship between MYC and HCF-1N at these sites is intimate ( Figure 5E and F and and Figure 5—figure supplement 1B ) , and sites of co-binding tend to have higher signals for MYC ( Figure 5G ) and HCF-1 ( Figure 5—figure supplement 1C ) than instances where each protein binds alone . As expected from the strong coalescence of MYC and HCF-1N binding events , the properties of shared MYC–HCF-1N peaks are very similar to those of HCF-1N alone , in terms of promoter-proximity ( Figure 5—figure supplement 1D ) , GO enrichment categories ( Figure 5—figure supplement 1E ) , and motif representation ( Figure 5—figure supplement 1F ) . We conclude that most HCF-1N binding sites on chromatin in Ramos cells occur at promoter proximal sites and that the majority of these sites are also bound by MYC . We previously reported that WDR5 has an important role in recruiting MYC to chromatin at a cohort of genes overtly linked to protein synthesis , including more than half of the ribosomal protein genes ( Thomas et al . , 2019 ) . To determine whether these genes are also bound by HCF-1 , we compared our HCF-1N and MYC ChIP-Seq data to those we generated for WDR5 in this setting . Interestingly , there is little overlap of binding sites for MYC , WDR5 , and HCF-1N in Ramos cells , with just ~5% of MYC–HCF-1N co-bound sites also being bound by WDR5 ( Figure 5H ) . Moreover , of the 88 sites bound by all three proteins , only three of these are sites where WDR5 has a functional role in MYC recruitment . Thus , despite the fact that both WDR5 and HCF-1 are often members of the same protein complex ( Cai et al . , 2010 ) , and despite both of them having links to key aspects of protein synthesis gene expression , the two proteins associate with MYC at distinct and separate regions of the genome . Finally , we overlaid the physical location of MYC and HCF-1N on chromatin with gene expression changes we had monitored in earlier experiments . Looking at genes displaying promoter proximal binding of MYC and HCF-1N—where clear gene assignments can be made—we see that approximately one-third of these genes are differentially regulated in the presence of either the 4A or VP16 HBM MYC mutants , and that this is significantly more than that predicted by chance alone ( Figure 5I and Figure 5—figure supplement 1G ) . For the 4A mutant , a slightly greater proportion of co-bound genes are downregulated , while the opposite is true for the VP16 HBM mutant ( Figure 5—figure supplement 1G ) . A relatively small cohort of MYC–HCF-1N co-bound genes are oppositely impacted by the 4A and VP16 HBM mutants ( Figure 5—figure supplement 1G; ‘4A/VP’ ) , but comparing these with those deregulated by depletion of HCF-1N ( Figure 5J ) , we again see that a majority are connected to ribosome biogenesis and mitochondria , and that most are positively regulated by HCF-1 and the MYC–HCF-1 interaction . Together , these data strongly support the notion that ribosome biogenesis and mitochondrially connected genes are direct targets of the MYC–HCF-1 interaction . It has been reported that deletion of MbIV from N-MYC reduces the ability of MYC:MAX dimers to bind DNA ( Cowling et al . , 2006 ) . This phenotype is unrelated to the MYC HBM , however , as we determined that neither the 4A nor the VP16 HBM mutations have an overt impact on the binding of recombinant MYC:MAX dimers to DNA in vitro ( Figure 6—figure supplement 1A and B ) . To further explore whether the MYC–HCF-1 interaction influences the ability of either protein to engage its chromatin binding sites in cells , we performed ChIP-Seq for HCF-1N and MYC-HA in our switchable MYC cells that were treated with 4-OHT for 24 hr . Binding of HCF-1N to chromatin is largely unaffected by the 4A or VP16 HBM mutations ( Figure 6A and Figure 6—figure supplement 1C ) , demonstrating that MYC does not recruit HCF-1 to chromatin . Binding of MYC is subtly altered by both the 4A and VP16 HBM mutations ( Figure 6B and Figure 6—figure supplement 1D ) , but these changes are widespread and for the most part shared between the loss-of-function and gain-of-function MYC mutants ( Figure 6—source data 1 ) . Indeed , focusing on the top 140 significant changes , we see that the 4A and the VP16 HBM mutants both tend to have increased or expanded chromatin binding , compared to WT MYC ( Figure 6C ) . Visual inspection of the ChIP-Seq data ( Figure 6D and Figure 6—figure supplement 1E ) confirms the subtlety of these effects and reinforces the concept that the binding of MYC ( and HCF-1 ) to chromatin is not impacted in opposite ways by the 4A and VP16 HBM mutations . We further verified by ChIP-PCR—at genes flagged as direct targets in Figure 5 —that binding of MYC ( Figure 6E ) and HCF-1 ( Figure 6—figure supplement 1F ) is largely insensitive to the 4A and VP16 HBM mutations . Thus , although the HBM may play a modest role in chromatin targeting by MYC in cells , this is likely to be independent of the MYC–HCF-1 interaction . We conclude that MYC and HCF-1 interact to control the expression of ribosome biogenesis and mitochondrially connected genes through a co-recruitment-independent mechanism . The ability of MYC to regulate ribosome ( van Riggelen et al . , 2010 ) and mitochondrial ( Morrish and Hockenbery , 2014 ) biogenesis are core aspects of its tumorigenic repertoire . We would expect , therefore , that disrupting the MYC–HCF-1 interaction would have a significant impact on the ability of Ramos lymphoma cells to establish and maintain tumors in vivo . To address this expectation , we tested the impact of the 4A MYC mutant on tumorigenesis in mice . Because this is such an aggressive tumor model ( Thomas et al . , 2019 ) , we did not test the gain-of-function VP16 HBM mutant . In these experiments , we included a second , independent clone carrying the switchable 4A mutation ( 4A-1 and 4A-2 ) ; we also included a switchable ∆264 mutant ( Thomas et al . , 2019 ) , which deletes residues in the carboxy-terminal half of MYC required for its nuclear localization , as well as interaction with WDR5 , HCF-1 , and MAX . First , we assayed tumor engraftment by switching the engineered cells in culture and then injecting into the flanks of nude mice ( Figure 7A ) . As expected , the WT to WT switched cells develop tumors rapidly in vivo ( Figure 7B , Figure 7—figure supplement 1A , and Figure 7—source data 1 ) , resulting in all mice reaching humane endpoints and being euthanized by 21 days post-injection ( Figure 7C ) . In contrast , 4A-1 , 4A-2 , and ∆264 switched cells are significantly delayed , both in tumor growth ( Figure 7B , Figure 7—figure supplement 1A , and Figure 7—source data 1 ) and mortality ( Figure 7C ) . Although 4A-1 , 4A-2 , and ∆264 switched cells did form tumors , these appear to originate from the outgrowth of unswitched cells in the injected populations , as ~75% of cells in these tumors are in their unswitched state ( Figure 7D ) . In this assay , therefore , the MYC–HCF-1 interaction is required for tumor growth , and there is little if any difference between disruption of the MYC–HCF-1 interaction and disabling the majority of the nuclear functions of MYC . Next , we injected unswitched cells into the flanks of mice , allowed tumors to form , and then switched to each of the MYC variants by injecting mice with tamoxifen ( Figure 7E ) . As we observed previously ( Thomas et al . , 2019 ) , the ‘WT to WT’ tumors continue to grow rapidly after switching ( Figure 7F , Figure 7—figure supplement 1B , and Figure 7—source data 1 ) , and all mice had to be euthanized before 30 days ( Figure 7G ) . For the 4A switches , however , tumors rapidly regressed ( Figure 7F ) , and all mice survived—and were tumor free—for the 60-day duration of the experiment ( Figure 7—figure supplement 1B and Figure 7—source data 1 ) . Regression of the 4A tumors occurs at a pace that is virtually indistinguishable from the ∆264 mutant ( Figure 7—figure supplement 1B and Figure 7—source data 1 ) , and like the ∆264 scenario , is accompanied by high levels of apoptosis , as measured by Annexin V staining ( Figure 7H ) , caspase activity ( Figure 7—figure supplement 1C ) , and sub-G1 DNA content ( Figure 7—figure supplement 1D ) . We conclude that the interaction of MYC with HCF-1 is essential for tumor maintenance in this context . Finally , we performed RNA-Seq on tumor material excised 48 hr after switching ( Figure 7—figure supplement 1E and Figure 7—source data 2 ) . Thousands of differentially expressed genes were identified in the mutant switch tumors , many of which are shared between the 4A and ∆264 mutants—for both the decreased ( Figure 7—figure supplement 1F ) and induced ( Figure 7—figure supplement 1G ) directions . Common reduced transcripts are enriched in those connected to ribosome biogenesis , translation , and mitochondrial envelope ( Figure 7—figure supplement 1H ) , whereas those that are induced are enriched for functions including transcription co-regulator activity , kinase binding , and the vacuole ( Figure 7—figure supplement 1I ) . Many of these gene expression changes are likely due to indirect effects of tumor regression . To tease these changes apart from those more closely connected to the MYC–HCF-1 interaction , we overlaid tumor RNA-Seq with that generated for the 4A MYC mutant in vitro ( Figure 3 ) . Responses that were only observed in the tumor regression model were linked to fairly broad categories such as ‘protein binding’ or ‘mRNA metabolism’ ( Figure 7—figure supplement 1J and K ) . Responses that were shared between the in vitro and in vivo systems were , in contrast , more extensive and specific . Indeed , more than 70% of the genes repressed in the 4A cell line are also repressed in either 4A-1 or 4A-2 tumors , and there is a common set of 942 genes that are shared between all three data sets ( Figure 7I ) . These genes coalesce on those connected to ribosome biogenesis , translation , and the mitochondria ( Figure 7J ) . The overlap of induced genes was less pronounced ( 30%; Figure 7K ) and these genes are less clustered , although we do observe modest enrichment in categories connected to metabolism , chromatin binding , and transcription coregulator activity ( Figure 7L ) . The recurring connections we observe between the MYC–HCF-1 interaction and ribosome biogenesis and mitochondrial function , both in vitro and in vivo , strongly support the notion that a major function of this interaction is to stimulate ribosome production and mitochondrial vigor , and that these actions are central for the ability of MYC to drive tumor onset and maintenance .
The wealth of MYC interaction partners provides a rich resource for the discovery of novel ways to eventually inhibit MYC in the clinic . Unfortunately , the complexity of the MYC interactome also presents a barrier to prioritizing which co-factors to pursue . The highest priority co-factors should be those that play a critical role in the core tumorigenic functions of the protein , and where there is proof-of-concept that disrupting interaction with MYC would provide a therapeutic benefit in the context of an existing malignancy . Here we provide this information for HCF-1 . We show that MYC interacts with HCF-1 via a non-canonical HBM , identify roles for the MYC–HCF-1 interaction in the control of genes involved in ribosome biogenesis , translation , and mitochondrial function , and show that loss of the MYC–HCF-1 interaction promotes frank and irreversible tumor regression in vivo . Although we do not yet know if a therapeutic window exists for targeting MYC through HCF-1 , and we do not know if our findings will extend to other tumor types , this work , together with our previous study ( Thomas et al . , 2016 ) , highlights HCF-1 as a critical MYC co-factor and one worth pursuing as means to inhibit MYC in cancer . Through the use of mutations that bidirectionally modulate the interaction between MYC and HCF-1 , inducible degradation of HCF-1N , and ChIP-Seq analyses , we identified a relatively small set of genes that we posit are direct targets of the MYC–HCF-1 interaction . It is possible that the multi-pronged strategy we took excludes some bonafide MYC–HCF-1 target genes . But this approach is important because we cannot exclude the possibility that MbIV interacts with multiple factors besides HCF-1 . Indeed , we observed instances where the 4A and VP16 HBM mutations produce similar effects on MYC behavior , suggesting that they impact some common aspect of MYC function that is independent of HCF-1 . Nonetheless , the genes surviving our stringent criteria can be considered high-confidence targets , and are particularly interesting because of their ( i ) biological clustering , ( ii ) connections to core pro-tumorigenic MYC activities , and ( iii ) ability to account for many of the overt consequences of modulating the MYC–HCF-1 interaction on metabolism and tumorigenesis . What is also interesting about this set of genes is that they appear to be regulated by MYC and HCF-1 through a co-recruitment-independent mechanism . Among the direct targets of the MYC−HCF-1 interaction are genes that catalyze rate-limiting steps in both ribosome biogenesis ( rDNA transcription by POLR1A ) and translation ( initiator tRNA binding to start codon by EIF2S1 [EIF-2α] ) ( Hershey , 1991; Laferté et al . , 2006 ) . MYC regulates ribosome biogenesis through controlling and coordinating the transcription of ribosomal DNA , ribosomal protein genes , and components in the processing and assembly of ribosomes ( van Riggelen et al . , 2010 ) . Furthermore , regulation of tRNA ligases by MYC is considered an essential contributor to MYC-driven cell growth in Drosophila ( Zirin et al . , 2019 ) . Interestingly , even with the direct MYC–HCF-targets classified as mitochondrially connected , we see that links to protein synthesis—MARS2 , MRPS12 , and MRPL32 , for example—are specifically involved in the synthesis of mitochondrial proteins , including those required for oxidative phosphorylation . We conclude that , in the context of its relationship with MYC , HCF-1 is dedicated to promoting multiple aspects of biomass accumulation . The notion that there are process-specific co-factors for MYC is not widely appreciated . MIZ-1 ( Vo et al . , 2016 ) and WDR5 ( Thomas et al . , 2019 ) are perhaps the best examples to date; our work here on HCF-1 further reinforces this concept . There are some particularly interesting parallels between the actions of WDR5 and HCF-1 as process-specific co-factors for MYC: both interact with MYC through conserved MYC boxes , both control the expression of relatively small cohorts of genes , both are required to interact with MYC to establish and maintain tumors , and both have clear links to biomass accumulation . Beyond this point , however , these parallels break down . There is little overlap of target genes regulated by the MYC–WDR5 and MYC–HCF-1 interactions , and whereas MYC and HCF-1 associate with predominantly control ribosome biogenesis , MYC and WDR5 work together to stimulate the expression of genes encoding structural components of the ribosome ( Thomas et al . , 2019 ) . This division of labor between HCF-1 and WDR5 in different aspects of protein synthesis is particularly intriguing given that HCF-1 and WDR5 work together as part of multi-protein chromatin modifying complexes ( Tyagi et al . , 2007; Wysocka et al . , 2003 ) . Yet MYC interacts with each protein through separate MYC boxes , and indeed the way in which MYC interacts with WDR5 likely precludes WDR5 from assembling into canonical HCF-1-containing complexes ( Thomas et al . , 2015 ) . We believe the separation of interaction surfaces allows MYC to access non-canonical functions of both WDR5 and HCF-1 , and may have evolved to permit discrete regulation of the constituents of the ribosomes versus the factors required to assemble these constituents into ribosome particles . Indeed , because HCF-1 ( Mazars et al . , 2010 ) , MYC ( Chou et al . , 1995 ) , and the MYC–HCF-1 interaction ( Itkonen et al . , 2019 ) all have ties to the metabolic sensor OGT ( Swamy et al . , 2016 ) , it is possible that this separation of function allows for rapid modulation of ribosome assembly by HCF-1 when the metabolic state of the cell declines , while at the same time ensuring that ribosome subunits are present for rapid recovery when metabolism ramps up . The concept that HCF-1 is a biomass-specific co-factor for MYC can also account for our discovery that intracellular amino acid levels are increased when the MYC–HCF-1 interaction is disrupted . Impairment of ribosome biogenesis and translation can lead to an accumulation of amino acids and compensatory changes in the expression of their transporters ( Guan et al . , 2014; Scott et al . , 2014 ) . Of note , the way in which amino acid levels almost globally respond to changes in the MYC–HCF-1 interaction can also provide a simple explanation for how the MYC–HCF-1 interaction contributes to the glutamine addiction of Ramos cells . If MYC partners with HCF-1 to drive biomass accumulation , the net effect of this interaction will be to increase the demand for intracellular amino acids , with glutamine playing a particularly important role in the biosynthesis of multiple non-essential amino acids , including glutamate , arginine , proline , and alanine ( Hosios et al . , 2016 ) . The ability of MYC to drive glutamine-addiction is a defining characteristic of its tumorigenic repertoire ( Tansey , 2014 ) , and is thought to result from the ability of MYC to promote glutaminolysis and induce the expression of amino acid transporters ( Wise et al . , 2008 ) . But the accumulation of glutamine we observe upon disruption of the MYC–HCF-1 interaction , together with the specific role HCF-1 plays in MYC function , suggests that glutamine addiction might also be fueled by the well-established ability of MYC to stimulate protein synthesis ( Iritani and Eisenman , 1999 ) . Our discovery that MYC and HCF-1 work together to regulate transcription via a co-recruitment-independent mechanism is surprising , but not without precedent . E2F transcription factors interact with HCF-1 through a canonical HBM to control the expression of genes connected to cell cycle progression , yet depletion of these proteins has little effect on the recruitment of either to chromatin ( Iwata et al . , 2013; Parker et al . , 2014; Tyagi et al . , 2007 ) . Similarly , Drosophila Myc and HCF interact to activate transcription and control growth , but their respective interactions with chromatin are independent of one another ( Furrer et al . , 2010 ) . A co-recruitment-independent mechanism may thus be a common way in which transcription factors interact with HCF-1 to modulate transcription . In imagining how this might occur , it is interesting to note that the biochemical context in which HCF-1 exists can be influenced by client proteins; its interaction with E2F1 drives association with SETD1A complexes , whereas its interaction with E2F4 favors binding to the SIN3A HDAC complex ( Tyagi et al . , 2007 ) . Perhaps , therefore , the post-recruitment interaction of MYC and HCF-1 modulates transcription by promoting the ejection of inhibitory proteins like SIN3A from , or recruiting activating proteins like SETD1A to , HCF-1 . Further experimentation will be required to determine the mechanism of action . Finally , there are two additional cancer-relevant connections worth mentioning . First , the COSMIC database ( Tate et al . , 2019 ) identifies three cancer-associated mutations within the MYC HBM , all of which convert the subprime ‘QHNY’ motif to ‘EHNY’ , which matches the perfect HBM consensus and closely resembles the gain-of-function VP16 HBM we used in this study . It is possible , therefore , that these rare mutations in MYC could contribute to disease progression via enhancement of the MYC–HCF-1 interaction . This is an intriguing idea , and one worth testing in a less aggressive in vivo model where increases in the tumorigenic potential of MYC can be visualized . Second , our demonstration that disrupting the MYC–HCF-1 interaction in the context of an existing tumor promotes its regression provides compelling proof-of-concept for the idea that inhibitors of this interaction could have utility as anticancer agents . Switching WT MYC to the 4A mutant caused rapid and widespread induction of apoptosis and was associated with changes in the expression of genes connected to ribosome biogenesis , translation , and the mitochondria , consistent with the idea that reduced expression of these MYC–HCF-1 target genes triggers the regression process ( Hanahan and Weinberg , 2011; Pelletier et al . , 2018 ) . The small and well-defined nature of the HBM suggests that if structural information becomes available for the HCF-1 VIC domain , it could be possible to develop small molecule inhibitors that block the MYC–HCF-1 interaction . The most obvious concern with this strategy is that HCF-1 is not a MYC-specific co-factor , and that its interactions with other transcription factors may limit or prevent attainment of a therapeutic window . To our knowledge , MYC proteins and E2F3a are the only transcription factors that interact with HCF-1 via an ‘imperfect’ HBM , which we show here is sub-optimal for robust HCF-1 association . It might be possible to develop a therapeutic window by exploiting the non-canonical nature of the HBM in MYC , with the expectation that this interaction will be more sensitive than others that carry higher affinity HBM motifs . We note , however , that many of the factors with which HCF-1 interacts via an HBM are inherently pro-proliferative , with the E2F proteins in particular playing a predominant role in cancer initiation , maintenance , and response to therapies ( Kent and Leone , 2019 ) . We also note that HCF-1 has been reported to be overexpressed in cancer , and its overexpression can correlate with poor clinical outcomes ( Glinsky et al . , 2005 ) . It is possible , therefore , that on-target collateral consequences of inhibiting the MYC–HCF-1 interaction could also have therapeutic benefit against malignancies .
See Supplementary file 1 for primer sequences . PCRs were performed using either Q5 DNA Polymerase ( NEB , Ipswich , Massachusetts , M0491 ) or OneTaq DNA Polymerase ( NEB M0480 ) . Gibson assemblies were performed using Gibson Assembly Cloning Kit ( NEB E5510 ) . More specific details about cloning steps can be found in the relevant sections . 293T cells were maintained in DMEM with 4 . 5 g/l glucose , L-Glutamine , and sodium pyruvate ( Corning , Corning , New York , 10–013-CV ) , and supplemented with 10% fetal bovine serum ( FBS , Denville Scientific , Metuchen , New Jersey , FB5001-H ) , and 1% penicillin/streptomycin ( P/S , Gibco , Waltham , Massachusetts , 15140122 ) . Ramos cells were obtained directly from the ATCC ( Manassas , Virginia , CRL-1596 ) and maintained in RPMI 1640 with L-glutamine ( Corning 10–040-CV ) , and supplemented with 10% FBS ( Denville Scientific FB5001-H ) , and 1% P/S ( Gibco 15140122 ) . All cell lines used were confirmed as mycoplasma-negative; 293T cells were authenticated by STR profiling . Rabbit anti-HCF1C polyclonal ( Bethyl Laboratories , Montgomery , Texas , Cat# A301-399A ) ; rabbit anti-HCF1N polyclonal ( Machida et al . , 2009 ) ; rabbit anti-c-MYC ( Y69 ) monoclonal ( Abcam , Cambridge , United Kingdom , Cat# ab32072 ) ; rabbit anti-HA ( C29F4 ) monoclonal ( Cell Signaling Technology , Danvers , Massachusetts , Cat# 3724 ) ; rabbit anti-WDR5 ( D9E1I ) monoclonal ( Cell Signaling Technology Cat# 13105 ) ; rabbit anti-IgG polyclonal ( Cell Signaling Technology Cat# 2729 ) ; mouse anti-FLAG ( M2 ) monoclonal , HRP-conjugated ( Sigma-Aldrich , St . Louis , Missouri , Cat# A8592 ) ; mouse anti-T7 monoclonal , HRP conjugated ( Millipore , Burlington , Massachusetts , Cat# 69048 ) ; mouse anti-GAPDH ( GA1R ) monoclonal , HRP conjugated ( Thermo Fisher Scientific , Waltham , Massachusetts , Cat# MA5-15738-HRP ) ; rabbit anti-GAPDH ( D16H11 ) monoclonal , HRP conjugated ( Cell Signaling Technology Cat# 8884 ) ; goat anti-rabbit IgG polyclonal , HRP conjugated ( Thermo Fisher Scientific Cat# 31463 ) ; mouse anti-rabbit IgG monoclonal , light chain specific , HRP conjugated ( Jackson ImmunoResearch Laboratories , West Grove , Pennsylvania , Cat# 211-032-171 ) ; rabbit anti-histone H3 ( D1H2 ) XP monoclonal , HRP conjugated ( Cell Signaling Technology Cat# 12648 ) ; rabbit anti-α-Tubulin ( 11H10 ) monoclonal , HRP conjugated ( Cell Signaling Technology Cat# 9099 ) ; goat anti-rabbit IgG polyclonal , Alexa Fluor 594 conjugated ( Thermo Fisher Scientific A11012 ) . Q5 site-directed mutagenesis of MYC-WT targeting vector from Thomas et al . , 2019 was used to create MYC-4A ( 4A_F and 4A_R ) and MYC-VP16 HBM ( VP16 HBM_F and VP16 HBM_R ) . The pGuide plasmid described by Thomas et al . , 2019 was used as a backbone to introduce the sgRNA sequence GCTACGGAACTCTTGTGCGTA ( pGuide-MYC1 ) by Q5 site directed mutagenesis with the primers GUIDE MYC-1A and GUIDE MYC-1B . For the generation of switchable cells , 10 million Ramos cells stably expressing CRE-ERT2 ( Thomas et al . , 2019 ) were electroporated ( BioRad Gene Pulser II , 220 V and 950 μF ) with 10 μg of relevant targeting vector ( MYC-4A or MYC-VP16 HBM ) , 15 μg pGuide-MYC1 , and 15 μg pX330-U6-Chimeric_BB-CBh-hSpCas9 ( gift from Feng Zhang , AddGene plasmid #42230 ) ( Cong et al . , 2013 ) . WT and ∆264 cell lines were the same as those used in Thomas et al . , 2019 . Cells were treated with 150 ng/ml puromycin ( Sigma-Aldrich P7255 ) and 100 μg/ml hygromycin ( Corning 30240CR ) , selecting for the switchable MYC cassette and CRE-ERT2 recombinase , respectively . Following selection , single cells , stained using propidium iodide ( PI , Sigma-Aldrich P4864 ) for viability , were sorted by the Vanderbilt Flow Cytometry Shared Resource using a BD FACSAria III flow cytometer into a 96-well plate to generate clonal cell lines under puromycin and hygromycin selection . Individual clones were expanded , and initially validated by switching for 24 hr using 20 nM ( Z ) −4-Hydroxytamoxifen ( 4-OHT , Tocris , Minneapolis , Minnesota , 3412 ) , and flow cytometry ( see below ) for GFP expression . Further validation was performed by western blotting after 24 hr 20 nM ±4 OHT ( see below ) and by sSouthern blotting ( see below ) . The 4A cell line ( 4A-1 ) used for the majority of experiments is haploinsufficient for part of chromosome 11 , approximately between co-ordinates 118 , 685 , 194 and 134 , 982 , 408 . This region of the genome was excluded from genomic analyses . For all experiments , switching was performed by treatment with 20 nM 4-OHT for 2 or 24 hr ( see relevant method or figure legend ) . To create pCRIS-mCherry-FLAG-dTAG-HCFC1 , pCRIS-PITChv2-Puro-dTAG ( BRD4 ) ( gift from James Bradner , AddGene plasmid #91793 ) ( Nabet et al . , 2018 ) was first modified to remove the BRD4 homology arms and replace the 2XHA tags with a FLAG tag . This was done by Gibson assembly of the vector ( Q5 amplification using pCRIS-HCFC1N_F and pCRIS_R ) , puromycin cassette ( Q5 amplification using Puro-1 and Puro-FLAG_R ) , and FKBP12FV ( Q5 amplification using FLAG-FKBP_F and pCRIS-FKBP_R ) . The resulting vector was again modified using Gibson assembly by combining the vector ( Q5 amplification using pCRIS-HCFC1N_F and pCRIS-HCFC1N_R ) , an upstream 271 bp HCFC1 5’ homology arm ( hg19 chrX:153236265–153236535 , OneTaq amplification using HCFC1N_F and HCFC1N-5’Hom_R ) , mCherry ( Q5 amplification using mCherry_F and mCherry_R ) , FKBP12FV ( Q5 amplification using HCFC1-mCherry-FKBP_F and HCFC1-mCherry-FKBP_R ) , and a downstream 800 bp HCFC1 3’ homology arm ( hg19 chrX:153235465–153236264 , OneTaq amplification using HCFC1N-3’Hom_F and HCFC1N-3’Hom_R ) . The pGuide plasmid described by Thomas et al . , 2019 was used as a backbone to introduce the guide RNA sequence CAGAAGCACCGCTGGCAAGT ( pGuide-HCFC1-N ) by Q5 site-directed mutagenesis with the primers HCFC1N-sgRNA_F and HCFC1N-sgRNA_R . Fifteen micrograms of pGuide-HCFC1-N and 10 μg of pCRIS-mCherry-FLAG-dTAG-HCFC1 were electroporated ( BioRad , Hercules , California , Gene Pulser II , 220 V and 950 μF ) into Ramos cells , alongside 15 μg pX330-U6-Chimeric_BB-CBh-hSpCas9 ( gift from Feng Zhang , AddGene plasmid #42230 ) ( Cong et al . , 2013 ) into 10 million Ramos cells . Because HCFC1 is on the X chromosome and Ramos cells are XY ( Klein et al . , 1975 ) , only a single copy of HCFC1 is present for targeting using CRISPR/Cas9 . Following electroporation , cells were expanded and a population of mCherry-positive cells , stained using Zombie NIR viability dye ( BioLegend , San Diego , California , 423105 ) , was sorted by the Vanderbilt Flow Cytometry Shared Resource using a FACSAria III flow cytometer ( Becton Dickinson ( BD ) , Franklin Lakes , New Jersey ) . This population of cells was expanded further before validation by western blotting . All experiments were conducted using this population and were treated with either DMSO ( Sigma-Aldrich D2650 ) or 500 nM dTAG-47 , which was synthesized by the Vanderbilt Institute of Chemical Biology Synthesis Core . Q5 site-directed mutagenesis of pFLAG-MYC WT ( Thomas et al . , 2015 ) was used to generate the following plasmids: pFLAG-MYC 4A ( 4A_F and 4A_R ) , pFLAG-MYC H307G ( H307G_F and H307G_R ) , pFLAG-MYC VP16 HBM ( VP16 HBM_F and VP16 HBM_R ) , and pFLAG-MYC VP16 HBM H307G ( VP16 HBM H307G_F and VP16 HBM H307G_R ) . pFLAG-MYC WBM is from Thomas et al . , 2015 . Plasmid ( 19 μg ) was prepared with 0 . 25 M CaCl2 , incubated for 10 min with 1× HBS ( 2× HBS: 140 mM NaCl , 1 . 5 mM Na2HPO4 , 50 mM HEPES , pH 7 . 05 ) , and then applied drop-wise to 293T cells . Cells were grown for 2 days before harvesting ( see below ) . Cell lysates for western blotting or immunoprecipitation ( IP ) were prepared by rinsing cells twice in ice-cold 1× PBS , and harvesting in Kischkel Buffer ( 50 mM Tris pH 8 . 0 , 150 mM NaCl , 5 mM EDTA , 1% Triton X-100 ) + protease inhibitor cocktail ( PIC , Roche , Basel , Switzerland , 05056489001 ) . Cells were sonicated at 25% power for 10 s ( Cole-Parmer , Vernon Hills , Illinois , GE 130PB-1 ) , debris removed by centrifugation , and protein concentration determined using Protein Assay Dye ( BioRad 500–0006 ) against a bovine serum albumin ( BSA ) standard . For western blotting , lysate was diluted in 5× Laemmli Buffer ( 375 mM Tris pH 6 . 8 , 40% glycerol , 10% SDS , bromophenol blue , 2-Mercaptoethanol ) . For IP , the concentration of samples was balanced using Kischkel Buffer . Antibody or anti-FLAG M2 Affinity Gel ( Millipore A2220 ) was added , and samples rotated overnight at 4°C . For unconjugated antibodies , 20 μl bed volume of Protein A Agarose ( Roche 11134515001 ) was added the following day to each sample , and rotated for 2–4 hr at 4°C . Samples were then washed 4× with Kischkel buffer ( 2× 4°C and 2× at room temperature ) , and incubated in Laemmli buffer for 5 min at 95°C . Protein from lysates and IPs were separated out by SDS-polyacrylamide gel electrophoresis ( PAGE ) in running buffer ( 25 mM Tris , 192 mM glycine , 0 . 2% SDS ) . Wet transfer to PVDF ( PerkinElmer , Waltham , Massachusetts , NEF1002 ) was carried out in Towbin Transfer Buffer ( 25 mM Tris , 192 mM glycine , and 10% methanol ) . Membrane was blocked in 5% milk in TBS-T ( 20 mM Tris pH 7 . 6 , 140 mM NaCl , 0 . 1% Tween-20 ) , hybridized overnight in primary antibody ( or 1 hr for HRP-conjugated ) , and for 1 hr in HRP-conjugated secondary antibody ( if required ) . ECL substrates , SuperSignal West Pico ( Pierce , Waltham , Massachusetts , 34080 ) , Pico+ ( Pierce 34580 ) , and Femto ( Pierce 34095 ) were used in various combinations for detection of bands by exposure to film . Chromatin fractionation was performed , with slight modification , as described by Méndez and Stillman , 2000 . Switchable MYC Ramos cells that had been treated for 24 hr with 20 nM 4-OHT were washed in ice-cold PBS , resuspended in 200 µl Buffer A ( 10 mM HEPES pH 7 . 9 , 10 mM KCl , 1 . 5 mM MgCl2 , 0 . 34 M sucrose , 10% glycerol , 1 mM DTT ) + PIC + PMSF + 0 . 1% Triton X-100 , and incubated on ice for 10 min . The resulting lysate was centrifuged 1300 × g for 5 min at 4°C , with the pellet ( P1 ) containing the nuclei . The supernatant ( S1 ) was centrifuged at 20 , 000 × g for 10 min at 4°C , giving pellet P2 and supernatant S2 . P2 was discarded and S2 corresponding to the soluble portion of the total cell extract was diluted out in Laemmli buffer . P1 was gently washed in Buffer A , resuspended by pipetting up and down in Buffer B ( 3 mM EDTA , 0 . 2 mM EGTA , 1 mM DTT ) + PIC + PMSF , and incubated on ice for 30 min . The lysed nuclei were centrifuged at 1700 × g for 5 min at 4°C to give soluble nuclear proteins ( S3 ) and chromatin-bound proteins ( P3 ) . S3 was diluted out in Laemmli buffer . P3 was gently washed in Buffer B , resuspended in Laemmli buffer , and sonicated for 15 s at 25% power . All samples were incubated at 95°C for 3 min . Proteins were separated out by SDS-PAGE and transferred to PVDF , as described above , and probed for HCF-1C , HA , tubulin , and H3 . Following a 24 hr treatment with 20 nM 4-OHT , 105 switchable MYC Ramos cells were attached to slides by CytoSpin ( 800 RPM , 3 min ) , fixed for 10 min with 3% methanol-free formaldehyde ( Thermo Fisher Scientific 28908 ) diluted in PBS , and washed three times with PBS . Cells were permeabilized for 10 min with Permeabilization Solution ( 0 . 1% Triton X-100 in PBS ) , blocked for 1 hr with Blocking Solution ( 2 . 5% BSA in Permeabilization Solution ) , and incubated with Blocking Solution containing anti-HA antibody ( 1:500 , Cell Signaling 3724 ) for 1 hr . Cells were washed three times with PBS and incubated with Blocking Solution containing Goat anti-Rabbit IgG Alexa Fluor 594 antibody ( 1:350 , Thermo Fisher Scientific A11012 ) for 1 hr . Cells were then washed three times with PBS , and coverslips mounted with ProLong Gold Antifade Mountant with DAPI ( Thermo Fisher Scientific P36941 ) . Slides were imaged by wide-field fluorescent microscopy on a Nikon Eclipse Ti equipped with a Nikon Plan Apo λ 100×/1 . 45 Oil objective , Nikon DS-Qi2 camera , and Excelitas X-Cite 120LED illuminator using identical settings for each sample and representative images shown . Cells were filtered into 35 μm nylon mesh Falcon round bottom test tubes for flow cytometry , which was performed in the Vanderbilt Flow Cytometry Shared Resource . Single cells were gated based on side and forward scatter using the stated instruments . To determine the proportion of switching of the switchable MYC allele Ramos cell lines , cells were fixed in 1% formaldehyde ( FA , Thermo Fisher 28908 ) in PBS for 10 min at room temperature . The number of GFP-positive cells was determined using a BD LSR II flow cytometer . For cell cycle analysis of the switchable MYC allele Ramos cell lines , cells were treated with 4-OHT for 2 hr , at which point the media was replaced . Cells were maintained for 7 days , at which point 1 × 106 cells were collected and fixed in 1% FA ( Thermo Fisher 28908 ) in PBS for 10 min at room temperature , then washed 2× with PBS . Permeabilization and staining was done using cell cycle staining buffer ( PBS , 10 μg/ml PI [Sigma-Aldrich P4864] , 100 μg/ml RNAse A , 2 mM MgCl2 , 0 . 1% Triton X-100 ) for 25 min at room temperature , then stored overnight at 4°C . PI staining of at least 10 , 000 single cells for each of the GFP-negative ( GFP− ) and GFP-positive ( GFP+ ) populations was measured using a BD LSR Fortessa , and cell cycle distribution determined using BD FACSDIVA software . For cell cycle analysis of the FKBPFV-HCF-1N Ramos cells , cells were treated with DMSO or 500 nM dTAG for 24 hr . One million cells were collected and fixed in ethanol overnight . After fixation , cells were stained overnight at 4°C in cell cycle staining buffer ( PBS , 10 μg/ml PI [Sigma-Aldrich P4864] , 100 μg/ml RNAse A , 2 mM MgCl2 ) . PI staining of at least 10 , 000 single cells was measured using a BD LSR Fortessa , and cell cycle distribution determined using BD FACSDIVA software . See section 'In vivo studies: Tumor formation and maintenance assays' for details regarding flow cytometry experiments conducted on cells extracted from mice tumors . Genomic DNA ( gDNA ) was prepared from parental and unswitched MYC switchable cells ( WT , 4A , and VP16 HBM ) ( Miller et al . , 1988 ) . Briefly , cells were rinsed in ice-cold 1× PBS and resuspended in DNA extraction buffer ( 10 mM Tris pH 8 . 1 , 400 mM NaCl , 10 mM EDTA , 1% SDS , 50 μg/ml proteinase K [PK , Macherey-Nagel , Düren , Germany , 740506] ) . Lysis was performed overnight in a rotisserie at 56°C , before gDNA was extracted using ethanol precipitation . Southern blot was performed similar to that described by Southern , 1975 . gDNA ( 10 μg ) was digested using XbaI ( NEB R0145 ) and run out on a 1% agarose gel . DNA was transferred overnight to Hybond-N+ nylon membrane ( GE Healthcare , Chicago , Illinois , RPN303B ) by capillary action in transfer buffer ( 0 . 5 M NaOH , 0 . 6 M NaCl ) . The following day the membrane was immersed in neutralization buffer ( 1 M NaCl , 0 . 5 M Tris pH 7 . 4 ) , UV cross-linked , and pre-hybridized overnight at 42°C in pre-hybridization buffer ( 50% formamide , 5× SSCPE [20× SSCPE: 2 . 4 M NaCl , 0 . 3 M Na citrate , 0 . 2 M KH2PO4 , 0 . 02 M EDTA] , 5× Denhardt’s solution ( Invitrogen , Waltham , Massachusetts , 750018 ) , 0 . 5 mg/ml salmon sperm DNA [Agilent , Santa Clara , California , 201190] , 1% SDS ) . Templates for probe generation were prepared by Q5 amplification from MYC-WT targeting vector using primers GFP_F and GFP_R ( GFP template ) and from parental Ramos cell gDNA using primers 5’_F and 5’_R , followed by gel purification . Probes were prepared by random priming of corresponding PCR products ( 5’ and GFP templates ) in the presence of [αP32]CTP ( PerkinElmer BLU513H100UC ) . Unincorporated nucleotides were removed using a Sephadex G-50 column ( GE Healthcare 28-9034-08 ) . The membrane was incubated overnight at 42°C with probe in hybridization buffer ( 50% formamide , 5× SSCPE , 5× Denhardt’s solution [Invitrogen 750018] , 0 . 1 mg/ml salmon sperm DNA [Agilent 201190] , 1% SDS , 10% Dextran solution ) . Membrane was washed three times in 2× SSC/0 . 1% SDS ( 20× SSC: 3 M NaCl , 0 . 3 M Na Citrate , pH 7 . 0 ) and twice in 0 . 2× SSC/0 . 1% SDS , then exposed to a phosphor screen and developed using a phosphorimager ( GE Healthcare Typhoon ) . Chromatin immunoprecipitation ( ChIP ) was performed , with slight modification , as described by Thomas et al . , 2015 . Cells were first treated with 20 nM 4-OHT for 24 hr , then cross-linked in 1% methanol-free FA ( Thermo Fisher 28908 ) for 10 min and quenched using 0 . 125 mM glycine . The cells were then rinsed twice in ice-cold 1× PBS , and lysed in formaldehyde lysis buffer ( FALB: 50 mM HEPES pH 7 . 5 , 140 mM NaCl , 1 mM EDTA , and 1% Triton X-100 ) + 1% SDS + PIC ( Roche 05056489001 ) . Sonication was performed in a BioRuptor ( Diagenode , Denville , New Jersey ) for 25 min , 30 s on/30 s off , and debris removed by centrifugation . To enable ChIP efficiency to be determined by qPCR , a 1:50 ( 2% ) sample of chromatin was removed ( input ) prior to antibody addition . For anti-HA , anti-rabbit IgG or anti-HCF1N ChIP , antibody was added to chromatin prepared from 12 × 106 Ramos cells , and samples rotated overnight at 4°C . Protein A Agarose ( Roche 11134515001 ) , blocked with 10 μg BSA , was added to each sample and rotated at 4°C for 2–4 hr . Washes were performed with Low Salt Wash Buffer ( 20 mM Tris pH 8 . 0 , 150 mM NaCl , 2 mM EDTA , 1% Triton X-100 ) , High Salt Wash Buffer ( 20 mM Tris pH 8 . 0 , 500 mM NaCl , 2 mM EDTA , 1% Triton X-100 ) , Lithium Chloride Wash Buffer ( 10 mM Tris pH 8 . 0 , 250 mM LiCl , 1 mM EDTA , 1% Triton X-100 ) , and twice with TE ( 10 mM Tris pH 8 . 0 , 1 mM EDTA ) . Input and ChIP sample crosslinking were reversed at 65°C overnight in 50 μl TE + 200 mM NaCl + 0 . 1% SDS + 20–40 μg PK ( Macherey-Nagel 740506 ) . For sequencing , each biological replicate was generated by combining two to three independent ChIPs ( anti-IgG from CRE-ERT2 parental cells , anti-HCF1N MYC-WT , 4A , or VP16 HBM cells , or anti-HA from CRE-ERT2 parental cells , MYC-WT , 4A , or VP16 HBM cells ) —performed using the same antibody on the same chromatin . DNA was extracted with phenol:chloroform:isoamyl alcohol , followed by ethanol precipitation in the presence of glycogen ( Roche 10901393001 ) or GlycoBlue ( Invitrogen AM9516 ) . Libraries were prepared using NEBNext Ultra II DNA Library Prep Kit for Illumina ( NEB E7645S ) and NEBNext Multiplex Oligos for Illumina ( NEB Set 1 E7335 , Set 2 E7500S , NEB Unique Dual Index E6440S ) . Additional AMPure clean-ups at the start and the end of library preparation were included . Sequencing was carried out by Vanderbilt Technologies for Advanced Genomics using 75 bp paired-end sequencing on Illumina NextSeq 500 for anti-IgG and anti-HCF1N ChIPs or 150 bp paired-end sequencing on Illumina NovaSeq for anti-HA . The total number of sequencing reads for each replicate is shown in Supplementary file 2 . For qPCR , samples ( either input or ChIP ) were brought up to a final volume of 200 μl using TE . Each reaction was performed in a final volume of 15 μl , containing 2× SYBR FAST qPCR Master Mix ( Kapa , Wilmington , Massachusetts , KK4602 ) , 300 nM of each primer , and 2 μl of diluted sample . Three technical replicates were performed for each sample , and the mean Ct of these was used for calculating percent input . The mean Ct value for input was adjusted using the equation Ct ( input ) -log2 ( 50 ) . Percent input was then calculated using the equation 100 × 2^ ( adjustedCt − Ct ( ChIP ) ) . Three biological replicates of ChIP-qPCR were performed using primers to amplify across the genes EXOSC5 , UTP20 , EIF2S1 , POLR1A , EIF4G3 , and HBB ( β-Globin ) . In vitro binding assays pSUMO-MYC WT-FLAG containing 6XHis- and FLAG-tagged full-length MYC ( Thomas et al . , 2015 ) was used for Q5 site-directed mutagenesis to substitute in the 4A ( 4A_F and 4A_R ) and VP16 HBM ( VP16 HBM_F and VP16 HBM_R ) mutations . Rosetta cells Millipore 70954 were transformed with these plasmids , grown overnight , and induced the following day for 3 hr with 1 mM isopropyl β-d-1-thiogalactopyranoside ( IPTG ) at 30°C . Resulting cell pellets were resuspended in Buffer A ( 100 mM NaH2PO4 , 10 mM Tris , 6 M GuHCl , 10 mM imidazole ) + PIC ( Roche 05056489001 ) . Cells were sonicated 3× at 25% power for 10 s ( Cole-Parmer GE 130PB-1 ) , and debris removed by centrifugation; 150 μl bed volume Ni-NTA agarose ( QIAGEN , Venlo , Netherlands , 30210 ) was added to the samples and rotated for 2 hr at 4°C . Beads were sequentially washed 1× with Buffer A , 1× with Buffer A/TI ( 1:3 Buffer A:Buffer TI ) , 1× with Buffer TI ( 25 mM Tris-HCl pH 6 . 8 , 20 mM imidazole ) , and 1× with SUMO wash buffer ( 3 mM imidazole , 10% glycerol , 1× PBS , 2 mM DTT ) + PIC ( Roche 05056489001 ) . Recombinant MYC was sequentially eluted twice from the beads using SUMO elution buffer ( 250 mM imidazole , 10% glycerol , 1× PBS , 2 mM DTT ) + PIC ( Roche 05056489001 ) . The concentration of recombinant MYC was determined by resolving on a 10% SDS-PAGE gel alongside a BSA standard , and staining using Coomassie ( 50% methanol , 10% acetic acid , 0 . 1% w/v Coomassie Brilliant Blue ) . HCF-1VIC ( residues 1–380 ) from pCGT-HCF1VIC ( Thomas et al . , 2016 ) was cloned into pT7-IRES His-N ( Takara 3290 ) using BamHI-HF ( NEB R3136 ) and SalI-HF ( NEB R3138 ) , and Q5 site-directed mutagenesis was used to add an N-terminal T7 tag using the primers T7-HCF1_F and T7-HCF1_R . pT7-IRES His-T7-HCF-1VIC was in vitro transcribed/translated using the TnT Quick Coupled Transcription/Translation System ( Promega , Madison , Wisconsin , L1171 ) . Two milligrams of recombinant MYC and 12 μl of T7-HCF-1VIC were rotated overnight at 4°C in Kischkel buffer + PIC ( Roche 05056489001 ) . Anti-FLAG M2 Affinity Gel ( Sigma-Aldrich ) , blocked with 10 μg BSA , was added to each sample and rotated for 2 hr at 4°C . Beads were washed 4× in Kischkel buffer + 2 μg/ml Aprotinin ( VWR , Radnor , Pennsylvania , 97062–752 ) + 1 μg/ml Pepstatin ( VWR 97063–246 ) + 1 μg/ml Leupeptin ( VWR 89146–578 ) , and incubated in 1× Laemmli buffer for 5 min at 95°C . MYC:MAX dimers were purified and prepared as described by Farina et al . , 2004 . pRSET-6XHis-MYC WT , a gift from Ernest Martinez , was used as a template for Q5 site directed mutagenesis to substitute in the 4A ( 4A_F and 4A_R ) and VP16 HBM ( VP16 HBM_F and VP16 HBM_R ) mutations . The resulting plasmids , or pET-His-MAX , also from Ernest Martinez , were transformed into Rosetta cells ( Millipore 70954 ) , grown overnight , and induced the following day for 3 hr with 1 mM IPTG at 30°C . Resulting bacterial cell pellets were washed 1× with ice-cold wash buffer ( 10 mM Tris pH 7 . 9 , 100 mM NaCl , 1 mM EDTA ) , then resuspended in lysis buffer ( 20 mM HEPES pH 7 . 9 , 500 mM NaCl , 10% glycerol , 0 . 1% NP-40 , 10 mM BME , 1 mM PMSF ) , and sonicated . Centrifugation was used to separate out the insoluble ( pellet ) and soluble ( supernatant ) fractions . For MAX , the recombinant protein is present in the supernatant and was then purified using Ni-NTA agarose ( see below ) . For MYC samples , the supernatant was discarded , and the insoluble pellet was resuspended in E-buffer ( 50 mM HEPES pH 7 . 9 , 5% glycerol , 1% NP-40 , 10% Na-DOC , 0 . 5 mM BME ) , lysed using a Dounce homogenizer and B-pestle , and centrifuged . The pelleted inclusion bodies were lysed overnight in S-buffer ( 10 mM HEPES pH 7 . 9 , 6 M GuHCl , 5 mM BME ) by shaking at 25°C , and debris cleared by centrifugation . Both the MAX supernatant and MYC supernatant from lysed inclusion bodies were adjusted to 5 mM imidazole . MYC and MAX were then bound to 75 μl bed volume Ni-NTA agarose ( QIAGEN 30210 ) . Successive washes were performed for 5 min each at 4°C: 3× with S-buffer + 5 mM imidazole , 3× with BC500 ( 20 mM Tris pH 7 . 9 , 20% glycerol , 500 mM KCl , 0 . 05% NP-40 , 10 mM BME , 0 . 2 mM PMSF ) + 7 M urea + 5 mM imidazole , 1× with BC100 ( 20 mM Tris pH 7 . 9 , 20% glycerol , 100 mM KCl , 0 . 05% NP-40 , 10 mM BME , 0 . 2 mM PMSF ) + 7 M urea + 15 mM imidazole , 1× with BC100 + 7 M urea + 30 mM imidazole . Elution was performed using BC100 + 7 M urea + 300 mM imidazole . Concentration of recombinant MYC and MAX was determined by running samples out on a 12% acrylamide gel alongside a BSA standard , and staining using Coomassie stain ( 50% methanol , 10% acetic acid , 0 . 1% w/v Coomassie Brilliant Blue ) . For renaturation , 1 . 5 μg recombinant MAX was combined with 15 μg recombinant MYC ( 1:3 molar ratio ) , and brought up to a final volume of 150 μl using BC100 + 7 M urea . Dialysis was performed using a ‘Slide-A-Lyzer’ ( Thermo Fisher 66383 ) , with each dialysis step done for 2 hr stirring in the following solutions: BC500 + 0 . 1% NP-40 + 4 M urea at room temperature , BC500 + 0 . 1% NP-40 + 2 M urea at room temperature , BC500 + 0 . 1% NP-40 + 1 M urea at room temperature , BC500 + 0 . 1% NP-40 + 0 . 5 M urea at room temperature , BC500 at 4°C , and twice BC100 at 4°C . The product was centrifuged to remove debris , and BSA was added to a final concentration of 500 ng/μl . Double-stranded labeled E-box probe ( biotin group at the 3’ end ) and unlabeled competitors were prepared with dsDNA buffer ( 30 mM Tris pH 7 . 9 , 200 mM KCl ) , and incubated at 95°C for 5 min . The E-box sequence used was 5′-GCTCAGGGACCACGTGGTCGGGGATC-3′ and the mutant E-box sequence used was 5′-GCTCAGGGACCAGCTGGTCGGGGATC-3′ ( IDT ) . Double-stranded probe and the specific and non-specific competitors were prepared by combining 25 μM of each strand in dsDNA buffer ( 30 mM Tris pH 7 . 9 , 200 mM KCl ) . For the probe , the forward strand carried a 3’ biotin group; 20 fmol of labeled probe was bound to 0 . 55 pmol MYC:MAX or 0 . 06 pmol MAX:MAX dimers in the presence of 20 ng poly ( dI-dC ) ( Thermo Fisher 20148E ) in binding reaction buffer ( 15 mM Tris pH 7 . 9 , 15% glycerol , 100 mM KCl , 0 . 15 mM EDTA , 0 . 075% NP-40 , 7 . 5 mM BME , 375 ng/μl BSA ) for 30 min at room temperature . For reactions involving unlabeled specific or nonspecific competitor , these were included in the binding reaction at a 100-fold excess over the biotinylated probe . EMSA gel loading solution ( Thermo Fisher 20148K ) was added to each sample and these were loaded onto a pre-run 6% polyacrylamide gel in 0 . 5× TBE ( 45 mM Tris , 45 mM boric acid , 1 mM EDTA ) . The gel was transferred to Hybond-N+ nylon membrane ( GE Healthcare RPN303B ) in 0 . 5× TBE for 30 min at 100 V . The remainder of the protocol was performed using LightShift Chemiluminescent EMSA Kit ( Thermo Fisher 20148 ) according to manufacturer’s instructions . Cell pellets were resuspended in 1 ml TRIzol ( Invitrogen 15596026 ) , and RNA was prepared according to the manufacturer’s instructions . For switchable MYC allele Ramos cells , cells were treated with 20 nM 4-OHT for 24 hr and harvested . Prepared RNA was submitted to GENEWIZ ( South Plainfield , NJ ) for DNAse treatment , rRNA depletion , library preparation , and 150 bp paired-end sequencing on Illumina HiSeq . For untagged or FKBPFV-HCF-1N Ramos cells , cells were treated with DMSO or 500 nM dTAG-47 for 3 hr , and prepared RNA was DNAse treated prior to submission to Vanderbilt Technologies for Advanced Genomics for rRNA depletion , library preparation , and 150 bp paired-end sequencing on Illumina NovaSeq 6000 . The total number of sequencing reads for each replicate is shown in Supplementary file 2 . For validation of RNA-seq by reverse transcriptase qPCR ( RT-qPCR ) , RNA was prepared as above and 1 μg converted to cDNA using M-MLV reverse transcriptase ( Promega M1701 ) in the presence of random hexamers ( Invitrogen N8080127 ) , RNase inhibitor ( Thermo Fisher Scientific N8080119 ) , and dNTPs ( NEB N0446S ) . The resulting cDNA was brought up to a final volume of 160 μl using water . qPCR was performed in a final volume of 15 μl , containing 2× SYBR FAST qPCR Master Mix ( Kapa ) , 300 nM of each primer , and 2 μl of diluted sample . Three technical replicates were performed for each sample , and the mean Ct of these was used for calculating FC . The mean Ct value for the gene of interest ( GOI ) was normalized to GAPDH ( ΔCt ) using the equation Ct ( GOI ) -Ct ( GAPDH ) . For switchable MYC allele Ramos cells , ∆∆Ct was calculated between treated ( +4-OHT ) and untreated ( −4-OHT ) cells . For FKBPFV-HCF-1N Ramos cells , ΔΔCt was calculated between dTAG-47-treated and DMSO-treated cells . FC was then calculated using the equation 2^ ( −∆∆Ct ) . Three biological replicates of RT-qPCR were performed . Primer sequences used are listed in Supplementary file 1 . After adapter trimming by Cutadapt ( Martin , 2011 ) , RNA-Seq reads were aligned to the hg19 genome using STAR ( Dobin et al . , 2013 ) and quantified by featureCounts ( Liao et al . , 2014 ) . Differential analysis were performed by DESeq2 ( Love et al . , 2014 ) , which estimated the log2 FCs , Wald test p-values , and adjusted p-value ( false discovery rate , FDR ) by the Benjamini–Hochberg procedure . The significantly changed genes were chosen with the criteria FDR < 0 . 05 . ChIP-Seq reads were aligned to the hg19 genome using Bowtie2 ( Langmead et al . , 2009 ) after adapter trimming . Peaks were called by MACS2 ( Feng et al . , 2012 ) with a q-value of 0 . 01 . ChIP read counts were calculated using DiffBind ( Stark and Brown , 2011 ) and differential peaks were determined by DESeq2 ( Love et al . , 2014 ) . Peaks were annotated using Homer command annotatePeaks , and enriched motifs were identified by Homer command findMotifsGenome ( http://homer . ucsd . edu/homer/ ) . All genomics data were deposited at GEO with the accession number GSE152385 . Reviewers may access these data with the token ‘enanoauyrpmfrmb’ . Details for the referenced MYC ChIP-Seq experiments from Thomas et al . , 2019 are available in Tansey et al . , 2019 ( samples: GSM3593604–GSM3593606 and GSM3593616–3593618 ) . To generate a growth curve and determine doubling time of the MYC mutants , switchable MYC cells were treated with 20 nM 4-OHT for 16 hr , and then grown for a further 24 hr . Cells were then plated at a density of 20 , 000 cells/ml , and counted 3 and 6 days later . Growth rate ( GR ) was determined through the equation GR = ln ( N ( 7 ) − N ( 1 ) ) /144 , where N ( 7 ) is the number of cells per milliliter on day 7 , N ( 1 ) is the number of cells per milliliter on day 1 , and 144 is the number of hours that elapsed between the two measurements . Doubling time ( DT ) was then determined by the equation DT = ln ( 2 ) /GR . To measure the impact of HCF-1N degradation on cell growth , FKBPFV-HCF-1N Ramos cells were plated at a density of 20 , 000 cells/ml with either DMSO or 500 nM dTAG-47 . Cells were then counted every 24 hr for the following 4 days , without replacement of the compound or changing of the media . To measure the impact of altering the MYC−HCF-1 interaction on cell growth , switchable MYC allele Ramos cell lines were treated with 20 nM 4-OHT for 2 hr to create a 50/50 mix of GFP-negative ( WT ) to GFP-positive ( mutant ) cells . Cells were sampled 24 hr later , and every 3 days following . Sampled cells were fixed in 1% FA in PBS for 10 min at room temperature , and the proportion of GFP-positive cells was determined using a BD LSR II flow cytometer at the Vanderbilt Flow Cytometry Shared Resource . To account for variation in the proportion of GFP-positive cells , each replicate was normalized to the proportion of GFP-positive cells at 24 hr post-treatment with 4-OHT . To measure the impact of altering the MYC−HCF-1 interaction on glutamine dependence , switchable MYC allele Ramos cell lines were treated with 20 nM 4-OHT for 2 hr , and allowed to recover for 3 days . Cells were then split into RPMI 1640 without L-glutamine ( Corning 15–040-CV ) , supplemented with 10% dialyzed FBS ( Gemini Bio , West Sacramento , California , 100–108 ) , and 1% P/S ( Gibco 15140122 ) , and grown for 16 hr with or without Supplementary file 2 mM glutamine ( Gibco 25030081 ) . Glutamine was added back to the cells that were deprived , grown for 3 days , and fixed in 1% FA in PBS for 10 min at room temperature . The proportion of GFP-positive cells was determined using a BD LSR II flow cytometer , and normalized to the proportion of GFP-positive cells prior to being grown with or without supplemental glutamine . For the 4A and VP16 HBM cells , the proportion of GFP-positive cells was normalized to that in WT for glutamine supplementation ( Gln+ ) or deprivation ( Gln− ) . Classification of annotated metabolites was extracted from HMDB ( Wishart et al . , 2018 ) and LIPID MAPS ( Fahy et al . , 2009 ) . Metabolite pathway analyses were performed using MetaboAnalyst 4 . 0 ( Chong et al . , 2019 ) , and GO analyses using Metascape ( Zhou et al . , 2019 ) or DAVID ( Huang et al . , 2009a; Huang et al . , 2009b ) . Chord diagrams were created using Circos ( Krzywinski et al . , 2009 ) , pathways using Cytoscape ( Shannon et al . , 2003 ) , and bubble plots using ggplot2 ( Wickham , 2016 ) . Unless otherwise stated , all experiments were conducted with at least three independent , biological replicates , and statistical tests were carried out using PRISM 8 ( GraphPad ) . Materials and resources mentioned here are available upon request . | Tumours form when cells lose control of their growth . Usually , cells produce signals that control how much and how often they divide . But if these signals become faulty , cells may grow too quickly or multiply too often . For example , a group of proteins known as MYC proteins activate growth genes in a cell , but too much of these proteins causes cells to grow uncontrollably . With one third of all cancer deaths linked to excess MYC proteins , these molecules could be key targets for anti-cancer drugs . However , current treatments fail to target these proteins . One option for treating cancers linked to MYC proteins could be to target proteins that work alongside MYC proteins , such as the protein HCF-1 , which can attach to MYC proteins . To test if HCF-1 could be a potential drug target , Popay et al . first studied how HCF-1 and MYC proteins interacted using specific cancer cells grown in the laboratory . This revealed that when the two proteins connected , they activated genes that trigger rapid cell growth . When these cancer cells were then injected into mice , tumours quickly grew . However , when the MYC and HCF-1 attachments in the cancer cells were disrupted , the tumours shrunk . This suggests that if anti-cancer drugs were able to target HCF-1 proteins , they could potentially reduce or even reverse the growth of tumours . While further research is needed to identify drug candidates , these findings reveal a promising target for treating tumours that stem from over-abundant MYC proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2021 | MYC regulates ribosome biogenesis and mitochondrial gene expression programs through its interaction with host cell factor–1 |
Tumors frequently exhibit aberrant glycosylation , which can impact cancer progression and therapeutic responses . The hexosamine biosynthesis pathway ( HBP ) produces uridine diphosphate N-acetylglucosamine ( UDP-GlcNAc ) , a major substrate for glycosylation in the cell . Prior studies have identified the HBP as a promising therapeutic target in pancreatic ductal adenocarcinoma ( PDA ) . The HBP requires both glucose and glutamine for its initiation . The PDA tumor microenvironment is nutrient poor , however , prompting us to investigate how nutrient limitation impacts hexosamine synthesis . Here , we identify that glutamine limitation in PDA cells suppresses de novo hexosamine synthesis but results in increased free GlcNAc abundance . GlcNAc salvage via N-acetylglucosamine kinase ( NAGK ) is engaged to feed UDP-GlcNAc pools . NAGK expression is elevated in human PDA , and NAGK deletion from PDA cells impairs tumor growth in mice . Together , these data identify an important role for NAGK-dependent hexosamine salvage in supporting PDA tumor growth .
Altered glycosylation is frequently observed in malignancies , impacting tumor growth as well as immune and therapeutic responses ( Akella et al . , 2019; Mereiter et al . , 2019; Munkley , 2019 ) . Several types of glycosylation , including O-GlcNAcylation and N-linked glycosylation , are dependent on the glycosyl donor uridine diphosphate N-acetylglucosamine ( UDP-GlcNAc ) , which is synthesized by the hexosamine biosynthesis pathway ( HBP ) . The HBP branches off from glycolysis with the transfer of glutamine’s amido group to fructose-6-phosphate ( F-6-P ) to generate glucosamine-6-phosphate ( GlcN-6-P ) , mediated by the rate limiting enzyme glutamine—fructose-6-phosphate transaminase ( GFPT1/2 ) . The pathway further requires acetyl-CoA , ATP , and uridine triphosphate ( UTP ) to ultimately generate UDP-GlcNAc . O-GlcNAcylation , the addition of a single N-acetylglucosamine ( GlcNAc ) moiety onto a serine or threonine residue of intracellular proteins , is upregulated in multiple cancers ( Akella et al . , 2019 ) . Targeting O-GlcNAcylation suppresses the growth of breast , prostate , and colon cancer tumors ( Caldwell et al . , 2010; Ferrer et al . , 2017; Gu et al . , 2010; Guo et al . , 2017; Lynch et al . , 2012 ) . Similarly , highly branched N-glycan structures are sensitive to HBP flux and are upregulated in malignant tissue ( Lau et al . , 2007 ) , and targeting the relevant Golgi GlcNAc transferase enzymes can limit tumor growth and metastasis in vivo ( Granovsky et al . , 2000; Li et al . , 2008; Zhou et al . , 2011 ) . Thus , improved understanding of the regulation of the HBP in cancer could point towards novel therapeutic strategies . Pancreatic ductal adenocarcinoma ( PDA ) is a deadly disease with a 5 year survival rate of 9% and a rising number of annual deaths ( Rahib et al . , 2014 ) ( ACS Cancer Facts and Figures 2019 , NIH SEER report 2019 ) . Mutations in KRAS occur in nearly all cases of human PDA and drive extensive metabolic reprogramming in cancer cells . Enhanced flux into the HBP was identified as a primary metabolic feature mediated by mutant KRAS in PDA cells ( Ying et al . , 2012 ) . Hypoxia , a salient characteristic of the tumor microenvironment ( Lyssiotis and Kimmelman , 2017 ) , was shown to further promote expression of glycolysis and HBP genes in pancreatic cancer cells ( Guillaumond et al . , 2013 ) . Notably , the glutamine analog 6-diazo-5-oxo-l-norleucine ( DON ) , which inhibited the HBP , suppressed PDA metastasis , and sensitized PDA tumors to anti-PD1 therapy ( Sharma et al . , 2020 ) . DON has also been reported to sensitize PDA cells to the chemotherapeutic gemcitabine in vitro ( Chen et al . , 2017 ) . Additionally , a recently developed inhibitor targeting the HBP enzyme phosphoacetylglucosamine mutase 3 ( PGM3 ) enhances gemcitabine-mediated reduction of xenograft tumor growth in vivo ( Ricciardiello et al . , 2020 ) . Thus , the HBP may represent a therapeutic target in PDA , although the regulation of UDP-GlcNAc synthesis and the optimal strategies to target this pathway for therapeutic benefit in PDA remain poorly understood . An outstanding question is the impact of the tumor microenvironment on UDP-GlcNAc synthesis . The HBP has been proposed as a nutrient-sensing pathway since its rate-limiting step , mediated by GFPT1/2 , requires both glutamine and the glycolytic intermediate fructose-6-phosphate ( Denzel and Antebi , 2015 ) . In hematopoietic cells , glucose deprivation limits UDP-GlcNAc levels and dramatically reduces levels of the N-glycoprotein IL3Rα at the plasma membrane in a manner dependent on the HBP ( Wellen et al . , 2010 ) . Similarly , O-GlcNAcylation of certain nuclear-cytosolic proteins , including cancer-relevant proteins such as Myc and Snail , has been demonstrated to be nutrient sensitive , impacting protein stability or function ( Housley et al . , 2008; Park et al . , 2010; Swamy et al . , 2016 ) . Yet , the PDA tumor microenvironment is thought to be particularly nutrient poor , owing to its characteristic dense stroma ( Halbrook and Lyssiotis , 2017 ) . This raises the question of how nutrient deprivation impacts the synthesis of UDP-GlcNAc and its utilization for glycosylation . Understanding how PDA cells regulate these processes under nutrient limitation could identify therapeutic vulnerabilities . In this study , we investigated the impact of nutrient deprivation on the HBP and glycosylation in PDA cells , identifying a key role for hexosamine salvage through the enzyme N-acetylglucosamine kinase ( NAGK ) in PDA tumor growth .
To examine the effects of nutrient deprivation on glycosylation , we cultured cells under glucose or glutamine limitation and examined O-GlcNAc levels and cell surface phytohemagglutinin-L ( L-PHA ) binding , a readout of N-acetylglucosaminyltransferase 5 ( MGAT5 ) -mediated cell surface N-glycans ( Figure 1—figure supplement 1A , B ) , which are highly sensitive to UDP-GlcNAc availability ( Lau et al . , 2007 ) . We focused on glucose and glutamine because of their requirement to initiate the HBP ( Figure 1A ) . First , as a positive control , we examined HCT-116 and SW480 colon cancer cells , previously documented to have glucose-responsive O-GlcNAcylation ( Park et al . , 2010; Steenackers et al . , 2016 ) , which we also confirmed in HCT-116 cells ( Figure 1B ) . Indeed , L-PHA binding was suppressed by glucose restriction in SW480 cells and by glutamine restriction in both colon cancer cell lines ( Figure 1C ) . Next , to test whether glycans were sensitive to nutrient restriction in PDA cells , we examined L-PHA binding and O-GlcNAc levels under nutrient deprivation conditions in a panel of human PDA cell lines , including PANC-1 , MIA PaCa-2 , AsPC-1 , and HPAC . Across these cell lines , no consistent changes in L-PHA binding were observed under glucose or glutamine limitation ( Figure 1D , E , Figure 1—figure supplement 1C ) . We also examined L-PHA binding in PDA cells under oxygen- or serum-deprived conditions and observed minimal changes ( Figure 1—figure supplement 1D , E ) . O-GlcNAcylation was minimally altered by culture in low glutamine and exhibited variable changes in response to glucose limitation ( Figure 1F ) , consistent with stress-induced regulation of this modification ( Taylor et al . , 2008 ) . Since glycosylation may be maintained through either sustained ability to add the modifications or through changes in turnover , we assayed active O-GlcNAcylation by inhibiting O-GlcNAcase with Thiamet G ( TMG ) . TMG treatment resulted in equivalently elevated O-GlcNAcylation levels in high and low glutamine conditions ( Figure 1G; Figure 1—figure supplement 1F ) , indicating that glutamine restriction does not limit the capacity of cells to add the O-GlcNAc modification . Mia-PaCa-2 cells exhibited some cell death in low glutamine , though this was not exacerbated by TMG treatment ( Figure 1—figure supplement 1F ) . Thus , under a variety of nutrient stress conditions , neither L-PHA binding nor O-GlcNAcylation were consistently suppressed in pancreatic cancer cell lines . Glutamine restriction in particular had remarkably little impact on O-GlcNAcylation and L-PHA binding , raising the question of how UDP-GlcNAc is generated during nutrient limitation . We therefore next asked whether the abundance of HBP metabolites is impacted by nutrient limitation . We measured HBP metabolites after glucose or glutamine restriction using HPLC-MS ( Guo et al . , 2016b ) . In low glutamine conditions , GlcN-6-P levels were potently decreased relative to glutamine-replete conditions in PANC-1 cells , while UDP-GlcNAc abundance was maintained ( Figure 2A ) . In MIA PaCa-2 cells , UDP-GlcNAc abundance actually increased upon glutamine restriction ( Figure 2—figure supplement 1A ) . These data indicate that UDP-GlcNAc might be generated through mechanisms other than de novo synthesis . Glycolytic intermediates were minimally impacted by low glutamine conditions , and TCA cycle intermediates such as α-KG and malate decreased as expected ( Figure 2A , Figure 2—figure supplement 1A ) . In contrast to that in glutamine restriction , UDP-GlcNAc abundance declined in 5 mM or 0 . 1 mM relative to 10 mM glucose conditions ( Figure 2—figure supplement 1B ) , suggesting that glutamine limitation specifically may trigger an adaptive response to sustain UDP-GlcNAc pools . We sought to understand how UDP-GlcNAc pools are sustained during glutamine restriction . In addition to de novo synthesis of UDP-GlcNAc via the HBP , free GlcNAc in the cell can also be phosphorylated via N-acetylglucosamine kinase ( NAGK ) to produce GlcNAc-P and then regenerate UDP-GlcNAc ( Figure 2B ) . However , NAGK’s roles in physiology and cancer biology have been minimally studied . To investigate the possibility that UDP-GlcNAc is generated through mechanisms other than its synthesis from glucose , we first designed a stable isotope labeling strategy to quantify the fraction of the glucosamine ring that is synthesized de novo in glutamine-replete versus -restricted conditions . Since multiple components of UDP-GlcNAc [glucosamine ring , acetyl group , uridine ( both the uracil nucleobase and the ribose ring ) ] can be synthesized from glucose , UDP-GlcNAc isotopologs up to M+16 can be generated from glucose ( Moseley et al . , 2011; Figure 2C ) . In order to measure the glucose carbon incorporated into GlcNAc-P and UDP-GlcNAc via the HBP , all isotopologs containing a fully labeled glucosamine ring are added together ( % labeled GlcN indicates sum of M+6 , M+8 , M+11 , and M+13 for UDP-GlcNAc and sum of M+6 and M+8 for GlcNAc-P ) ( Figure 2C ) . After 48 hr of glutamine restriction , cells were incubated with fresh low glutamine medium containing [U-13C]-glucose to track the incorporation of glucose carbons into hexosamine intermediates . Across multiple PDA cell lines , the fractional labeling of the glucosamine ring in both GlcNAc-P and UDP-GlcNAc pools was markedly suppressed by glutamine restriction , indicating decreased de novo synthesis in low glutamine conditions ( Figure 2D , Figure 2—figure supplement 1C–E ) . Notably , labeling into the ribose component of UDP-GlcNAc was also suppressed ( % labeled ribose indicates sum of isotopologs containing M+5 [i . e . , M+5 , M+7 , M+11 , and M+13]; Figure 2D , Figure 2—figure supplement 1C ) . Consistently , incorporation of 13C glucose into UTP was suppressed upon glutamine restriction ( Figure 2—figure supplement 2A ) , even though UTP levels were maintained or increased ( Figure 2A , Figure 2—figure supplement 1A ) , suggesting a role for nucleoside salvage in maintaining nucleotide pool in these conditions . This is consistent with previous reports demonstrating that autophagy/ribophagy is a source of nucleosides in amino acid-deprived conditions ( Guo et al . , 2016a; Wyant et al . , 2018 ) . Indeed , silencing of either of the uridine salvage enzymes uridine kinase 1 or 2 ( UCK1/2 ) resulted in decreased UDP , UTP , and UDP-GlcNAc levels ( Figure 2—figure supplement 2B , C ) , indicating that nucleoside salvage contributes to maintaining uridine phosphate and UDP-GlcNAc pools . Thus , glutamine restriction suppresses the de novo synthesis of both GlcNAc-P and UTP , both of which are required to produce UDP-GlcNAc . We noted that GlcNAc-P and UDP-GlcNAc pools labeled from glucose with similar but not identical kinetics . While this is potentially due to limitations in detection since GlcNAc-P is much less abundant than UDP-GlcNAc , we considered whether GlcNAc-P-independent pathways may also have minor contributions to glucose-dependent UDP-GlcNAc labeling . Although pathways through which glucose can feed into UDP-GlcNAc’s glucosamine ring independent of the HBP have not been described in mammalian cells to our knowledge , we nevertheless tested the two major metabolic branch points diverging from UDP-GlcNAc , which mediate UDP-GalNAc and sialic acid synthesis . UDP-galactose-4-epimerase ( GALE ) interconverts UDP-GlcNAc and UDP-GalNAc , and UDP-GlcNAc-2-epimerase/ManAc kinase ( GNE ) initiates sialic acid biosynthesis . Silencing of neither GALE nor GNE reduced UDP-GlcNAc labeling from glucose , however , indicating that these enzymes are unlikely to facilitate a bypass pathway ( Figure 2—figure supplement 2D , E ) . Although a minor contribution from another unknown pathway cannot be ruled out , the slight apparent differences in timing of GlcNAc-P and UDP-GlcNAc labeling most likely reflect technical limitations . Regardless , the data clearly indicate that UDP-GlcNAc abundance is maintained despite reduced de novo hexosamine synthesis from glucose ( Figure 2D ) . As mentioned , UDP-GlcNAc can be generated via phosphorylation of free GlcNAc by NAGK-generating GlcNAc-6-P ( Figure 2B ) . When supplemented , GlcNAc is salvaged into the UDP-GlcNAc pool ( Ryczko et al . , 2016; Wellen et al . , 2010 ) . Endogenous sources of GlcNAc may include removal of O-GlcNAc protein modifications or breakdown of glycoconjugates and extracellular matrix components . Notably , intracellular levels of GlcNAc increase upon glutamine restriction ( Figure 3A ) . Yet , the significance of GlcNAc salvage to maintenance of UDP-GlcNAc pools has been little studied , and the proportion of UDP-GlcNAc generated via the NAGK-dependent salvage pathway is unknown . NAGK mRNA expression increased in PDA cell lines in low glutamine conditions and in some cell lines also in low glucose ( Figure 2—figure supplement 3A , B ) . GFPT1 expression was also induced in both low glucose conditions , consistent with a prior report ( Moloughney et al . , 2016 ) , and in low glutamine conditions ( Figure 2—figure supplement 3A ) , even though de novo synthesis is suppressed when glutamine is limited . Protein levels of NAGK did not increase in concordance with mRNA at these time points , however , although a mobility shift potentially indicative of post-translational modification was apparent when protein lysates were run on a gel using a large electrophoresis system ( see Materials and methods; Figure 2—figure supplement 3C , D ) . Removal of the phosphatase inhibitor Na3VO4 from the sample buffer prevented the mobility shift , suggesting that NAGK may be phosphorylated on one or more residues in low glutamine conditions ( Figure 2—figure supplement 3D ) . Taken together , these data indicate that under low glutamine conditions , GlcNAc availability for salvage increases and the salvage enzyme NAGK is subject to regulation . These findings prompted us to investigate the role of NAGK in UDP-GlcNAc synthesis in PDA cells . We functionally examined the role of NAGK in PDA cell lines by using CRISPR-Cas9 gene editing to generate NAGK knockout ( KO ) PANC-1 and MiaPaCa-2 clonal cell lines ( Figure 3—figure supplement 1A , B ) . N-[1 , 2-13C2]acetyl-d-glucosamine ( 13C GlcNAc ) was efficiently salvaged in control cells , and this was suppressed by NAGK deletion , as evidenced by reduced fractional labeling of GlcNAc-P and UDP-GlcNAc ( Figure 3B ) . Since we did not observe any residual protein expression , we hypothesized that the N-acetylgalactosamine ( GalNAc ) salvage enzyme GalNAc kinase ( GALK2 ) might be responsible for the remaining GlcNAc salvage in the absence of NAGK . Indeed , silencing of GALK2 further suppressed incorporation of 13C GlcNAc into GlcNAc-P and UDP-GlcNAc in the NAGK KO cells ( Figure 3—figure supplement 1C ) . We hypothesized that knockout cells would conversely conduct increased de novo UDP-GlcNAc synthesis . To test this , we incubated cells with [U-13C]-glucose and examined incorporation into GlcNAc-P and UDP-GlcNAc . Indeed , in the absence of NAGK , we observed increased glucose-dependent fractional labeling of the glucosamine ring of UDP-GlcNAc and GlcNAc-P , but not the ribose component of UDP-GlcNAc ( Figure 3C , D , Figure 3—figure supplement 1D , E ) . This effect was also observed with knockdown of NAGK by shRNA , though to a lesser extent ( Figure 3—figure supplement 2A , B ) . Incorporation of glucose into F-6-P did not change ( Figure 3C ) , and the proportion of UDP-GlcNAc containing an M+5 ribose ring was also unchanged in knockout cells ( Figure 3D ) , as expected . Thus , when NAGK is deleted and GlcNAc salvage is suppressed , de novo hexosamine synthesis increases . We next assessed changes in the levels of hexosamine intermediates in control and NAGK KO cell lines . In PANC-1 KO cells in 4 mM glutamine , GlcN-P increased significantly , consistent with increased de novo synthesis in the absence of NAGK ( Figure 3E ) . GlcNAc-P was modestly reduced in KO cells , though UDP-GlcNAc levels were maintained ( Figure 3E ) . In MIA PaCa-2 cells , GlcNAc-P was markedly suppressed in the absence of NAGK , though UDP-GlcNAc was not ( Figure 3—figure supplement 2C ) . We also measured HBP metabolites under glutamine restriction , where we expected NAGK would play a more significant role in UDP-GlcNAc generation . We were only able to measure metabolites accurately in PANC-1 KO cells because MIA PaCa-2 NAGK KO cells began to die quickly in low glutamine , which will be discussed further in the next section . GlcN-P decreased in control and KO cells , consistent with reduced de novo hexosamine synthesis ( Figure 3E ) . GlcNAc-P levels decreased in low glutamine in control cells and decreased further in cells lacking NAGK , consistent with contributions from both de novo synthesis and salvage ( Figure 3E ) . Reciprocally , GlcNAc abundance was elevated upon glutamine limitation in both control and NAGK KO cells ( Figure 3E ) . UDP-GlcNAc abundance was modestly reduced in NAGK KO cells relative to controls under glutamine restriction , though levels were still comparable to that in high glutamine ( Figure 3E ) , possibly reflecting changes in utilization . GALK2 silencing did not further suppress UDP-GlcNAc in NAGK KO cells ( Figure 3—figure supplement 2D ) , suggesting that GALK2 may not have a major role in physiological GlcNAc salvage . Cumulatively , the data demonstrate that GlcNAc is salvaged into UDP-GlcNAc pools in PDA cells in a manner dependent at least in part on NAGK . To test the role of NAGK in cell proliferation , we first monitored growth of NAGK KO cells compared to controls in 2D and 3D culture in 4 mM glutamine , finding minimal differences ( Figure 4A , Figure 4—figure supplement 1 ) . We hypothesized that NAGK KO cell proliferation would be impaired in 0 . 05 mM glutamine , where de novo UDP-GlcNAc synthesis is suppressed . Indeed , MIA PaCa-2 KO cells died more quickly in 0 . 05 mM glutamine than did control cells ( Figure 4A ) . PANC-1 KO cells did not show this effect ( Figure 4A ) , but we hypothesized that NAGK loss might have a stronger effect in vivo where tumor growth can be constrained by nutrient availability . To gain initial insight into whether NAGK is likely to play a functional role in PDA progression in vivo , we queried publicly available datasets . From analysis of publicly available microarray data ( Pei et al . , 2009 ) and gene expression data from the Cancer Genome Atlas ( TCGA ) , we indeed found NAGK expression to be increased in tumor tissue relative to adjacent normal regions of the pancreas or to pancreas GTEx data ( Figure 4B , Figure 4—figure supplement 1B ) . GFPT1 expression was also increased in tumor tissue ( Figure 4B , Figure 4—figure supplement 1B ) , consistent with its regulation by mutant KRAS ( Ying et al . , 2012 ) . Two other HBP genes , PGM3 and UAP1 , did not show significantly increased expression in PDA tumors in these datasets ( Figure 4B , Figure 4—figure supplement 1B ) . We then studied the role of NAGK in tumor growth in vivo by injecting NAGK CRISPR KO cells into the flank of NCr nude mice . Final tumor volume and weight were markedly reduced in the absence of NAGK ( Figure 4C–D ) . Of note , initial tumor growth was comparable between control and KO cells , but the NAGK knockout tumors either stopped growing or shrank while control tumors continued to grow larger ( Figure 4—figure supplement 1C ) . Interestingly , KO tumor samples showed increased L-PHA signal ( Figure 4—figure supplement 1D ) , indicating that NAGK deficiency results in altered glycosylation within tumors . This could possibly reflect either elevated de novo synthesis in the small tumors that form or differences in cellular composition . For example , activated fibroblast marker α-smooth muscle actin ( α-SMA ) was more abundant in the NAGK KO tumors ( Figure 4—figure supplement 1D ) . Residual NAGK signal in whole tumors also presumably reflects expression in other cell types ( Figure 4—figure supplement 1D ) , since NAGK was undetectable in the clonal cell lines used for injections ( Figure 3—figure supplement 1B ) . Taken together , these data are consistent with the notion that NAGK is dispensable when nutrients are abundant but becomes more important as the tumors outgrow their original nutrient supply and become more dependent on scavenging and recycling and indicate that NAGK-mediated hexosamine salvage supports tumor growth in vivo .
In this study , we identify a key role for NAGK in salvaging GlcNAc for UDP-GlcNAc synthesis in PDA cells . We show that glutamine deprivation suppresses de novo hexosamine biosynthesis , which is reciprocally increased upon NAGK deletion . Glutamine deprivation also results in increased availability of GlcNAc for salvage . NAGK expression is elevated in human PDA tumors , and NAGK deficiency suppresses GlcNAc salvage in cells and tumor growth in mice . This work raises several key questions for future investigation . First , the sources of GlcNAc salvaged by NAGK remain to be fully elucidated . GlcNAc may be derived from recycling of GlcNAc following O-GlcNAc removal or breakdown of glycoconjugates . Additionally , GlcNAc may be recovered from the environment . Nutrient scavenging via macropinocytosis is a key feature of PDA ( Commisso et al . , 2013; Kamphorst et al . , 2015 ) . Macropinocytosis has mostly been associated with scavenging of protein to recover amino acids , but lysosomal break down of glycoproteins may also release sugars including GlcNAc . Additionally , ECM components , including hyaluronic acid ( HA ) , which is a polymer of GlcNAc and glucuronic acid disaccharide units , may be additional sources of GlcNAc for salvage in the tumor microenvironment . Indeed , in a manuscript co-submitted with this one , Kim et al . , 2020 identify HA as a major source of scavenged GlcNAc . Our manuscript and the Kimet al . ’s manuscript together indicate that NAGK may take on a heightened importance in the context of high GlcNAc availability and nutrient deprivation , a situation that is likely to occur within the tumor microenvironment . Furthermore , the key fates of UDP-GlcNAc that support tumor growth remain to be elucidated . Sufficient UDP-GlcNAc is required for protein glycosylation to maintain homeostasis and prevent ER stress , particularly in a rapidly dividing cell . Additionally , a wide range of cancers exhibit elevated O-GlcNAc , which could contribute to driving pro-tumorigenic transcriptional and signaling programs . In PDA specifically , the glycan CA19-9 is currently used as a biomarker for disease progression and recent studies point to a functional role for CA19-9 in tumorigenesis ( Engle et al . , 2019 ) . UDP-GlcNAc is also required for HA synthesis , which is present in low amounts in normal pancreas but increases in PanIN lesions and PDA ( Provenzano et al . , 2012 ) . PDA cells are capable of producing HA in vitro ( Mahlbacher et al . , 1992 ) . Depletion of fibroblasts in an autochthonous PDA mouse model results in a decrease in collagen I , but not HA in the tumor microenvironment , indicating that HA must be generated by another cell type , possibly the tumor cells themselves ( Özdemir et al . , 2014 ) . Previous studies demonstrated that treatment of PDA with exogenous hyaluronidase can increase vascularization and improve drug delivery to the tumor ( Jacobetz et al . , 2013; Provenzano et al . , 2012 ) , although a phase III clinical trial reported no improvement in overall patient survival when combining pegylated hyaluronidase with nab-paclitaxel plus gemcitabine ( Van Cutsem et al . , 2020 ) . Recently , it was shown that inhibiting the HBP by treatment with DON depletes HA and collagen in an orthotopic mouse model . DON treatment also increased CD8 T-cell infiltration into the tumor , sensitizing the tumor to anti-PD1 therapy ( Sharma et al . , 2020 ) . Thus , targeting the HBP holds promise for improving the efficacy of other therapeutics . The findings of the current study suggest that in addition to de novo hexosamine synthesis , targeting of hexosamine salvage warrants further investigation in terms of potential for therapeutic intervention . Of note , an inhibitor targeting PGM3 , which converts GlcNAc-6-P to GlcNAc-1-P and is thus required for both de novo UDP-GlcNAc synthesis and GlcNAc recycling , showed efficacy in treating gemcitabine-resistant patient-derived xenograft PDA models ( Ricciardiello et al . , 2020 ) , as well as in breast cancer xenografts ( Ricciardiello et al . , 2018 ) . Finally , almost nothing is currently known about the role of NAGK and GlcNAc salvage in normal physiology . Even in non-cancerous IL-3-dependent hematopoietic cells , a substantial proportion of the UDP-GlcNAc pool remains unlabeled from 13C-glucose ( Wellen et al . , 2010 ) , suggesting that salvage may contribute to UDP-GlcNAc pools in a variety of cell types . However , while GFPT1 is required for embryonic development in mice , NAGK knockout mouse embryos are viable ( Dickinson et al . , 2016 ) . NAGK deficiency has not yet been characterized in postnatal or adult mice . Perhaps GlcNAc salvage is dispensable when nutrients are available and cells are not dividing , as in most healthy tissues . However , in a tumor , in which cells are proliferating and nutrients are spread thin , NAGK and GlcNAc salvage may become more important in feeding UDP-GlcNAc pools . Related questions include elucidating the mechanisms regulating NAGK gene expression and putative post-translational modification , as well as understanding the role of GALK2 in hexosamine salvage . In sum , we report a key role for NAGK in feeding UDP-GlcNAc pools in PDA cells and in supporting xenograft tumor growth . Further investigation will be needed to elucidate the physiological functions of NAGK , as well as the mechanisms through which it supports tumor growth and its potential role in modulating therapeutic responses .
Cells were cultured in DMEM high glucose ( Gibco , 11965084 ) with 10% calf serum ( Gemini GemCell U . S . Origin Super Calf Serum , 100–510 ) , unless otherwise noted . Glucose- or glutamine-restricted media was prepared using glucose , glutamine , and phenol red-free DMEM ( Gibco , A1443001 ) supplemented with glucose ( Sigma-Aldrich , G8769 ) , glutamine ( Gibco , 25030081 ) , and dialyzed fetal bovine serum ( Gemini , 100–108 ) . For all glutamine restriction experiments except S2 . 3 D , cells were plated 2–3× more densely for the nutrient-restricted condition samples to achieve similar confluency at the experiment endpoint . One percent oxygen levels were achieved by culturing cells in a Whitley H35 Hypoxystation ( Don Whitley Scientific ) . ATCC names and numbers for the cell lines used in this study are as follows: MIA PaCa-2 ( ATCC# CRL-1420 ) , PANC-1 ( ATCC# CRL-1469 ) , HPAC ( ATCC# CRL-2119 ) , AsPC-1 ( ATCC# CRL-1682 ) , BxPC-3 ( ATCC# CRL-1687 ) , HCT 116 ( ATCC# CCL-247 ) , and SW480 ( ATCC# CCL-228 ) . All cells were routinely tested for mycoplasma and authenticated by short tandem repeat ( STR ) profiling using the GenePrint 10 System ( Promega , B9510 ) . Generation of CRISPR cell lines sgRNA sequences targeting NAGK or Mgat5 from the Brunello and Brie libraries ( Doench et al . , 2016 ) was cloned into the lentiCRISPRv2 vector ( Sanjana et al . , 2014 ) . Lentivirus was produced in 293 T cells according to standard protocol . Cells were then infected with the CRISPR lentivirus and selected with puromycin . Cells were plated at very low density into 96-well plates to establish colonies generated from single-cell clones . Mgat5 gene disruption was validated by qPCR and L-PHA binding . NAGK gene disruption was validated by qPCR , western blot , and 13C-GlcNAc tracing . Seven NAGK knockout clonal cell lines established from three different sgRNAs , four in PANC-1 cells and three in MIA PaCa-2 cells , were chosen for use in the study . Please see table at end of methods for primer sequences of guides used . For protein extraction from cells , cells were kept on ice and washed three times with PBS , then scraped into PBS and spun down at 200 g for 5 min . The cell pellet was resuspended in 50–100 µL RIPA buffer ( 1% NP-40 , 0 . 5% deoxycholate , 0 . 1% SDS , 150 mM NaCl , 50 mM Tris plus protease inhibitor cocktail [Sigma-Aldrich , P8340] and phosSTOP [Sigma-Aldrich , 04906845001] ) , and lysis was allowed to continue on ice for 10 min . Cells were sonicated with a Fisherbrand Model 120 Sonic Dismembrator ( Fisher Scientific , FB120A110 ) for three pulses of 20 s each at 20% amplitude . Cell lysate was spun down at 15 , 000 g for 10 min at 4°C , and supernatant was transferred to a new tube . For protein extraction from tissue , the sample was resuspended in 500 µL RIPA buffer and homogenized using a TissueLyser ( Qiagen , 85210 ) twice for 30 s at 20 Hz . Following incubation on ice for 10 min , the same procedure was followed as for cells . For both cells and tissue , lysate samples were stored at −80°C until analysis by immunoblot . All blots were developed using a LI-COR Odyssey CLx system . Antibodies used in this study were as follows: O-GlcNAc CTD110 . 6 ( Cell Signaling 9875S ) , tubulin ( Sigma T6199 ) , HSP60 ( Cell Signaling 12165S ) , NAGK ( Atlas Antibodies , HPA035207 ) , and PARP ( Cell Signaling 9532 ) . For blots showing the mobility shift for NAGK in low glutamine , samples were prepared in lysis buffer containing 50 mM Tris pH 8 . 0 , 150 mM NaCl , 0 . 5% IGEPAL CA-630 ( Sigma , I3021 ) , 1 mM PMSF , 1 . 5 µM aprotinin , 84 µM leupeptin , 1 µM pepstatin A , ±10 mM NaF , and 20 mM Na3VO4 as indicated . To visualize the NAGK mobility shift in response to low glutamine , 20 µg total protein per sample was separated across 12 . 5 cm of 11% SDS-PAGE resolving space under reducing conditions using the large electrophoresis systems available from C . B . S . Scientific until approximately 3 cm of separation was obtained between the 25 and 37 kDa protein standards ( Bio-Rad; 1610375 ) . Using electrophoresis , proteins were transferred ( 30 V , 4°C , overnight ) to 0 . 45 µM pore size nitrocellulose membrane ( Amersham , 10600002 ) . The primary antibodies used were NAGK ( Proteintech , 15051–1-AP ) , AKT phosphoS473 ( Cell Signaling Technology , 4060 ) , and Vinculin ( Sigma , V9264 ) . Membranes were developed using the LI-COR Odyssey CLx system . For RNA extraction from cells , cells were put on ice , washed with PBS , and scraped into PBS . Samples were then spun down at 200 g for 5 min and resuspended in 100 µL Trizol ( Life Technologies ) . For RNA extraction from tissue , samples were resuspended in 500 μL Trizol and homogenized using a TissueLyser twice for 30 s at 20 Hz . For both cells and tissue , RNA was extracted following the Trizol manufacturer protocol . cDNA was prepared using high-capacity RNA-to-cDNA master mix ( Applied Biosystems , 4368814 ) according to kit instructions . cDNA was diluted 1:20 and amplified with PowerUp SYBR Green Master Mix ( Applied Biosystems , A25778 ) using a ViiA-7 Real-Time PCR system . Fold change in expression was calculated by the ΔΔCt method using HPRT as a control . Please see table at end of chapter for primer sequences . Cells were put on ice , washed with PBS , and then scraped into PBS . Samples were then spun down at 200 g for 5 min and resuspended in 3% BSA with fluorophore-conjugated lectin added 1:1000 ( Vector Labs FL-1111–2 ) . Samples were covered and incubated on ice for 30 min at room temperature , then spun down , and resuspended in PBS before analysis with an Attune NxT Flow Cytometer ( Thermo Fisher Scientific ) . Data was further analyzed using FlowJo 8 . 7 . For all metabolite quantitation experiments , each sample was collected from a 10 cm sub-confluent plate of cells . To achieve similar confluency and protein content at the experiment end point , cells were initially plated more densely for the nutrient-deprived samples than for the nutrient-replete samples . For low glutamine experiments , PANC-1 cells were plated 3 × 105 for 4 mM glutamine samples and 5 . 5 × 105 for 0 . 05 mM samples . MIA PaCa-2 cells were plated 3 × 105 for 4 mM samples and 1 . 2 × 106 for 0 . 05 mM samples . Samples were prepared according to Guo et al . , 2016b . Briefly , cells were put on ice and washed 3× with PBS . Then , 1 mL of ice cold 80% methanol was added to the plate , and cells were scraped into solvent and transferred to a 1 . 5 mL tube . For quantitation experiments , internal standard containing a mix of 13C labeled metabolites was added at this time . Samples were then sonicated and spun down , and the supernatants were dried down under nitrogen . The dried samples were then resuspended in 100 µL of 5% sulfosalicylic acid and analyzed by liquid chromatography–high-resolution mass spectrometry as reported ( Guo et al . , 2016b ) with the only modification that the LC was coupled to a Q Exactive-HF with a heated ESI source operating in negative-ion mode alternating full scan and MS/MS modes . The [M-H]− ion of each analyte and its internal standard was quantified , with peak confirmation by MS/MS . GlcNAc quantification was done on a triple quadropole instrument exactly as described ( Guo et al . , 2016b ) . Data analysis was conducted in Thermo XCalibur 3 . 0 Quan Browser and FluxFix ( Trefely et al . , 2016 ) . For quantitation experiments , samples were normalized first to peak integrations of 13C-labeled internal standard components and then to protein content in the sample , measured by BCA assay . Relative quantification was then calculated by normalizing to the control condition in each experiment . For glucose labeling experiments , cells were cultured in DMEM without glucose , glutamine , or phenol red supplemented with 10 mM [U-13C]-glucose ( Cambridge Isotopes , CLM-1396–1 ) , 4 mM glutamine , and 10% dialyzed fetal bovine serum . Cells were incubated for the indicated time , and samples were prepared as above . For GlcNAc labeling experiments , cells were cultured in DMEM without glucose , glutamine , or phenol red supplemented with 10 mM N-[1 , 2–13 C2]acetyl-d-glucosamine ( 13C GlcNAc ) ( Omicron Biochemicals , GLC-006 ) , 4 mM glutamine , 10 mM glucose , and 10% dialyzed fetal bovine serum . Cells were incubated for the indicated time , and samples were prepared as above . Cells were trypsinized and counted using a Bright-Line hemacytometer ( Sigma , Z359629 ) . The bottom agar layer was prepared by adding Bacto Agar ( BD Bioscience , 214050 ) to cell culture media for a final concentration of 0 . 6% . Two milliliter bottom agar was added to each well of a six-well tissue culture plate . Once bottom agar solidified , top layer agar was prepared by combining trypsinized cells with the bottom agar mix for a final concentration of 0 . 3% Bacto Agar . One milliliter top layer agar was added to each well with a bottom layer of agar . Cells were plated 2 . 5 × 104 per well . 0 . 5 mL DMEM high glucose with 10% calf serum was added to cells every 7 days . Images were taken after 3 weeks . Images were blinded , and colonies per image were counted using ImageJ ( Schneider et al . , 2012 ) . Cells were plated 3 . 5 × 104 per well of a six-well plate . For each day that counts were recorded , three wells were trypsinized and cells were counted twice using a hemocytometer ( Sigma , Z359629 ) . The average of the two counts was recorded for each well , and the average count of the three wells was used to graph the data . For proliferation assays in 0 . 05 mM glutamine , trypan blue was used during cell counts . The PDAC expression profiling dataset ( GEO accession GSE16515 , Pei et al . , 2009 ) from NCBI GEO Profile database ( Edgar et al . , 2002 ) was used to compare the expression level between human normal and PDAC tumor samples . The dataset consists of 52 samples , in which 16 samples are matched tumor and normal tissues , and 20 samples are only tumor tissues . The statistical analysis was conducted by one-way ANOVA; the level of significance was evaluated by p<0 . 01 and plotted in box-and-whisker diagram . Comparison of HBP gene expression between tumor ( TCGA PAAD dataset ) and normal tissue ( GTEx ) was also conducted using GEPIA2 ( Tang et al . , 2019 ) . 3 × 106 PANC-1 NAGK CRISPR cells were injected with 1:1 Matrigel ( Corning , CB354248 ) into the flanks of NCr nude mice and measured with calipers once per week for 22 weeks . At the experiment end point ( 22 weeks or when tumor reached 20 mm in length ) , mice were euthanized with CO2 and cervical dislocation . Tumors were removed , weighed , cut into pieces for analysis , and frozen . All animal experiments were approved by the University of Pennsylvania and the Institutional Animal Care and Use Committee ( IACUC ) . | Inside tumors , cancer cells often have to compete with each other for food and other resources they need to survive . This is a key factor driving the growth and progression of cancer . One of the resources cells need is a molecule called UDP-GlcNAc , which they use to modify many proteins so they can work properly . Because cancer cells grow quickly , they likely need much more UDP-GlcNAc than healthy cells . Many tumors , including those derived from pancreatic cancers , have very poor blood supplies , so their cells cannot get the nutrients and other resources they need to grow from the bloodstream . This means that tumor cells have to find new ways to use what they already have . One example of this is developing alternative ways to obtain UDP-GlcNAc . Cells require a nutrient called glutamine to produce UDP-GlcNAc . Limiting the supply of glutamine to cells allows researchers to study how cells are producing UDP-GlcNAc in the lab . Campbell et al . used this approach to study how pancreatic cancer cells obtain UDP-GlcNAc when their access to glutamine is limited . They used a technique called isotope tracing , which allows researchers to track how a specific chemical is processed inside the cell , and what it turns into . The results showed that the pancreatic cancer cells do not make new UDP-GlcNAc but use a protein called NAGK to salvage GlcNAc ( another precursor of UDP-GlcNAc ) , which may be obtained from cellular proteins . Cancer cells that lacked NAGK formed smaller tumors , suggesting that the cells grow more slowly because they cannot recycle UDP-GlcNAc fast enough . Pancreatic cancer is one of the most common causes of cancer deaths and is notable for being difficult to detect and treat . Campbell et al . have identified one of the changes that allows pancreatic cancers to survive and grow quickly . Next steps will include examining the role of NAGK in healthy cells and testing whether it could be targeted for cancer treatment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2021 | Glutamine deprivation triggers NAGK-dependent hexosamine salvage |
Effective perceptual decisions rely upon combining sensory information with knowledge of the rewards available for different choices . However , it is not known where reward signals interact with the multiple stages of the perceptual decision-making pathway and by what mechanisms this may occur . We combined electrical microstimulation of functionally specific groups of neurons in visual area V5/MT with performance-contingent reward manipulation , while monkeys performed a visual discrimination task . Microstimulation was less effective in shifting perceptual choices towards the stimulus preferences of the stimulated neurons when available reward was larger . Psychophysical control experiments showed this result was not explained by a selective change in response strategy on microstimulated trials . A bounded accumulation decision model , applied to analyse behavioural performance , revealed that the interaction of expected reward with microstimulation can be explained if expected reward modulates a sensory representation stage of perceptual decision-making , in addition to the better-known effects at the integration stage .
A central issue for models of perceptual decision-making is how information about the value associated with different choices is combined with sensory information to influence the final decision . Balancing sensory evidence with the prospect of reward must take place on a case-by-case basis for every trial in which a decision is made . When sensory information is used simply to signal the availability of reward in value-based decisions , the activity of lateral intraparietal ( LIP ) neurons correlates with the relative reward an animal can expect for a particular choice ( Platt and Glimcher , 1999; Sugrue et al . , 2004 ) . In perceptual decision-making , however , making a correct decision requires the discrimination of one version of the stimulus from another . Experimentally , different perceptual choices can be arranged to receive greater or lesser reward , contingent upon a correct decision . Such manipulations affect perceptual decision-making , for example by improving performance or biasing decisions ( e . g . , Summerfield and Koechlin 2010; Weil et al . 2010 ) . However , the neural mechanisms by which reward signals influence perceptual decision-making are not fully understood . Perceptual decision-making has been successfully modelled as a bounded evidence-accumulation process in which sensory evidence about the visual stimulus is represented in sensory cortex and then integrated over time in sensorimotor structures such as lateral intraparietal cortex ( area LIP ) ( Shadlen and Newsome , 2001; Roitman and Shadlen , 2002; Mazurek et al . , 2003; Huk and Shadlen , 2005; Gold and Shadlen , 2007 ) ( Figure 1 ) . Previous human neuroimaging and animal electrophysiology studies found that when perceptual decision-making is modelled in this way , information about reward is reflected only at the integration stage , represented in sensorimotor structures such as area LIP ( Rorie et al . , 2010; Summerfield and Koechlin , 2010; Mulder et al . , 2012 ) . But does reward simply affect the accumulation of sensory evidence and the motor response , or does it also affect neuronal representations of the sensory stimulus , for example in visual cortical area V5/MT ( Figure 1 ) ? 10 . 7554/eLife . 07832 . 003Figure 1 . Schematic illustration of a visual perceptual decision-making pathway with V5/MT microstimulation and reward . Primary visual cortex ( area V1 ) comprises the initial stage of visual information processing but neurons here can respond to visual sensory signals that do not reach perception ( Cumming and Parker , 1997 ) . On the other hand , activity of motion- and disparity-selective neurons in visual area V5/MT have been closely linked to animals' subjective perception during discrimination of structure-from-motion ( SFM ) visual stimuli ( Dodd et al . , 2001 ) . In the present study , we artificially activate motion- and disparity-selective neurons in visual area V5/MT with electrical microstimulation . Microstimulation biases perceptual choices towards visual interpretations that match the tuning of the stimulated neurons ( Krug et al . , 2013 ) . In visual discrimination tasks , sensory evidence is integrated over time into a decision variable ( DV ) , represented in the activity of neurons in lateral intraparietal cortex ( area LIP; see Gold and Shadlen 2007 for review ) . Since visually evoked and electrically evoked signals both influence behaviour , they are presumably integrated together to influence the DV . The results of evidence integration also affect activity in the frontal eye fields ( FEF ) that represent the planning of eye movements ( saccades ) , which animals use to indicate their perceptual decision in the discrimination task ( Figure 2 ) . Previous studies have established an influence of reward on neural representations in sensorimotor regions such as area LIP ( Platt and Glimcher , 1999; Sugrue et al . , 2004; Rorie et al . , 2010 ) . However , it is not known whether reward can also affect sensory or perceptual signals represented in visual cortex during perceptual decision-making . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 003 There is some indication that expected reward can affect sensory representations . Blood-oxygen level dependent ( BOLD ) signals in sensory cortex are modulated by reward probability , size and delivery ( Pleger et al . , 2008; Serences and Saproo , 2010; Weil et al . , 2010 ) . However , the BOLD signal is an indirect measure of neural activity that averages the responses of many neurons , including those that are excited by the sensory stimulus and those that are inhibited by it ( Logothetis et al . , 2001; Logothetis and Wandell , 2004 ) , and may be differentially affected by modulatory signals in comparison to neuronal spiking ( Boynton , 2011 ) . A stronger indication comes from a study in which neuronal responses from area V1 were electrophysiologically recorded while animals performed a saccadic curve-tracing task ( Stanisor et al . , 2013 ) . This study showed that V1 activity was predicted by the reward value of a stimulus relative to other stimuli , implying that reward signals have the potential to reach the earliest stages of sensory cortical processing ( Figure 1 ) . Experimental interference with the activity of sensory cortex may directly induce changes in the subject's behaviour or their perceptual report , and provides a tool with which to dissect the perceptual decision-making pathway . Such interventions provide strong evidence for a causal role of that region in generating that behaviour or report ( Parker and Newsome , 1998 ) . Electrical microstimulation of small populations of neurons in relevant cortical regions , combined with sensory stimulation , can produce reliable changes in perceptual reports , which are specific to the stimulus preferences of the neurons at the microstimulated site ( Salzman et al . , 1990 , 1992 ) . When the stimulated neurons encode combinations of visual cues , microstimulation adds a signal that is specific to that combination , as found in cortical area V5/MT for visual motion and binocular disparity ( Krug et al . , 2013 ) . Using microstimulation to bias perceptual choices under different reward conditions is a novel approach with which to explore the interaction between expected reward and sensory signals that contribute directly to perceptual decisions . We investigated the interaction between electrical microstimulation of visual cortical area V5/MT and available reward size for a correct choice during visual discrimination of a structure-from-motion ( SFM ) stimulus . The available reward size for a correct choice was varied according to animals' preceding behavioural performance . When electrical microstimulation is applied in a sensory area during a perceptual decision task , any effect of reward that acts exclusively at the integration stage , by which time visually and electrically evoked sensory signals have been combined ( Figure 1 ) , would not differentially affect visually evoked and electrically evoked signals . On the other hand , reward could have a differential effect on neurons that have been specifically excited by either visual or electrical microstimulation . We found that the size of the microstimulation effect on perceptual choices was modulated by the size of the available reward for a correct choice . In the context of a simplified bounded evidence-accumulation model , these results were best explained by an effect of reward on neuronal representations of sensory evidence , in addition to an effect on the integration stage .
In agreement with the existing literature about reward and perceptual decisions , available reward size affected perceptual choices . Trials from psychophysical ( non-microstimulated ) experimental blocks were separated according to available reward size and pooled across blocks and animals . There was a significant interaction between the slope of the fitted psychometric function and available reward size , indicating better performance in large reward trials ( χ2 likelihood-ratio test of nested models Equations 2a , b , ‘Materials and methods’: p = 0 . 001; Figure 3 ) . Since the reward schedule was dependent on performance history , fluctuations in animals' performance over the course of a block could have contributed to a correlation between large reward trials and high performance . However , the control analyses for performance fluctuations ( see below ) show that such fluctuations do not explain the main results of the study . 10 . 7554/eLife . 07832 . 006Figure 3 . Large reward size is associated with improved performance . Animals performed two blocks of the visual discrimination task at each site , prior to the introduction of electrical microstimulation . Trials were separated according to available reward size ( small or large ) and pooled across all sites over both animals . Psychometric functions are fit with cumulative Gaussians . There is a significant interaction between reward size and steepness of the slope ( s . d . of the fitted function; χ2 likelihood-ratio test of nested models Equations 2a , b: p = 0 . 001 ) . For large reward trials the slope is steeper , indicating better performance accuracy . Error bars show the standard error of the mean ( s . e . m . ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 006 It is not well understood at which stages of cortical processing for perceptual decision-making neural signals for reward are integrated ( Figure 1 ) . However , the neural mechanisms of perceptual decision-making per se have been modelled successfully by a bounded evidence-accumulation process in which representations of sensory information are noisily integrated over time towards decision bounds that represent the possible perceptual outcomes ( Gold and Shadlen , 2007; Ratcliff and McKoon , 2008 ) ( Figure 4A ) . Considering the possible effects of reward on different parameters in the bounded accumulation model can help us to interpret how expected reward influences perceptual decisions . 10 . 7554/eLife . 07832 . 007Figure 4 . Schematic illustration of possible interactions between microstimulation , reward and the parameters of the bounded accumulation model of perceptual decision-making . ( A ) Momentary evidence , in favour of choice ‘A’ over choice ‘B’ , is accumulated over time . The time-varying accumulation of evidence is termed the decision variable ( DV; black line ) . ( B ) The bounded accumulation model applied to the cylinder task with microstimulation . The momentary evidence ( e ) is the difference in activity between two neuronal pools , one selective for CW rotation and one for CCW rotation . Parameter k represents the sensitivity of these sensory representations to the cylinder disparity C ( Palmer et al . , 2005 ) . In each trial , the DV follows a stochastic ‘drift-diffusion’ path with an average drift rate dependent on the mean of e = kC . Parameter B represents the distances to the CW and CCW decision bounds , which are assumed to be equal . Microstimulation contributes to e as additional evidence in favour of the sensory preference of the stimulated V5/MT multi-unit site ( CCW in this example ) . The perceptual decision depends on which bound was first reached , or , if neither is reached , which bound the DV is nearest to at the end of stimulus viewing . ( C ) Model simulations illustrate that the slope of the psychometric function , fitted by a simple logistic model of bounded accumulation ( Equation 3a ) , can be affected by changes in either parameter k or B; for example , an increase in either parameter will steepen the slope , indicating improved performance . ( D ) Model simulations illustrate how insertion of additional sensory evidence by electrical microstimulation in visual cortex biases perceptual choices toward the preferred ( PREF ) cylinder disparity of the stimulated multi-unit ( MU ) , which is revealed as a shift of the psychometric function . ( E ) An increase in parameter B steepens the slope of the psychometric functions , but does not change the size of microstimulation shift . This is because by the evidence-accumulation stage , visually evoked and electrically evoked signals are combined and cannot be differentially affected ( see panel B ) . ( F ) An increase in parameter k also steepens the slope of the psychometric function . However , improved perceptual sensitivity to the stimulus affects only the visually-evoked sensory representations ( see panel B ) , so the relative contribution of the electrically-evoked signals is decreased , which results in a decrease in the microstimulation-induced shift of the psychometric function . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 007 In the bounded accumulation model , sensory evidence about the cylinder stimulus is represented as the difference in activity of two pools of sensory neurons , where the conjoint motion and disparity tuning of each pool represents one of the two possible interpretations of cylinder rotation direction ( Mazurek et al . , 2003; Palmer et al . , 2005; Gold and Shadlen , 2007; Krug et al . , 2013 ) ( Figure 4B ) . Sensory evidence is integrated over time to influence the drift-diffusion of the decision variable ( DV ) ( Huk and Shadlen , 2005 ) . The DV starts at a particular distance from the two decision bounds and , during stimulus viewing , drifts stochastically towards one or other decision bound at a rate proportional to the sensory evidence favouring the prospective choices . Upon stimulus offset , the animal makes a saccade towards the target that corresponds to its perceptual choice , which depends upon either the first bound that was crossed or , if no bound was crossed , the bound nearer to the DV final position ( Kiani et al . , 2008; Fetsch et al . , 2014 ) ( Figure 4B ) . The distance between the DV starting point and the decision bounds predicts the reaction time distribution in a reaction time task; in the present fixed viewing duration task , it offers predictions of decision times that are shorter than the stimulus duration and , critically for our data , it predicts performance accuracy ( Kiani et al . , 2008; Fetsch et al . , 2014 ) . According to the bounded accumulation model , an increase in either the stimulus sensitivity ( represented by parameter k ) or the distance between the DV starting point and the bounds ( represented by parameter B , Equation 3a , ‘Materials and methods’; Figure 4B ) can improve performance ( Figure 4C; compare to Figure 3 ) . This is because increasing the perceptual sensitivity to the stimulus allows smaller stimulus disparity signals to have a greater effect on the perceptual decision; on the other hand , increasing the diffusion distance to the decision bounds decreases the relative impact of spurious noise on the final decision ( Bogacz et al . , 2006; Ratcliff and McKoon , 2008 ) . The contributions of these parameters cannot be differentiated based upon these perceptual choice data alone because they both improve performance . However , these parameters make different predictions about what should happen to visual cortical microstimulation when performance improves . Electrical microstimulation biases perceptual decisions towards the PREF direction of the stimulated MU site ( Figure 4D; Krug et al . , 2013 ) . If performance in large reward trials improves as a result of an increase in B , this will not affect the size of the microstimulation shift , because B acts at an integration stage that is positioned after the location at which visually evoked and electrically evoked signals have been combined ( Figure 4E; see also Figure 1 ) . If performance improves as a result of an increase in k , this will decrease the relative influence of microstimulation on perceptual choices ( illustrated by a smaller horizontal shift between microstimulated and non-microstimulated psychometric functions ) because k acts specifically on visually evoked sensory representations before they are combined with electrical microstimulation signals ( Figure 4F ) . Therefore , the insertion of the electrical microstimulation signal at the level of representation of sensory evidence ( area V5/MT ) allows us to separate the effects of parameters k and B according to this model . We fit the bounded accumulation model of decision-making ( Equation 3a , ‘Materials and methods’ ) to behavioural choice data combined across significant PREF microstimulation sites ( sites with a significant microstimulation effect in their PREF cylinder rotation direction ) , separately for each animal ( Figure 5 ) . Prior to pooling , cylinder disparity values were normalised by the maximum disparity from the range used at each site . The size of the microstimulation effect - the horizontal shift between microstimulated and non-microstimulated psychometric functions—was decreased in large reward trials ( dashed lines ) compared to small reward trials ( smooth lines ) , implicating a change in stimulus sensitivity ( parameter k ) by reward ( see Figure 4F ) . Using a χ2 likelihood-ratio test , we compared the goodness-of-fit of the full bounded-accumulation model in which all parameters were allowed to vary by reward ( Equation 3a , ‘Materials and methods’ ) to a nested model in which the parameter in question ( either k or B ) was frozen , that is , not allowed to vary by reward ( Equations 3b , c , respectively ) . 10 . 7554/eLife . 07832 . 008Figure 5 . Reward modulates the effect of visual cortical microstimulation on perceptual decisions . Trials were pooled across significant PREF microstimulation sites separately for monkeys Fle ( A ) and Ica ( B ) . Psychometric functions were fit with the bounded accumulation model ( Equation 3a ) . In both cases , the effect of microstimulation ( horizontal shift between red and black psychometric functions ) was smaller in large reward trials ( dashed lines ) compared to small reward trials ( smooth lines ) . For both animals , the best-fitting model allowed both parameter k ( stimulus sensitivity ) and parameter B ( distance to decision bounds ) to be affected by reward condition ( χ2 likelihood-ratio tests of nested models Equations 3b , c , p < 0 . 05 in all cases , see main text for details ) . This suggests that reward can affect sensory representations as well as evidence integration during perceptual decision-making . Error bars show s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 008 For both animals , the overall best-fitting model was the full model in which both parameters B and k were affected by reward condition ( Equation 3a ) . For both animals , parameter k was significantly increased in large compared to small reward trials ( χ2 likelihood-ratio test of nested models Equations 3 a , b: Fle: p < 0 . 001; Ica: p < 0 . 001; Table 1 ) . This result supports the interpretation that the improvement in performance seen in large reward trials is in part due to an increase in parameter k , that is , increased sensitivity of sensory representations to the visual stimulus when reward is large . The estimated distance between the starting point of DV integration and the decision bounds , represented by parameter B , was also significantly affected by reward , decreasing in large compared to small reward trials ( χ2 likelihood-ratio test of nested models Equations 3a , c: Fle: p = 0 . 007; Ica: p < 0 . 001; Table 1 ) . Altogether these results indicate that expected reward affects perceptual decision-making both at the sensory representation stage ( parameter k ) and the integration stage ( parameter B ) . Statistical results are unaffected when data were pooled across all microstimulation sites . 10 . 7554/eLife . 07832 . 009Table 1 . Parameter values for bounded accumulation model fit to data combined over significant PREF microstimulation sitesDOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 009AnimalParameterEstimated value ( small reward ) Estimated value ( large reward ) % Change in estimate by reward−logL* for frozen parameter model−logL* for full modelχ-value of model comparisonp value of χ2-test†Flek1 . 742 . 41+38 . 33045 . 03034 . 321 . 40 . 00002Icak1 . 943 . 50+80 . 42398 . 92389 . 219 . 40 . 00001FleB1 . 771 . 40−20 . 73037 . 93034 . 37 . 20 . 0073IcaB1 . 290 . 79−38 . 42394 . 82389 . 211 . 20 . 0008*Negative log likelihoods ( −logL ) reported for each model fit , that is , with all parameters allowed to vary by reward condition ( Equation 3a ) compared with each parameter frozen ( Equations 3b , c ) . †p values were obtained from a χ2 likelihood-ratio test comparing the fit of the full bounded evidence-accumulation model to a restricted version of the model in which the specified parameter was not allowed to change value for different reward conditions ( nested models Equations 3a , b , c , ‘Materials and methods’; Figure 5A , B ) . An independent site-by-site analysis confirmed that the effect of microstimulation was significantly decreased in large reward trials compared to small reward trials ( Figure 6 ) . For each significant PREF microstimulation site , psychometric functions were fit with a pair of cumulative Gaussian functions ( Equation 1a , ‘Materials and methods’ ) , whose means were allowed to differ by microstimulation condition but whose standard deviation was constrained to be the same ( Krug et al . , 2013 ) . We illustrate these psychometric functions for example site fle302 . Here , MU neurons were selective for a CCW cylinder rotation direction ( Figure 6—figure supplement 1A ) . The effect of electrical microstimulation was equivalent to adding binocular disparity of −0 . 023° in magnitude to the stimulus dots , biasing choices toward the PREF cylinder direction CCW ( Figure 6—figure supplement 1B ) . We then separated trials according to whether available reward was large or small ( Figure 6—figure supplement 1C , D ) . Microstimulation always biased the animal's response toward CCW rotation , the PREF rotation direction at this site . However , when reward size was large , the effect of electrical microstimulation was equivalent to addition of −0 . 019° of binocular disparity ( Figure 6—figure supplement 1C ) , representing a reduction of about 30% compared to when reward size was small , where the microstimulation current was equivalent to addition of −0 . 027° ( Figure 6—figure supplement 1D ) . 10 . 7554/eLife . 07832 . 010Figure 6 . A site-by-site analysis confirms that visual cortical microstimulation is less effective at biasing perceptual choices in trials with large expected reward compared to trials with small expected reward . This is significant over both animals for microstimulation shifts normalised by discrimination threshold at each site ( A; Wilcoxon sign-rank test: p < 0 . 001 ) and also for the raw microstimulation shifts ( B; p < 0 . 001 ) . This is also significant for each animal separately ( see ‘Results’ for details ) . Raw microstimulation shifts ( B ) are plotted on a log-scale; please note that the effect of reward is comparable across different microstimulation effect sizes ( see Figure 8 ) . Black lines indicate the identity relationship . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 01010 . 7554/eLife . 07832 . 011Figure 6—figure supplement 1 . Effect of reward on visual cortical microstimulation at an example V5/MT site , fle302 . ( A ) At this site , the MU tuning had a preference for counter-clockwise rotating cylinders ( PREF direction = CCW ) . Filled circles: response to cylinder stimulus . Open circle: baseline ( blank screen ) . ( B ) Electrical microstimulation consistently increased the proportion of choices in the PREF direction . The proportions of PREF responses with and without microstimulation were fitted with a pair of cumulative Gaussians , constrained so that both curves had the same standard deviation ( s . d . ) but different means . The two-mean fit was better than a single function describing both microstimulated and non-microstimulated data , indicating a significant microstimulation effect at this site ( χ2 likelihood-ratio test of nested models Equations 1a , b , p < 0 . 001 ) . The horizontal shift between the fitted curves is a measure of microstimulation effect size . Here the effect is equivalent to adding a binocular disparity of −0 . 023° to the stimulus , such that the cylinder would appear to rotate more strongly in the PREF direction , CCW . Division by the s . d . gives the normalized microstimulation effect ( here , 1 . 298 ) . ( C ) In trials where available reward was large , microstimulation was decreased , equivalent to adding −0 . 019° of disparity ( normalized shift = 1 . 165 ) . ( D ) In trials where available reward was small , microstimulation was equivalent to adding −0 . 027° of binocular disparity ( normalized shift = 1 . 542 ) . All error bars show s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 011 Over all significant PREF microstimulation sites over both animals , we considered both the unit-less normalised microstimulation-induced shift ( divided by the cylinder threshold at the site; Figure 6A ) , and the raw shift measured in degrees of visual angle ( Figure 6B ) . The normalised microstimulation shift was significantly smaller for large expected reward trials ( median = 0 . 67 ) compared to small expected reward trials ( median = 0 . 88; Wilcoxon sign-rank test of normalised shift values , two-sided: n = 28 , p < 0 . 001 ) and also when considered separately for each animal Fle ( median large reward trials = 0 . 94; small reward trials = 1 . 37; n = 13 , p = 0 . 005 ) and Ica ( median large reward trials = 0 . 54; small reward trials = 0 . 68; n = 15 , p = 0 . 022 ) . Similarly , the raw microstimulation shift was significantly smaller for large expected reward trials ( median = 0 . 0064° ) compared to small expected reward trials ( median = 0 . 0093°; Wilcoxon sign-rank tests of raw shift values: n = 28 , p < 0 . 001 ) and also when considered separately for each animal Fle ( median large reward trials = 0 . 0099°; small reward trials = 0 . 024°; n = 13 , p = 0 . 013 ) and Ica ( median large reward trials = 0 . 0025°; small reward trials = 0 . 0071°; n = 15 , p = 0 . 005 ) . Statistical results were unaffected when data were pooled across all stimulation sites . We tested the possibility that the animals detect some subtle difference between trials with and without electrical microstimulation , and accordingly adjust their decision criterion to bias their choices against the microstimulation effect - and that they might do this more in large reward trials to reduce their error rate . This seems unlikely , as in principle the animals could then have applied a different criterion to all microstimulation trials and thereby eliminated the effect of microstimulation altogether , resulting in far fewer errors overall and thereby improving their acquisition of rewards . Nevertheless , to examine the possibility of such a strategy , we performed a set of psychophysical control experiments with monkey Ica that reproduced the effect of electrical microstimulation using a visual stimulus perturbation . This variation omitted the electrical microstimulation itself , but retained the cognitive cues that might be delivered to the animal during electrical microstimulation . Using the same experimental parameters as for electrical microstimulation sessions , we inserted an additional binocular disparity signal , ‘∆dx’ , to the cylinder stimulus pseudo-randomly on 50% of trials ( see Salzman et al . 1992; Fetsch et al . 2014 for a similar manipulation of visual motion signals in a motion direction discrimination task ) . In our experiment , the trials with the ∆dx signal were overtly flagged by a change in the appearance of the cylinder stimulus from all black dots to all white dots ( Figure 7 ) . This change in cylinder appearance is highly salient to human observers and we assume it is also salient for the monkey . Importantly , the animal was only rewarded for perceptual choices that would have been correct had the additional disparity signal not been present . This creates a form of ‘microstimulation’ experiment in which the presence of the extra disparity cue ∆dx is clearly signalled by a change in stimulus colour from black to white . The animal was first trained over 10 days and 45 , 000 trials on this task . Unlike the electrical microstimulation experiments , which contained a mixture of microstimulation biases across different V5/MT sites , favouring sometimes CW and sometimes CCW choices , the added disparity signal ∆dx always gave a consistent shift towards CCW perceptual choices . This increased the chance that the animal would engage in a strategy that exploited all available information to reduce its error rate and gain more reward . 10 . 7554/eLife . 07832 . 012Figure 7 . Effect of reward on visual cortical microstimulation cannot be explained by a change in the animal’s strategy . A control experiment was run with animal Ica , in which microstimulation was simulated by insertion of an additional disparity signal ( ‘∆dx’ ) into the cylinder stimulus . The animal was rewarded only for correct choices with respect to cylinder disparity before insertion of ∆dx . Inset panels illustrate the colour change of cylinder dots from all black in trials with no additional disparity added ( A ) to all white dots in ∆dx trials ( B ) . Performance in the ∆dx control was significantly better in large reward compared to small reward trials ( χ2 likelihood-ratio test of nested models Equations 2a , b: p < 0 . 001 ) , but there was no significant effect of reward on the shift induced by additional ∆dx disparity ( χ2 likelihood-ratio test of nested models Equations 2a , c: p > 0 . 05 ) . Therefore , even when trials that contain additional disparity were clearly signalled , the animal did not make an adjustment to counter this additional signal more on large relative to small reward trials . By contrast , under the same model , reward significantly reduced the electrical microstimulation shift for both Fle and Ica ( see main text ) . This suggests that the change in microstimulation shift with reward was not likely to be due to the animals adopting a decision criterion that depended both on reward size and on detecting microstimulation . Error bars show s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 012 Despite the fact that trials with ∆dx were clearly signalled , the psychometric function for ∆dx trials was horizontally shifted towards the response direction predicted by the additional disparity signal , successfully emulating the horizontal shift of the psychometric function due to V5/MT electrical microstimulation ( Figure 7 ) . As observed in the electrical microstimulation experiments , there was a significant improvement in performance in trials where a larger reward was available for a correct choice ( χ2 likelihood-ratio test of nested models Equations 2a , b: Ica-∆dx: p < 0 . 001 ) . However , there was no significant interaction between reward condition and ∆dx-induced shift in the ∆dx psychometric functions ( χ2 likelihood-ratio test of nested models Equations 2a , c: Ica-∆dx: p > 0 . 05; Figure 7 ) , in contrast to the results when the same test was applied to the electrical microstimulation datasets ( χ2 likelihood-ratio test of nested models Equations 2a , c: Fle: p < 0 . 001; Ica: p < 0 . 001 ) . Therefore , there is no evidence to support the view that microstimulation induces a different percept that the animal discounts more effectively when reward is higher , since the animal does not make an adjustment in large reward trials when the presence of additional disparity is clearly signalled . We explored further trends in the data to evaluate the potential relationship between the reward modulation of visual cortical microstimulation , stimulus eccentricity , discrimination threshold , and raw overall microstimulation effect at each site . Visual area V5/MT contains a retinotopic map of contralateral visual space ( Albright and Desimone , 1987 ) and the receptive field ( RF ) location of each microstimulated MU varies in eccentricity , that is , distance from the fovea ( Desimone and Ungerleider , 1986 ) . The cylinder stimulus was matched to the MU RF location at each microstimulation site . Both animals' cylinder discrimination thresholds significantly correlated with stimulus eccentricity , worsening as eccentricity increased ( Figure 8A ) . The raw overall microstimulation effect also significantly correlated with the cylinder discrimination threshold at each site ( Figure 8B ) . The reward modulation of the raw microstimulation shift effect was significantly correlated with stimulus eccentricity and with the raw overall microstimulation effect size for both animals considered together , and for one animal considered alone ( Ica ) , such that the biggest raw effects of reward on microstimulation occurred for sites with largest eccentricity ( where thresholds were high ) and the biggest raw overall microstimulation effect ( Figure 8C , D ) . However , these correlations disappeared for reward modulation of the microstimulation shift effect when normalised by cylinder discrimination threshold at each site ( Figure 8E , F ) . This shows that the reward effect is constant when microstimulation is expressed as a fraction of the discrimination threshold . 10 . 7554/eLife . 07832 . 013Figure 8 . Effect of reward on visual cortical microstimulation is constant over stimulus eccentricity and over raw overall microstimulation effect size . ( A ) Cylinder discrimination threshold ( s . d . of the cumulative Gaussian fitted to the psychometric function ) significantly correlates with stimulus eccentricity , which is determined by the MU receptive field ( RF ) location at the microstimulation site ( Fle: p = 0 . 001; Ica: p = 0 . 001 ) . ( B ) Raw overall microstimulation effect at each site is significantly correlated with discrimination threshold ( Ica: p < 0 . 001; Fle: p < 0 . 001 ) . ( C ) Effect of reward on the raw microstimulation shift ( i . e . the raw microstimulation shift in large reward trials minus the raw shift in small reward trials ) significantly correlates with stimulus eccentricity for both animals considered together ( p = 0 . 018 ) and also for Ica alone ( p < 0 . 001 ) . ( D ) Effect of reward on the raw microstimulation shift significantly correlates with the raw overall microstimulation effect for both animals considered together ( p = 0 . 028 ) and also for Ica alone ( p < 0 . 001 ) . ( E ) When the effect of reward is considered over microstimulation shifts normalised by discrimination threshold , the correlation with stimulus eccentricity disappears ( p > 0 . 05 in all cases ) . ( F ) When the effect of reward is normalised by discrimination threshold at each site , the correlation with raw overall microstimulation effect disappears ( p > 0 . 05 in all cases ) . ( E , F ) show that the reward effect is constant when expressed as a fraction of discrimination threshold . All correlation tests use Pearson's product–moment correlation coefficient . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 01310 . 7554/eLife . 07832 . 014Figure 8—figure supplement 1 . Fluctuations in performance do not explain effect of reward on visual cortical microstimulation . ( A ) Time-series analysis of performance over the microstimulation session at example site fle300 . Smoothed performance accuracy ( proportion correct choices ) was calculated over a sliding time window of 30 trials . The horizontal line indicates mean performance accuracy at the site . Smoothed performance accuracy fluctuates above and below this mean , perhaps reflecting fluctuations in task engagement . The amount of performance fluctuation at each site was calculated as the area between the smoothed performance curve and the horizontal line , normalised by total number of trials . Epochs of good and bad performance were identified , marked in green and red respectively . ( B ) There is no correlation between normalised reward effect size and the amount of performance fluctuation across microstimulation sites , either for animals separately or together ( Pearson's product–moment correlation: p >> 0 . 05 in all cases ) , suggesting that performance fluctuation cannot explain the reward effect on microstimulation . Next , we tested whether the effect of reward on microstimulation remained in good epochs ( C ) and bad epochs ( D ) , combined across sites and animals . For both types of epoch , the effect of microstimulation was significantly smaller in large compared to small expected reward trials ( χ2-test of nested models Equations 2a , c: good epochs: p < 0 . 001; bad epochs: p < 0 . 001 ) , showing that the effect of expected reward on microstimulation remains present in different states of task engagement . Although it might be expected that there may be fewer difficult ( small disparity ) trials in good epochs than bad epochs , it can be seen that all disparities are represented in both epoch types , allowing the psychometric functions to be fitted . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 01410 . 7554/eLife . 07832 . 015Figure 8—figure supplement 2 . Waning of microstimulation does not explain reward effect . The microstimulation effect may wane over time in some experiments ( Salzman et al . , 1992 ) . Microstimulation might be less effective near the end of each experimental session , simultaneously with the occurrence of a greater proportion of large reward trials near the end of the session – for example , owing to a training effect that increases the number of consecutive correct responses . Such a co-variation could explain the difference in microstimulation effect by reward . A running average of the proportion of large reward trials was calculated for each experimental session at each significant PREF microstimulation site ( significant shift in the PREF stimulus direction ) , across both animals . The normalised reward effect on microstimulation shift is plotted against the gradient of the best-fit line of the running-average proportion of large reward trials over time , for each site . A positive gradient indicates an increased proportion of large reward trials towards the end of the microstimulation session . There was no correlation between reward effect and proportion of large reward trials at the end of the session over the two animals ( Spearman's rank correlation: n = 28 , p > 0 . 05; Pearson's linear correlation: n = 28 , p > 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 01510 . 7554/eLife . 07832 . 016Figure 8—figure supplement 3 . Evaluation of model fits with quantile–quantile comparisons . Pair-wise quantile–quantile comparison plots between the electrical microstimulation data ( pooled over significant PREF sites ) , fitted cumulative Gaussian model ( Equation 2a ) , and fitted bounded accumulation model ( Equation 3a ) for Fle ( A , B , C ) and Ica ( D , E , F ) . ‘Quantile’ values are the proportions of PREF choices for each stimulus disparity data point , taken from the real datasets . For each such quantile value , the corresponding cylinder disparity value was extracted from the fitted psychometric function for each model . These model-simulated disparity values were plotted against the disparities obtained from the microstimulation data , and against each other . A linear quantile–quantile relationship indicates a reasonable fit of the model to the microstimulation data for model-data comparisons , and a reasonable correspondence between model fits for model–model comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 07832 . 016 In these experiments , available reward size for a correct choice depended upon animals' performance in the preceding trials , increasing gradually when animals performed correctly over consecutive trials ( Figure 2 ) . Sequences of correct trials are more likely when the animal is more engaged in the task , perhaps due to attentional or motivational states . In an alternative interpretation , slower fluctuations in task engagement , rather than expected reward size , are the cause of the observed effects . Variations in task engagement would be expected to occur at least an order of magnitude slower ( over the space of around 30 trials ) than differential valuations of upcoming reward ( which occur over the space of two or three trials; Figure 2 ) . We therefore address this concern with a time-series analysis of task engagement . For the experimental session at each microstimulation site for each animal , a smoothed performance average was calculated using a sliding time window of size 30 trials ( Figure 8—figure supplement 1A ) . The extent of performance fluctuation over the experimental session was estimated as the area between the smooth performance curve and average ( normalised by total number of trials ) . If fluctuations in task engagement underlie the observed association between reward and microstimulation shift , there should be a correlation between the extent of performance fluctuation and size of reward effect . However , there was no correlation for either animal considered separately or together ( p > 0 . 05 in all cases; Figure 8—figure supplement 1B ) . Furthermore , the effect of expected reward on microstimulation shift remained when trials were separated according to whether they occur in ‘good’ or ‘bad’ epochs of task engagement ( χ2 likelihood-ratio test of nested models Equations 2a , c , data pooled across sites and animals: good epochs: p < 0 . 001; bad epochs: p < 0 . 001; Figure 8—figure supplement 1C , D ) . Therefore , the modulation of cortical microstimulation by available reward size does not seem to be associated with or dependent upon slower fluctuations in task engagement . In further controls , we found that the effect of reward could not be explained by the waning of microstimulation effect over time during an experimental session , combined with a propensity to have a greater proportion of large reward trials at the end of a microstimulation session ( Figure 8—figure supplement 2 ) . Also , there were no significant differences in proportion of microstimulated trials or mean absolute disparity between reward conditions ( Supplementary file 1 ) .
Our results could not be explained by an adaptive change in the animals' strategy based upon recognizing the presence of microstimulation . In a control experiment , in which an additional visual disparity ( ∆dx ) signal that mimicked the microstimulation effect was clearly cued by a change in stimulus colour , the effect of ∆dx on choices was not affected by expected reward size , even though overall task performance was better in large reward trials as seen in the electrical microstimulation sessions . This indicates that the animal did not apply a strategy to compensate for the ∆dx signal specifically on large reward trials , even when trials that contained additional visual disparity were clearly identifiable . This experiment controls for the situation in which microstimulation produces a different percept that the animal is able to discount more effectively when available reward is higher . However , this psychophysical experiment cannot simulate possible noise or other aberrance in inputs associated with electrically evoked sensory neuronal activity , which , for example , might be weighed less compared to more ‘normal’ , visually evoked inputs; weights that could change based on expected reward . We discuss this issue further in the next section . In the present study , reward depended on animals' performance in the preceding trials , increasing gradually when animals performed correctly over consecutive trials , which is more likely to happen when the animal is more engaged in the visual task . Therefore , instead of expected reward size , fluctuations in task engagement ( for example due to attentional or motivational states ) could result in epochs of good and bad performance , which are inevitably associated with more large reward or small reward trials respectively . These fluctuations could therefore potentially underlie the observed association between large reward trials and better performance , and the observed effects on the microstimulation shift . However , a time-series analysis of performance over the course of the microstimulation experimental sessions showed that there was no correlation between the performance fluctuation and the differential effect of microstimulation by reward . Furthermore , when trials in each microstimulation session were separated according to whether they occurred in a ‘good epoch’ ( a window of better-than-average performance ) or ‘bad epoch’ ( poorer-than-average performance ) , the modulation of microstimulation shift by reward size remained within both types of epoch . These analyses , which monitor closely engagement with the task over time , provide evidence against the hypothesis that the differential microstimulation effect can be accounted for purely by fluctuations in task engagement . Future experiments can directly rule out this possibility by varying reward size independently of performance , for example by using a block design or trial-by-trial cueing of the available reward size . In our interpretation of the results , we assume that electrically evoked signals in visual area V5/MT are thereafter integrated into the perceptual decision-making process in a manner identical to visually evoked evidence ( Figure 5 ) . This interpretation is supported by the behavioural shift induced by electrical microstimulation in perceptual decisions . However , the circuit mechanisms of the effect of electrical microstimulation are incompletely understood , and microstimulation has been reported to have unanticipated inhibitory effects on downstream cortical targets ( Logothetis et al . , 2010; Sultan et al . , 2011 ) . Such observations have generally been made in anaesthetised animals , without the concurrent presentation of a visual stimulus expected to activate the same brain areas as microstimulation , and at stimulation currents orders of magnitude higher than used in this study ( 250–1000 μA , compared to our 20 μA ) . Therefore , it is difficult to interpret these effects in the context of our experimental protocols . A remaining possibility is that odd or aberrant inputs from electrically evoked sensory neurons might be weighed less by decision-making areas compared with more ‘normal’ that is , visually evoked inputs , and that these weights could change based upon expected reward . Our ∆dx experiment cannot control for this case because we added the additional stimulus disparity signal via visual stimulation that by this argument cannot mimic the artificially induced microstimulation signal . However , we argue that there is currently little evidence to suggest that the brain applies differential weights to electrically and visually evoked sensory signals in this way . It is important to note that microstimulation affects behaviour in both large and small reward trials , suggesting that it is not an aberrant signal but a clear signal that affects perceptual choices in all cases . When combined with relevant visual stimulation , electrical microstimulation in many visual cortical areas significantly and reliably affects behaviour in a variety of visual perceptual tasks ( Cicmil and Krug , 2015 ) . If the microstimulation signal could affect final motor output via a neural path independent from visually evoked signals , it would be expected that microstimulation would be detectable and/or processed by the brain in some qualitatively different way . But a recent study that modelled in detail the relationship between microstimulation , perceptual choices and decision confidence found that their data did not support the view that microstimulation induces a change in neural activity that is processed qualitatively differently from visually evoked activity ( Fetsch et al . , 2014 ) , and our ∆dx control shows that the animal cannot use an obvious difference in the stimulus percept to discard some of the visual information provided . Electrophysiological recordings from stimulus-relevant neurons during visual perceptual decision-making are necessary to directly show whether expected reward modulates the neuronal firing rates and tuning functions at a specific visual cortical site in the manner predicted by an increase in stimulus sensitivity under large reward . Yet , the unique contribution of a microstimulation intervention to this question is to show that reward modulations at a level prior to the integration stage have a causal effect on the perceptual decision , rather than representing a top-down echo of decision-formation that might not be causally relevant to the perceptual choice itself . The form of the bounded accumulation model considered here is a simplified case of the race model , in which evidence for each choice is integrated separately by two decision variables ( Usher and McClelland , 2001 ) . We used the assumption that the accumulator inputs are perfectly negatively correlated , which corresponds to the action of one accumulator only ( Figure 5A , B ) ( Palmer et al . , 2005; Ratcliff and McKoon , 2008 ) . To avoid unnecessary assumptions and over-fitting , we chose the simplest model framework that could explain our findings . The microstimulation signal was modelled as additional sensory evidence towards the cylinder rotation direction preferences of the microstimulated neurons . The null choice bias , whereby animals apply an overall bias towards the non-preferred direction of the stimulated neurons in order to match overall reward proportions for different choices ( Salzman et al . , 1992 ) , was modelled as an offset to the overall psychometric function ( Equation 3a , ‘Materials and methods’ ) . The parameters of the bounded accumulation model are commonly used to explain both choice and reaction time . However , bounded integration underlies perceptual decisions even when stimulus viewing duration is dictated by the environment ( Kiani et al . , 2008 ) , and the model can be fit usefully in these circumstances ( Fetsch et al . , 2014 ) . Although we employed a fixed viewing duration task , differential effects of reward on parameter k , coding for stimulus scaling sensitivity , and parameter B , coding for distance-to-bound , could be differentiated in our model due to the insertion of the microstimulation signal in visual area V5/MT ( Figure 4 ) . The lack of reaction-time measurements in the present experiment leaves open the possibility that changes to parameter B may also arise from an overall increase in the gain of accumulation of sensory evidence . The consensus from electrophysiological studies ( Rorie et al . , 2010 ) , human neuroimaging ( Summerfield and Koechlin , 2010; Mulder et al . , 2012 ) , and psychophysical studies ( Diederich and Busemeyer , 2006; Diederich , 2008 ) is that reward effects on the DV accumulation process occur through changes to the distance-to-bound; we therefore follow this interpretation . A microstimulation-reaction time task , or detailed measurement of neuronal firing rates in sensorimotor structures such as LIP that represent the integration stage , would be necessary to resolve this issue . Regardless , the specific interpretation of parameter B's effects does not affect the main conclusions of this study . We have argued that the most parsimonious explanation for our observations is that larger available reward increases the sensitivity of visually evoked sensory representations about the visual stimulus . This bears strong similarity to the effect on neuronal responses when attention is allocated to a specific spatial location or to a PREF stimulus feature ( Treue and Maunsell , 1996 , 1999; McAdams and Maunsell , 2000; Corbetta and Shulman , 2002; Saenz et al . , 2002 ) . Indeed , manipulations of reward and attention are often difficult to separate ( Maunsell , 2004 ) . However , these effects cannot trivially explain our findings . With regard to feature attention , our trial-by-trial reward schedule was balanced so that both perceptual interpretations of the cylinder were equally rewarded , if correctly identified . Feature attention was therefore not directed in favour of one perceptual interpretation of cylinder rotation direction . Our experimental task always directed spatial attention to the location of the visual stimulus , because the animal responded to the visual stimulus in every trial to obtain a reward , whether small or large . Thus , results from spatial attention studies , which measure differences in neuronal activity when attention is directed into and away from their RF ( Treue and Maunsell , 1996 , 1999; McAdams and Maunsell , 2000 ) , cannot directly be applied to our results . The definition of spatial attention could be extended to encompass gradations of attention applied to a particular spatial location according to expected reward size , which may be manipulated by experiments such as ours . In that case , the present study would be the first to provide data demonstrating such an effect during perceptual decision-making , and to integrate such modulations of sensory representations coherently with reward manipulation and decision model theory ( Smith and Ratcliff , 2009 ) . In summary , we found that the effect of V5/MT electrical microstimulation on perceptual decisions was less effective when available reward was larger . In the context of a bounded evidence-accumulation model , this suggests that reward modulates sensory representations during perceptual decision-making , in addition to altering the integration of sensory evidence into the DV in sensorimotor structures ( Gold and Shadlen , 2007 ) . Further electrophysiological research should record in area LIP to show directly how the electrically introduced signal in V5/MT is integrated into the perceptual decision under different reward states , and should record in visual area V5/MT to show how reward modulates the signals of sensory evidence in visual cortex during perceptual decisions .
Microstimulation experiments were conducted in one cortical hemisphere in two adult male rhesus macaque monkeys ( Macaca mulatta ) . Prior to the experiments , each monkey was surgically implanted under general anaesthesia with a head-holding device and a recording chamber placed over a craniotomy above the occipital lobe . For monkey Fle , scleral magnetic search coils were implanted in both eyes under general anaesthesia , to monitor eye position and vergence . Monkey Ica had one scleral magnetic search coil only . Monkeys were trained to fixate on a binocularly presented marker and to perform the visual task that required the animals to discriminate the motion and binocular disparity information in a rotating SFM cylinder stimulus ( Figure 2 ) . Experiments were conducted at two locations: monkey Fle was tested at Oxford University , UK; monkey Ica was tested at the National Institutes of Health , Bethesda , USA ( henceforth referred to as Oxford and NIH , respectively ) . Except where specifically described , the protocols were identical in the two locations . At Oxford , all procedures complied with United Kingdom Home Office regulations on animal experimentation . At NIH , all procedures complied with US Public Health Service policy on the humane care and use of laboratory animals , and all protocols were approved by the National Eye Institute Animal Care and Use Committee . The visual stimulus for the discrimination task was a rotating SFM cylinder made up of two transparent surfaces of random dots moving in opposite directions ( random dot kinematogram ) ( Treue et al . , 1991 ) ( Video 1 ) . The dots had a sinusoidal velocity profile as would be expected in an orthographic projection of a three-dimensional rotating cylinder . When the two dot surfaces are presented at the same binocular disparity , the cylinder rotation direction is bistable and over time will undergo spontaneous fluctuations in the perceived direction of rotation ( Wallach and O'Connell , 1953; Ullman , 1979 ) . Adding opposite horizontal disparity separates the front and back walls of the cylinder in depth , thereby disambiguating the direction of rotation . The disparity of each dot was scaled sinusoidally to match the velocity profile , with the maximal disparity difference between the two dot surfaces at the central axis of the cylinder . Equal but opposite disparities were added to the front and back surfaces of the cylinder , so that its principal axis remained in the plane of fixation . Below , we use the term ‘disparity’ to refer to this maximal difference in disparity between dots moving in opposite directions . We used the following convention: when the front wall of dots moved to the left with respect to cylinder orientation axis , this corresponded to a CW rotating cylinder as viewed from above; when the front wall moved to the right , this corresponded to a CCW rotating cylinder . To allow pooling across microstimulation sites , quantitative behavioural analysis was carried out with reference to PREF choices . For the pooled data , positive disparities refer to the stimulus rotation direction for which the site was selective . Cylinder stimuli consisted of 50% black dots and 50% white dots presented at maximum contrast ( 99% of full contrast ) on a mid-grey background at a dot density of 25% . Half the dots ( randomly selected ) moved in one direction and the remaining dots moved in the opposite direction . When a dot reached the edge of the cylinder , it reversed direction . In Oxford , stimuli were displayed binocularly using a Wheatstone stereoscope configuration with two monitors ( Eizo FlexScan 78 , Eizo , Bracknell , UK ) . A pair of mirrors positioned in front of the monkey reflected the images from each of the two monitor screens separately into the left and right eyes . Frame rate was 72 Hz . Monitor screens were positioned 84 cm from the monkey , and covered approximately 21° × 17° of the visual field . Mean luminance was 42 cd/m2 and dot diameter was 0 . 2° . The dot refresh rate was 2% , that is , on each video frame , 2% of the dots were relocated to a randomly chosen location on the cylinder surface , so that dots had a mean lifetime of 615 ms in each location . At NIH , stimuli were displayed binocularly using two DLP projectors ( Projection Design evo2sx+ , projectiondesign®/Barco , Inc . , Xenia , OH ) with polarizing filters . The image was projected onto a polarisation-preserving screen ( Filmscreen 150 , Stewart Filmscreen , Amelia , OH ) . Frame rate was 60 Hz . The screen was 112 cm away from the monkey , and covered approximately 41° × 32° . Mean luminance of the display viewed through polarized filters was 17 . 5 cd/m2 and dot diameter was 0 . 18° . Dot refresh rate was 1% . The psychophysical task was a two-alternative forced choice discrimination of the direction of rotation of the SFM cylinder . Stimulus disparities were matched to the psychophysical threshold of the animal in order to obtain a full psychometric function . To initiate a trial , the animal had to acquire the fixation point . The visual stimulus was then presented for 2000 ms ( Figure 2 ) . At stimulus offset , the fixation point also disappeared and two choice targets corresponding to the two possible directions of cylinder rotation appeared . In the Oxford , targets were located to the left and right side of the fixation point . In the NIH set-up , targets were also located opposite one another on either side of the fixation point but on an axis normal to the cylinder orientation . The animals indicated their perceptual decision by making a saccadic eye-movement to one of the two targets . A saccade to the correct target resulted in a fluid reward . If the choice was incorrect , the animal received no reward and there was a short time-out before the start of the next trial . For ambiguous cylinders ( zero-disparity trials ) , 50% of the trials were rewarded at random . If the animal broke fixation during stimulus presentation , no reward was given . Animals worked on the task to gain fluid rewards to meet their daily requirements . Reward size available for a correct choice on each trial depended upon the number of immediately preceding consecutive correct responses that the animal had made , increasing in two steps up to a maximum ( Figure 1 ) . Choices in zero-disparity trials were rewarded 50% of the time at random , and for the purposes of calculating the reward sequence they were discounted . For monkey Ica , reward size was 0 . 08 ml on the first and second consecutive correct choices after an error , 0 . 12 ml for the third consecutive correct choice , and 0 . 2 ml on the fourth and all subsequent consecutive correct trials . For monkey Fle , reward size was 1/3 of maximum for the first correct choice , 2/3 of maximum for the second , and reached maximum size ( usually 0 . 18 ml ) for the third and all subsequent consecutive correct choices . Thus reward size increased more quickly for Fle than Ica as a function of consecutive correct responses . For both animals , we categorized trials into two conditions: maximal ( large ) reward size and sub-maximal ( small ) reward size , where for both animals the average size of the two sub-maximal reward sizes was half that of the maximal reward size . Animals were familiar with their respective reward schedule because it was used throughout all training and recording sessions with the discrimination task; it was not introduced only for the microstimulation experiments . Recording from MU neuronal clusters was carried out to characterize and select the cortical microstimulation sites . Single tungsten microelectrodes coated with polyimide tubing were used ( 0 . 1–0 . 3 MΩ impedance at 1 kHz; MicroProbe , Inc . , Gaithersburg , MD ) . A hydraulic microdrive mounted on the recording cylinder advanced the electrode through a guide tube . Electrical signals from the electrode were filtered , amplified and displayed through visual and audio monitors , and stored to computer disk . Online classification of activity and pre-processing steps were done with the DataWave Discovery system ( DataWave Technologies , Loveland , CO ) at Oxford and Spike 2 ( Cambridge Electronic Design Ltd . , Cambridge , UK ) data acquisition program at NIH . Area V5/MT was approached posteriorly through the recording chamber in incremental steps of about 100 µm . Area V5/MT was identified by established physiological criteria: ( 1 ) the approach through the grey/white matter pattern comprising the striate cortex , lunate sulcus and the posterior bank of the superior temporal sulcus ( STS ) ( Zeki , 1974 ) ; ( 2 ) by neurons' characteristic direction selectivity and binocular disparity selectivity and its clustering in V5/MT ( Van Essen et al . , 1981; Maunsell and Van Essen , 1983; Albright and Desimone , 1987; Bradley et al . , 1995 , 1998; DeAngelis et al . , 1998; DeAngelis and Newsome , 1999 ) ; ( 3 ) the known relationship between the size and eccentricity of V5/MT receptive fields ( Desimone and Ungerleider , 1986 ) ; and ( 4 ) from penetration to penetration , the relationship between RF positions and known topography of V5/MT ( Albright and Desimone , 1987 ) . Monkeys maintained fixation on a central fixation point while direction selectivity was established for an MU cluster at a particular electrode location , using kinetic dots moving coherently in different directions . The direction evoking the greatest response on average was assigned to be the PREF motion direction for the MU cluster . The RF boundaries were then mapped using a patch of dots moving in the PREF motion direction . After quantitative confirmation of direction tuning within these boundaries , the binocular disparity preference of the MU cluster was measured by presenting a plane of moving dots in the PREF direction at different binocular disparities around the fixation plane . Only MU sites with discernible motion and disparity tuning were further explored . The directions of motion of the two transparent planes of dots that made up the cylinder were aligned with the PREF and the opposite ( NULL ) directions of the MU cluster and the MU cylinder preference was calculated online over a range of disparities . When the direction- and cylinder disparity-preference of MU cluster activity were found to remain constant over at least 300 µm of cortex , the electrode was retracted to the middle of that stretch . At the proposed stimulation site , the MU RF size , direction selectivity and cylinder preference were quantitatively re-measured . The cylinder stimulus position , size , orientation , and dot speed were carefully matched to the preferences of the MU at the stimulation site . The monkey first performed 50 to 140 trials of the psychophysical task without microstimulation , prior to the microstimulation experiment , in order to allow the selection of a range of stimulus disparities near threshold at that site . The total number of cortical sites at which microstimulation was applied was 20 in monkey Fle and 28 in monkey Ica . This number of biological replicates ( different stimulation sites ) was chosen to be as close as possible to the number of microstimulation sites used in previous studies of this cortical area ( V5/MT; in particular , DeAngelis et al . 1998 ) . Stimulation sites were selected freshly on each day of recording and the choice of sites for stimulation was independently determined on each occasion of performing the experiment , by reference to the selection criteria presented above . Electrical microstimulation was applied during the 2000 ms of stimulus presentation on half of the trials , pseudo-randomly interleaved with non-microstimulated trials . Stimulation consisted of 20 µA biphasic pulses of 200 µs cathodal stimulation followed by 200 µs anodal stimulation , delivered at 200 Hz . Only cortical microstimulation sites with at least 10 microstimulation trials and 10 non-stimulation trials at each of at least 5 different stimulus disparities were included in subsequent analyses . All included sites showed significant tuning to disparity in the cylinder stimulus ( one-way ANOVA , p < 0 . 05 ) . For each microstimulation site , we plotted the proportion of choices made by the monkey towards the PREF rotation direction , at each stimulus disparity , separately for microstimulated and non-microstimulated trials . Behavioural data were fitted for each site with a two-mean cumulative Gaussian distribution: ( 1a ) PPREF ( C ) =1/2 ( 1+erf ( ( C− ( µ0 + βµ1 ) ) /σ2 ) ) , where C is the cylinder disparity ( positive in PREF direction ) ; µi is the mean of the distribution; σ is the s . d . ; erf is the error function; β = 1 for microstimulated trials and β = 0 for non-microstimulated trials . PPREF corresponds to the probability of making a decision in the PREF direction of the microstimulated site . To test whether microstimulation significantly shifted the psychometric function , the two-mean model was compared to a model in which mean was not allowed to vary by microstimulation ( Krug et al . , 2013 ) : ( 1b ) PPREF ( C ) =1/2 ( 1+erf ( ( C−µ0 ) /σ2 ) ) . A χ2 likelihood-ratio test was used to ascertain whether the full model fitted the data significantly better than the nested model ( p < 0 . 05; 1 degree of freedom as models differed by 1 parameter ) . Maximum likelihood estimate model fitting was performed with MATLAB's fminsearch function and was repeated multiple times from a wide range of starting parameter values , in order that results would not be driven by local minima . The microstimulation effect is quantified by taking the estimated value of parameter µ1; this is equivalent to a horizontal shift in the psychometric curve . The size of the electrical microstimulation effect can thus be expressed in terms of the size of binocular disparity that would need to be added to the stimulus to shift perceptual behaviour to the same extent ( Krug et al . , 2013 ) . A cumulative Gaussian function was used to evaluate the interaction between reward , task performance and microstimulation for psychometric data pooled across multiple microstimulation sites . Prior to pooling , we normalized stimulus disparity values at each site by dividing them by the maximum disparity value of the site . This ensured that the maximum normalized disparity was 1 and minimum was −1 , with the other disparity values lying in between . In the full model both the mean µ and standard deviation ( s . d . ) σ are allowed to vary by microstimulation condition and reward condition , ( 2a ) PPREF ( C ) =1/2 ( 1+erf ( ( C− ( µ0+βµ1+αµ2+αβµ3 ) ) / ( σ0+ασ1+βσ2 ) 2 ) ) , where C is the cylinder disparity ( positive in PREF direction ) ; µi represents the mean; σi represents s . d . , α = 1 in trials were available reward is large , otherwise α = 0; β = 1 in trials where additional stimulation is introduced , otherwise β = 0; erf is the error function . PPREF corresponds to the probability of making a decision in the PREF direction of the additional stimulation introduced . To test whether reward affected task performance , this full model was compared with a nested model in which s . d . σ ( a measure of performance threshold ) was not allowed to vary by reward condition: ( 2b ) PPREF ( C ) =1/2 ( 1+erf ( ( C− ( µ0+βµ1+αµ2+αβµ3 ) ) / ( σ0+βσ2 ) 2 ) ) . A χ2 likelihood-ratio test was used to ascertain whether the full model fitted the data significantly better than the nested model ( p < 0 . 05; 1 degree of freedom as models differed by 1 parameter ) . Similarly , the effect of reward on microstimulation was evaluated with a χ2 likelihood-ratio test to compare the full model with a nested model in which the change in mean by microstimulation , represented by βµi , was not allowed to vary by reward condition: ( 2c ) PPREF ( C ) =1/2 ( 1+erf ( ( C− ( µ0+βµ1+αµ2 ) ) / ( σ0+ασ1+βσ2 ) 2 ) ) . Model fitting was performed as described in the previous section . We report uncorrected p values; however , all statistically significant results survive Bonferroni correction for multiple comparisons across the two animals ( where relevant ) . A logistic regression model representing the effect of cylinder stimulus disparity , microstimulation and trial reward condition on the psychometric functions was derived from the simple , one-dimensional drift-diffusion model of perceptual decision-making ( Palmer et al . , 2005; Gold and Shadlen , 2007 ) ( see also Figure 1 ) . The full bounded evidence-accumulation model explains perceptual choices in the PREF direction ( PPREF ) according to the following equation: ( 3a ) PPREF ( C ) = 1 / ( 1 + e^ ( − ( ( B0 + αB1 ) ( ( k0+αk1+βk2 ) C+β ) +x0+αx1 ) ) ) , where C is the ( normalized ) cylinder disparity ( positive in PREF direction ) ; α = 1 for large reward trials and α = 0 for small reward trials; β = 1 for microstimulated trials and β = 0 for non-microstimulated trials . The relation between stimulus strength C and the aspect of DV drift that is driven by visually evoked sensory evidence is represented by parameter ki ( see also Figure 4A , B ) . The distance to the decision bounds from the starting point of the drift-diffusion process and the overall gain of the drift rate are represented by parameter Bi . Behavioural null bias in non-stimulated trials ( Salzman et al . , 1990 , 1992 ) is represented by parameter xi . To test whether reward affected parameter ki , this full model was compared with a nested model in which ki was not allowed to vary by reward condition: ( 3b ) PPREF ( C ) = 1 / ( 1 + e^ ( − ( ( B0 + αB1 ) ( ( k0+βk2 ) C+β ) +x0+αx1 ) ) ) . A χ2 likelihood-ratio test was used to ascertain whether the full model fitted the data significantly better than the nested model ( p < 0 . 05; 1 degree of freedom as models differed by 1 parameter ) . Similarly , the effect of reward on parameter Bi was evaluated with a χ2 likelihood-ratio test to compare the full model with a nested model in which Bi was not allowed to vary by reward: ( 3c ) PPREF ( C ) = 1 / ( 1 + e^ ( − ( ( B0 ) ( ( k0+αk1+βk2 ) C+β ) +x0+αx1 ) ) ) . Model fitting was performed as described in previous sections . We report uncorrected p values; however , all statistically significant results survive Bonferroni correction for multiple comparisons across the two animals ( where relevant ) . Quantile–quantile comparisons between the fit of this decision model and the cumulative Gaussian model described in the previous section indicate a reasonable correspondence between model fits ( Figure 8—figure supplement 3 ) . At each microstimulation site , trials were divided according to whether the available reward size for a correct choice in the trial was large or small . The size of the horizontal shift of the psychometric function induced by electrical microstimulation was re-calculated as described in the previous section , separately for large and small reward trials . To normalise the effect of microstimulation across sites with different associated threshold due to stimulus eccentricity ( see Figure 8 ) , the horizontal shift of the psychometric function in each reward condition was divided by the psychometric threshold for that reward condition that is , the standard deviation ( s . d . ) of the fitted cumulative Gaussian ( Equation 1a; Uka and DeAngelis 2006 ) . A Wilcoxon sign-rank test across sites ( p < 0 . 05 , two-sided ) was used to ascertain whether there was an overall significant difference in the size of the microstimulation shift between the large and small reward conditions . A non-parametric test was used because although the normalized shifts were normally distributed ( Lilliefors test , p > 0 . 05 ) , the underlying distribution of raw shift values was not normally distributed ( Lilliefors test , p = 0 . 001 ) . We report uncorrected p values; however , all statistically significant results survive Bonferroni correction for multiple comparisons across the two animals ( where relevant ) . We performed a control experiment on one monkey , Ica , in which we used a visual stimulus to mimic the psychophysical effect of microstimulation . We matched stimulus and task parameters to the microstimulation experiment but no electrical microstimulation took place . Instead , we added a signal of +0 . 005° disparity , termed ‘∆dx’ , to the stimulus in half of the trials , pseudo-randomly selected ( see Salzman et al . 1992; Fetsch et al . 2014 for a similar manipulation in a visual motion task ) . This extra signal was not included in the determination of the correct response , that is , we rewarded the monkey's perceptual choices according to the stimulus disparity without the addition of ∆dx . The ∆dx trials were overtly cued by a change in the colour of the cylinder stimulus , from 100% white dots to 100% black dots ( Figure 7A , B ) . This allowed us to explore the behavioural strategy adopted by the animal when trials with and without an additional , non-rewarded visual disparity signal are clearly cued . We kept reward schedules the same . Therefore , for maximal reward accumulation , the monkey's optimal strategy would to bias its perceptual reports away from the fake signal cylinder disparity on the trials that were overtly cued as ∆dx . Over 5 days , 15 experimental sessions of psychophysical data was collected , with all experimental parameters remaining constant . There were approximately 1000 to 2000 trials in each experimental session . By contrast , there were on average 500 trials in each experimental block at the electrical microstimulation sites . The ∆dx control experiment comprised 12 , 149 trials in total , whilst the Fle and Ica electrical microstimulation experiments comprised 9 , 411 and 10 , 592 trials in total , respectively , over all microstimulation sites . Since our objective was to ascertain whether the animal learns to treat the ∆dx trials differently from non-∆dx trials , we kept the direction of the ∆dx signal the same throughout the control experiment . If the animal fails to adjust its criterion on a reward-size basis in this situation when it is clearly apparent when the additional visual disparity signal is added , it seems very unlikely that they could do so when the signal introduced by microstimulation varied between experiments ( as it did for the real microstimulation ) . Performance accuracy ( proportion of correct choices ) was calculated over a sliding time window 30 trials wide . To estimate the amount of performance fluctuation over the microstimulation session at a given site , the area between the smoothed performance curve and the horizontal line indicating the mean overall performance over that session was calculated using trapezoidal numerical integration implemented by MATLAB's trapz function ( see Figure 8—figure supplement 1A ) . Since different sites had microstimulation sessions of different lengths , this area was normalised across sites by division by the total number of trials in the microstimulation session at each site . ‘Good’ and ‘bad’ performance epochs were identified as stretches of smoothed performance that were above or below mean site performance , respectively , for at least the length of the smoothing time window ( i . e . at least 30 trials ) . To evaluate whether the reward effect on microstimulation remained within good and bad epochs considered individually , trials from good and bad epochs were pooled across sites and animals and fit with cumulative Gaussians to test whether there was significant interaction between reward and microstimulation shift ( χ2-test of nested models Equations 2a , c , p < 0 . 05 , as described in previous sections ) . | Identifying how an object is moving in three-dimensional ( 3D ) space depends upon a brain region known as V5/MT . The neurons that make up area V5/MT form groups that each have a ‘preference’ for a particular direction of movement and a particular 3D depth . If a group of neurons detects its preferred direction of movement and 3D depth , it will become highly active . The brain can assess which groups of neurons are active , in a process known as integration . This information can then be used to work out the object's movement in space . The process of integration can be influenced by whether a rewarding outcome is expected to result from identifying the 3D movement correctly . This allows the brain to increase its likelihood of success in situations where a large reward is on offer . Until now , it was thought that the activity in area V5/MT , which takes place before integration , was not affected by the likelihood of receiving a reward . As well as being ‘naturally’ stimulated by moving objects , the V5/MT neurons can also be ‘artificially’ activated by a technique called microstimulation , which uses a tiny electrode to electrically stimulate groups of neurons . Microstimulation can bias visual perception towards the movement and 3D depth ‘preference’ of the artificially activated neurons . If the V5/MT neurons do receive information about potential rewards from other areas of the brain , we would expect rewards to affect naturally and artificially stimulated neural activity in different ways . On the other hand , if the V5/MT neurons do not receive any information about reward , then it will not matter whether their activity is natural or artificial; the signal that they produce will be the same . Cicmil et al . gave two monkeys a task in which they could receive rewards for correctly identifying a three-dimensional cylinder's direction of rotation , and applied microstimulation to specific groups of V5/MT neurons on some of the trials . When a larger reward was available , microstimulation was less able to bias the monkeys' choices about the rotation direction of the 3D cylinders . Overall , Cicmil et al . 's results suggest that the V5/MT neurons are able to incorporate information about reward , before integration occurs . The next step will be to record the activity of area V5/MT to investigate exactly how this happens . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Reward modulates the effect of visual cortical microstimulation on perceptual decisions |
Myelination is a biosynthetically demanding process in which mTORC1 , the gatekeeper of anabolism , occupies a privileged regulatory position . We have shown previously that loss of mTORC1 function in Schwann cells ( SCs ) hampers myelination . Here , we genetically disrupted key inhibitory components upstream of mTORC1 , TSC1 or PTEN , in mouse SC development , adult homeostasis , and nerve injury . Surprisingly , the resulting mTORC1 hyperactivity led to markedly delayed onset of both developmental myelination and remyelination after injury . However , if mTORC1 was hyperactivated after myelination onset , radial hypermyelination was observed . At early developmental stages , physiologically high PI3K-Akt-mTORC1 signaling suppresses expression of Krox20 ( Egr2 ) , the master regulator of PNS myelination . This effect is mediated by S6K and contributes to control mechanisms that keep SCs in a not-fully differentiated state to ensure proper timing of myelination initiation . An ensuing decline in mTORC1 activity is crucial to allow myelination to start , while remaining mTORC1 activity drives myelin growth .
Myelin is a lipid-rich membrane specialization of Schwann cells ( SCs ) and oligodendrocytes ( OLs ) winding around axons to accelerate the propagation of action potentials ( Hartline and Colman , 2007 ) . The geometrical properties of each myelin segment , such as its thickness and length , are essential to achieve fast and accurate conduction along myelinated fibers ( Waxman , 1980 ) . Conversely , myelin geometry is deranged in numerous diseases of the nervous system , including neuropathies and multiple sclerosis . To produce myelin segments with correct thickness and length , the plasma membrane of myelinating cells undergoes a remarkable , highly regulated expansion . It was estimated that the SC membrane has to expand several thousand fold during myelination ( Chrast et al . , 2011; Webster , 1971 ) and that OLs produce myelin equal to three times their own weight every day ( Norton and Poduslo , 1973 ) . Consequently , myelination is an anabolically demanding task . The mechanistic target-of-rapamycin ( mTOR ) , a serine-threonine protein kinase , is the core component of mTOR complex 1 ( mTORC1 ) and mTOR complex 2 ( mTORC2 ) . Among them , mTORC1 is the gatekeeper of multiple anabolic reactions , including mRNA translation and biosynthesis of lipids , purines , and pyrimidines ( Ben-Sahra et al . , 2013; Ben-Sahra et al . , 2016; Porstmann et al . , 2008; Thoreen et al . , 2012 ) . The mTORC1 pathway is tightly connected to PI3K-Akt signaling downstream of growth factors . Upon growth factor stimulation , Akt activates mTORC1 by phosphorylating inhibitors of mTORC1 , the TSC complex and PRAS40 ( Inoki et al . , 2002; Manning et al . , 2002; Sancak et al . , 2007; Vander Haar et al . , 2007 ) . The TSC complex is composed of the subunits TSC1 , TSC2 , and TBC1D7 ( Dibble et al . , 2012 ) , and functions as a GTPase-activating unit towards the small GTPase Rheb , a potent activator of mTORC1 when loaded with GTP ( Tee et al . , 2003 ) . Accordingly , disruption of the TSC complex causes mTORC1 hyperactivation in a wide range of tissues and cell types ( Byles et al . , 2013; Castets et al . , 2013; Kwiatkowski et al . , 2002 ) . Consistent with the intense anabolic challenge posed by myelination , the PI3K-Akt-mTORC1 axis has emerged as a fundamental player in PNS and CNS myelination ( Taveggia , 2016; Wood et al . , 2013 ) . Genetic disruption of mTORC1 in SCs or OLs impaired myelination , demonstrating that mTORC1 function is required for this process ( Bercury et al . , 2014; Lebrun-Julien et al . , 2014; Norrmén et al . , 2014; Sherman et al . , 2012; Wahl et al . , 2014; Zou et al . , 2014 ) . However , studies aimed at examining the consequences of increased mTORC1 activity , predicted to augment myelin growth , have yielded conflicting results . Overexpression of constitutively active Akt or hyperactivation of the PI3K-Akt pathway by deletion of PTEN in OLs caused hypermyelination ( Flores et al . , 2008; Goebbels et al . , 2010 ) , while deleting TSC1 was detrimental to OL myelination ( Jiang et al . , 2016; Lebrun-Julien et al . , 2014 ) . In SCs , hyperactivation of the PI3K-Akt pathway with various approaches did not lead to univocal results ( Cotter et al . , 2010; Domènech-Estévez et al . , 2016; Flores et al . , 2008; Goebbels et al . , 2012 ) . Thus , we aimed here at elucidating the functional roles of mTORC1 activation in SCs during various steps of development , in homeostasis , and after injury . Based on our results and reconciling also previous reports , we propose a model suggesting that mTORC1 signaling exerts multiple distinct roles at different stages of SC differentiation .
Using a loss-of-function approach , we have previously discovered that mTORC1 – but not mTORC2 – promotes myelin growth in the PNS ( Norrmén et al . , 2014 ) . To further define the role of mTORC1 in myelination , we have now pursued the converse approach , hyperactivating mTORC1 by conditional ablation of TSC1 , a strategy that destabilizes the TSC complex ( Dibble et al . , 2012 ) . SC-specific TSC1 mutants were generated by crossing mice harboring a floxed allele of Tsc1 ( Kwiatkowski et al . , 2002 ) with mice expressing a Cre transgene under control of regulatory sequences of the Dhh gene ( Jaegle et al . , 2003 ) ( DhhCre:Tsc1KO ) . We first confirmed , by western blot analysis of postnatal day 5 ( P5 ) sciatic nerves , that TSC1 was successfully depleted ( Figure 1a , Figure 1—figure supplement 1a ) . We then assessed mTORC1 activity by analyzing the phosphorylation levels of its downstream effectors . As expected , phosphorylation of two well-established mTORC1 targets , S6K and 4EBP1 ( Hay and Sonenberg , 2004 ) , was increased in TSC1-mutant nerves , together with phospho-S6S235/236 levels , a target of S6K ( Figure 1a , b , Figure 1—figure supplement 1a ) . As additional evidence of mTORC1 hyperactivation , we found that cultured mutant SCs isolated from either dorsal root ganglia ( DRG ) or postnatal nerves were enlarged , consistent with the prominent role of this pathway in cell size control ( Lloyd , 2013 ) ( Figure 1c , Figure 1—figure supplement 2a , b ) . Next , we assessed the extent of myelination by electron microscopy ( EM ) . Surprisingly , P5 DhhCre:Tsc1KO nerves exhibited a strong reduction in myelinated fibers due to an arrest of most SCs at the promyelinating stage ( Figure 1d , e ) . No overt defect in radial sorting was evident , thus indicating a bona fide impairment in the onset of myelination . The percentage of myelinated fibers progressively increased with time , almost doubling by P14 . By P60 , most fibers were ultimately myelinated , although occasional promyelinating SCs were still present ( Figure 1d , e ) . Additionally , the myelinated nerve fibers were hypomyelinated , presumably as a consequence of delayed onset of myelination ( Figure 1d; for quantification , see Figure 6l ) . Impaired SC differentiation was reflected in reduced levels of myelin protein P0 , while cJun and Oct6 – both highly expressed in promyelinating SCs – were upregulated ( Figure 1—figure supplement 2c , d ) . Consistent with a failure of mutant cells to promptly differentiate , we also detected an increase in proliferating Sox10-positive SCs and , consequently , we found overall more SCs ( P3; Figure 1f–h ) . Non-mTORC1 related functions of the TSC complex have been reported ( Neuman and Henske , 2011 ) . Thus , we assessed whether the phenotype of DhhCre:Tsc1KO mice was genuinely due to high mTORC1 activity by treating mice with the mTORC1-inhibiting drug rapamycin . Rapamycin was administered at P3 and P4 and the mice analyzed at P5 . Despite the short course of treatment , we found a robust increase in myelinated fibers in rapamycin-treated , compared to vehicle-only treated , mutant nerves ( Figure 1i , j ) . The morphological rescue was paralleled by a decrease in cJun and an increase in P0 , roughly back to the levels of control nerves , together with the expected suppression of S6K phosphorylation at the mTORC1-sensitive site T389 ( Figure 1k , Figure 1—figure supplement 1b ) . In line with the in vivo results , we found that in vitro myelination of TSC1 mutant-derived DRG explant cultures was strongly defective , but could be remarkably improved by acute treatment with rapamycin , consistent with a PNS-autonomous origin of the in vivo rescue ( Figure 1—figure supplement 2e , f ) . Collectively , our data show that high mTORC1 activity following deletion of TSC1 in SCs has , paradoxically , a detrimental effect on PNS myelination by delaying the transition from promyelinating to myelinating SCs . The PI3K-Akt pathway is a key upstream driver of mTORC1 and , in turn , mTORC1 activation dampens the PI3K-Akt pathway through multiple inhibitory feedback loops ( Efeyan and Sabatini , 2010 ) . To explain the discrepancy between the effects of high mTORC1 signaling in SCs lacking TSC1 and the positive role generally attributed to PI3K-Akt signaling in PNS myelination ( Taveggia , 2016 ) , we reasoned that overactive mTORC1 may suppress the PI3K-Akt pathway , and consequently SC differentiation , via the aforementioned feedback loops . To test this hypothesis , we assessed phosphorylation of Akt and the upstream receptor ErbB2 in DhhCre:Tsc1KO nerves . Since the MAPK pathway can also be subjected to feedback inhibition by mTORC1 ( Carracedo et al . , 2008 ) , and considering the crucial functions of this pathway in SC biology ( Ishii et al . , 2013; Newbern et al . , 2011; Sheean et al . , 2014 ) , we also examined the phosphorylation status of Erk1/2 . No major changes in ErbB2 or Erk1/2 phosphorylation could be detected ( Figure 2a , Figure 2—figure supplement 1a ) . By contrast , we observed a strong reduction in Akt phosphorylation at both T308 and S473 , together with a global decrease of Akt substrates phosphorylation ( Figure 2a , Figure 2—figure supplement 2a and Figure 2—figure supplement 1a ) . Moreover , pharmacological inhibition of the PI3K-Akt axis , but not the Mek-Erk axis , abolished S6 phosphorylation upon neureugulin-1 stimulation of primary SCs , although some minor contribution of the Mek-Erk axis to mTORC1 activation was also apparent in this experimental setting ( Figure 2—figure supplement 2b ) . Together , these results demonstrate that in SCs mTORC1 is mainly activated by the PI3K-Akt pathway , and that , in turn , high mTORC1 activity suppresses Akt activation without significantly affecting the MAPK pathway or activation of ErbB2-ErbB3 receptors . To address whether the observed Akt suppression is responsible for the defective myelination due to TSC1 deletion in SCs , we sought to restore Akt activity in TSC1 mutants by simultaneously deleting the phospholipid phosphatase PTEN which limits activation of the PI3K-Akt pathway ( Song et al . , 2012 ) ( Figure 2b ) . Given the known tumor suppressor role of both TSC1 and PTEN ( Inoki et al . , 2005 ) , we used the more restricted MpzCre mouse line ( Feltri et al . , 1999 ) to generate suitable single or double mutants , referred to as MpzCre:Tsc1KO , MpzCre:PtenKO , and MpzCre:Tsc1KO:PtenKO . Both proteins could be successfully depleted ( Figure 2—figure supplement 2c , d ) . As anticipated , co-deletion of TSC1 and PTEN released the previously observed suppression on the PI3K-Akt pathway upon single deletion of TSC1 , indicated by restored phosphorylation of Akt at T308 ( Figure 2c , Figure 2—figure supplement 1b ) . We also noticed , based on the resulting levels of phospho-S6S235/236 , that MpzCre:Tsc1KO:PtenKO nerves contained higher mTORC1 activity compared to MpzCre:Tsc1KO ( Figure 2c , Figure 2—figure supplement 1b ) . To explain this finding , we hypothesized that the high Akt activity in MpzCre:Tsc1KO:PtenKO may further increase mTORC1 activity via phosphorylation and subsequent inhibition of PRAS40 , an inhibitory protein associated with mTORC1 ( Sancak et al . , 2007 ) ( Figure 2b ) . In agreement , MpzCre:Tsc1KO:PtenKO nerves displayed increased PRAS40 phosphorylation at the Akt-dependent phospho-site T246 compared to MpzCre:Tsc1KO nerves ( Figure 2c , Figure 2—figure supplement 1b ) . On the morphological level , P5 MpzCre:Tsc1KO:PtenKO nerves looked profoundly larger than those of controls or single mutants ( Figure 2d ) . Nevertheless , EM analysis revealed no detectable improvement in SC differentiation compared to single mutants , but rather an aggravation of the phenotype with SCs in contact with sorted axons arrested at the promyelinating stage and virtually no myelinated fibers ( Figure 2e , f ) . Unexpectedly , closer inspection of P5 MpzCre:PtenKO mice revealed also a substantial decrease in myelinated fibers compared to controls due to SCs arrested at the promyelinating stage . Consistent with the morphological findings , cJun levels were comparably increased in both single knockout nerves compared to controls , and significantly more increased in double knockout nerves compared to single mutants ( Figure 2g , Figure 2—figure supplement 1c ) . These findings indicated that reduced Akt signaling due to mTORC1 hyperactivation in MpzCre:Tsc1KO is likely not responsible for the observed impaired SC differentiation . On the other hand , the data also suggested that hyperactivation of the PI3K-Akt pathway early in development is per se detrimental to SC differentiation , potentially via mTORC1 . In line with these conclusions , we found that SC-specific expression of constitutively active ( myristoylated ) Akt1 , Akt2 , or Akt3 in DRG-explant cultures inhibited myelination ( Figure 2h , Figure 2—figure supplement 2e ) . To corroborate our findings and to determine the aforementioned potential role of high mTORC1 signaling in PTEN mutants , we generated PTEN knockout mice using DhhCre-mediated recombination ( DhhCre:PtenKO ) and treated the resulting animals with vehicle or rapamycin ( treatments at P3 and P4 , analyses at P5 ) . As observed in MpzCre:PtenKO mice previously ( Figure 2f ) , also vehicle-treated DhhCre:PtenKO animals showed a decrease in myelinated fibers compared to controls , and rapamycin treatment effectively improved the DhhCre:PtenKO phenotype ( Figure 2i , j ) . In line with the morphological rescue , rapamycin also reduced expression of Oct6 and cJun ( Figure 2k , Figure 2—figure supplement 1d ) . To elucidate the molecular basis of how hyperactive mTORC1 impairs SC differentiation , we compared the transcriptomes of P5 DhhCre:Tsc1KO and DhhCre:PtenKO nerves , both displaying high mTORC1 activity , with age-matched nerves in which mTORC1 activity has been eliminated by conditional ablation of Raptor in SCs ( DhhCre:RptorKO ) ( Norrmén et al . , 2014 ) , together with controls . Unbiased hierarchical clustering showed that the replicates within each genotype group closely clustered together , thus validating comparisons . Notably , DhhCre:Tsc1KO and DhhCre:PtenKO samples formed a distinct cluster compared to the other groups , confirming also at a global transcriptional level the previously described similarities ( Figure 3a ) . Gene ontology ( GO ) analyses of DhhCre:Tsc1KO and DhhCre:PtenKO data sets revealed a particular upregulation of transcripts associated with cell proliferation compared to controls ( Figure 3—figure supplement 1a , b ) , consistent with our previous observations ( Figure 1f , g ) and with other reports ( Goebbels et al . , 2010 ) . In addition , downregulated lipid biosynthesis-associated mRNAs were overrepresented in DhhCre:RptorKO compared to controls ( Figure 3—figure supplement 1c ) , in line with our previous findings ( Norrmén et al . , 2014 ) . Given our focus on cell differentiation , we scrutinized our datasets to identify transcription factors ( TFs ) potentially regulated by mTORC1 . To this end , we selected two subsets out of significantly changed mRNAs ( fold change >1 . 2 , FDR < 0 . 05 ) : ( 1 ) Upregulated in both DhhCre:Tsc1KO and DhhCre:PtenKO ( i . e . high mTORC1 activity ) , but downregulated in DhhCre:RptorKO ( i . e . no mTORC1 activity ) or ( 2 ) Downregulated in DhhCre:Tsc1KO and DhhCre:PtenKO , but upregulated in DhhCre:RptorKO . Applying these criteria , we identified 110 putative mTORC1-induced genes ( subset 1 ) and 131 potential mTORC1-repressed genes ( subset 2 ) , among which 7 and 16 encoded TFs , respectively ( Figure 3b , c ) . Among the mTORC1-repressed TFs emerged Krox20 ( also known as Egr2 ) , a crucial TF for the differentiation of myelinating SCs ( Svaren and Meijer , 2008; Topilko et al . , 1994 ) . Thus , we reasoned that Krox20 may link mTORC1 activity to the onset of SC myelination . First , we confirmed the RNA-sequencing results with qRT-PCRs for Krox20 and other TFs with well-established roles in PNS myelination ( cJun , Oct6 , Brn2 , Sox10 , Sox2 , Id2 ) . Only Krox20 displayed a pattern consistent with stringent mTORC1-dependent regulation , being downregulated in both DhhCre:Tsc1KO and DhhCre:PtenKO nerves and mildly , but significantly upregulated in DhhCre:RptorKO nerves at P5 and P8 ( Figure 3d , Figure 3—figure supplement 1d ) . Furthermore , protein levels of Krox20 were also increased in Raptor mutants and decreased in TSC1 mutants ( P5; Figure 3e , Figure 3—figure supplement 2a ) . Krox20 downregulation correlated with the extent of mTORC1 hyperactivation: Krox20 levels were comparably reduced in mice with single deletion of TSC1 or PTEN compared to controls and barely detectable in double knockout mice at P5 ( Figure 3f , Figure 3—figure supplement 2b ) . In further support of the critical relationship between Krox20 and the PI3K-Akt-mTORC1 axis , we found that acute inhibition of mTORC1 either with rapamycin or torin-1 , an ATP-competitive inhibitor of mTOR , strikingly increased basal Krox20 expression in cultured SCs , as did pharmacological inhibition of PI3K with LY294002 ( Figure 3g , h ) . Suppression of mTORC1 or PI3K also enhanced cyclic-AMP induced upregulation of Krox20 ( Figure 3i ) . We then examined the roles of the classical mTORC1 targets , 4EBPs and S6K , in regulating Krox20 . Overexpression of a constitutively active version of 4EBP1 ( i . e . cannot be phosphorylated by mTORC1 ) did not significantly affect Krox20 expression ( Figure 3j ) . In contrast , inhibition of S6K with LYS6K2 moderately increased basal Krox20 expression and partially rescued the defective myelination in DRG-explant cultures from DhhCre:Tsc1KO animals ( Figure 3k , l , Figure 3—figure supplement 1e ) . In conclusion , this set of data provides converging evidence that Krox20 is an mTORC1-regulated transcript , and that S6K is part of the mechanisms linking mTORC1 activity to Krox20 expression . Our results indicate that high mTORC1 activity suppresses expression of Krox20 , the main driver of SC myelination , thereby inhibiting onset of myelination . Our results so far show a peculiar inhibitory role of high mTORC1 signaling in SC differentiation . In this light , mTORC1 activity in physiological conditions would be expected to be high before onset of myelination and to decrease as SCs differentiate into myelinating SCs . Thus , we evaluated mTORC1 and Akt activities throughout mouse nerve development . Consistently , we found high activity of mTORC1 , reflected by the phosphorylation status of S6 , from embryonic day 17 . 5 ( E17 . 5 ) to P1 , while Akt activation peaked at P1 . In both cases , highest levels preceded detection of P0 and MBP ( Figure 4a , b ) , in line with recent data obtained from postnatal rats ( Heller et al . , 2014 ) . Conversely , as mTORC1 activity declined between E17 . 5 and P5 , the levels of myelin proteins increased , as did the levels of Krox20 mRNA ( Figure 4c ) , consistent with negative regulation of Krox20 expression by mTORC1 . To examine mTORC1 activity in early nerve development at cellular resolution , we performed immunohistochemistry at P1 . The majority of SCs highly expressing phospho-S6 were not yet myelinating ( 80 . 09 ± 2 . 35% , mean ±s . e . m . , n = 4 mice ) ( Figure 4d , inset 1 ) , while weaker phospho-S6 staining was found in association with myelinated fibers ( Figure 4d , inset 2 ) , consistent with previous cell culture data ( Heller et al . , 2014 ) . High mTORC1 activity was often observed in arrangements reminiscent of axon bundles . These assemblies consist of SCs surrounding multiple axons and extending cytoplasmic processes to sort large caliber axons prior to myelination in a process called radial sorting . Consistently , the temporal span of high mTORC1 activity coincides with the period of intense radial sorting ( Figure 4a , b ) . Radial sorting was impaired when mTORC1 was disrupted in DhhCre:RptorKO nerves ( Norrmén et al . , 2014 ) , indicating that the high mTORC1 activity observed in early nerve development is required in this process . Thus , we examined P5 MpzCre:Tsc1KO:PtenKO nerves in which we had observed the highest mTORC1 activity ( Figure 2c ) for radial sorting alterations . We found that bundles often contained fewer axons than in control nerves ( Figure 4e ) . Coherent with this finding , the number of sorted axons was significantly higher compared to controls ( Figure 4f ) , indicative of increased radial sorting . Collectively , we conclude that: ( 1 ) In normal nerve development , high mTORC1 activity is present before the onset of myelination and declines as SCs start myelinating; ( 2 ) The decrease in mTORC1 activity is physiologically required to allow SCs to differentiate into myelinating SCs , based on the findings that SC differentiation is impaired by sustained activation of mTORC1 in TSC1 and/or PTEN mutants; ( 3 ) High mTORC1 activity inhibits the differentiation of myelinating SCs , but promotes radial sorting , possibly to ensure that the differentiation program of myelination is activated only after radial sorting is completed . According to our timeline analysis , the activity of the PI3K-Akt-mTORC1 axis in adult nerves is substantially lower than in early development ( Figure 4a , b ) . Thus , we analyzed and compared systematically the effects of differently increased mTORC1 and/or PI3K-Akt signaling in adult SCs . To this end , we crossed our floxed mice with mice carrying a MpzCreERT2 transgene , thus allowing inducible SC-specific ablation of TSC1 and/or PTEN ( referred to as MpzCreERT2:Tsc1KO , MpzCreERT2:PtenKO , and MpzCreERT2:Tsc1KO:PtenKO ) . Tamoxifen was administered to young adult mice and three months later ( months post-tamoxifen , mpt ) TSC1 and/or PTEN protein levels were substantially reduced in the corresponding nerves ( Figure 5a–c , Figure 5—figure supplement 1a–c ) . Western blot analyses for phospho-S6KT389 and phospho-AktT308 showed that , like in development , deletion of TSC1 in adult nerves resulted also in hyperactivation of mTORC1 and suppression of Akt activation , while deletion of PTEN or combined deletion of TSC1 and PTEN hyperactivated both PI3K-Akt and mTORC1 signaling . Notably , MpzCreERT2:Tsc1KO:PtenKO mice displayed higher mTORC1 levels than either single knockout animals , reminiscent of development ( Figure 5d , Figure 5—figure supplement 1d ) . Macroscopically , MpzCreERT2:PtenKO and MpzCreERT2:Tsc1KO:PtenKO nerves were enlarged compared to controls and MpzCreERT2:Tsc1KO nerves at 3 mpt ( Figure 5e ) . At the ultrastructural level , we found no signs of demyelination in controls or single mutants , while MpzCreERT2:Tsc1KO:PtenKO nerves contained small numbers of demyelinated fibers ( 174 ± 45 per sciatic nerve cross section , mean ±s . e . m . , n = 3 mice per genotype ) . Next , we evaluated myelin thickness expressed as g-ratio , i . e . the ratio between the axon diameter and the fiber diameter ( axon plus myelin ) . All mutant nerves showed radial hypermyelination ( i . e . increased myelin thickness without visible structural abnormalities ) as indicated by lower g-ratios , with MpzCreERT2:Tsc1KO:PtenKO animals being significantly more hypermyelinated than single knockouts ( Figure 5f , g ) . In TSC1 mutants , radial hypermyelination was more pronounced in fibers with an axonal diameter smaller than 2 μm , whereas in MpzCreERT2:PtenKO and MpzCreERT2:Tsc1KO:PtenKO nerves , axons of all sizes were significantly affected ( Figure 5h ) . We did not observe substantial major shifts in axonal size distributions , pointing to a genuine increase in myelin thickness as reason for reduced g-ratios ( Figure 5i ) . The observed radial hypermyelination was persistent at 6 mpt in MpzCreERT2:Tsc1KO and MpzCreERT2:PtenKO nerves ( Figure 5—figure supplement 2a–d ) . Intriguingly , at 6 mpt , PTEN mutants were more hypermyelinated than TSC1 mutants and showed a stronger reduction in g-ratio compared to 3 mpt ( Figure 5—figure supplement 2b , f ) . In addition to radial hypermyelination , MpzCreERT2:PtenKO nerves exhibited various myelin abnormalities , including infoldings , outfoldings , and tomacula , both at 3 mpt and 6 mpt ( Figure 5f , j , Figure 5—figure supplement 2e ) , consistent with a previous study ( Goebbels et al . , 2012 ) . In MpzCreERT2:Tsc1KO mice such alterations were slightly , but not significantly , increased , while MpzCreERT2:Tsc1KO:PtenKO nerves contained more myelin abnormalities than single mutants . Finally , many Remak bundles in PTEN and double mutants displayed redundant SC membrane wrapping around non-myelinated axons , while such features were undetectable in controls or TSC1 knockouts ( Figure 5k , Figure 5—figure supplement 3a ) . These findings are in line with previous reports in PTEN mutants ( Goebbels et al . , 2010 ) , as is our confirmatory observation of occasional redundant membranes wrapped around collagen fibers in these mice ( Figure 5—figure supplement 3b ) . Together , our data show that high mTORC1 signaling following deletion of TSC1 and/or PTEN in adult differentiated SCs is capable of reactivating radial myelin growth . This effect is proportional to the levels of mTORC1 hyperactivation . Additionally , the subtly different phenotypes of TSC1 and PTEN mutants , including the presence of abnormal wrapping of Remak fibers in PTEN and double mutants , but not in TSC1 mutants , are consistent with potential mTORC1-independent functions of the PI3K-Akt pathway in membrane wrapping and myelin growth . In marked contrast to the outcome in early PNS development , hyperactivation of the PI3K-Akt-mTORC1 axis in adulthood promotes myelin growth . These two opposing effects point to a dual role of mTORC1 signaling in normal SC myelination: Inhibition of SC differentiation before the onset of myelination , but promotion of myelin production after myelination has started . This dual role model predicts that the response of SCs to mTORC1 hyperactivation depends on the SC differentiation status . Thus , we performed two additional sets of experiments , in which either SCs in adult nerves with hyperactive mTORC1 were forced to dedifferentiate and redifferentiate , or mTORC1 was hyperactivated in SCs that had just differentiated into myelinating SCs . First , we induced SC dedifferentiation by carrying out nerve crush injuries in TSC1- or PTEN-conditional knockout backgrounds . Upon nerve injury , SCs dedifferentiate , engage with regrowing axons , and remyelinate them in a highly regulated sequence of events that recapitulates many aspects of developmental myelination ( Chen et al . , 2007; Scherer et al . , 1994; Zorick et al . , 1996 ) , but also involves a special repair program ( Jessen and Mirsky , 2016 ) . We reasoned that , if our model is valid , inducible deletion of TSC1 or PTEN should lead to impaired remyelination when mice undergo a nerve crush injury . Nerve surgeries were performed in MpzCreERT2:Tsc1KO and MpzCreERT2:PtenKO animals at 2 mpt and nerves were analyzed morphologically at 5 , 12 , and 30 days post-crush ( dpc ) ( Figure 6a ) . Quantification of intact-appearing myelin profiles ( defined as non-discontinuous and non-collapsed myelin rings ) at 5 dpc revealed no major differences in demyelination between mutants and controls ( Figure 6b , c ) allowing comparative analysis of subsequent remyelination ( although a subtle influence of the maintenance phenotypes of MpzCreERT2:Tsc1KO and MpzCreERT2:PtenKO mice cannot be excluded ) . At 12 dpc , when initial remyelination occurs , both MpzCreERT2:Tsc1KO and MpzCreERT2:PtenKO nerves showed a marked reduction in remyelinated fibers . In TSC1 knockouts , a minor reduction was persistent at 30 dpc , when remyelination initiation in controls was virtually completed ( Figure 6b , d ) . In line with defective remyelination initiation , the percentage of SCs expressing Krox20 was reduced at 12 dpc in MpzCreERT2:PtenKO mice , while the overall proportions of SCs were not changed ( Figure 6e–g ) . Consistently , Oct6 levels were increased in MpzCreERT2:Tsc1KO nerves at the same time point ( Figure 6h , Figure 6—figure supplement 1a ) . In the second set of experiments , we genetically hyperactivated mTORC1 during developmental myelination , but only after most SCs have started myelinating . According to our model , constitutive activation of mTORC1 in cells that have just differentiated to myelinating SCs should lead to radial hypermyelination . To achieve this goal , we crossed Tsc1-floxed mice with mice carrying a Plp1CreERT2 transgene ( yielding Plp1CreERT2:Tsc1KO ) . We induced recombination by administering 4-hydroxytamoxifen ( 4OH-TMX ) ( Zuchero et al . , 2015 ) from P8 to P12 since most SCs have started myelinating at this time ( Arroyo et al . , 1998 ) , and compared these inducible mutant mice with the previously described DhhCre:Tsc1KO mice , in which recombination occurs before onset of myelination ( Figure 6i ) , together with controls . Plp1CreERT2-driven deletion of TSC1 was successful and led to constitutive mTORC1 activation , as indicated by increased phosphorylation of S6K at T389 ( Figure 6j , Figure 6—figure supplement 1b ) . Consistent with our model , P60 Plp1CreERT2:Tsc1KO nerves were radially hypermyelinated , in contrast to the hypomyelinated DhhCre:Tsc1KO nerves ( Figure 6k , l ) .
Our results reveal that PI3K-Akt-mTORC1 signaling directs PNS myelination at multiple critical levels , including onset of myelination and myelin growth . We started our studies by conditionally deleting TSC1 to explore the effects of hyperactivating mTORC1 in developing SCs . Unexpectedly , the resulting constitutively elevated mTORC1 activity profoundly delayed SC differentiation to myelinating cells . Searching for an explanation , we investigated whether inhibitory feedback of mTORC1 on PI3K-Akt signaling , shown in other cases to account for paradoxical results upon hyperactivation of mTORC1 ( Byles et al . , 2013; Tang et al . , 2014; Yecies et al . , 2011 ) , might be responsible . However , the phenotype of TSC1 mutants was even aggravated after restoring Akt activity in SCs via simultaneous PTEN deletion . Moreover , we found that single deletion of PTEN leading to high Akt activity in vivo , or expression of constitutively active Akt in cell cultures , inhibited the onset of myelination per se . Notably as in TSC1 mutants , inhibition of SC differentiation by hyperactive PI3K-Akt due to loss of PTEN was mediated by mTORC1 since rapamycin treatment effectively rescued this effect . These data unveiled a direct myelination-inhibitory function of the PI3K-Akt-mTORC1 axis . In support of this paradigm , we discovered that mTORC1 negatively regulates Krox20 , the main TF required for onset of SC myelination , via S6K . Coherently , we found that mTORC1 activity is physiologically high in not-yet myelinating SCs and declines with their commitment to myelination . In this light , the impaired onset of myelination in TSC1 and/or PTEN mutants can be viewed as a failure of mutant SCs to promptly downregulate mTORC1 activity when myelination has to start . What is then the physiological function of high mTORC1 during early nerve development ? We have shown previously that genetic inhibition of mTORC1 impairs radial sorting ( Norrmén et al . , 2014 ) and describe now that strong hyperactivation of mTORC1 can accelerate this crucial process in myelination . We infer that high mTORC1 activity during early nerve development may drive radial sorting , concomitant with transiently inhibiting further SC differentiation to myelinating cells , thereby serving as a key component of the regulatory program that coordinates completion of radial sorting and onset of myelination ( Feltri et al . , 2016 ) . Intriguingly , recent studies on melanocytes , which share a common developmental origin with SCs , revealed parallels between mTORC1 functions in these cell types . Analogous to the differentiation-inhibiting role of mTORC1 in SCs and the negative regulation of Krox20 by mTORC1 , mTORC1 hyperactivation in melanocytes impaired melanogenesis , while inhibition of mTORC1 increased levels of MITF , a TF essential for melanin production ( Cao et al . , 2017; Ho et al . , 2011 ) . Hence , it will be appealing to explore whether an inhibitory role of mTORC1 activity on cell differentiation is a frequent feature of neural crest-derived cells . Our results show that physiologically high activity of the PI3K-Akt-mTORC1 axis before the ordinary developmental onset of myelination has a myelination-inhibitory function . However , it has been previously reported that conditional ablation of PTEN in the Schwann cell lineage increased the number of myelinated fibers in 3 months-old mice , most likely due to aberrant myelination of normally non-myelinated C fibers ( Goebbels et al . , 2010 ) . Thus , chronic hyperactivation of PI3K-Akt signaling may be able to , in the long term , drive Remak cells to a myelinating-like phenotype ( see also below ) . In contrast to the inhibitory effect on SC differentiation , we found that deletion of TSC1 and/or PTEN in adult SCs was able to reactivate myelin growth , proportionally to the levels of mTORC1 activity . However , in spite of comparable mTORC1 hyperactivity , we noticed subtle phenotypical differences between MpzCreERT2:Tsc1KO and MpzCreERT2:PtenKO nerves . First , MpzCreERT2:PtenKO , but not MpzCreERT2:Tsc1KO , displayed increased radial hypermyelination at 6 mpt compared to 3 mpt , indicating that specifically PTEN ablation led to continuous radial myelin growth . Second , in MpzCreERT2:PtenKO and MpzCreERT2:Tsc1KO:PtenKO , but not in MpzCreERT2:Tsc1KO , redundant processes of non-myelinating SCs repeatedly wrapped around axons in Remak bundles , analogous to previous observations ( Domènech-Estévez et al . , 2016; Goebbels et al . , 2010 ) . These findings indicate that the PI3K-Akt pathway is likely to serve also mTORC1-independent functions in controlling SC myelination . In particular , the redundant wrapping of SC membranes upon loss of PTEN , but not of TSC1 , prompts us to speculate that , also during physiological myelination , driving of SC membrane wrapping might involve PI3K-Akt-dependent/mTORC1-independent mechanisms . We envisage that the PI3K-Akt pathway serves converging scopes during myelin growth , recruiting ( 1 ) mTORC1 to activate the synthesis of myelin building blocks , and ( 2 ) mTORC1-independent targets for membrane wrapping . The existence of such a PI3K-Akt-dependent/mTORC1-independent ‘wrapping force’ in myelination is supported by recent reports ( Domènech-Estévez et al . , 2016; Mathews and Appel , 2016 ) and may underlie the regulation of the mechanistically required cytoskeletal dynamics . In agreement , recent work demonstrated that constitutively active Akt induces Rac1 activity in SCs ( Domènech-Estévez et al . , 2016 ) , and remodeling of the actin cytoskeleton is an essential prerequisite for membrane expansion during myelination ( Nawaz et al . , 2015; Novak et al . , 2011; Zuchero et al . , 2015 ) . The formation of myelin abnormalities – a hallmark of numerous neuropathies ( Dyck and Thomas , 2005 ) – may also involve mTORC1-independent targets of the PI3K-Akt pathway , given that such alterations accumulated substantially in MpzCreERT2:PtenKO , but less so in MpzCreERT2:Tsc1KO . Nevertheless , high mTORC1 signaling probably supports their growth , since myelin abnormalities were more abundant in MpzCreERT2:Tsc1KO:PtenKO with higher mTORC1 activity compared to single MpzCreERT2:PtenKO . In accord , treatment with rapamycin has been shown to diminish the load of myelin abnormalities in PTEN mutant mice ( Goebbels et al . , 2012 ) . Taken together , also supported by data from others ( Beirowski et al . , 2017 ) , our results support a model in which the PI3K-Akt-mTORC1 axis fulfills multiple major roles in SC myelination ( Figure 7 ) . In early SC development , high mTORC1 signaling maintains SCs in a non-differentiated state by suppressing Krox20 expression and , at the same time , promotes radial sorting . After axonal sorting is completed , a physiological decline in mTORC1 activity releases the suppression on Krox20 expression and allows myelination to start . When this turning point has been passed , the residual mTORC1 activity drives myelin production in concert with mTORC1-independent targets of PI3K-Akt . Downstream of mTORC1 , both lipid synthesis via activation of the RXRγ-SREBP1c axis ( Norrmén et al . , 2014 ) and protein synthesis ( Sheean et al . , 2014 ) are key processes supporting myelin growth . Our model is especially supported by the opposing outcomes of mTORC1 hyperactivation in developing non-differentiated SCs versus in adult myelinating SCs , revealing that SCs reacted differently depending on their differentiation status . To confirm this notion , we forced adult differentiated SCs to dedifferentiate due to nerve crush injuries , followed by axonal regeneration and SC redifferentiation . As expected in this experimental setting , redifferentiation and remyelination were delayed in MpzCreERT2:Tsc1KO and MpzCreERT2:PtenKO mutants . Conversely , deletion of TSC1 when most SCs had just started myelinating enhanced radial myelin growth . Our conceptual model raises a number of questions for future investigations including: Are different upstream receptors mediating the differentiation-inhibiting and myelin-growth promoting roles of mTORC1 ? Is a single upstream receptor system accounting for both , but through different ligands ( e . g . neuregulin-1 isoforms ) , or through different concentrations of the same ligand ? What is responsible for the physiological decline in mTORC1 activity that allows SCs to start myelinating ? What are the postulated mTORC1-independent targets of PI3K-Akt ? We can only speculate in this context , but we have shown here that neuregulin-1 signaling is a major activator of the PI3K-Akt-mTORC1 axis in SCs , in line with previously reported findings of strongly reduced phospho-S6 levels in SCs co-cultured with DRG neurons missing neuregulin-1 ( Heller et al . , 2014 ) . Neuregulin-1 isoforms are known to exert different effects on SCs in various experimental settings involving several signaling pathways ( Mei and Nave , 2014 ) . Thus , signaling through different neuregulin-1 isoforms might contribute to the different functions of mTORC1 in SC biology , potentially in concert with additional signals integrating other adaxonal and also abaxonal cues ( Ghidinelli et al . , 2017; Heller et al . , 2014; Herbert and Monk , 2017; Monk et al . , 2015; Pereira et al . , 2012 ) . Modulation of the pathways upstream of mTORC1 has been recently explored as a promising therapeutic approach in animal models of hereditary peripheral neuropathies ( Bolino et al . , 2016; Fledrich et al . , 2014; Goebbels et al . , 2012; Nicks et al . , 2014 ) . Based on our ‘dual-role’ model , we expect that future studies will benefit from monitoring mTORC1 activity in conjunction with the differentiation status of SCs to enhance efficacy . Moreover , inhibition of mTORC1 should be considered as a valuable therapeutic strategy in hereditary or acquired peripheral neuropathies in which SC differentiation is defective .
Mice harboring floxed alleles of Tsc1 ( STOCK Tsc1<tm1Djk>/J , RRID:IMSR_JAX:005680 ) and Pten ( C;129S4-Pten<tm1Hwu>/J , RRID:IMSR_JAX:004597 ) were obtained from The Jackson Laboratory . Mice harboring floxed alleles of Rptor ( Bentzinger et al . , 2008; Polak et al . , 2008 ) and mice carrying a Cre transgene under control of the Dhh ( RRID:IMSR_JAX:012929 ) or Mpz promoter ( RRID:IMSR_JAX:017927 ) , or a CreERT2 transgene under control of the Plp1 or Mpz promoters have been described ( Feltri et al . , 1999; Jaegle et al . , 2003; Leone et al . , 2003 ) . To generate non-inducible conditional deletion of TSC1 , PTEN , or Raptor , floxed mice were crossed with DhhCre- or MpzCre-positive mice . To generate inducible conditional deletion of TSC1 or PTEN , floxed mice were crossed with MpzCreERT2-positive mice , and at 8–10 weeks of age 2 mg of tamoxifen ( Sigma-Aldrich , Saint Louis , MO , USA ) in 10% ethanol/sunflower seed oil ( Sigma-Aldrich ) were injected intraperitoneally once a day on five consecutive days in both mutant and control animals . Developmental inducible deletion of TSC1 was achieved crossing floxed mice with Plp1CreERT2-positive mice and administering intraperitoneally to both mutant and control pups 75 μg per gram of body weight of 4-hydroxytamoxifen ( Sigma-Aldrich ) dissolved in Kolliphor EL ( Sigma-Aldrich ) once a day on five consecutive days from P8 to P12 . Experimental animals were on a hybrid background between C57B6 and the background of origin of the floxed mice ( 2–4 backcrosses with C57B6 for all TSC1 , PTEN , or TSC1/PTEN mutants and their controls; 7–8 backcrosses with C57B6 for Raptor mutants and their controls ) . Cre-negative animals ( floxed homozygous or heterozygous ) were used as controls in the experiments . To control for variations in background , littermate controls were used . Wild-type mice were on a C57B6 background . Mice of either sex were used in the experiments . For in vivo rescues , rapamycin ( Millipore , Billerica , MA , USA ) was dissolved in a solution of 5% PEG-400 , 5% Tween-80 , and 4% ethanol , and 5 μg rapamycin per gram of body weight were administered intraperitoneally once a day at P3 and P4 . Genotypes were determined through genomic PCR using the following primers: Cre forward 5’-accaggttcgttcactcatgg-3’ , reverse 5’-aggctaagtgcctcttctaca-3’; TSC1 forward 5’-gtcacgaccgtaggagaagc-3’ , reverse 5’-gaatcaaccccacagagcat-3’; PTEN forward 5’-caagcactctgcgaactgag-3’ , reverse 5’-aagtttttgaaggcaagatgc-3’ , Raptor forward 5’-atggtagcaggcacactcttcatg-3’ , reverse 5’-gctaaacattcagtccctaatc-3’ . Mice were housed with a maximum number of five animals per cage , kept in a 12 hr light-dark cycle , and fed standard chow ad libitum . Mice were subjected to unilateral sciatic nerve crush injury two months after tamoxifen administration . After inducing anesthesia with isofluorane inhalation , the sciatic nerve was exposed through blunt dissection of the thigh muscles and compressed for 30 s . For analgesia 0 . 1 mg kg−1 buprenorphine ( Temgesic , Reckitt Benckiser , UK ) were injected intraperitoneally once prior to the surgery , and thereafter every 12 hr for two days . Morphological analysis and immunostaining of crushed nerves were performed on sections 3 mm distal to the injury site . For biochemical analysis , the whole nerve distal to the injury site was used . Immediately after dissection , sciatic nerves were fixed with 3% glutaraldehyde and 4% paraformaldehyde in 0 . 1 M phosphate buffer . Sciatic nerves were further treated with 2% osmium tetroxide ( EMS , Hatfield , PA , USA ) , dehydrated over a series of acetone gradients and embedded in Spurrs resin ( EMS ) . Semithin sections ( 650 nm ) were stained with 1% toluidine blue and used for qualitative analysis and for quantifying intact myelin profiles after nerve crush . Ultrathin sections ( 65 nm ) were imaged with a FEI Morgagni 268 TEM , and random 5 × 5 MIA ( multiple image alignment ) fields were acquired for qualitative analysis and for quantifying the percentage of myelinating SCs . G-ratio measurements and quantification of myelin abnormalities were performed on EM reconstructions of the entire sciatic nerve , obtained from additional sections ( 99 nm ) collected on ITO coverslips ( Optics Balzers , Germany ) and imaged with either a Zeiss Gemini Leo 1530 FEG or Zeiss Merlin FEG scanning electron microscopes attached to ATLAS modules ( Zeiss , Germany ) . To calculate the g-ratio , the axon diameter was derived from the axon area , while the fiber diameter was calculated adding to the axon diameter twice the average of the myelin thickness measured at two different locations of the myelin ring . At least 200 fibers per nerve were analyzed . The following primary antibodies were used: TSC1 ( CST , #6935 , 1:1 , 000 ) , phospho-S6KT389 ( Cell Signaling Technology Cat# 9234 also 9234L , 9234P , 9234S RRID:AB_2269803 , 1:1 , 000 ) , S6K ( Cell Signaling Technology Cat# 9202S RRID:AB_331676 , 1:1 , 000 ) , phospho-4EBP1T37/46 ( Cell Signaling Technology Cat# 2855 also 2855L , 2855S , 2855P RRID:AB_560835 , 1:1 , 000 ) , 4EBP1 ( Cell Signaling Technology Cat# 9452S RRID:AB_10693791 , 1:1 , 000 ) , phospho-S6S235/236 ( Cell Signaling Technology Cat# 4857S RRID:AB_2181035 , 1:1 , 000 ) , S6 ( Cell Signaling Technology Cat# 2317S RRID:AB_10694551 , 1:1 , 000 ) , cJun ( Cell Signaling Technology Cat# 9165 RRID:AB_2130165 , 1:1 , 000 ) , phospho-ErbB2Y1248 ( Cell Signaling Technology Cat# 2247S RRID:AB_331725 , 1:1 , 000 ) , ErbB2 ( Cell Signaling Technology Cat# 2165S RRID:AB_10692490 , 1:1 , 000 ) , phospho-AktT308 ( Cell Signaling Technology Cat# 9275 also 9275S , 9275L RRID:AB_329828 only for the western blot in Figure 4a , Cell Signaling Technology Cat# 4056 also 4056S , 4056L RRID:AB_331163 for the other western blots , both used 1:1 , 000 ) , phospho-AktS473 ( Cell Signaling Technology Cat# 4051 also 4051S , 4051L RRID:AB_331158 , 1:1 , 000 ) , Akt ( Cell Signaling Technology Cat# 9272 also 9272S RRID:AB_329827 , 1:1 , 000 ) , phospho-ERK1/2 T202/Y204 ( Cell Signaling Technology Cat# 9106 also 9106L , 9106S RRID:AB_331768 , 1:1 , 000 ) , ERK1/2 ( Cell Signaling Technology Cat# 9102 also 9102L , 9102S RRID:AB_330744 , 1:1 , 000 ) , phospho-PRAS40T246 ( CST , #13175 , 1:1 , 000 ) , PRAS40 ( Cell Signaling Technology Cat# 2691S RRID:AB_2225033 , 1:1 , 000 ) , PTEN ( Cell Signaling Technology Cat# 9559S RRID:AB_10695541 , 1:1 , 000 ) , Akt phospho-substrates ( Cell Signaling Technology Cat# 10001S RRID:AB_10950819 , 1:1 , 000 ) , α-tubulin ( Sigma-Aldrich Cat# T5168 RRID:AB_477579 , 1:1 , 000 ) , β-actin ( Sigma-Aldrich Cat# A5316 RRID:AB_476743 , 1:1 , 000 ) , GAPDH ( Hytest Cat# 5G4-9B3 RRID:AB_1616725 , 1:10 , 000 ) , NF ( Sigma-Aldrich Cat# N5264 RRID:AB_477278 , used for immunocytochemistry , 1:200 ) , NF ( Aves Labs Cat# NF-M RRID:AB_2313554 , used for immunohistochemistry , 1:1 , 000 ) , MBP ( Bio-Rad/AbD Serotec Cat# MCA409S RRID:AB_325004 , 1:200 ) , Sox10 ( R&D Systems , AF2864 , 1:200 ) , S100 ( Dako Cat# Z0311 RRID:AB_10013383 , 1:200 ) , laminin-α2 ( Sigma-Aldrich Cat# L0663 RRID:AB_477153 , 1:200 ) . Primary antibodies against Oct6 and Krox20 ( both used 1:1 , 000 ) were generous gifts from Dr . Dies Mejer . HRP- , AP- , and fluorophore-conjugated secondary antibodies were purchased from Jackson ImmunoResearch and used 1:200 for immunostainings or 1:10 , 000 for western blots . Alexa488-coupled phalloidin ( used 1:100 ) was from Life Technologies ( Carlsbad , CA , USA ) ( RRID:AB_2315147 ) . The following chemical inhibitors were used in in vitro experiments: rapamycin ( Sigma-Aldrich ) , LY294002 ( Sigma-Aldrich ) , CI-1040 ( Sigma-Aldrich ) , torin-1 ( Tocris , UK ) , LYS6K2 ( Focus Biomolecules , Plymouth Meeting , PA , USA ) . Immediately after dissection , sciatic nerves were placed in ice-cold PBS , the epineurium was removed with fine forceps , and the nerves were snap-frozen in liquid nitrogen and stored at −80°C until further processing . To prepare lysates , nerves were ground on dry ice , mixed with PN2 lysis buffer ( 25 mM Tris-HCl pH 7 . 4 , 95 mM NaCl , 10 mM EDTA , 2% SDS , protease and phosphatase inhibitors ( Roche , Switzerland ) ) , boiled , and spun for 15 min . 10–30 μg of proteins per sample were mixed 1:4 with sample buffer ( 200 mM Tris-HCl pH6 . 8 , 40% glycerol , 8% SDS , 20% β-mercaptoethanol , 0 . 4% bromophenol blue ) , run on 4–15% polyacrylamide gradient gels ( Biorad , Hercules , CA , USA ) , and blotted onto PVDF membranes ( Millipore ) . After blocking with 5% milk in TBS-T , membranes were incubated overnight with primary antibodies diluted in 5% BSA in TBS-T . Chemiluminescent signals were generated using HRP- or AP-conjugated secondary antibodies and ECL ( GE Healthcare ) , ECL prime ( GE Healthcare ) , or CDP-Star ( Roche ) , and detected using Fusion FX7 ( Vilber Lourmat , Germany ) . If required , after signal detection membranes were stripped with a buffer containing β-mercaptoethanol ( 1 . 52 g Tris , 2 g SDS , 700 μl β-mercaptoethanol in 100 ml water , pH 6 . 8 ) for 15 min at 55°C and reprobed with antibodies . Quantification of band intensities was performed with ImageJ ( version 1 . 50i ) . α-tubulin , β-actin , or GAPDH were used as normalization controls . For cropped blots , the control bands are displayed immediately below the other bands that belong to the same membrane . Full-length blots are shown in Supplementary files 1–8 . Size markers refer to All Blue Precision Protein Standards ( Biorad ) . Protein size is expressed as apparent molecular weight in kDa . Immediately after dissection , sciatic nerves were fixed for one hour in 4% paraformaldehyde , cryopreserved in 10% sucrose for one hour , 20% sucrose overnight , and then embedded in OCT ( TissueTek , Torrance , CA , USA ) . 8 μm thick cryosections were cut , blocked with blocking buffer ( 1% BSA , 10% goat or donkey serum , 0 . 1% triton X-100 in PBS ) for one hour , incubated overnight with primary antibodies , incubated for one hour with fluorophore-conjugated secondary antibodies , counterstained with DAPI ( Life Technologies ) , and mounted with Vectashield ( Vector Laboratories , Burlingame , CA , USA ) . The same procedure was used for cell culture stainings . For cell proliferation experiments , mouse pups were injected with 50 μg EdU ( Life Technologies ) per gram of body weight and sacrificed one hour later . EdU staining was performed using the Click-iT EdU Alexa647 kit ( Life Technologies ) as per manufacturer’s instructions . Immunostainings were imaged using an epifluorescence microscope ( Zeiss Axio Imager . M2 ) equipped with a monochromatic CCD camera ( sCMOS , pco . edge ) , or with a confocal microscope ( Leica TCS SP8 ) . One representative section per sample was imaged and analyzed . Immediately after dissection , sciatic nerves were placed in ice-cold PBS , the epineurium was removed with fine forceps , and the nerves were snap-frozen in liquid nitrogen and stored at −80°C until needed . Total RNA was extracted using RNeasy kit ( Qiagen , Germany ) for developmental samples or Qiazol ( Qiagen ) for adult samples and cells as per manufacturer’s instructions . 50–200 ng of total RNA were reverse transcribed using Maxima First Strand cDNA Synthesis Kit ( Thermo Fisher Scientific , Waltham , MA , USA ) as per manufacturer’s instructions . qPCR reactions were performed using FastStart Essential DNA Green Master ( Roche ) and Light Cycler 480 II ( Roche ) . The sequences of the primers used and their specificity are the following: Krox20 ( mouse and rat ) forward 5’-acagcctctacccggtggaagac-3’ , reverse 5’-cagagatgggagcgaagctactcggata-3’; cJun ( mouse and rat ) forward 5’-gccaagaactcggaccttctcacgtc-3’ , reverse 5’-tgatgtgcccattgctggactggatg-3’; Oct6 ( mouse ) forward 5’-gagcactcggacgaggatg-3’ , reverse 5’-cacgttaccgtagagggtgc-3’; Brn2 ( mouse ) forward 5’-tcaaatgccctaagccctcg-3’ , reverse 5’-cgggaggggtcatccttttc-3’; Sox10 ( mouse ) forward 5’-ccgaccagtaccctcacct-3’ , reverse 5’-tcaatgaaggggcgcttgt-3’; Sox2 ( mouse ) forward 5’-ggaaagggttcttgctgggt-3’ , reverse 5’-acgaaaacggtcttgccagt-3’; Id2 ( mouse ) forward 5’-catcagcatcctgtccttgc-3’ , reverse 5’-ttctcctggtgaaatggctgat-3’; GAPDH ( mouse and rat ) forward 5’-ggtgaaggtcggtgtgaacggatttgg-3’ , reverse 5’-ggtcaatgaaggggtcgttgatggcaac-3’; β-actin ( mouse ) forward 5’-gtccacacccgccacc-3’ , reverse 5’-ggcctcgtcacccacatag-3’; α-tubulin ( mouse ) forward 5’-tcttagttgtcgggaacggt-3’ , reverse 5’-ggagatgcactcacgcatgata-3’ . Relative mRNA fold changes for each gene were obtained by using the 2-ΔΔCt method after normalization to GAPDH , β-actin , or α-tubulin . To generate lentiviral vectors for overexpression in SCs , the hPGK promoter of the pCCLsin . PPT . hPGK . PRE lentiviral backbone was replaced by a 1 . 1 kb fragment of the rat P0 promoter , as previously described ( Norrmén et al . , 2014 ) . myrAkt constructs were obtained from Addgene ( #9008 , #9016 , #9017 ) ( Ramaswamy et al . , 1999 ) , PCR-amplified , and inserted between the AgeI and SalI restriction sites of the modified pCCLsin . PPT . hPGK . PRE vector . The 4EBP1-4xA construct was obtained from Addgene ( #38240 ) ( Thoreen et al . , 2012 ) and subcloned between the BamHI and NheI restriction sites of the pcDNA3 . 1 vector . As control , the eGFP coding sequence from the pCCLsin . PPT . hPGK . PRE vector was subcloned between the NheI and EcoRI restriction sites of the pcDNA3 . 1 vector . All constructs were sequence-verified prior to usage . To prepare embryonic SCs , E13 . 5 mouse DRGs were isolated , digested with trypsin-EDTA 0 . 25% ( Life Technologies ) for 30 min , resuspended in N2-medium ( Advanced DMEM:F12 plus N2 supplement ( Life Technologies ) ) with 50 ng ml−1 7S NGF ( Millipore ) , and plated on uncoated dishes . After one week , the SC-neuron network was mechanically detached from the underlying fibroblast layer , digested with 0 . 25% trypsin ( Sigma-Aldrich , T9201 ) and 0 . 1% collagenase ( Sigma-Aldrich , C0130 ) , and replated on PLL-coated dishes in N2-medium devoid of NGF . After reaching confluency , cells were trypsinized , centrifuged , and resuspended in flow buffer ( PBS plus 2% FCS ( Life Technologies ) ) . Flow cytometry experiments were performed with LSR Fortessa ( BD Biosciences , Franklin Lakes , NJ , USA ) using between 105 and 106 cells per genotype , and the data were analyzed with the FlowJo software ( RRID:SCR_008520 , version 10 . 0 . 7 ) . To account for differences in the number of recorded events for each sample , the flow cytometry profiles were normalized to their modes . To prepare neonatal mouse SCs , sciatic nerves were dissected from P2 mouse pups , the epineurium was removed , and the nerves were digested with 1 . 25 mg ml−1 trypsin ( Sigma-Aldrich , T9201 ) and 2 mg ml−1 collagenase ( Sigma-Aldrich , C0130 ) for one hour . Cells were then centrifuged , resuspended in D-medium ( DMEM-Glutamax plus 10% FCS ( Life Technologies ) ) , and seeded on PLL-coated 12 mm glass coverslips . After overnight incubation , cells were fixed with 4% paraformaldehyde and stained . To prepare rat SCs , sciatic nerves were dissected from P2 Sprague-Dawley rat pups , the epineurium was removed , and nerves were digested with 1 . 25 mg ml−1 trypsin ( Sigma-Aldrich , T9201 ) and 2 mg ml−1 collagenase ( Sigma-Aldrich , C0130 ) for one hour . After centrifugation , the cells were resuspended in D-medium ( DMEM-Glutamax plus 10% FCS ( Life Technologies ) ) and plated on PLL-coated dishes . Contaminating fibroblasts were eliminated with 10 μM Ara-C treatment ( Sigma-Aldrich ) for 48 hr , followed by complement-mediated removal of fibroblasts using an anti-Thy1 . 1 antibody ( Bio-Rad/AbD Serotec Cat# MCA04G RRID:AB_322809 , 1:50 ) . After a 10 min incubation with the anti-Thy1 . 1 antibody , rabbit complement ( Millipore ) was added for 40 min . SCs were then further expanded in SC growth medium ( DMEM-Glutamax ( Life Technologies ) , 10% FCS ( Life Technologies ) , 5 μg ml−1 bovine pituitary extract ( Life Technologies ) , 2 μM forskolin ( Sigma-Aldrich ) ) , tested for mycoplasma contamination , aliquoted , and stored in liquid nitrogen . For differentiation experiments , SCs were transferred to defined medium ( 100 μg ml−1 of holotransferrin ( Sigma-Aldrich ) , 60 ng ml−1 progesterone ( Sigma-Aldrich ) , 1 μg ml−1 insulin ( Sigma-Aldrich ) , 16 μg ml−1 putrescine ( Sigma-Aldrich ) , 400 ng ml−1 L-thyroxin ( Sigma-Aldrich ) , 160 ng ml−1 sodium selenite ( Sigma-Aldrich ) , 10 ng ml−1 triiodothyronine ( Sigma-Aldrich ) , 38 ng ml−1 dexamethasone ( Sigma-Aldrich ) , and 300 μg ml−1 BSA ( Sigma-Aldrich ) in DMEM:F12-Glutamax ( Life Technologies ) ) , and differentiation was induced with 1 mM dbcAMP ( Sigma-Aldrich ) for 24 hr . For experiments with neuregulin stimulation , SCs were serum-starved for 6 . 5–8 hr in DMEM-Glutamax with 0 . 05% FCS ( Life Technologies ) and subsequently treated with 10 ng ml−1 recombinant human EGF domain of neuregulin-1 β1 ( R&D Systems , Minneapolis , MN , USA ) . For all other experiments , SCs were kept in SC growth medium . Transfection of rat SCs was performed using Lipofectamine 3000 ( Life Technologies ) as per manufacturer’s instructions . Only SCs lower than passage five were used for experiments . DRGs were isolated from E13 . 5 embryos , digested for 45 min with trypsin-EDTA 0 . 25% ( Life Technologies ) , and seeded on 12 mm coverslips coated with 10% Matrigel ( Corning , NY , USA ) at a density of 1 . 2 DRGs per coverslip . DRGs from embryos with the same genotype were pooled before seeding . Cultures were kept in supplemented NB-medium ( Neurobasal ( Life Technologies ) , B27 supplement ( Life Technologies ) , 4 g l−1 D-glucose ( Sigma-Aldrich ) , 2 mM L-glutamine ( Life Technologies ) , 50 ng ml−1 7S NGF ( Millipore ) ) for the first five days and then in C-medium ( MEM-Glutamax ( Life Technologies ) , 10% FCS ( Life Technologies ) , 4 g l−1 D-glucose ( Sigma-Aldrich ) , 50 ng ml−1 7S NGF ( Millipore ) ) supplemented with 50 μg ml−1 ascorbic acid ( Sigma-Aldrich ) to induce myelination for an additional 8–10 days . Both supplemented NB- and C-medium were changed every other day . Before staining , DRG-cultures were fixed in 4% paraformaldehyde for 15 min and subsequently permeabilized with ice-cold methanol for 10 min . Images of whole coverslips were acquired using an epifluorescence microscope ( Zeiss Axio Imager . M2 ) equipped with a monochromatic CCD camera ( sCMOS , pco . edge ) and an automated stage . Between 3 and 6 coverslips per condition were imaged and analyzed . To quantify the extent of myelination , four random fields per coverslip were selected , the MBP-positive and the NF-positive areas per field were measured with ImageJ ( version 1 . 50i ) after thresholding , and the MBP area/NF area ratio per coverslip was calculated as average of the MBP area/NF area ratios of the different fields . The same threshold was applied to all samples and conditions . HEK293T cells were from ATCC ( Manassas , VA , USA ) , were not further authenticated , and were regularly monitored to assure lack of mycoplasma contamination . Subconfluent HEK293T cells were transfected in 10 cm dishes with lentiviral vectors and the packaging plasmids psPAX2 and pCMV-VSV-G using Lipofectamine 2000 as per manufacturer’s instructions ( Life Technologies ) . Non-concentrated viruses were collected 48 hr after transfection , aliquoted , and stored at −80°C until needed . For infection of DRG-explant cultures , non-concentrated viruses were mixed 1:1 with supplemented NB-medium and added to the cultures for 24 hr from DIV 3 to 4 . The quantity and quality of isolated RNA was determined with a Qubit ( 1 . 0 ) Fluorometer ( Life Technologies ) and a Bioanalyzer 2100 ( Agilent , Santa Clara , CA , USA ) . The TruSeq Stranded mRNA Sample Prep Kit ( Illumina , San Diego , CA , USA ) was used in the succeeding steps . Briefly , total RNA samples ( 150 ng ) were ribosome depleted and then reverse-transcribed into double-stranded cDNA with actinomycin added during first-strand synthesis . The cDNA samples were fragmented , end-repaired and polyadenylated before ligation of TruSeq adapters . The adapters contain the index for multiplexing . Fragments containing TruSeq adapters on both ends were selectively enriched with PCR . The quality and quantity of the enriched libraries were validated using Qubit ( 1 . 0 ) Fluorometer and the Bioanalyzer 2100 ( Agilent ) . The product is a smear with an average fragment size of approximately 360 bp . The libraries were normalized to 10 nM in Tris-Cl 10 mM , pH 8 . 5 with 0 . 1% Tween-20 . The TruSeq SR Cluster Kit v4-cBot-HS or TruSeq PE Cluster Kit v4-cBot-HS ( Illumina ) was used for cluster generation using 8 pM of pooled normalized libraries on the cBOT . Sequencing was performed on the Illumina HiSeq 2500 paired-end at 2 × 126 bp or single-end 126 bp using the TruSeq SBS Kit v4-HS ( Illumina ) . The raw reads were first cleaned by removing adapter sequences , trimming low quality ends , and filtering reads with low quality ( phred quality <20 ) using Trimmomatic ( Bolger et al . , 2014 ) . Sequence alignment of the resulting high-quality reads to the Mus musculus reference genome ( build GRCm38 ) and quantification of gene level expression was carried out using RSEM ( version 1 . 2 . 22 ) ( Li and Dewey , 2011 ) . To detect differentially expressed genes we applied count based negative binomial model implemented in the software package EdgeR ( R version: 3 . 2 . 2 , edgeR_3 . 12 . 0 ) ( Robinson et al . , 2010 ) . The differential expression was assessed using an exact test adapted for over-dispersed data . Genes showing altered expression ( fold change >1 . 2 ) with adjusted ( Benjamini and Hochberg method ) p-value<0 . 05 ( indicated as false discovery rate , FDR ) were considered differentially expressed . Within this set of genes , downregulated and upregulated genes were separately subjected to gene ontology analysis of biological processes using the online tool Enrichr ( http://amp . pharm . mssm . edu/Enrichr/ ) . Data processing and statistical analyses were performed using GraphPad Prism ( RRID:SCR_002798 , version 7 . 0a ) and Microsoft Excel ( version 15 . 27 ) . Data distribution was assumed to be normal and variances were assumed to be equal , although this was not formally tested due to low n number . Sample sizes were chosen according to sample sizes generally employed in the field . The investigators were blinded to the genotypes during analysis of morphological and immunohistochemical data , except for those cases in which mutant mice exhibited an obvious phenotype . No randomization methods were used . Two-tailed unpaired Student’s t-test was used if only two conditions or genotypes were compared . In all other cases , one- or two-way ANOVAs followed by Tukey’s , Dunnett’s , or Sidak’s multiple comparisons tests were employed , as indicated in the figure legends . p<0 . 05 was considered to be statistically significant . No samples or data were omitted during the analyses . RNA-sequencing data have been deposited in the ENA database under accession number PRJEB20661 . | Neurons transmit electrical impulses throughout our body along cable-like structures called axons . Similar to electric cables , the axons are enveloped in an insulating sheath called myelin , which makes the impulses travel faster down the axon . If myelin does not form correctly or gets lost later in life , it can lead to muscle weakness and numbness . Myelin is produced in nerves by specialized cells called the Schwann cells , which wrap around the axons several times to create a thick myelin sheath . The signaling complex called mTORC1 plays an important role in this process . A study in 2014 showed that when mTORC1 was inactive , the myelin sheath was abnormally thin . However , it was not known whether increased mTORC1 would force Schwann cells to produce more myelin . Now , Figlia et al . – including some of the researchers involved in the 2014 study – used genetically modified mice to manipulate two proteins known to control the activity of mTORC1 . When both proteins were removed , either individually or in combination , mTORC1 activity was higher than normal . However , in developing nerve cells , high levels of mTORC1 did not cause the young Schwann cells to produce more myelin , but rather stopped them to become the specialized cells that wrap around axons . Figlia et al . then increased mTORC1 levels after Schwann cells had already started wrapping around axons . In this case , a high activity of mTORC1 resulted in thicker myelin . This suggests that a normal development of a nerve and ultimately the thickness of the myelin sheath depend on when and how much of mTORC1 is available . This discovery could help to develop new therapies for myelin diseases . Increasing the activity of mTORC1 in Schwann cells after they have started wrapping may boost myelin production in diseases in which the myelin sheath is too thin . Conversely , inhibiting mTORC1 could help in situations when the Schwann cells cannot develop properly . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2017 | Dual function of the PI3K-Akt-mTORC1 axis in myelination of the peripheral nervous system |
Understanding how humans represent others’ pain is critical for understanding pro-social behavior . ‘Shared experience’ theories propose common brain representations for somatic and vicarious pain , but other evidence suggests that specialized circuits are required to experience others’ suffering . Combining functional neuroimaging with multivariate pattern analyses , we identified dissociable patterns that predicted somatic ( high versus low: 100% ) and vicarious ( high versus low: 100% ) pain intensity in out-of-sample individuals . Critically , each pattern was at chance in predicting the other experience , demonstrating separate modifiability of both patterns . Somatotopy ( upper versus lower limb: 93% accuracy for both conditions ) was also distinct , located in somatosensory versus mentalizing-related circuits for somatic and vicarious pain , respectively . Two additional studies demonstrated the generalizability of the somatic pain pattern ( which was originally developed on thermal pain ) to mechanical and electrical pain , and also demonstrated the replicability of the somatic/vicarious dissociation . These findings suggest possible mechanisms underlying limitations in feeling others’ pain , and present new , more specific , brain targets for studying pain empathy .
A fundamental feature of social interactions is our capacity for vicarious experience—the ability to perceive another person’s affective state , reference it to ourselves , and generate an emotional response . This ability provides the foundation for empathy and cooperative behavior ( Smith , 1759 ) by allowing us to recognize and respond to suffering in others ( Batson et al . , 1981 ) , and learn from their experiences ( Olsson and Phelps , 2007 ) . Vicarious experiences of others’ pain , in particular , aid in representing others’ distress with sufficient vividness and importance that we are moved to action . But how do we represent others’ pain , and how vivid and automatic are such representations ? Theories differ markedly on this point . One theory suggests that vicarious pain involves shared experience , activating neural circuits that represent somatic pain in the perceiver . Such theories are based largely on overlapping activity in the dorsal anterior cingulate cortex ( dACC ) and anterior insular ( aINS ) cortices when both experiencing pain and observing pain in others ( Corradi-Dell'Acqua et al . , 2011; 2016; Fan et al . , 2011; Jackson et al . , 2005; 2006a; Lamm et al . , 2007; 2011; Ogino et al . , 2007; Singer et al . , 2004 ) . Such overlaps have also been found in primary and secondary somatosensory areas [i . e . , SI and SII] ( Jackson et al . , 2005; Keysers et al . , 2004; Lamm et al . , 2011; Singer et al . , 2004 ) . Witnessing others in pain can increase one’s own pain ( Langford et al . , 2006; Loggia et al . , 2008 ) , and placebos given for somatic pain can also reduce vicarious ‘pain’ ( Rütgen et al . , 2015 ) . Together , these findings have been taken as brain evidence for shared representation of pain . In particular , dACC and aINS activation are thought to reflect a kind of shared experience that has been described in terms of 'neural resonance' ( Zaki et al . , 2012 ) , and as 'automatic' ( Singer et al . , 2004 ) . By this account , empathy is 'built bottom up from relatively simple mechanisms of action production and perception' ( Iacoboni , 2009 ) . Another theory suggests that vicarious pain may be primarily a reflective , cognitive experience whose experiential qualities are hard to mimic or simulate directly ( Hooker et al . , 2008; Loewenstein , 1996; Morley and Morley , 1993; Zaki , 2014 ) . Others’ pain need not activate somatic pain representations to be aversive ( Chang et al . , 2015 ) , and empathy may work by engaging other emotional systems apart from ‘pain in oneself’ ( Hooker et al . , 2008 ) . Unlike somatic pain , vicarious pain is all too easy to ignore . For example , many participants will inflict substantial pain on an innocent person when instructed to do so by authority figures ( Cheetham et al . , 2009; Milgram , 1963 ) . We also ignore our own past and future pain ( Gilbert et al . , 1998; Loewenstein , 1996; Van Boven and Loewenstein , 2003 ) , making decisions in line with our goals when pain is distant but often reversing those choices when pain is imminent . Brain evidence also suggests that vicarious pain may rely on circuits specialized for representing the thoughts and intentions of others ( Cheetham et al . , 2009; Zaki et al . , 2007; 2009 ) . A network encompassing the dorso-medial prefrontal cortex [dmPFC] , posterior cingulate cortex [PCC] , temporal-parietal junction [TPJ] , and superior temporal sulcus [STS; ( Frith and Frith , 2006; Van Overwalle , 2009; Zaki et al . , 2012 ) ] is reliably involved in ‘mentalizing’—thinking about others’ thoughts ( Saxe and Kanwisher , 2003 ) , preferences ( Mitchell et al . , 2006 ) , and intentions ( Hampton et al . , 2008 ) —and in imagining one’s own and others’ responses to painful stimuli ( Jackson et al . , 2006a; Lamm et al . , 2011 ) . This ‘mentalizing’ network is distinct from both pain-related circuitry and circuitry thought to underlie shared affective experiences and actions , including the dACC and aINS ( Frith and Frith , 2006; Iacoboni , 2009; Shamay-Tsoory et al . , 2009; Singer et al . , 2004; Van Overwalle and Baetens , 2009; Zaki et al . , 2007; 2012 ) . A critical question in assessing shared representation is whether the overlapping brain activity observed really reflects shared pain-related processes . Several recent lines of evidence suggest this may not be the case . In a meta-analysis of over 3500 neuroimaging studies available from Neurosynth . org , activation in the aINS and dACC were among the most frequently observed findings across all kinds of tasks ( Yarkoni et al . , 2011 ) , suggesting that activation in these regions frequently has nothing to do with pain . It is possible that activity in portions of the dACC is preferentially related to pain on average; for example , Lieberman and Eisenberger , 2015 used Neurosynth . org to identify a statistical association between studies using the term 'pain' and activation of the dACC . However , this does not mean that dACC activity is sufficient to infer pain , as the dACC responds to a variety of cognitive and emotional events that are not painful ( Wager et al . , 2016 ) . For instance , electrophysiological and optogenetic studies have identified neurons engaged during foraging behavior , attention , emotion , reward expectancy , skeletomotor and visceromotor activity , and other functions ( Davis et al . , 2005; Picard and Strick , 1996; Shidara and Richmond , 2002 ) . ‘Pain-encoding’ portions of the dACC can be activated in individuals with a congenital insensitivity to pain ( Salomons et al . , 2016 ) . Only a small minority of dACC neurons are pain-related ( Davis et al . , 2005; Hutchison et al . , 1999 ) , and the dACC encodes emotional events , including rejection and general negative emotion , in a way that is distinct from how it encodes pain ( Chang et al . , 2015; Woo et al . , 2014 ) . More fine-grained analyses of population-level representations are required to make inferences about pain based on activity in the dACC and other regions . In this study , we attempted to identify representations ( or markers ) for somatic and vicarious pain—both within the dACC and aINS and across the brain—and test their similarity . For a marker ( e . g . , a multivariate brain pattern ) to serve as a representation of pain , it should satisfy three criteria . It should: a ) closely track pain experience ( be sensitive to the presence of pain ) ; b ) not respond to experiences that are defined as non-painful ( be pain-specific ) ; and c ) generalize across multiple forms of pain ( Woo and Wager , 2015 ) . Only to the degree that brain patterns represent somatic and vicarious pain does testing their similarity bear on shared representation theories . Previous studies [e . g . , ( Corradi-Dell'Acqua et al . , 2011; Corradi-Dell'Acqua et al . , 2016 ) ] have compared the similarity of self-pain and other-pain patterns , but these patterns have not shown strong sensitivity to pain experience [criterion ( a ) above] . Indeed , prior work has suggested that sensitivity to pain and emotional experiences requires combining signals across brain regions and networks ( Chang et al . , 2015; Kassam et al . , 2013; Kragel and Labar , 2013; Nummenmaa et al . , 2014; Wager et al . , 2015 ) , whereas previous studies of vicarious pain have focused only on local patterns within a single region at a time . In addition , the patterns identified as shared across self- and other-pain have responded to other types of negative emotion ( e . g . , emotional pictures , disgust , and unfairness ) , and so do not satisfy criterion ( b ) for pain representations . Finally , no vicarious pain studies have tested generalizability of a specific brain pattern across types of pain [criterion ( c ) ] . Here , in Study 1 , we used a 3 ( stimulation level ) × 2 ( body site ) × 2 ( pain modality; i . e . , somatic vs . vicarious ) experimental design , with pain reports after each trial , to identify patterns that predicted the level of reported somatic and vicarious pain in response to painful heat and observation of pain in others , respectively . Between-participant machine learning analyses were used to identify patterns that test sensitivity [i . e . , satisfy criterion ( a ) above] for both pain modalities ( i . e . , somatic vs . vicarious ) , and to provide an unbiased test of similarity and cross-prediction across somatic and vicarious pain predictive patterns . In addition , we tested somatotopy ( upper versus lower limb ) within sensory cortices and other systems for both pain modalities , and compared the somatotopic representations . The identification of patterns that can be generalized across participants allowed us to test the specificity [criterion ( b ) ] , and generalizability [criterion ( c ) ] of these patterns in additional studies . Previous work has identified a pattern that is , thus far , sensitive and specific [criteria ( a ) and ( b ) above] for somatic pain across multiple studies . We used this pattern , called the Neurologic Pain Signature [NPS; ( Wager et al . , 2013 ) ] , in our primary analyses , and attempted to identify a parallel pattern for vicarious pain . The NPS has over 90% sensitivity and specificity in predicting somatic pain relative to several other salient states , including non-painful warmth , anticipated pain , pain recall ( Wager et al . , 2013 ) , social rejection ( Woo et al . , 2014 ) , and general negative emotion ( Chang et al . , 2015 ) . In the current paper , Study 2 and Study 3 tested the generalizability of the NPS to mechanical and electrical pain , respectively [addressing criterion ( c ) above] . Study 3 also provided a replication of the sensitivity , specificity , and similarity of somatic and vicarious pain predictive patterns . Our primary goal was to identify—and test the similarity of—whole-brain patterns for somatic and vicarious pain that generalize across individuals and can be tested prospectively in multiple studies . However , we also conducted analyses focused on the dACC and aINS specifically , and on pain-predictive patterns identified for each individual participant . In addition , we conducted ‘searchlight’ analyses designed to test the maximal similarity of local brain patterns related to somatic and vicarious pain across the brain , and provide inferences on whether distributed , whole-brain patterns are necessary to capture representations of both somatic and vicarious pain .
In Study 1 , for the somatic pain condition , participants were scanned using functional Magnetic Resonance Imaging ( fMRI ) while they experienced three levels of thermal pain ( 46 , 47 or 48°C; Figure 1A ) on their left volar forearm ( the ‘upper limb or UL’ site ) and left dorsal foot ( the ‘lower limb or LL’ site ) , and rated the intensity of pain with their right hand following every trial . In the vicarious pain condition , participants were scanned while they viewed images depicting injury to others’ right hands ( the ‘upper limb or UL’ site ) and feet ( the ‘lower limb or LL’ site ) and engaged in perspective-taking ( Jackson et al . , 2006a; 2005 ) —participants imagined that the injuries were happening to their own bodies—to actively reference the observed pain to their own bodies . 10 . 7554/eLife . 15166 . 003Figure 1 . Experimental paradigm and behavioral results . ( A ) Trial timeline for somatic and vicarious pain sessions ( differing only by stimulation type and fixation jitter , see 'Materials and methods' for more details ) ; ( B ) Averaged behavioral ratings for somatic and vicarious pain across levels of stimulation and body site ( i . e . , upper limb and lower limb; including within-participant standard error of the mean ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 003 Participants reported increased intensity with increasing levels of somatic pain for both upper limb and lower limb sites with no significant difference between body sites ( tUL ( 27 ) = 12 . 93 p<0 . 0001 , tLL ( 27 ) = 10 . 36 , p<0 . 0001 , tUL-LL ( 27 ) = 1 . 08 , n . s . ; Figure 1B ) . Participants also reported increasing intensity with increasing levels of vicarious pain for both upper limb and lower limb sites , with no significant difference between sites ( tUL ( 27 ) = 11 . 99 , p<0 . 0001 , tLL ( 27 ) = 11 . 37 , p<0 . 0001 , tUL-LL ( 27 ) = 1 . 49 , n . s . ; Figure 1B ) . Intensity ratings were robust and comparable for both somatic and vicarious pain conditions , with no significant difference between mean behavioral ratings for both types of experiences ( tSom-Vic ( 27 ) = −0 . 05 , n . s ) . To test whether activity in the NPS pattern ( Figure 2A ) tracked both somatic and vicarious pain intensity , we calculated the NPS pattern response [the weighted average activation; ( Wager et al . , 2013 ) ] for each single-participant activation map ( regression parameter estimate maps from single-participant general linear models ) for each condition ( 3 stimulation levels × 2 body sites × 2 pain modalities [i . e . , somatic vs . vicarious] ) . This provided a measure of the NPS pattern activation in each condition for each participant , which we analyzed for effects of stimulation level within each body site and for differences across body sites and pain modalities . 10 . 7554/eLife . 15166 . 004Figure 2 . Neurologic pain signature ( NPS ) pattern and analyses . ( A ) Between-participant thresholded ( False Discovery Rate [FDR] q<0 . 05 ) Neurologic Pain Signature ( NPS ) pattern ( Wager et al . , 2013 ) , ( all voxels within the NPS were used in the analyses ) , examples of unthresholded patterns are presented in the insets; small squares indicate voxel weights , black squares indicate empty voxels located outside of the NPS pattern , and red-outlined squares indicate significance at FDR q<0 . 05; ( B ) Signature responses computed as the dot product of the NPS pattern weights and estimated activation maps for each participant ( including within-participant error bars ) ; ( C ) Receiver Operating Characteristic ( ROC ) curves for two-choice forced-alternative accuracies for somatic and vicarious pain , high and low somatic pain , and high and low vicarious pain; ( D ) Participant-Wise NPS Responses for ( I ) somatic and vicarious pain ( accuracySom-Vic = 100% , p<0 . 0001 ) , ( II ) high and low somatic pain ( accuracyHsom-Lsom = 100% , p<0 . 0001 ) , and ( III ) high and low vicarious pain ( accuracyHvic-Lvic = 32% , n . s . ) , showing the direction of response for each participant . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 004 The NPS response increased monotonically for each level of somatic pain for both upper limb and lower limb sites ( tUL ( 27 ) = 9 . 08 , p<0 . 0001 , accuracyHigh-Low= 100% , p<0 . 0001; tLL ( 27 ) = 8 . 88 , p<0 . 0001 , accuracyHigh-Low = 100% , p<0 . 0001; see Figure 2B and C . The NPS response was slightly reduced on the lower limb site ( tUL-LL ( 27 ) = 2 . 13 , p<0 . 05; see Figure 2B and C ) , but this difference was not significant after controlling for rated intensity , implying that the NPS response magnitude is consistent with the reported intensity across both sites . Importantly , the NPS response did not increase across levels of vicarious pain on either site ( tUL ( 27 ) = −1 . 57 , n . s . , tLL ( 27 ) = −1 . 99 , n . s . , tUL-LL ( 27 ) = 1 . 71 , n . s . ; see Figure 2B and C ) . Individual participants also showed the same pattern of NPS responses for both somatic and vicarious pain ( Figure 2D ) . There was a strong de-activation in the NPS response for all vicarious pain conditions , with levels below zero . This decrease was driven by picture-induced activation of regions negatively predictive of pain in the NPS ( e . g . , ventral occipital cortex , superior temporal sulcus , ventromedial prefrontal cortex; see Appendix and Appendix 1—figure 1 for additional details ) . Additional analyses re-training a new somatic pain pattern on this dataset showed similar results ( see Appendix and Appendix 1—figure 2 for more details ) but the NPS is preferred because it was defined a priori and its specificity was validated across multiple datasets . We next sought to identify a distributed pattern of fMRI activity that predicts the intensity of vicarious pain experience ( i . e . , Vicarious Pain Signature or VPS ) . To parallel the development of the NPS ( Wager et al . , 2013 ) , we used LASSO-PCR ( Least Absolute Shrinkage and Selection Operator-regularized Principal Components Regression ) to predict the intensity of reported vicarious pain during pain observation ( Wager et al . , 2013 ) . We used a leave-one-participant-out cross-validation ( see 'Materials and methods' for details; Figure 3A shows the thresholded VPS map using a bootstrap procedure ) to get an unbiased test of responses to both somatic and vicarious pain in held-out individuals . 10 . 7554/eLife . 15166 . 005Figure 3 . Vicarious pain signature ( VPS ) pattern and analyses . ( A ) Between-participant LASSO-PCR ( Least Absolute Shrinkage and Selection Operator-regularized Principal Components Regression ) derived pattern for the Vicarious Pain Signature ( VPS ) , and bootstrap thresholded at FDR q<0 . 05 for display purposes ( all voxels within the VPS were used in the analyses ) , examples of unthresholded patterns are presented in the insets; small squares indicate voxel weights , and red or blue-outlined squares indicate significance at FDR q<0 . 05; ( B ) Cross-validated signature responses computed from the VPS as the dot product of the VPS pattern weights with the estimated activation maps for each participant ( including within-participant standard error of the mean ) ; ( C ) Receiver Operating Characteristic ( ROC ) curves for two-choice forced-alternative accuracies for vicarious and somatic pain , high and low somatic pain , and high and low vicarious pain; ( D ) Participant-Wise VPS Responses for ( I ) vicarious and somatic pain ( accuracyVic-Som = 89% , p<0 . 0001 ) , ( II ) high and low vicarious pain ( accuracyHvic-Lvic = 100% , p<0 . 0001 ) , and ( III ) high and low somatic pain ( accuracyHsom-Lsom = 64% , n . s . ) , showing the direction of response for each participant . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 005 The VPS responded strongly and monotonically to increases in vicarious pain for both upper limb and lower limb sites ( tUL ( 27 ) = 7 . 42 , p<0 . 0001 , accuracyHigh-Low = 89% , p<0 . 0001; tLL ( 27 ) = 10 . 44 , p<0 . 0001 , accuracyHigh-Low = 100% , p<0 . 0001; Figure 3B and C ) , with a reduced response for the lower limb site ( tUL-LL ( 27 ) =2 . 83 , p<0 . 05 ) . Importantly , the VPS showed near-zero responses for all levels of somatic pain , and did not differentiate between somatic pain levels ( tUL ( 27 ) = 1 . 39 , n . s . , tLL ( 27 ) = 0 . 84 , n . s . , tUL-LL ( 27 ) = 0 . 33 , n . s . ; see Figure 3B and C ) . Individual participants also showed the same pattern of VPS responses for both somatic and vicarious pain ( Figure 3D ) . To test how strongly the VPS’s sensitivity and specificity depended on occipital activation—which might be related to enhanced sensory attention—we re-trained the VPS excluding the occipital cortex , with qualitatively identical results ( see Appendix and Appendix 1—figure 3 for more details ) . The signature response monotonically increased for each level of vicarious pain for both the upper limb and lower limb sites ( tUL ( 27 ) = 5 . 75 , p<0 . 0001 , tLL ( 27 ) = 5 . 45 , p<0 . 0001 , accuracyHvic-Lvic = 93% , p<0 . 0001 ) , but did not differentiate between high and low levels of somatic pain ( tUL ( 27 ) = 1 . 56 , n . s . , tLL ( 27 ) = 1 . 23 , n . s . ; tUL-LL ( 27 ) = 0 . 03 , n . s . , accuracyHsom-Lsom = 64% , n . s . ) . Additionally , we examined whether the observed differences between the NPS and VPS patterns were due to differences in the lateralization of stimuli presentation . As in the studies used to define and validate the NPS , somatic pain was administered on the left limbs . Likewise , as in previous literature , the visual stimuli used to evoke vicarious pain showed injuries to the right limbs . This mirrors the side of painful stimulation from the observer’s point of view in an allocentric reference frame , but not an egocentric one; thus , laterality could conceivably play a role . To test this , we repeated the signature response analysis with a left-right flipped version of the VPS pattern ( excluding the occipital cortex ) , which preserves the pattern but with opposite laterality . The results remained the same: The flipped VPS pattern did not track somatic pain intensity ( tUL ( 27 ) = 1 . 47 , n . s . , tLL ( 27 ) = 0 . 04 , n . s . , accuracyHsom-Lsom = 50% , n . s . ) , but did track vicarious pain intensity ( tUL ( 27 ) = 2 . 68 , p<0 . 05 , tLL ( 27 ) = 2 . 53 , p<0 . 05 , accuracyHvic-Lvic = 79% , p<0 . 005 ) . These results indicate that the laterality of the VPS , and by extension the laterality of the stimuli it was trained on , are not an important determinant of its functional properties . As described above , each of the two patterns we identified ( the NPS and VPS ) was influenced by only one type of ‘pain’ induction: Somatic pain induced intensity-dependent responses in the NPS only , and vicarious pain induced intensity-dependent responses in the VPS only . This pattern of results shows separate modifiability of these patterns , a strong inferential criterion by which two processes are functionally independent if experimental manipulations can modify each pattern without affecting the other one ( Sternberg , 2001 ) . It rules out common influences of general shared processes such as increased allocation of attention [e . g . , ( Woo et al . , 2014 ) ] . In addition , the two patterns were anatomically distinct . Regions in the NPS that most reliably predicted somatic pain included aINS , dACC , dorso-posterior insula ( dpINS ) , and SII . By contrast , VPS regions encoding vicarious pain included the dmPFC , amygdala , PCC , and TPJ . Moreover , the predictive weights in the NPS and VPS were orthogonal , showing near-zero spatial correlations across the brain ( r = −0 . 03; 104 , 360 voxels in the NPS & VPS ) and within key regions ( see below ) . Additional analyses showed that both the NPS and VPS were activated in response to the somatic or vicarious pain stimulus specifically during the stimulation period , with the expected time course . Both the NPS and VPS responded only during somatic and vicarious pain events , respectively , with no anticipatory activity ( Figure 4A and B ) . There were small differences in the time courses of responses to somatic and vicarious pain , which are expected because somatic pain summates across time and peaks after stimulus offset ( Koyama et al . , 2004 ) , whereas vicarious pain does not have this physiological property . Critically , we found no evidence for ‘off-target’ responses ( VPS responses to somatic pain or NPS responses to vicarious pain ) at any point during the trial . 10 . 7554/eLife . 15166 . 006Figure 4 . Trial-Level finite-impulse response ( FIR ) analyses . ( A ) Stimulus-related activity for the NPS and VPS for somatic and vicarious pain ( including within-participant standard error of the mean ( shading ) , where the green bar shows stimulus duration; ( B ) Anticipatory ( cue-locked ) activity for the NPS and VPS for somatic and vicarious pain ( including within-participant error ( shading ) , where the green bars show stimulus duration after a jittered pre-stimulus anticipatory fixation . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 006 Another issue that may make the NPS and VPS responses appear more dissimilar is differential habituation or sensitization across trials . To test this , we estimated NPS and VPS amplitudes for each trial using a ‘beta series’ approach ( Mumford et al . , 2012 ) and examined their stability across trials . We did not see any evidence for systematic variation [e . g . , sensitization or habituation; ( Jepma and Wager , 2013 ) ] across trials ( see Appendix and Appendix 1—figure 4 ) . Additionally , we tested whether the VPS pattern was similar to whole-brain patterns we identified as related to ( a ) romantic rejection ( Woo et al . , 2014 ) and ( b ) negative emotion induced by affective pictures ( Chang et al . , 2015 ) , two other types of aversive experience . The VPS was uncorrelated with either pattern ( r = −0 . 03 and r = 0 . 03 , respectively ) , demonstrating that the signature for vicarious pain is not a marker for general negative emotion , unlike previous results ( Corradi-Dell'Acqua et al . , 2011 ) , and supporting the specificity of the VPS for pain empathy . Theories of pain empathy have focused on the dACC and aINS as critical for shared representations of pain affect , and emphasize overlapping activity in spite of differences in the sensory system involved ( e . g . , somatic vs . visual ) . Like previous work ( Corradi-Dell'Acqua et al . , 2011; Lamm et al . , 2011; Singer et al . , 2004 ) , we found strong activation in these regions in both somatic and vicarious pain in standard univariate analyses ( Figure 5A[I] and A[II] ) . We identified regions of overlapping activation in the dACC and aINS ( Figure 5B ) , and tested the similarity and separate modifiability of the pain-predictive patterns only within these ‘shared’ regions . 10 . 7554/eLife . 15166 . 007Figure 5 . Univariate general linear model analyses and multivariate pattern comparisons . ( A ) group level general linear model ( GLM ) analyses for somatic and vicarious pain . ( I . ) : GLM results for somatic pain against baseline thresholded at FDR q<0 . 01; ( II . ) : GLM results for vicarious pain against baseline thresholded at FDR q<0 . 01; ( B ) General Linear Model ( GLM ) results for somatic pain ( in orange ) and vicarious pain ( in purple ) against baseline , thresholded at FDR q<0 . 01 with overlap in the statistically significant regions ( FDR q<0 . 01 ) between somatic and vicarious pain shown in yellow ( these overlapping regions were used for further within-participant cross-prediction analyses ) ; ( C ) Pattern comparison within the anterior insula ( aINS ) and dorsal anterior cingulate cortex ( dACC ) for the NPS and VPS ( black squares in the pattern mask indicate empty voxels not included for the analysis ) ; correlation values are computed using all overlapping voxels in the dACC and bilateral aINS voxels ( only right aINS shown for display ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 007 The NPS and VPS pattern weights within the aINS and dACC were negatively correlated ( aINS: r = −0 . 31 [349 Voxels]; dACC: r = −0 . 41 [1 , 125 Voxels]; see Figure 5C ) , indicating a lack of positive shared representation . Additionally , we used data from each region of interest ( ROI ) to train local patterns predictive of somatic and vicarious pain intensity within individual participants . These patterns were also spatially uncorrelated within each region ( aINS: mean r = 0 . 0185 ± 0 . 0092 , n . s . ; dACC: mean r = 0 . 0350 ± 0 . 0195 , n . s . ) . Further analyses indicated that both dACC and aINS successfully tracked somatic pain ( cross-validated accuracyHigh-Low for aINS: 89% , p<0 . 0001; dACC: 75% , p<0 . 01 ) . Vicarious pain , however , showed marginal within-modality classification ( aINS: 71% , p<0 . 05; dACC: 57% , n . s . ) , suggesting that local patterns may be insufficient to represent vicarious pain . Importantly , cross-prediction analyses showed that the somatic pain pattern did not predict vicarious pain intensity in both aINS and dACC , and vice versa ( all accuracy values < 54% , n . s . ) , demonstrating separate modifiability . Together , these results support the independence of somatic and vicarious pain representations in the aINS and dACC . In addition , they suggest that vicarious pain representation is distributed across regions , and can best be captured in whole-brain but not local analyses . Finally , we used within-individual searchlight analysis ( 8 mm sphere ) to test ( a ) whether any local region of the brain was highly predictive of somatic or vicarious pain , and ( b ) whether there was strong shared representation in any local region ( Kriegeskorte et al . , 2006 ) . The within-modality patterns exhibited a similar spatial topography to our whole-brain analyses ( see Figure 6A–D ) , but no results for vicarious pain survived false discovery rate ( FDR ) correction for multiple testing . Importantly , the effect sizes even for the most predictive regions were an order of magnitude smaller than those observed in our whole brain analyses: The maximal within-modality effect size out of 181 , 129 local regions tested was r2 = 0 . 06 for somatic pain and r2 = 0 . 018 for vicarious pain . In addition , no cross-prediction results survived FDR correction , and the distribution of cross-prediction results across regions was centered on zero for both somatic and vicarious pain ( see Figure 6A–D ) . Again , effect sizes across the brain were small: The largest effects were r2 = 0 . 011 ( somatic to vicarious ) and r2 = 0 . 017 ( vicarious to somatic ) , indicating that local regions are insufficient by themselves to accurately predict ratings , and distributed patterns are required . Together , these results suggest that though there may be weak shared pattern information ( Corradi-Dell'Acqua et al . , 2011; 2016 ) , local patterns in the aINS and dACC are insufficient to serve as representations for vicarious pain . In addition , separate cross-validated prediction analyses within-individuals revealed that whole brain spatial patterns predictive of somatic and vicarious pain were spatially uncorrelated ( mean r = 0 . 0055 ± 0 . 0065 S . E . , n . s . ; see Appendix and Appendix 1—figure 5 ) . 10 . 7554/eLife . 15166 . 008Figure 6 . Within-participant whole-brain and local searchlight analyses . ( A—D ) Distribution of correlations between actual behavioral ratings and estimated behavioral ratings for within and cross-modality predictions from whole-brain and local searchlight analyses . Histograms: Mean outcome correlation from local searchlight analysis computed across participants ( dark shades represent voxels with top 5% of correlation values ) ; Stars with Error bar: Mean outcome correlation computed across participants from whole-brain analysis ( error bars represent standard error of the mean [SEM] ) ; A and B Whole brain and local searchlight results from cross-validated within-modality ( somatic to somatic; vicarious to vicarious ) predictions . Brain maps show voxels with the top 5% correlations; C and D Whole brain and local searchlight results from cross-modality ( somatic to vicarious; vicarious to somatic ) prediction with brain map showing voxels with the top 5% correlations . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 008 The results presented above demonstrate that both the NPS and VPS generalize across upper and lower limb sites . However , our experimental design also allowed us to identify body-site specific representations and provide a preliminary comparison of the somatotopic organization of somatic and vicarious pain . Shared somatotopic organization ( e . g . , if vicarious pain on the upper limb activated upper limb-specific somatosensory regions ) would provide evidence for shared representation , while divergent somatotopy would provide further evidence for differential brain representation . We trained between-participant support vector machine ( SVM ) classifiers to differentiate between upper and lower limb sites separately for somatic and vicarious pain ( we also performed within-participant SVM classification for completeness , with the same results; see Appendix for details ) . We then compared the body site-predictive maps across somatic and vicarious pain . The leave-one-participant-out SVM classifier successfully discriminated somatic pain on upper limb ( UL ) versus lower limb ( LL ) sites ( see Figure 7A; tSom . Pain ( 27 ) = 6 . 90 , p<0 . 0001; accuracyUL-LL= 93% , p<0 . 0001 ) . The regions that made reliable contributions to classification ( as tested with a bootstrap procedure; see 'Materials and methods' for details ) paralleled the somatotopy identified in previous literature , including specific regions of contralateral SI , mid INS , and dpINS ( Baumgärtner et al . , 2010; Brooks et al . , 2005; Hua et al . , 2005; Picard and Strick , 1996 ) . For vicarious pain , SVM classification also discriminated UL versus LL ( see Figure 7A; tVic . Pain ( 27 ) = 6 . 12 , p<0 . 0001; accuracyUL-LL= 93% , p<0 . 0001; within-participant average accuracyUL-LL= 100% , p<0 . 0001 ) . Crucially , the somatotopic patterns for vicarious pain did not include the expected topography in contralateral SI , SII , mid-INS or dpINS , in either hemisphere ( Figure 7B and C ) . Instead , vicarious pain somatotopy was represented in other brain areas , including the supplementary motor area , anterior cingulate ( ACC ) , and medial prefrontal ( mPFC ) cortices ( see Appendix and Appendix 1—figure 6 for additional analyses ) . 10 . 7554/eLife . 15166 . 009Figure 7 . Upper limb versus lower limb multivariate patterns for somatic and vicarious pain . ( A ) Accuracy statistics for upper limb versus lower limb weight maps for somatic and vicarious pain; Receiver Operating Characteristic ( ROC ) curves for two-choice forced alternative tests for upper versus lower limb sites; ( B ) Support Vector Machine ( SVM ) derived pattern for upper limb versus lower limb sites for somatic and vicarious pain , bootstrap thresholded at FDR q<0 . 05 for display purposes , showing somatotopy for these sites in the primary somatosensory cortex ( schematic of homunculus shown for reference ) ; ( C ) SVM derived pattern for upper limb versus lower limb sites for somatic and vicarious pain , bootstrap thresholded at FDR q<0 . 05 for display purposes , showing somatotopy for these sites in the mid- and dorsal-posterior insular cortex ( warm colors indicate upper limb regions and cool colors indicate lower limb sites ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 009 Thus , somatotopy was apparent for both somatic and vicarious pain and strongly predictive of upper versus lower limb stimulation , but the fMRI patterns were qualitatively distinct in each condition . Somatotopy is the primary way of identifying somatosensory cortical representations ( Penfield and Rasmussen , 1950 ) ; thus , our preliminary finding of the expected somatotopy for somatic pain but not vicarious pain suggests that vicarious pain does not involve re-activation of somatosensory representations . Rather , body site-specific representations of vicarious pain may be accomplished using the mPFC ‘mentalizing’ system and perhaps other ideomotor systems . To test the generalizability of the NPS across different types of noxious input , we analyzed data from two additional fMRI studies ( Study 2 and Study 3 ) that used other types of somatic pain . Study 2 ( N = 28 ) used mechanical pressure pain at two intensities—4 kg/cm2 and 6 kg/cm2—applied on the right thumbnail . The NPS responded more strongly to high versus low pressure , demonstrating sensitivity to mechanical pain ( tHigh-Low ( 27 ) = 3 . 12 , p<0 . 005; accuracyHigh-Low = 71% , p<0 . 05; see Figure 8A ) . The discrimination accuracy is limited here by the inclusion of relatively few trials ( 5 per condition ) . Nevertheless , these results show that the NPS successfully predicts intensity in the somatic pain regardless of the site of the stimulation or type of somatic pain . 10 . 7554/eLife . 15166 . 010Figure 8 . NPS and VPS responses for Study 2 and Study 3 . ( A ) Study 2 – NPS responses for mechanical ( pressure pain ) delivered at low and high intensities , ( i . e . , 4 kg/cm2 and 6 kg/cm2 ) computed as the dot product of the NPS pattern weights and estimated activation maps for each participant ( including within-participant standard error of the mean; total of five trials per level of stimulation per participant; ( B ) Study 3 – NPS ( right ) and VPS ( left ) responses for electrodermal ( electrical or shock ) pain and observed pain , computed as the dot product of the respective pattern weights for NPS ( right ) and VPS ( left ) and estimated activation maps for each participant ( including within-participant standard error of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 15166 . 010 Study 3 ( N = 15 ) used electrodermal ( electrical or shock ) pain delivered to the left ankle , and also included an observed pain condition with visual stimuli depicting pain on upper and lower limbs . However , perspective-taking was not employed in this study—the instructions emphasized observation only rather than engage in perspective-taking—so VPS activation was expected to be weaker than in Study 1 . The NPS responded more to shock pain than observed pain ( tShock-Obs ( 14 ) = 5 . 58 , p<0 . 0001; accuracyShock-Obs = 100% , p<0 . 0001; see Figure 8B , left ) . Conversely , the VPS responded to more to observed pain than shock pain ( tObs-Shock ( 14 ) = 3 . 05 , p<0 . 005; accuracyObs-Shock = 73% , n . s . ; see Figure 8B , right ) . Thus , the findings from Study 3 show that the NPS and VPS dissociate somatic versus observed pain in an independent sample .
Most previous studies of vicarious pain point out its similarity with somatic pain , and it is thus widely believed that the two experiences rely on the same systems . Why are our findings and conclusions different ? There are three main reasons . First , many previous studies focused only on the points of similarity—mainly identified in two isolated brain regions , dACC and aINS—ignoring dissimilarities [but cf . ( Lamm et al . , 2011 ) , who write , 'our results reveal more differences than similarities…' p . 2500] . Here , we aimed for an unbiased assessment of the two processes . Second , most previous work has identified ‘pain-related’ activation by contrasting pain with loosely matched control conditions ( e . g . , neutral or innocuous stimuli ) . Overlapping activity in such contrasts may be caused by many processes that are not pain , including general negative affect , attention , and arousal [see ( Wager et al . , 2016 ) for discussion] . Engagement of these processes may be responsible for the similar activation in previous work . By contrast , in our study , we attempted to isolate pain-relevant patterns that predicted the magnitude of experienced intensity , and examined the similarity of those patterns . Third , most previous studies focused on voxel-wise activation ‘blobs , ’ rather than on multivariate patterns , which can be sensitive to information at finer spatial scales ( Shmuel et al . , 2010 ) , possibly even below the intrinsic resolution determined by the voxel size ( Kamitani and Tong , 2005 ) . This property , combined with our experimental approach targeting within-person variations in pain intensity , suggests that the multivariate patterns we identified are more likely to reflect specific representations contained in meso-scale neural circuits . Based on our findings , we infer that the overlapping activation in the dACC , aINS , and other areas is not related to shared pain experience . Interestingly , on close reading , the few previous multivariate pattern-based studies agree broadly with this interpretation . The brain patterns they identified as shared across somatic and vicarious pain were not specific to ‘pain , ’ as these patterns were also activated by other , non-painful types of negative affect ( Corradi-Dell'Acqua et al . , 2016; Zaki et al . , 2016 ) . As in our study , in these studies what is shared does not seem to be particular to pain per se . Somatic and vicarious pain experiences , by definition , are similar in some ways and different in others . They are similar in that they are both aversive , salient , and attention-capturing . They differ in their sensory input modalities ( somatosensory versus visual ) and likely many of the cognitive processes involved . Brain activity , including both regional activation and multivariate patterns , may reflect any or all of these processes . For this reason , it is necessary to test the similarity of the brain patterns that most closely encode pain experiences of each type . We have proposed three criteria for identifying patterns that may be used to make inferences about pain representation . Such patterns should be ( a ) sensitive to pain , and thus respond in proportion to its experience; ( b ) specific to pain , and not activated by non-painful events; and ( c ) generalizable across varied instances of the type of pain . For example , a brain pattern that encodes ‘pain affect’—whether common to self- and other-pain or not—should increase as pain affect increases , and not respond to non-painful forms of negative emotion such as disgust ( Zaki et al . , 2016 ) . Previous studies have identified both shared and dissimilar patterns of activity ( Corradi-Dell'Acqua et al . , 2011; 2016 ) in local dACC and aINS regions . Though an important step , these patterns have not been shown to track pain experience with meaningfully large effect sizes , and our findings here suggest that local regions are insufficient to do so . In addition , they have not identified a pattern that tracks ‘pain affect’ specifically and does not respond to other negative emotions . Our findings provide an advance because they identify patterns that are strongly predictive of pain experience . Though they do not test many other kinds of emotions , the patterns are demonstrably specific to either pain for oneself or another . In addition , a lack of correlations between the NPS , VPS , and patterns predictive of rejection and negative emotion further suggest that these patterns are not sensitive to general emotional arousal . Fortunately , the between-participant patterns we identified here can be prospectively tested for generalizability to other emotions and paradigms in future studies . Indeed , in an initial step towards broader generalizability , we showed that the sensitivity and specificity of both patterns generalize to new studies , and that our putative somatic pain representation , the NPS , generalizes across three types of somatic pain . The fact that capturing vicarious pain experience requires identifying distributed patterns across large-scale networks is consistent with prior work suggesting that accurate predictions of pain and emotional experiences requires combining signals across brain regions and networks ( Brodersen et al . , 2012; Cecchi et al . , 2012; Chang et al . , 2015; Kassam et al . , 2013; Kragel and Labar , 2013; Nummenmaa et al . , 2014; Wager et al . , 2015 ) . This work fits in with evidence for population coding in motor processing ( Georgopoulos et al . , 1992 ) , spatial processing ( Fried et al . , 1997 ) , and other domains ( Davis and Poldrack , 2013; Haxby et al . , 2014; Panzeri et al . , 2015 ) . It also converges with views of somatic pain as represented by patterns across populations of neurons distributed across macroscopic brain regions ( Coghill et al . , 1999; Leknes and Tracey , 2008; Melzack and Wall , 1967 ) . Thus , it may be useful to explicitly consider population coding hypotheses in future studies , at both the fMRI and neural levels of analyses . Our approach is not biased towards or against finding common or distinct representations . It is grounded in widely accepted approaches to testing cross-stimulus type prediction ( Bruneau et al . , 2013; Etzel et al . , 2008; Kourtzi and Kanwisher , 2001; Parkinson et al . , 2011 ) . However , our approach is not guaranteed to be maximally sensitive to shared representations . Optimizing for shared representation would require using machine-learning approaches to predict pain in a modality-nonspecific fashion . Another important issue is individual differences in the experience of vicarious pain . Having established pain-predictive brain patterns as we do here , subsequent studies could include stimuli that are externally matched , and brain responses in the 'vicarious pain signature' are correlated meaningfully with empathy ratings across participants . In addition , the VPS could potentially be used to test clinical populations: for example , whether patients in chronic pain show greater ( or lesser ) responses to others’ suffering . These questions are beyond the scope of the present study , but could be undertaken in future work . Our study used conditions designed to maximize the neural overlap between vicarious and somatic pain , by using ‘painful’ pictures previously demonstrated to induce stronger somatosensory activity—and are thus potentially more somatic pain-like—than other , ‘cued empathy’ paradigms ( Lamm et al . , 2011 ) . In addition , we instructed participants to adopt a perspective-taking stance previously shown to enhance activation in dACC , aINS , and somatosensory cortex ( Jackson et al . , 2006a ) . Further studies are needed to assess the impact of instructions , mental stance , and which aspects of pain ( e . g . , sensory , affective ) are being evaluated on the activation of the VPS ( Zaki et al . , 2016 ) , and to assess generalizability across different types of vicarious pain paradigms . This work provides a specific target signature , the Vicarious Pain Signature ( VPS ) , which can be used prospectively for empathy-related activity in future studies . The VPS may help us understand factors that promote or impair vicarious pain experience , and its impact on prosocial behavior ( Craig et al . , 2010 ) . Perceiving others’ pain does not appear to recruit the same neural circuitry as experiencing the pain ourselves . Rather than recruiting our somatosensory system to understand another’s pain , we use processes involved in representing another’s mental state . The lack of direct representation of others’ pain in somatic pain systems provides a mechanism for understanding why we might systematically under-weigh others’ painful experiences , including their suffering ( Gilbert et al . , 1998; Loewenstein , 1996; Van Boven and Loewenstein , 2003 ) , and substantiates Adam Smith’s insight from 250 years ago that our moral sentiments are grounded in our cognitive rather than sensory faculties . | The ability to experience others’ pain is a cornerstone of empathy , and binds us together in times of hardship . However , we have not yet fully understood the complex interactions in the brain that make people empathetic to others’ suffering . One possibility is that we experience others’ pain through the activation of the same brain regions as those that enable us to experience physical pain ourselves . To test this idea , Krishnan et al . compared patterns of brain activity in human volunteers as they experienced pain ( from heat being applied to their forearm or foot ) or watched images of others’ hands or feet being injured . While watching these images , the volunteers were asked to try to imagine that the injuries were happening to their own bodies . The patterns of brain activity that arose when the volunteers observed someone else in pain did not overlap with the patterns produced when the volunteers experienced pain themselves . Instead , seeing someone else in pain activated regions involved in taking another person’s perspective . This process , which is known as mentalizing , involves thinking about the other person’s thoughts , intentions and preferences . Thus within the brain , the experience of observing someone else in pain is distinct from that of experiencing physical pain in oneself . The results presented by Krishnan et al . raise new questions about how the brain regions involved in empathy help us to relate to other people when they experience different types of pain . Future studies should explore the factors that influence our ability to adopt another’s perspective , and whether it might be possible to improve this ability . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"neuroscience"
] | 2016 | Somatic and vicarious pain are represented by dissociable multivariate brain patterns |
Stress induced by cytoplasmic protein aggregates can have deleterious consequences for the cell , contributing to neurodegeneration and other diseases . Protein aggregates are also formed within the endoplasmic reticulum ( ER ) , although the fate of ER protein aggregates , specifically during cell division , is not well understood . By simultaneous visualization of both the ER itself and ER protein aggregates , we found that ER protein aggregates that induce ER stress are retained in the mother cell by activation of the ER Stress Surveillance ( ERSU ) pathway , which prevents inheritance of stressed ER . In contrast , under conditions of normal ER inheritance , ER protein aggregates can enter the daughter cell . Thus , whereas cytoplasmic protein aggregates are retained in the mother cell to protect the functional capacity of daughter cells , the fate of ER protein aggregates is determined by whether or not they activate the ERSU pathway to impede transmission of the cortical ER during the cell cycle .
Asymmetric cell division is a mechanism that generates cells with different properties . Specifically , in Saccharomyces cerevisiae , asymmetric cell division allows daughter cell rejuvenation while ensuring that cellular damage is left behind in the mother cell ( Henderson and Gottschling , 2008; Kaganovich et al . , 2008; Spokoini et al . , 2012; Gallagher et al . , 2014; Nystrom and Liu , 2014; Zhou et al . , 2014 ) . Recent studies have revealed that cytoplasmic protein aggregates are retained in the mother cell , although the underlying mechanism ( s ) that establishes such an asymmetric mode of inheritance remains to be fully elucidated ( Abbas et al . , 2013 ) . Furthermore , little is known about whether such asymmetric division is regulated during the cell cycle . The endoplasmic reticulum ( ER ) is a gateway for the secretory pathway in eukaryotic cells . Proteins that are secreted or reside within the organelles of the secretory pathway initiate their journey when they are translocated into the membrane or lumen of the ER . In the unique oxidizing environment of the ER , nascent polypeptides undergo chaperone assisted folding and modifications , such as glycosylation and formation of disulfide bonds , to become mature active proteins before exiting from the ER ( Mori , 2000; Rutkowski and Kaufman , 2004; Ron and Walter , 2007 ) . In addition , the ER is a major site for lipid synthesis and storage of intracellular calcium ( Mcmaster , 2001 ) . Many of these ER functions must work in concert to satisfy cellular demands ( Oakes and Papa , 2014 ) . The unfolded protein response ( UPR ) -signaling pathway is a conserved response to ER stress and plays a critical role in maintaining ER function by up-regulating the transcription of genes coding for ER chaperones and protein-folding components ( Cox et al . , 1993; Mori et al . , 1993; Ron and Walter , 2007; Wu et al . , 2014 ) . Importantly , the ER cannot be synthesized de novo and is generated only from existing ER . Given the critical function of the ER , it seems likely that cell cycle regulatory mechanisms must exist to ensure inheritance of a fully functional ER during cell division . Recently , we reported the existence of a cell cycle surveillance mechanism or ‘checkpoint’ in S . cerevisiae that safeguards the inheritance of functional ER by the daughter cell ( Bicknell et al . , 2007; Babour et al . , 2010 ) . Upon ER stress induction , activation of this ER Stress Surveillance ( ERSU ) pathway results in re-localization of the cytokinesis-associated septin complex away from the bud neck , leading to a block in ER inheritance and cytokinesis . We showed that the ERSU pathway is independent of the UPR and is mediated by the Slt2 Mitogen-Activated Protein Kinase ( MAPK ) . In the absence of Slt2 , cells do not exhibit the block in ER inheritance and the septin ring remains at the bud neck following exposure to ER stress , similar to normally dividing , unstressed cells . Ultimately , however , slt2Δ cells are not able to sustain their growth due to the transmission of the stressed ER into the daughter cell . In fact , preventing ER transmission into slt2Δ daughter cells by genetic or pharmacological inhibition of actin polymerization can restore growth . Importantly , while Slt2 MAPK is known to play a role in the cell wall integrity ( CWI ) pathway , we found that the ERSU and CWI pathways are completely distinct ( Babour et al . , 2010; Levin , 2011 ) . The discovery of the ERSU pathway thus not only identified a novel cell cycle checkpoint that ensures the inheritance of functional ER but also raised a number of important questions about the underlying mechanisms . Furthermore , it is also unclear how the ER contents , including misfolded proteins , are segregated during the cell cycle . Under normal growth conditions , terminally misfolded proteins in the ER are retro-translocated into the cytoplasm and degraded by proteasomes in a process known as ER-associated degradation ( ERAD ) ( Hampton , 2002; Bukau et al . , 2006; Vembar and Brodsky , 2008; Smith et al . , 2011; Thibault and Ng , 2012 ) . When misfolded ER proteins are overexpressed or the ERAD function is diminished , the damaged proteins accumulate into large foci within the ER lumen . A recent study proposed that these large ‘aggregate’-like foci are selectively retained in the mother cell via a mechanism that depends on the lateral ER diffusion barrier established by the septin ring at the bud neck ( Clay et al . , 2014 ) . Such lateral diffusion barriers between the mother and daughter yeast cells have been proposed to play pivotal roles in preventing undesirable materials , such as protein aggregates , from transferring to the daughter cells . While the exact mechanisms that establish the mother–daughter diffusion barrier remain to be elucidated , the barrier was reported to be formed as soon as the new bud emerges and depends on the bud site selection component GTPase , Bud1 ( Clay et al . , 2014 ) . This study thus presented an attractive model suggesting that ER protein aggregate inheritance is regulated similarly to that of large protein aggregates in the cytoplasm , such as Q-bodies , JUNQ ( juxta-nuclear quality control compartment ) and IPOD ( insoluble protein deposit ) , which are actively retained in the mother to protect the daughter cell from toxicity of the protein aggregates ( Kaganovich et al . , 2008 ) . However , a potentially unique feature of ER protein aggregate inheritance is that it could be affected by inheritance of the ER itself . To further our understanding of how ER protein aggregates are divided between mother and daughter cells , we investigated the distribution of ER protein aggregates in relation to the inheritance of the ER .
To investigate the distribution of both the ER and ER protein aggregates between the mother and daughter cell , we monitored the distribution of a mutant form of the vacuolar protein carboxypeptidase Y ( CPY* ) fused to mRFP in cells also expressing Hmg1-GFP , a well-characterized ER marker ( Finger et al . , 1993; Nishikawa et al . , 2001; Spear and Ng , 2005; Clay et al . , 2014 ) . A single amino acid change in CPY* ( G255R ) leads to improper folding , and the protein accumulates in the ER ( Finger et al . , 1993 ) . Expression of CPY*-mRFP was placed under the control of the galactose ( GAL1 ) promoter and induced by incubation in galactose-containing media . After 2 hr of induction , CPY*-mRFP formed aggregate-like foci that co-localized with both the cortical ER ( cER ) and perinuclear ER ( pnER ) ( Figure 1A ) . We quantitated and evaluated the number of CPY*-mRFP foci in individual cells according to the daughter cell ( bud ) size ( Figure 1C ) . A small number ( <20% ) of cells with small bud size ( less than 2 μm in length; classified as class I cells ( Babour et al . , 2010 ) , transferred CPY* foci to the bud , while the majority of cells ( ∼80% ) contained foci only in the mother ( Figure 1C ) . In contrast , ∼50–60% of class II ( medium sized buds , larger than 2 μm in length ) and class III ( large buds with the nucleus and pnER in the bud ) cells transferred CPY* foci to the bud ( Figure 1C ) . 10 . 7554/eLife . 06970 . 003Figure 1 . Inheritance of CPY* and CFTR aggregates by daughter cells . ( A ) Wild-type ( WT ) cells expressing galactose ( Gal ) -inducible CPY*-mRFP and the endoplasmic reticulum ( ER ) marker Hmg1-GFP were grown in 2% ( w/v ) Gal for 2 hr and then visualized by microscopy . Note that some Hmg1-GFP foci co-localized with CPY*-mRFP foci . ( B ) Cells expressing copper-inducible GFP-CFTR and the ER marker DsRed-HDEL were grown in copper-containing medium for 2 hr and then visualized . Note that DsRed-HDEL also co-localized with Cystic Fibrosis Transmembrane conductance Regulator ( CFTR ) foci . ( C and D ) Quantification of daughter cells containing CPY*-mRFP ( C ) or GFP-CFTR ( D ) foci at different stages of the cell cycle ( small-budded cells , less than 2-µm length [class I]; medium-budded cells , greater than 2-µm length [class II] , and large-budded cells containing nuclear ER [class III] ) . Error bars represent the standard deviation ( SD ) and were generated from at least three independent experiments with n > 100 cells . ( E ) CPY*-mRFP and GFP-CFTR foci are detergent insoluble and are present in the pellet fraction after detergent extraction . ( T ) Total , ( S ) supernatant , ( P ) pellet . These tests were previously utilized to characterize protein aggregates in cells ( Alberti et al . , 2010 ) and thus , we termed CPY*-mRFP and GFP-CFTR foci as aggregates throughout our study . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 003 We also examined the inheritance of Cystic Fibrosis Transmembrane conductance Regulator ( CFTR ) , which has been shown to form foci within the yeast ER ( Fu and Sztul , 2003 ) . In mammalian cells , a large proportion of newly synthesized wild-type ( WT ) CFTR is not properly folded and is ultimately degraded ( Lukacs et al . , 1994; Ward and Kopito , 1994; Jensen et al . , 1995; Ward et al . , 1995; Moyer et al . , 1998; Gnann et al . , 2004; Younger et al . , 2006; Turnbull et al . , 2007 ) . Similarly , only a minor fraction of translated CFTR actually reaches the plasma membrane in yeast ( Kiser et al . , 2001; Zhang et al . , 2001 ) . GFP-CFTR foci were observed in the ER of cells expressing another well-characterized ER marker , DsRed-HDEL ( Figure 1B ) . Significantly , most daughter cells , regardless of their class , inherited GFP-CFTR foci ( Figure 1B , D ) . Foci formed after expression of CPY* in yeast are often referred to and treated as protein aggregates without biochemical characterization . Therefore , we subjected CPY* foci to a well-established detergent extraction test commonly used to characterize protein aggregates ( Alberti et al . , 2010 ) . The crude cell extracts prepared from CPY*-mRFP- or GFP-CFTR-expressing cells were treated with or without detergent and then fractionated by differential centrifugation . The majority of CPY*-mRFP was found in the insoluble protein pellet fraction regardless of detergent pre-treatment ( Figure 1E ) . Similar results were obtained for GFP-CFTR ( Figure 1E ) . These data indicate that the CPY* and CFTR foci observed here meet the definition of protein aggregates according to previous studies ( Simons et al . , 1995; Sondheimer and Lindquist , 2000; Nishikawa et al . , 2001; Alberti et al . , 2010 ) , and we therefore refer to CPY*-mRFP and GFP-CFTR foci as aggregates throughout this study . Because both CPY*-mRFP and GFP-CFTR form aggregates in the ER , we tested whether the difference in their transmission to daughter cells might lie in the different effects of the aggregates on ER function . Previous studies have reported that expression of CPY* , but not CFTR , induces the UPR ( Chaudhuri et al . , 1995; Zhang et al . , 2001 ) . Indeed , we found that induction of CPY*-mRFP resulted in the expression of a UPR reporter ( UPRE-GFP , Figure 2A , lane 3 ) , and this was further increased upon treatment of cells with the glycosylation inhibitor tunicamycin ( Tm ) , a well-characterized ER stress inducer ( Figure 2A , lane 4 ) ( Cox et al . , 1993; Mori et al . , 1993 ) . We also asked whether CPY* aggregates activate the ERSU pathway , which functions to ensure the inheritance of functional cER ( Babour et al . , 2010 ) . We found that a majority of the class I daughter cells did not inherit the cER under the ER stress condition evoked by CPY*-mRFP expression , and cER inheritance was also diminished in class II and III cells , but to a much lesser extent ( Figure 2B , C ) . Additionally , the magnitude of the cER inheritance block caused by CPY*-mRFP expression was less than that induced by Tm treatment ( Figure 2C , E , F ) . Finally , we found that 63% of class I , 73% of class II , and 60% of class III daughter cells containing the cER also contained CPY* aggregates ( Figure 2G; yellow vs gray bars ) . Taken together , these data indicate that for CPY*-mRFP-expressing cells , more than 65% of buds that inherited ER also contained at least one CPY* aggregate ( Figure 2—figure supplement 1A ) . Conversely , 57% of buds without aggregates also lacked the ER ( Figure 2G , light blue vs gray ) . 10 . 7554/eLife . 06970 . 004Figure 2 . Differential inheritance of CPY* and CFTR aggregates and cER by daughter cells . ( A ) CPY*-mRFP expression activates the unfolded protein response ( UPR ) pathway . WT cells expressing CPY*-mRFP and the UPR reporter UPRE-GFP were incubated alone or with 1 . 0 μg/ml Tm , and GFP expression was quantified in individual cells . N > 100 cells per experiment; error bars ( SD ) were generated from at least three independent experiments . Dex: dextrose control medium , Gal; galactose-containing medium . ( B ) Quantitation of cortical ER ( cER ) in the buds of unstressed cells or cells treated with 0 . 5 or 1 . 0 μg/ml Tm for 3 hr cER inheritance was grouped by bud size: small-budded cells ( class I ) , medium-budded cells ( class II ) , and large-budded cells containing nuclear ER ( class III ) . ( C ) CPY*-mRFP expression for 2 hr blocks cER inheritance . ( D ) GFP-CFTR expression for 2 hr in copper-containing media does not block cER inheritance . ( E and F ) Exposure to mild ER stress with 0 . 5 μg/ml Tm ( E ) blocks cER inheritance ( induces the ER Stress Surveillance ( ERSU ) pathway ) to a similar extent as does 1 . 0 μg/ml Tm ( F ) . ( G ) Distribution ( % ) of cells at different stages of the cell cycle in which the daughter cells contain both cER and CPY*-mRFP aggregates ( yellow ) , cER but not CPY*-mRFP aggregates ( gray ) , and neither cER nor CPY*-mRFP aggregates ( pale blue ) . Panel shows representative images of the most abundant cell types with CPY*-mRFP and Hmg1-GFP . ( H ) Same as G except that cells expressed GFP-CFTR and DsRed-HDEL . n > 100 cells were counted per experiment , repeated at least 3 times to generate error bars representing SD . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 00410 . 7554/eLife . 06970 . 005Figure 2—figure supplement 1 . Inheritance of the ER and CPY* aggregates at different stages of the cell cycle . ( A ) Distribution of CPY*-mRFP aggregates per bud in all ( Total ) , class I , and class II + III , and unstressed cells ( without Tm ) . ( B ) cER inheritance by different classes of WT cells expressing CPY*-mRFP and left untreated or subjected to additional ER stress with Tm ( 1 μg/ml , 2 hr ) . ( C ) Percentage of daughter cells containing CPY*-mRFP aggregates following incubation with or without Tm . ( D ) Percentage of daughter cells containing cER and CPY*-mRFP aggregates ( yellow ) , cER but no CPY*-mRFP aggregates ( gray ) , and neither cER nor CPY*-mRFP aggregates ( blue ) for each class of cells . ( E ) Distribution of CPY*-mRFP aggregates per bud in all ( Total ) , class I , and class II + III Tm-treated cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 00510 . 7554/eLife . 06970 . 006Figure 2—figure supplement 2 . Inheritance of cER and CFTR aggregates at different stages of the cell cycle . ( A ) cER inheritance by different classes of WT GFP-CFTR-expressing cells treated without or with Tm ( 1 μg/ml , 2 hr ) . GFP-CFTR expression was induced by incubation in copper-containing medium ( Cu ) . Graphs ( B ) and ( C ) are the same as those in Figure 2D , F but are shown again for comparison . ( D ) Percentage of daughter cells containing GFP-CFTR aggregates following incubation with or without Tm . ( E and F ) Percentage of daughter cells containing cER and GFP-CFTR aggregates ( yellow ) , cER but no GFP-CFTR aggregates ( gray ) , and neither cER nor GFP-CFTR aggregates ( blue ) for each class of cells . Cells were untreated ( E ) or treated with 1 μg/ml Tm ( F ) . ( E ) The graph shown is the same as Figure 2H and is presented for comparison . ( G and H ) Distribution of GFP-CFTR aggregates per bud in all ( Total ) , class I , and class II + III cells untreated ( G ) or treated with Tm ( H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 00610 . 7554/eLife . 06970 . 007Figure 2—figure supplement 3 . Colocalization of cER and CPY* or CFTR aggregates . ( A ) Two representative fields of cells expressing the ER marker Hmg1-GFP and treated with galactose ( Gal ) for 2 hr to induce expression of CPY*-mRFP . ( B ) Representative field of Hmg1-GFP-expressing cells treated with galactose ( Gal ) and Tm for 2 hr to induce expression of CPY*-mRFP . ( C ) Representative field of cells expressing the ER marker DsRed-HDEL and treated with 100 μM copper ( Cu ) and Tm for 2 hr to induce expression of GFP-CFTR . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 007 In contrast to the findings in CPY*-expressing cells , we found that cER inheritance was not affected by expression of GFP-CFTR ( Figure 2D ) unless the cells were also subjected to Tm treatment ( Figure 2—figure supplement 2A–C ) . Therefore , CFTR aggregation presented an opportunity to evaluate the transmission of ER protein aggregates independently of the ER inheritance block . Quantitation of the number of CFTR aggregates and cER in the mother and daughter cells showed that virtually all class I , II , and III daughter cells contained CFTR aggregates ( Figure 2H: yellow and Figure 2—figure supplement 2G ) , indicating no preferential retention of CFTR aggregates by the mother cells . Thus , although CPY* and CFTR both form ER protein aggregates in yeast cells , preferential retention of aggregates was only observed in the CPY*-expressing cells , which also displayed the cER inheritance block . Finally , the cER inheritance block and asymmetric distribution of CFTR aggregates was observed in GFP-CFTR-expressing cells after Tm treatment ( Figure 2—figure supplement 2D–F , H ) , strengthening the relationship between ER inheritance and ER protein aggregate distribution . Recently , it was reported that misfolded ER proteins were also retained in the mother cell when cells were exposed to relatively low levels of ER stress ( 0 . 5 μg/ml of Tm ) , as measured by Kar2sfGFP fluorescence recovery after photobleaching ( FRAP ) analysis ( Clay et al . , 2014 ) . In our experiments , the moderate expression level of UPRE-GFP revealed that CPY*-mRFP aggregates induced a moderate level of ER stress ( Figure 2A , lane 3 ) and also induced a mild block in ER inheritance compared with that induced by 0 . 5 μg/ml of Tm ( compare Figure 2C , E , F , and Figure 2—figure supplement 3A ) . Based on the previous report , we anticipated that CPY*-mRFP expression alone ( which induced low/medium levels of ER stress ) should result in retention of CPY* aggregates in the mother cells . We observed , however , that CPY*-mRFP aggregates were distributed in both mother and daughter cells ( Figure 2G , Figure 2—figure supplement 1A and Figure 2—figure supplement 3A ) . We also examined the effect of the compounded ER stress by treating CPY* aggregate-expressing cells with Tm ( Figure 2—figure supplement 1B , E , and Figure 2—figure supplement 3B ) . As anticipated , the combined ER stresses further decreased ER inheritance by the daughter cells ( Figure 2—figure supplement 1B ) . Notably , distribution of CPY* aggregates to the daughter cells was also further diminished in these cells ( Figure 2—figure supplement 1C–E ) and correlated with the reduced level of cER transmission ( Figure 2—figure supplement 1B ) . Likewise , distribution of CFTR aggregates to the daughter cells was only diminished in cells in which ER inheritance was blocked by Tm treatment ( Figure 2—figure supplement 2A , D , and Figure 2—figure supplement 3C ) . Collectively , these data therefore demonstrate that the magnitude of the ER inheritance block is affected by the level of ER stress , and that preferential retention of CPY* or CFTR aggregates in the mother cell is not an intrinsic property of the protein aggregates themselves , but rather , is dictated by ER inheritance . We reasoned that if the ERSU pathway-dependent ER inheritance regulates the distribution of protein aggregates to the daughter cell , then cells lacking the ERSU pathway should also show diminished retention of protein aggregates in the mother cell . To test this , we examined the distribution of CFTR and CPY* aggregates in slt2Δ cells , which are incapable of blocking ER inheritance in response to ER stress . We reported previously that Tm ( 1 μg/ml ) treatment does not block cER inheritance in slt2Δ cells ( Babour et al . , 2010 ) and this was also observed in slt2Δ cells when ER stress was induced by CPY* aggregate expression ( Figure 3A +Gal and Figure 3—figure supplement 1A; compare to Figure 2C for WT cells ) . Notably , entry of the cER into the daughter slt2Δ cells was paralleled by the entry of CPY* aggregates , which contrasts with the ERSU-dependent block in both cER and aggregate inheritance by WT daughter cells ( Figure 3D , G , Figure 3—figure supplement 1A for slt2Δ vs 2G , Figure 2—figure supplement 3A for WT ) . As shown above , CFTR-expressing WT cells only exhibited a block in ER inheritance and CFTR aggregate transmission when treated with Tm ( Figure 2—figure supplement 2A , D , and Figure 2—figure supplement 3C: Tm- vs Tm+ ) . However , Tm treatment of CFTR-expressing slt2Δ cells failed to block either ER inheritance ( Figure 3B , C , and Figure 3—figure supplement 1B vs Figure 3—figure supplement 1C ) or CFTR aggregate entry into the daughter cell ( Figure 3E , F , H , and Figure 3—figure supplement 1D ) . Taken together , these data revealed that ER inheritance , which is ultimately regulated by the ERSU pathway , regulates protein aggregate transmission into the daughter cell . 10 . 7554/eLife . 06970 . 008Figure 3 . Inheritance of ER protein aggregates is ERSU dependent . ( A ) cER inheritance was not blocked in slt2Δ cells upon ER stress induction with CPY*-mRFP for 2 hr . ( Compare to cER inheritance in CPY*-mRFP expressed WT cells [Figure 2C] ) . ( B ) GFP-CFTR expression for 2 hr has no impact on cER inheritance in slt2Δ cells . ( C ) GFP-CFTR expression for 2 hr in the presence of 1 µg/ml Tm does not block cER inheritance in slt2Δ cells . ( D ) Distributions ( % ) of slt2Δ cells that contain cER and CPY*-mRFP aggregates in daughter cells ( yellow ) , cER but not CPY* aggregates ( gray ) , and no cER and no aggregates ( light blue ) in different stages of cell cycle . ( E and F ) Distributions ( % ) of slt2Δ cells that contain cER and GFP-CFTR aggregates in daughter cells ( yellow ) , cER but not GFP-CFTR aggregates ( gray ) , and no cER and no aggregates ( light blue ) in different stages of cell cycle treated with ( F ) or without ( E ) Tm ( 1 . 0 µg/ml ) . ( G ) Representative images of slt2Δ cells in class I , II , and III with CPY*-mRFP and Hmg1-GFP to mark the ER . ( H ) Representative images of slt2Δ cells in class I , II , and III with GFP-CFTR and DsRed-HDEL ER marker . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 00810 . 7554/eLife . 06970 . 009Figure 3—figure supplement 1 . Colocalization of ER and CPY* or CFTR in slt2Δ cells . ( A ) Representative field of Hmg1-GFP-expressing slt2Δ cells treated with galactose ( Gal ) for 2 hr to induce expression of CPY*-mRFP . ( B ) Representative field of DsRed-HDEL-expressing cells treated with 100 μM copper ( Cu ) for 2 hr to induce expression of GFP-CFTR . ( C ) Representative field of DsRed-HDEL-expressing cells treated with 100 μM copper ( Cu ) and Tm for 2 hr to induce expression of GFP-CFTR . Note that slt2Δ cells do not exhibit a block in ER inheritance under ER stress . ( D ) Distribution of CPY*-mRFP aggregates per bud in all ( Total ) , class I , and class II + III cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 009 As described in the introduction , previous work has suggested that a diffusion barrier limiting transmission of the cER between mother and daughter cells is formed during bud emergence via the activity of the Ras-like GTP-binding protein , Bud1 ( Clay et al . , 2014 ) . Therefore , we examined whether Bud1 deficiency affected either cER inheritance or the distribution of ER protein aggregates . We found that the cER inheritance behavior of bud1Δ cells and WT cells subjected to ER stress by treatment with Tm ( Figure 4A ) or expression of CPY*-mRFP ( Figure 4B ) was similar . Furthermore , transmission of CPY* aggregates was also similar in the two strains ( Figure 4C ) . These data suggest that Bud1 is not involved in the distribution of ER protein aggregates ( see ‘Discussion’ ) . 10 . 7554/eLife . 06970 . 010Figure 4 . BUD1 deletion has no effect on inheritance of the ER and CPY* aggregates . ( A ) cER inheritance is blocked to similar extents in WT and bud1Δ cells upon ER stress induction ( 1 μg/ml Tm for 3 hr ) . ( B ) bud1Δ cells display a normal block in ER inheritance upon exposure to ER stress induced by CPY*-mRFP expression for 2 hr . ( C ) Percentage of bud1Δ daughter cells containing CPY*-mRFP aggregates in class I , II , and III cells . ( D ) Distribution ( % ) of cells at different stages of the cell cycle in which daughter cells contain cER and CPY*-mRFP aggregates ( yellow ) , cER but not CPY*-mRFP aggregates ( gray ) , and neither cER nor CPY*-mRFP aggregates ( light blue ) . ( E ) Distribution of CPY*-mRFP aggregates per bud in all , class I , and class II + III bud1Δ cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 010 In the experiments described above , we noted that a significant population of class II and III daughter cells contained the cER even after Tm treatment ( Figure 2B–F ) . We considered that these cells might be incapable of inducing the ER inheritance block or that they may be daughter cells that had already inherited the cER before induction of ER stress . In both cases , one would expect that the retention of the stressed ER , and thus , of ER protein aggregates , in the mother cell would be limited only to cells with small daughter cells . To investigate these possibilities , we synchronized yeast cells in G1 by incubation with α-factor . After washing to remove α-factor , the cells were allowed to proceed normally through the cell cycle for either 20 min ( Figure 5A; phase I cells ) or 50 min ( Figure 5B; phase II cells ) before ER stress was induced by addition of Tm . To unambiguously identify the original mother cells present at the time of α-factor arrest , we fluorescently labeled cells by incubation with Texas Red ( TR ) -conjugated ConA ( TR-ConA ) during the α-factor treatment ( Figure 5A , B ) . Thus , after washout of both α-factor and TR-ConA , newly emerging daughter cells will be TR-negative , while the mother cell remains TR-positive ( Figure 5A–C , and Figure 5—figure supplements 2 , 3 ) . 10 . 7554/eLife . 06970 . 011Figure 5 . Activation of the ERSU pathway varies with the cell cycle stage . ( A ) ER stress was induced by treating synchronized WT cells ( phase I: 20 min after α-factor release ) with 1 μg/ml Tm . Cells were incubated with Texas Red ( TR ) -ConA during the α-factor treatment and then washed before exposure to Tm . Mother cells ( TR-positive , arrows ) can thus be distinguished from daughter cells emerging after induction of ER stress ( TR-negative , arrowheads ) , as shown in the upper schematic . Cells were analyzed by DIC and fluorescence microscopy at the indicated times prior to and after addition of Tm . ( B ) As described for A , except that Tm was used to induce ER stress in phase II cells ( 50 min after α-factor release ) . Many cells underwent cytokinesis , as evident from the presence of unbudded TR-positive and TR-negative cells . ( C ) Quantification of TR-positive and TR-negative phase I ( purple ) and phase II ( gray ) cells at the time points indicated . ( D ) Quantification of cER inheritance by the daughters of phase I ( E ) and phase II ( F ) cells upon Tm treatment for the indicated times . ( E and F ) Hmg1-GFP-expressing phase I ( E ) and phase II ( F ) cells were treated as shown in A and B , and cER inheritance was evaluated at the indicated times . ( G and H ) UPR induction occurred regardless of the time of addition of Tm . HAC1 mRNA splicing was measured as an indicator of UPR induction in phase I ( G ) and phase II ( H ) cells . Northern blotting of HAC1 mRNA was performed at the indicated times after Tm treatment . Positions of the spliced and unspliced HAC1 mRNA are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 01110 . 7554/eLife . 06970 . 012Figure 5—figure supplement 1 . Phase I and phase II cells activate the ERSU pathway . ( A ) Percentage of phase I ( purple ) and phase II ( gray ) cells that showed one bud or two buds after Tm treatment . Error bars indicate SD from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 01210 . 7554/eLife . 06970 . 013Figure 5—figure supplement 2 . ER stress induction in cells at an early stage of the cell cycle . ( A ) A schematic diagram of the experimental set-up for examination of phase I cells . WT cells were synchronized with α-factor , washed , and treated with 1 μg/ml Tm 20 min later ( phase I: 20 min after α-factor release ) . ( B ) Representative fields of cells treated as in ( A ) and incubated with Texas Red ( TR ) -ConA during the α-factor treatment . Mother cells ( TR-positive ) can thus be distinguished from daughter cells emerging after induction of ER stress ( TR-negative ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 01310 . 7554/eLife . 06970 . 014Figure 5—figure supplement 3 . Activation of the ERSU pathway varies with the cell cycle stage . ( A ) A schematic diagram of the experimental set-up for examination of phase II cells . WT cells were synchronized with α-factor , washed , and treated with 1 μg/ml Tm 50 min later ( phase II: 50 min after α-factor release ) . ( B ) Representative fields of cells treated as in ( A ) and incubated with Texas Red ( TR ) -ConA during the α-factor treatment . Mother cells ( TR-positive ) can thus be distinguished from daughter cells emerging after induction of ER stress ( TR-negative ) . Many cells underwent cytokinesis , as evident from the presence of TR-positive and TR-negative cells at 110 and 230 min . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 01410 . 7554/eLife . 06970 . 015Figure 5—figure supplement 4 . Activation of the ERSU pathway in cells at an early stage of the cell cycle . ( A ) A schematic diagram of the experiment shown in Figure 5E , shown again here for clarity . ER stress was induced by treating α-factor-synchronized Hmg1-GFP-expressing WT cells ( phase I: 20 min after α-factor release ) with 1 μg/ml Tm , and cER inheritance was evaluated at the indicated time points . Two representative cells at each time point are shown . For the 200 min time point ( 3 hr after addition of Tm ) , representative daughter cells containing only the pnER ( 85% of total cells ) or both the pnER and cER ( 15% of total cells ) are shown for comparison . Another set of representative cells is shown in Figure 5E . ( B ) Two representative fields of cells are shown for each time point . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 01510 . 7554/eLife . 06970 . 016Figure 5—figure supplement 5 . Activation of the ERSU pathway in cells at a later stage of the cell cycle . ( A ) A schematic diagram of the experiment shown in Figure 5F , shown again here for clarity . ER stress was induced by treating α-factor-synchronized Hmg1-GFP-expressing WT cells ( phase II: 50 min after α-factor release ) with 1 μg/ml Tm , and cER inheritance was evaluated at the indicated time points . Cells underwent cytokinesis , as evident from the presence of unbudded cells at 110 min ( also see Figure 5B , C ) . Two representative cells at each time point are shown . For the 230 min time point ( 3 hr after addition of Tm ) , representative daughter cells containing only the pnER ( 80% of total cells ) or both the pnER and cER ( 20% of total cells ) are shown for comparison . Another set of representative cells is shown in Figure 5F . ( B ) Two representative fields of cells are shown for each time point . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 016 We found that phase I cells exhibited a cytokinesis block and did not undergo cell division at 80 min after α-factor release ( 1 hr after Tm addition; Figure 5A , C and Figure 5—figure supplement 2 ) . Even at 200 min after release ( 3 hr post-Tm ) , >85% of phase I cells remained undivided ( Figure 5A , C; purple bars and Figure 5—figure supplement 2 ) . Furthermore , in most phase I cells , the cER remained in the mother cell ( Figure 5D–E and Figure 5—figure supplement 4 ) . In contrast , at 50 min after α-factor release ( before induction of ER stress ) , almost all of the phase II daughter cells had already inherited the cER ( Figure 5D , F and Figure 5—figure supplement 5 ) . After Tm addition , these cells underwent cytokinesis ( cell division ) , and at 1 hr after Tm addition ( 110 min ) , ∼50% of cells were derived from TR-positive mother cells , and the remaining ∼50% were TR-negative and were derived from the first daughter cell that emerged after α-factor release ( Figure 5B , C , 110 min gray bars and Figure 5—figure supplement 3 ) . After division , the number of cells with the cER in the daughter cell was small ( Figure 5D , F; 110 min , and Figure 5—figure supplement 5 ) . The observed differences between phase I and phase II cells in cytokinesis and ER inheritance were not due to an inability of phase II cells to respond to Tm . Phase I and II cells showed similar degrees of UPR activation after ER stress , as reflected in the levels of spliced HAC1 mRNA resulting from activated Ire1 RNase-mediated mRNA cleavage ( Figure 5G , H ) . Intriguingly , phase II cells exhibited a block in both cytokinesis and cER inheritance block at the second round of division , and the daughter cell arising from the first cell division did not further divide . Instead , we observed cells with two daughter cells . This was also observed for phase I cells in which the second daughter cell started to emerge after 3 hr of Tm treatment . At this point ( Tm , 3 hr ) , ∼13% of phase II and ∼20% of phase I cells had two buds ( Figure 5—figure supplement 1 ) . These results therefore demonstrate that cells in which the cER is already in the daughter cell at the time of ER stress induction proceed through cytokinesis once , but display blocks in both cytokinesis and cER inheritance in the next cell cycle . Finally , we considered that if ER protein aggregates are preferentially retained in the mother cell independently of the ERSU pathway , then the ER stress levels in the mother cell should also be elevated relative to the daughter cells . To test this , we used a FRAP assay with WT cells expressing Kar2/BiP-sfGFP reporter ( a major ER chaperone fused to ‘superfolder’ GFP ) , which displays significantly better folding in the oxidizing ER luminal environment than GFP or EGFP ( Pedelacq et al . , 2006; Lai et al . , 2010; Aronson et al . , 2011; Lajoie et al . , 2012 ) . In both mammalian and yeast cells , Kar2/BiP binding to unfolded client proteins increases in response to ER stress , reducing its mobility within the ER lumen ( Snapp et al . , 2006 ) . This can be monitored by the reduced rate of FRAP . Small areas of the cER or pnER ( indicated by black rectangles in Figure 6 ) in the mother and daughter cell were photobleached and the rate of Kar2/BiP-sfGFP fluorescence recovery from the surrounding area was assessed . In the mother cell , the fluorescence recovery rate was significantly reduced by Tm treatment when compared with control DMSO-treated cells , and this was similar for both the cER and pnER ( Figure 6A , B ) . However , there were no significant differences between the mother and daughter cell in recovery rates in either the cER or pnER under control or Tm-treated conditions ( Figure 6A , B ) , indicating that the ER stress levels are identical in mother and daughter cells . Taken together , the data presented do not support the preferential retention of ER protein aggregates in the mother cell , but instead argue that ER inheritance regulates the fate of unfolded proteins and the inheritance of ER protein aggregates by the daughter cell . 10 . 7554/eLife . 06970 . 017Figure 6 . Mother and daughter cells display similar degrees of cortical and perinuclear ER stress . ( A ) Quantification of ER stress was performed by fluorescence recovery after photobleaching ( FRAP ) of cells labeled with the ER chaperone Kar2/BiP-sfGFP . Cells were exposed to DMSO or Tm ( 1 μg/ml ) and then discrete regions of cER ( indicated by the black rectangles ) in mother ( blue ) and daughter ( red ) cells were photobleached and recovery was monitored . ( B ) Cells were treated as in A , except that FRAP was monitored in the indicated regions of the perinuclear ER ( pnER; rectangles ) . cER and pnER stress , as indicated by the rate of FRAP , was comparable in untreated or Tm-treated mother and daughter cells . The results are the average of three independent experiments , each of which analyzed at least seven independent cells under both DMSO and Tm-treated conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 017
In budding yeast , the decision of whether or not to transfer specific cellular components or organelles to the daughter cell is critical for the health of the new generation . Limiting the transmission of potentially harmful components ensures that the functional capacity of the new daughter cell is reset ( Shcheprova et al . , 2008; Liu et al . , 2010; Peraza-Reyes et al . , 2010; Hughes and Gottschling , 2012; Longo et al . , 2012; Ouellet and Barral , 2012 ) . A recent report demonstrated that ER-resident misfolded protein aggregates , such as a mutant form of carboxypeptidase ( CPY* ) , were prevented from entering the daughter cell and were retained within the mother cell ( Clay et al . , 2014 ) . The expression of certain misfolded proteins or protein aggregates , including CPY* , in the ER lumen induces ER stress ( Spear and Ng , 2003 ) , as also shown here . In turn , ER stress activates the ERSU pathway , which blocks ER transmission into the daughter cell ( Babour et al . , 2010 ) . These observations raise important questions about the potential mechanism ( s ) underlying the lack of CPY* aggregates in the daughter cell . One explanation is that transmission is blocked concomitantly with the ERSU pathway-mediated block in ER inheritance . Alternatively , independent regulatory mechanisms may also control the segregation of ER protein aggregates . The results of our experiments described here unambiguously demonstrated that the ERSU pathway governs the location of ER protein aggregates . We found that CPY*-mRFP aggregates alone activate both the UPR and the ERSU pathway ( Figure 2A , C ) , although activation was significantly less robust than when induced by Tm ( even at 0 . 5 μg/ml ) and was more equivalent to a low/moderate level of ER stress ( Figure 2E , F ) . We found that a majority of class I cells ( 83% ) retained CPY*-mRFP aggregates in the mother and only 17% of daughter cells had inherited the aggregates . At face value , these numbers appear to indicate that CPY*-mRFP aggregates are retained in the mother cells . However , daughter cells lacking the inherited cER should not contain CPY*-mRFP aggregates ( Figure 2G ) . In fact , ER stress induced by CPY*-mRFP aggregate expression blocked cER inheritance in a large population of the class I daughter cells . Thus , when the analysis was restricted to daughter cells that contain the cER , we found that 63% of these cER-positive daughter cells also contained CPY*-mRFP aggregates ( Figure 2G ) . Even for cells with larger buds ( class II or III ) , we observed that CPY*-mRFP localization was dictated by the presence of the cER in the daughter cell . Thus , we conclude that the cellular distribution of CPY*-mRFP aggregates is determined by the location of the cER . Our conclusions are further supported by the observations with GFP-CFTR; we found that GFP-CFTR aggregates did not activate the ERSU pathway , in agreement with the lack of UPR activation ( Zhang et al . , 2001 ) . If a separate mechanism exists to retain ER protein aggregates in the mother cell irrespective or independently of the ER inheritance block , we would expect that GFP-CFTR aggregates should be retained in the mother cells . However , we found that GFP-CFTR aggregates entered the daughter cells effectively ( Figure 2H ) , and there was no evidence for preferential retention of aggregates in the mother cell . Finally , further support for the hypothesis that the distribution of ER protein aggregates is dictated by the ER inheritance status came from our study of ERSU-deficient slt2Δ cells . These cells do not undergo a block in ER inheritance in response to ER stress and accordingly , slt2Δ daughter cells contained higher levels of CPY*-mRFP aggregates than do WT CPY*-mRFP-expressing daughter cells ( Figure 3 ) . One notable difference in the behavior of CPY*-mRFP and GFP-CFTR aggregates was that only CPY*-mRFP aggregates induced the ERSU pathway ( Figure 2C vs Figure 2D ) . The aggregates had similar biochemical behavior , as indicated by their detergent insolubility ( Figure 1E ) . Interestingly , GFP-CFTR aggregates were present in almost all of the cER-positive small-budded cells ( class I: <2-μm diameter , gray vs yellow bars in Figure 2H ) , whereas CPY*-mRFP aggregates were present in only about half of cER-positive small-budded cells ( gray vs yellow bars in Figure 2G ) . A similar trend was found after treatment with Tm ( Figure 2—figure supplement 1D vs Figure 2G ) . These data suggest that GFP-CFTR aggregates behave like soluble proteins in the ER lumen , perhaps because of the lack of proteotoxicity . In this regard , analysis of GFP-CFTR aggregates could provide a unique opportunity to monitor the behavior of soluble non-toxic ER luminal proteins in live cells . We found that the behavior of CPY*-mRFP and GFP-CFTR aggregates in bud1Δ cells and WT cells was indistinguishable , indicating that the mother–daughter lateral diffusion barrier established by Bud1 does not play a significant role in either the cER inheritance block or the distribution of ER protein aggregates in response to ER stress . Our data differ from those in a previous study ( Clay et al . , 2014 ) , which showed an increase in CPY*-GFP foci in bud1Δ daughter cells and thus suggested an important role for Bud1 in the retention of CPY*-GFP foci in the mother cell . Paradoxically , that study observed only one or two CPY*-GFP foci in WT daughter cells and up to six CPY*-GFP foci in bud1Δ daughter cells . Interestingly , Barral and colleagues argued that the greater abundance of CPY*-GFP foci in the bud1Δ daughter cells was due to increased transfer of the soluble form of CPY*-GFP from the mother and subsequent formation of CPY*-GFP aggregates in the daughter cell ( Clay et al . , 2014 ) . However , there was no direct demonstration that soluble CPY*-GFP levels were in fact increased . During our analysis ( n > 100 cells per experiment with at least three independent repeats ) , we never observed significant differences in the average number of CPY*-mRFP aggregates in WT and bud1Δ daughter cells , or even in WT and bud1Δ mother cells ( data not shown ) . The reason for the different findings in the two studies is currently not clear . The ERSU pathway functions in at least two different yeast strain backgrounds ( W303 and BY4741 ) . Interestingly , bud1Δ mother cells appear to have a shorter replicative lifespan than WT mother cells under optimal stress-free growth conditions . If CPY* aggregates are retained in the WT mother cell in order to preserve ER proteostasis in the daughter cell and thus ensure its longevity , one may anticipate that bud1Δ mother cells , which contain fewer CPY* aggregates , would have better ER proteostasis and thus live longer than WT mother cells ( Clay et al . , 2014 ) . Together with other data presented here , our finding that CPY*-mRFP aggregate distribution correlates with the lack of the ER inheritance block in bud1Δ cells strengthens the hypothesis that the ERSU pathway regulates the distribution of ER protein aggregates in yeast daughter cells . Our results suggest a potentially significant difference in the manner in which cells cope with cytoplasmic vs ER protein aggregates . To date , several distinct cytoplasmic protein aggregates have been reported: ubiquitinated cytoplasmic proteins associate with JUNQ , whereas insoluble proteins associate with IPOD upon proteasome inactivation ( Kaganovich et al . , 2008 ) or in the presence of amyloid proteins such as Huntingtin with extended polyQ . In cells with functional proteasomes , misfolded proteins dynamically form inclusion bodies called Q-body protein aggregates ( Spokoini et al . , 2012; Roth and Balch , 2013; Zhou et al . , 2014 ) . An interesting feature of these cytoplasmic protein aggregates is their differential subcellular localization; JUNQ associates with the ER/nucleus , IPOD is found next to the vacuole ( Kaganovich et al . , 2008; Ogrodnik et al . , 2014; Polling et al . , 2014 ) , and Q-bodies are scattered throughout the cytoplasm ( Spokoini et al . , 2012; Escusa-Toret et al . , 2013; Zhou et al . , 2014 ) . These protein aggregates are selectively retained in the mother cell , and a recent study has revealed an intriguing mechanism for retention of Q-body protein aggregates through association with the ER and eventually with the mitochondria ( Zhou et al . , 2014 ) . We found that fluorescently tagged forms of both CPY* and CFTR aggregates appeared to be scattered throughout the ER . Ultimately , the ERSU pathway blocks ER inheritance and prevents transmission of ER protein aggregates into the daughter cell . The asymmetric distribution of cytosolic protein aggregates and large inclusions such as JUNQ and IPOD is protective and allows the daughter cells to be rejuvenated during each cell division . ER aggregates may similarly be prevented from entering the daughter cell . But during ER stress , ER aggregates are retained by the ERSU mechanism that prevents transfer of the stressed ER . Cytosolic protein aggregates have been reported to associate with the ER , raising the intriguing possibility that the ER functions as a central controller for the distribution of protein aggregates in the cell . Alternatively , the association of cytoplasmic protein aggregates with the ER may somehow induce ER stress and consequently , the ERSU pathway . In this scenario , the ERSU pathway may function as a master regulator for both the ER and cytoplasmic protein aggregates . Many of the cell cycle checkpoints that ensure accurate DNA replication and genome transmission are restricted to specific stages of the cell cycle ( Rhind and Russell , 2012; Yasutis and Kozminski , 2013 ) . By observing asynchronous populations of yeast cells , we found that both class II and class III cells exhibit the block in ER inheritance during ER stress , but it is less pronounced than in class I cells , as we described in our initial report . This finding suggests that ER stress must be recognized early in the cell cycle in order to induce the ERSU pathway . Significantly , ER transfer to the daughter cell had already taken place in many class II and III cells prior to encountering ER stress , and presumably , these populations contribute to the lower number of cells exhibiting the cER inheritance block . Using synchronized cells , we found that cells at later stages of the cell cycle , when the cER is already in the daughter cell at the time of exposure to ER stress , undergo cytokinesis for the first round of the cell cycle but exhibit a block in cER inheritance and cytokinesis during the second round . Thus , the ERSU pathway is effective only when ER stress is sensed at an early stage of the cell cycle and can be ignored until the second cell cycle if the cER was already present in the daughter cell . Such a mode of ERSU pathway regulation of ER inheritance is rather unusual when compared to cell cycle checkpoints that regulate DNA replication and transmission ( Vleugel et al . , 2012; Hayashi and Karlseder , 2013 ) . Failure to align chromosomes properly , for example , activates the spindle assembly checkpoint , leading to inhibition of the anaphase-promoting complex and induction of cell death ( Chang and Barford , 2014 ) . To our knowledge , the observation that class II and III cells with cER-containing daughter cells proceed normally through the first round of cytokinesis after ER stress , and that this ‘error’ is not corrected until the second round , is unprecedented . Results of the FRAP experiments indicate that ER stress was manifested within 30 min of stress induction in both class II or III daughter cells . Thus , these cells must somehow bypass the ERSU pathway-induced cytokinesis block in the first round of division but halt the cell cycle during the next round of cytokinesis . In the case of the replication checkpoints , it would not be possible to lose a chromosome at the first division and then recover during the second cell cycle . In contrast , a functionally stressed ER might be tolerated if the problem is resolved promptly in the next cell cycle . Currently , the molecular mechanisms dictating the decisions by class I , II , and III cells to proceed—or not—through cytokinesis are unknown; the answers to these and other questions raised here await further studies .
Cells expressing the Kar2/BiP-sfGFP reporter were grown to mid-log phase in filter-sterilized 0 . 5X YPD ( 0 . 5% yeast extract , 1% peptone , and 2% dextrose ) and treated with Dimethyl Sulfoxide ( DMSO ) or tunicamycin ( Tm 1 μg/ml ) for 3 hr at 30°C . Cells were transferred to 1 . 6% agarose pads made with 0 . 5× YPD ± 1 μg/ml Tm and the pads were maintained at 30°C for the duration of the experiment . Photobleaching was achieved with one 0 . 2-s pulse from a 488-nm argon laser set to 50% power using an Applied Precision optical sectioning microscope ( 100× 1 . 65 Apo objective , immersion oil n = 1 . 78 [Cargille Laboratories] ) with softWoRx version 3 . 3 . 6 ( Applied Precision , Issaquah , WA ) . To compare multiple FRAP events on a single graph , we calculated the fluorescence recovery by determining the relative intensity of the bleached region compared with the unbleached region and defining the bleaching event as 0 and complete recovery as 1 for each photobleached cell . The average fluorescence recovery curves were obtained by averaging the fluorescence recovery values at the same time points for each strain . Images were acquired immediately before and at 6-s intervals after the photobleaching event . Cells ( WT-MNY1037 , slt2Δ-MNY1043 , bud1Δ-MNY2112 , these and all other yeast strains used in this study are described in Table 1 ) expressing Hmg1-GFP were treated with DMSO or Tm ( 1 μg/ml ) unless otherwise indicated , for 3 hr during mid-log phase , imaged with fluorescence microscopy , and scored for the presence or absence of cER in class I , class II , and class III buds . An Axiovert 200M Carl Zeiss Micro-Imaging microscope with a 100× 1 . 3 NA objective was used as described previously ( Babour et al . , 2010 ) . 10 . 7554/eLife . 06970 . 018Table 1 . Yeast strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 018Strain nameGenotypeReferenceMNY1037MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1::HMG1-GFP:URA3 , ade2-1 , his3-11 , 15::UPRE-lacZ:HIS3 ( Babour et al . , 2010 ) MNY2215MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1 , ade2-1 , his3-11 , 15::HIS3 , bar1Δ::LEU2 , DsRed-HDEL::ADE2 ( Babour et al . , 2010 ) MNY1000MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1 , ade2-1 , his3-11 , 15 ( Cox et al . , 1993 ) MNY1002MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1::HMG1-GFP::URA3 , ade2-1 , his3-11 , bar1Δ::LEU2 ( Bicknell et al . , 2007 ) MNY2119MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1 , ade2-1 , his3-11 , 15::UPRE-lacZ:HIS3 KAR2sfGFP::KanMXThis studyMNY2702MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1::4xUPRE-GFP::URA3 , ade2-1 , his3-11 , 15This studyMNY1043MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1::HMG1-GFP:URA3 , ade2-1 , his3-11 , 15::UPRE-lacZ:HIS3 slt2Δ::KanMX ( Babour et al . , 2010 ) MNY2112MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1::HMG1-GFP:URA3 , ade2-1 , his3-11 , 15::UPRE-lacZ:HIS3 bud1Δ::KanMXThis studyMNY2825MATa , leu2-3 , 112 , trp1-1 , can1-100 , ura3-1 , ade2-1 , his3-11 , 15::HIS3 , bar1Δ::LEU2 , DsRed-HDEL::ADE2 slt2Δ::KanMXThis study S . cerevisiae strains ( WT-MNY1037 , slt2Δ-MNY1043 , bud1Δ-MNY2112 ) were transformed with pFJP10 ( pRS425-GAL1-CPY*-mRFP , this and all other plasmids used in this study are described in Table 2 ) . Cells were grown overnight on SCD-Leu with 4% raffinose . Cells were then diluted to OD 0 . 06 , grown to OD 0 . 25 , and then either 2% dextrose ( ±1 μg/ml Tm ) or 2% galactose ( ±1 μg/ml Tm ) was added . Cultures were further incubated at 30°C for 2 hr before cER inheritance , and CPY*-mRFP foci formation were analyzed by fluorescence microscopy . The yeast strains carrying DsRed-HDEL ( WT-MNY2215 , slt2Δ-MNY2825 ) were transformed with pCU426CUP1/EGFP-CFTR ( a gift from Dr Elizabeth Sztul [Fu and Sztul , 2003] ) . Cells were diluted to OD 0 . 06 , grown to OD 0 . 25 , and then 100 µM copper sulfate ±1 μg/ml Tm was added . Cultures were further incubated at 30°C for 2 hr before cER inheritance , and GFP-CFTR foci formation were analyzed by fluorescence microscopy . 10 . 7554/eLife . 06970 . 019Table 2 . Plasmids used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 06970 . 019Plasmid nameConstructReferencepFJP1pFA6a-sfGFP-HDEL::KanMX6This studypRH12094XUPRE-GFP::URA3 ( pJCI86-GFP ) ( Cox et al . , 1993 ) pCU426CUP1/EGFP-CFTRpRS426-CUP1-EGFP-CFTR ( Fu and Sztul , 2003 ) pFJP10pRS425-GAL1-CPY*-mRFPThis study pRH1209 ( 4xUPRE-GFP::URA3 ) plasmid ( a gift from Dr Randy Hampton ) ( Hampton et al . , 1996 ) was digested with restriction enzyme StuI and transformed into yeast strain MNY1000 for genomic integration at the URA3 locus to generate strain MNY2702 . MNY2702 was then transformed with pFJP10 to induce CPY*-mRFP expression as described above . Yeast strain MNY1002 cells were treated with 50 ng/ml α−factor for 2 . 5 hr , washed twice with an equal volume of fresh YPD containing 1 M sorbitol , diluted to OD 0 . 25 , and then allowed to recover for either 20 min ( phase I ) or 50 min ( phase II ) before the addition of 1 μg/ml Tm . Cells were imaged at the indicated time points after Tm addition . For staining of synchronized cells , 200 μg/ml Texas Red-conjugated concanavalin A ( TR-ConA , Sigma , St . Louis , MO ) was added to the culture for the final 30 min of α−factor treatment , and the culture was incubated in the dark at 30°C with shaking . The cells were washed to remove TR and α−factor and then treated as described above , except that the TR-labeled cell cultures were maintained in the dark . Cells were collected and imaged by fluorescence microscopy at the indicated time points . Cells were synchronized as described above . At the appropriate time points , 20-ml aliquots of cells were collected and flash frozen in liquid nitrogen . Total RNA was extracted as described before ( Chawla et al . , 2011 ) . Samples of 20 μg of total RNA were separated on a 4 . 5% agarose gel with 6 . 7% formaldehyde and transferred to a zeta-probe membrane ( Bio-Rad , Hercules , United States ) in 10× SCC overnight . After UV-crosslinking , membranes were probed with a radiolabeled HAC1 DNA probe as described in ( Chawla et al . , 2011 ) . pFA6a-GFP::KanMX6 ( a gift from Jurg Bahler and John Pringle , Addgene plasmid # 39292 ) ( Bahler et al . , 1998 ) was modified by replacing GFP with sfGFP-HDEL . sfGFP-HDEL was PCR amplified from plasmid psfGFP-HDEL ( a gift from Dr Erik Snapp ) and switched with GFP in pFA6a-GFP::Kanr to generate pFJP1 ( pFA6a-sfGFP-HDEL::KanMX6 ) . The plasmid was checked by sequencing . sfGFP-HDEL::KanMX6 was PCR amplified with primers FJP17 ( ATAAATTAACAACCTTGAAGCTTCCAGCAGCAAAAATTTTTAACTATTTTATgaattcgagctcgtttaaac ) and FJP37 ( CAGTCTCTATACTCTTCAATG ) to tag KAR2 at the genomic locus of strain MNY1004 to generate strain MNY2119 using the Longtine method ( Longtine et al . , 1998 ) . Aliquots of 20 ml of cells were collected after 2 hr of induction , washed once with water , and flash frozen in liquid nitrogen . Protein extracts were prepared and the following detergent solubility test was performed ( as described in [Alberti et al . , 2010] ) to characterize the CPY*-mRFP and GFP-CFTR foci observed under the microscope . The cell pellet was resuspended in 600 μl of lysis buffer ( 50 mM Tris , pH 7 . 5 , 150 mM NaCl , 2 mM ethylenediaminetetraacetic acid ( EDTA ) , 5% glycerol ) with protease inhibitors ( 5 mM phenylmethylsulphonyl fluoride [PMSF] , aprotinin , leupeptin , pepstatin A ) , 50 mM N-ethylmaleimide ( NEM ) , and 25 μM MG132 , and then lysed with glass beads at 4°C . A sample of 300 μl of the lysate was mixed with 300 μl of cold detergent-lysis buffer ( 50 mM Tris , pH 7 . 0 , 150 mM NaCl , 1% Triton X-100 , 0 . 5% deoxycholate , 0 . 1% sodium dodecyl sulfate , ( SDS ) and vortexed for 10 s . The remaining 300 μl of sample was diluted with an additional 300 μl of lysis buffer . The crude lysates were centrifuged for 2 min at 800 rcf ( 4°C ) to pellet the cell debris . Supernatant samples ( 250 μl each ) were centrifuged in a TLA 100 rotor for 30 min at 80 , 000 rpm and 4°C using a TL Beckman ultracentrifuge . The pellet from the RIPA buffer lysate was resuspended in 250 μl of RIPA buffer . The pellet from the lysis buffer only was resuspended in 250 µl of lysis buffer ( no detergent ) . Equal volumes of unfractionated ( total ) , supernatant ( S ) , and pellet ( P ) samples were incubated in sample buffer containing 2% SDS and 2% β-mercaptoethanol . CPY*-mRFP extracts were heated for 5 min at 95°C , and GFP-CFTR extracts were heated for 20 min at 37°C . Samples were analyzed by SDS-PAGE and Western blotting . CPY*-mRFP was detected with anti-mRFP rat monoclonal antibody ( 1:1000 , cat # ABIN334653 , Life Technologies , Carlsbad , CA , United States ) and secondary HRP-conjugated anti-rat antibody ( Bio-Rad ) . GFP-CFTR was detected with anti-GFP mouse monoclonal ( 1:1000 , cat # 11814460001 , Roche , Basel , Switzerland ) and secondary HRP-conjugated anti-mouse antibody ( Bio-Rad ) . | Many species of yeast form new cells by a process known as budding in which a small daughter cell ‘buds’ out of a larger mother cell . Mothers can only produce a limited number of buds before they die of old age . However , age is reset in the daughters to ensure that they are fully rejuvenated when born . Therefore , the mother cell needs to prevent the factors that cause aging and cell damage from entering the daughter . Inside cells , proteins are made and folded correctly in a structure called the endoplasmic reticulum . If proteins are not folded properly , they are normally rapidly destroyed . However , if a cell requires lots of proteins to be made quickly , this can sometimes overwhelm and ‘stress’ the endoplasmic reticulum . When this occurs , proteins start misfolding and clump up to form toxic aggregates , some of which collect inside the endoplasmic reticulum . The Endoplasmic Reticulum Stress Surveillance ( ERSU ) pathway monitors the health of the endoplasmic reticulum and prevents ‘stressed’ endoplasmic reticulum from entering daughter cells , which can cause them to die . By visualizing the endoplasmic reticulum and the aggregates contained within it during budding in the yeast species Saccharomyces cerevisiae , Piña and Niwa have now found that the ERSU pathway can also prevent these aggregates from entering daughter cells . However , if the ERSU pathway is not switched on—as may be the case if the level of endoplasmic reticulum stress is very low—then aggregates can enter the daughter cells . This is in contrast to protein aggregates that form elsewhere in the cell , which are normally always kept inside the mother cell due to their damaging effects . These results suggest that the ERSU pathway is responsible for preventing protein aggregates in the endoplasmic reticulum from entering daughter cells , but only does so when these aggregates stress the endoplasmic reticulum . Future research will aim to identify how the ERSU pathway senses protein aggregates and prevents the transmission of damaged endoplasmic reticulum . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | The ER Stress Surveillance (ERSU) pathway regulates daughter cell ER protein aggregate inheritance |
The microbialization of coral reefs predicts that microbial oxygen consumption will cause reef deoxygenation . Here we tested this hypothesis by analyzing reef microbial and primary producer oxygen metabolisms . Metagenomic data and in vitro incubations of bacteria with primary producer exudates showed that fleshy algae stimulate incomplete carbon oxidation metabolisms in heterotrophic bacteria . These metabolisms lead to increased cell sizes and abundances , resulting in bacteria consuming 10 times more oxygen than in coral incubations . Experiments probing the dissolved and gaseous oxygen with primary producers and bacteria together indicated the loss of oxygen through ebullition caused by heterogenous nucleation on algae surfaces . A model incorporating experimental production and loss rates predicted that microbes and ebullition can cause the loss of up to 67% of gross benthic oxygen production . This study indicates that microbial respiration and ebullition are increasingly relevant to reef deoxygenation as reefs become dominated by fleshy algae .
Anthropogenic pressures are shifting the composition of reef primary producer communities from calcifying to non-calcifying organisms globally ( Cinner et al . , 2016; Smith et al . , 2016 ) . These changes occur as fleshy algae gain competitive advantage over corals through their interaction with heterotrophic microbes ( Barott and Rohwer , 2012; Jorissen et al . , 2016 ) . High Dissolved Organic Carbon ( DOC ) release rates by fleshy algae stimulate overgrowth of heterotrophic bacteria ( Haas et al . , 2011; Kelly et al . , 2014; Nelson et al . , 2013 ) . The exacerbated heterotrophic growth creates hypoxic zones at the coral-algae interface that kill corals ( Haas et al . , 2013a; Haas et al . , 2014b; Roach et al . , 2017; Smith et al . , 2006 ) . At the reef-scale , algae dominance stimulates heterotrophic bacteria in the water above the reef and leads to an increase in microbial biomass and energetic demands , described as reef microbialization ( Haas et al . , 2016; McDole et al . , 2012; Silveira et al . , 2015; Silveira et al . , 2017 ) . Overgrowth of heterotrophic bacteria creates a feedback loop of coral death , opening space for more algae overgrowth and microbial biomass accumulation ( Dinsdale and Rohwer , 2011; Rohwer et al . , 2002 ) . While DOC-microbe relationships have been extensively described both experimentally and in situ , the oxygen fluxes within the microbialization feedback loop are not fully understood . Differences in microbial responses to coral and algae dominance stem from the physiology of these primary producers . Calcifying organisms , including scleractinian corals and crustose coralline algae ( CCA ) , invest 50% to 80% of their daily fixed carbon in respiration to sustain the energetic costs of calcification ( Hatcher , 1988; Houlbrèque and Ferrier-Pagès , 2009; Tremblay et al . , 2012 ) . Fleshy macroalgae allocate 10% to 30% to respiration , and release up to 60% of their primary production as dissolved organic carbon ( DOC ) in the water ( Cheshire et al . , 1996; Crossland , 1987; Jokiel and Morrissey , 1986; Peninsula et al . , 2007 ) . Fleshy algae also allocate a higher proportion of their daily synthesized carbon to biomass compared to corals , sustaining high herbivory pressure ( Duarte and Cebrián , 1996; Falkowski et al . , 1984; Tremblay et al . , 2016 ) . Together , these processes are predicted to increase DOC-to-O2 ratios in coral exudates compared to algae , but experimental data have shown the opposite to be true: the ratio between DOC and O2 released by fleshy algae in bottle incubations is higher compared to corals ( Haas et al . , 2013b; Haas et al . , 2011 ) . Heterotrophic microbes growing on algal exudates produce fewer cells per unit of carbon consumed compared to growth on coral exudates ( Haas et al . , 2011; Nelson et al . , 2013; Silveira et al . , 2015 ) . Microbes growing on algae-dominated reefs shift their metabolism from classic glycolysis ( Embden–Meyerhof–Parnas pathway , EMP ) towards Pentose Phosphate ( PP ) and Entner-Doudoroff ( ED ) pathways ( Haas et al . , 2016 ) . All three pathways consume glucose in a series of redox reactions that produce pyruvate , however they transfer electrons to distinct reduced coenzymes that act as intermediary electron transfer molecules ( Klingner et al . , 2015; Spaans et al . , 2015 ) . These reduced coenzymes have different fates in the cell , and as a result , the classic glycolic pathway generates more ATP and consumes more oxygen , while ED and PP leave a higher fraction of electrons accumulated as biomass ( Fuhrer and Sauer , 2009 ) . These differences predict that microbes found in microbialized reefs have different oxygen consumption rates when compared to healthy reefs ( Flamholz et al . , 2013; Stettner and Segrè , 2013 ) . Fleshy turf and macroalgae release up to three times more O2 in the surrounding seawater than calcifying organisms ( Haas et al . , 2011; Naumann et al . , 2010; Nelson et al . , 2013; Silveira et al . , 2015 ) . Yet , algae-dominated systems have consistently lower O2 standing stocks ( Calhoun et al . , 2017; Haas et al . , 2013b; Martinez et al . , 2012 ) . Fleshy algae can produce O2 bubbles through heterogeneous nucleation resulting from O2 super-saturation at the algae’s surface ( Kraines et al . , 1996 ) . The gaseous O2 in bubbles is not detected by dissolved O2 instruments , and may cause underestimation of autotrophy in oxygen-based primary productivity methods ( Atkinson and Grigg , 1984; Kraines et al . , 1996 ) . While several studies recognize bubble formation on algal surfaces in benthic ecosystems , the fraction of photosynthetic oxygen lost as bubbles from coral reef primary producers has never been quantified ( Freeman et al . , 2018; Odum and Odum , 1955 ) . Here we test the hypothesis that O2 loss in coral reefs is a result of the high microbial oxygen consumption and the biophysical loss through ebullition from algae surfaces . The analysis of organic carbon consumption pathways encoded in bacterial metagenomes from 87 reefs in the Atlantic and Pacific combined with experimental quantification of cell-specific DOC and oxygen consumption showed that fleshy macroalgae and microbes remove larger amounts of photosynthetic O2 by ebullition and respiration compared to corals . As a result , the incubation bottles have similar to lower oxygen concentrations , providing a mechanistic explanation for the lower dissolved O2 observed on reef ecosystems shifted to algal dominance .
Different carbon consumption pathways employed by bacteria are associated with varying levels of oxygen consumption rates ( Russell and Cook , 1995 ) . To investigate the carbon metabolic pathways selected among reef microbes , 87 metagenomes from reefs across a gradient in algal cover were analyzed for the relative abundance of genes encoding rate-limiting enzymes of central carbon metabolism ( Figure 1—source data 1 ) . On coral reefs , metagenomic data reflect the strain-level genomic adaptation that occurs within hours , the timescale of residence time depending on the tides and current regime ( Nelson et al . , 2011 ) . This selection occurs as offshore microbial communities are transported onto the reef environment and water masses get enriched with benthic exudates changing their biogeochemistry ( Nelson et al . , 2011; Kelly et al . , 2014 ) . In our dataset , microbial biomass in the reef boundary layer increased with fleshy algae cover ( linear regression , p=6 . 25 x 10-8 , R2 = 0 . 29 ) . A dimension-reduction random forest regression analysis was applied to genes encoding rate-limiting enzymes of central carbon metabolism , stress response and control genes , for a total of 23 variables ( Supplementary file 1 ) , using fleshy algae cover as predicted variable . Random forest is a non-parametric method that does not rely on normality , and along with other machine-learning approaches , is the method of choice for the identification of a robust subset of variables within complex meta-omics data ( Li , 2015; Liu et al . , 2019; Thompson et al . , 2019 ) . The relative abundance of these genes could explain 14 . 3% of the variance in fleshy algae cover . The genes with highest and significant explanatory power in the permutation test were phosphoenolpyruvate carboxylase , aspartate aminotransferase , oxoglutarate dehydrogenase , glucose 6-phosphate dehydrogenase and KDPG aldolase ( Figure 1 displays the enzyme genes with highest explanatory power according to their increasing mean squared error: 53 . 54% , 22 . 03% , 23 . 99% , 26 . 08% and 17 . 89% , and permutation test p-values = 0 . 009 , 0 . 01 , 0 . 03 , 0 . 01 and 0 . 03 , respectively; Figure 1—source data 1 ) . Phosphoenolpyruvate carboxylase and aspartate aminotransferase are involved in anaplerotic routes , reactions that replenish the Krebs cycle when its intermediaries are diverted for biosynthesis . Glucose 6-phosphate dehydrogenase shunts glucose to both Entner-Doudoroff and Pentose Phosphate pathways , and KDPG aldolase is unique to the Entner-Doudoroff pathway . These four enzyme genes had a positive relationship with fleshy algae cover ( identified by random forest by their increasing mean squared error , followed by permutation test with p-values p=0 . 009 , 0 . 01 , 0 . 01 and 0 . 03 , respectively ) , and participate in pathways of incomplete carbon oxidation , that is , lower oxygen consumption . Oxoglutarate dehydrogenase relative abundance had an inverse relationship with fleshy algae cover ( random forest increasing mean squared error , p=0 . 03 ) . This enzyme catalyzes an oxidative decarboxylation step in the Krebs cycle , a pathway that produces most of the NADH that fuels oxygen consumption by oxidative phosphorylation . These results suggest that the bacterial community is being increasingly selected for pathways related to overflow metabolism and incomplete carbon oxidation at high algal cover , which would predict a decoupling between organic carbon and oxygen consumption rates by microbes . The gene rpoB ( RNA polymerase ) was used as a control and did not vary in relative abundance across the algal cover gradient ( Figure 1—source data 1 ) . Genes involved in oxidative stress did not show significant relationships either . To test if the changes in enzyme gene relative abundances could be explained by taxonomic shifts of bacterial genera that consistently display larger genome sizes and duplications in the genes described above , the relationship between fleshy algae cover and taxonomic profiles at the genus and species levels was tested using random forest . The relative abundances of each taxon were calculated accounting for genome size ( Cobián Güemes et al . , 2016 ) . Seven species significantly changed in abundance across the algae cover gradient ( Figure 1—source data 1 , Prochlorococcus marinus , Fibrisoma limi , Prolixibacter bellariivorans , Pelagibacter ubique , Ruegeria mobilis , Phaeobacter italicus , and Psychrobacter pacificensis ) . At the genus level , five taxa ( Prochlorococcus , Thermotoga , Methanobacterium , Leuconostoc , and Rhodopirellula ) changed their abundance with increasing fleshy algae cover . We searched if the taxa varying in abundance consistently display copy number variations in the genes that significantly vary with algae cover . The only taxon that could potentially explain the abundance trends in enzyme genes was Prochlorococcus , which lacks an oxoglutarate dehydrogenase gene but encodes all the other four enzymes . However , there was no significant relationship between the abundances of Prochlorococcus and oxogluratae dehydrogenase ( p=0 . 26 in the linear regression of oxoglutarate dehydrogenase and log10-transformed Prochlorococcus abundance ) . Figure 1—source data 1 shows the relative abundances of the 20 most abundant species that together made up 97 . 8 ± 3 . 0% of the identified reads at species level in the metagenomes . To test the hypothesis of a decoupling between organic carbon and oxygen consumption by heterotrophic bacteria predicted by the metagenomic analysis , cell-specific carbon and O2 consumption rates were obtained by incubating reef heterotrophic bacteria with coral and algae exudates ( Figure 2—source data 1; Experiment one in Materials and methods ) . The amount of DOC consumed per cell was higher , that is , low cell yields per carbon consumed , in fleshy macroalgae treatments compared to coral and control treatments ( Figure 2A , Kruskal-Wallis χ2 ( 2 ) =13 . 7 , p=0 . 001; Wilcoxon , p=0 . 005 and 0 . 008 for pairwise comparisons of algae vs controls and corals , respectively ) . However , the amount of O2 consumed per cell was not different between treatments ( Figure 2B , Kruskal-Wallis χ2 ( 2 ) =0 . 5 , p=0 . 77 ) . Incomplete carbon oxidation and low oxygen consumption relative to organic carbon are hallmarks of increases in overflow metabolism and organic carbon accumulation as biomass . To test whether low O2-to-DOC consumption ratios in algae exudates are causing biomass accumulation , heterotrophic bacteria were incubated in coral , CCA and fleshy turf and macroalgae exudates and their biomass was quantified accounting for changes in both abundance and cell volume . The change in bacterial abundance in all primary producer bottles was higher compared to controls ( Figure 3A , Kruskal-Wallis χ2 ( 4 ) =20 . 6 , p=0 . 0003 , Wilcoxon p=0 . 04 , 0 . 02 . 0 . 02 . and 0 . 02 for pairwise comparisons of coral , CCA , turf and macroalgae against control , Figure 3—source data 1 ) . The change in abundance was higher in fleshy algae treatments compared to calcifying treatments ( Kruskal-Wallis χ2 ( 2 ) =19 . 1 , p=6 . 9×10−5 , Wilcoxon p=0 . 001 for pairwise comparison of fleshy vs calcifying ) . Microbial cell volume changed differently in treatments during incubations ( Figure 3B , Kruskal-Wallis χ2 ( 5 ) =1235 . 2 , p<2×10−16 , Figure 3—source data 2 ) . Bacteria growing in control bottles showed no change in volume ( Wilcoxon , p=0 . 56 for pairwise comparison of control vs T0h ) . Cell volume decreased when bacteria grew on coral and CCA exudates , but increased when growing on turf and macroalgae exudates , and showed no change in controls ( Wilcoxon p<2×10−16 for all pairwise comparisons vs T0h and vs control ) . When taking abundance and cell volume into account , the change in total bacterial biomass was significantly different among treatments ( Figure 3C , Kruskal-Wallis χ2 ( 4 ) =21 . 6 , p=0 . 0002 , Figure 3—source data 1 ) . Coral exudates decreased total bacterial biomass , while turf and macroalgae increased total bacterial biomass , and CCA lead to no change ( Wilcoxon , p<0 . 05 for pairwise comparisons of coral , turf and macroalgae vs control ) . The change in biomass in the two fleshy algae together was greater than in the calcifying organisms together ( Kruskal-Wallis χ2 ( 2 ) =19 . 1 , p=7 . 1×10−5; Wilcoxon , p=0 . 0005 for pairwise comparison of fleshy vs calcifying ) . This same pattern was observed in two independent experiments , one performed in the island of Curaçao , in the Caribbean ( Figure 3 ) and one performed at the Hawaiian Institute of Marine Biology ( HIMB , Figure 3—figure supplement 1 ) , with distinct sets of primary producers , suggesting that this result may be broadly applicable to other reefs ( Figure 3—source datas 1 and 2 ) . While bacteria became larger and more abundant when growing on algae exudates , their oxygen consumption was lower than that predicted by their DOC consumption: the theoretical value for full carbon oxidation through aerobic metabolism is 1:1 , changing due overflow metabolism , futile cycles , uncoupling , and other processes reviewed in Russell and Cook ( 1995 ) . To resolve the oxygen budget , heterotrophic bacteria were incubated together with the primary producers in POP ( Peripheral Oxygen Production ) incubation chambers that allow to quantify O2 in both dissolved and gas fractions . Net dissolved O2 production was significantly different among primary producer treatments ( Figure 4A , Kruskal-Wallis χ2 ( 4 ) =19 . 5 , p=0 . 0006 , Figure 4—source data 1 ) . Primary producers showed higher net dissolved O2 production compared to controls ( Wilcoxon with FDR-corrected pairwise comparisons p=0 . 004 , 0 . 01 , 0 . 01 , 0 . 004 for comparisons between controls vs coral , CCA , turf and macroalgae , respectively ) . Net dissolved O2 production was lower in macroalgae incubations , but not significantly different between fleshy and calcifying organisms ( 5 . 99 ± 2 . 29 and 6 . 33 ± 2 . 20 μmol cm−2 day−1 for fleshy and calcifying organisms , respectively , mean ± SE; Wilcoxon p=0 . 68 ) . There was a significant difference between gaseous O2 production observed in most primary producer bottles , while no gas was observed in control bottles ( Figure 4B , Kruskal-Wallis χ2 ( 4 ) =30 . 2 , p=4 . 361e-06 , Wilcoxon p=0 . 03 , 0 . 002 , 0 . 002 , and 0 . 002 for pairwise comparisons between controls vs corals , CCA , turf and macroalgae , respectively ) . Fleshy organisms produced significantly more gaseous O2 than calcifying organisms ( 0 . 42 ± 0 . 35 and 2 . 33 ± 1 . 35 μmol cm−2 day−1 for calcifying and fleshy , respectively , mean ± SE; Wilcoxon p=0 . 0001 for comparisons between fleshy vs controls and vs coral ) . Fleshy macroalgae had the highest fraction of net photosynthetic O2 ( sum of dissolved and gaseous O2 ) released in the form of gas ( 37 . 33 ± 8 . 34% ) , followed by turf algae ( 13 . 78 ± 1 . 33% ) , CCA ( 10 . 19 ± 2 . 88% ) , and corals ( 5 . 00 ± 5 . 55% ) ( Figure 4C ) . The difference in the fraction of O2 accumulated as gas was significant among treatments ( Kruskal-Wallis χ2 ( 4 ) =29 . 2 , p=7 . 067e-06 ) , being higher in fleshy organisms ( Kruskal-Wallis χ2 ( 2 ) =26 . 3 , p=1 . 904e-06; Wilcoxon p<9 . 5e-05 for fleshy vs calcifying ) . In the POP experiments , the oxygen in the gas phase was close to equilibrium with the dissolved oxygen , and in some of the fleshy algae and CCA , the dissolved oxygen concentration was above accuracy limits of the probe . During the course of the experiment , we observed that a headspace was formed by bubbles that were formed on the algae surfaces and rose to the top of the incubation bottles . To quantify ebullition rates and to test whether this effect would be observed in an open system , bubble production was quantified in open-tank experiments using image analysis ( Figure 5A—source data 1 ) . The green fleshy algae Chaetomorpha had the highest bubble production rates ( 10 . 3 ± 0 . 45 bubbles per min per dm2 , mean ± SE; Video 1 ) , followed by the fleshy red algae Gracilaria sp . ( 1 . 29 ± 1 . 21 bubbles per min per dm2 ) . Bubble production by the two coral species analyzed was close to zero ( 0 . 1 ± 0 . 14 and 0 . 006 ± 0 . 02 bubbles per min per dm2 for Favia sp . and Montipora sp . , respectively ) . Ebullition rate was significantly higher in fleshy compared to calcifying organisms ( Kruskal-Wallis χ2 ( 3 ) =349 . 46 , p =<2 . 2e-16; Wilcoxon p<2e-16 for fleshy vs calcifying ) . Bubbles were observed forming on the algal surfaces , detaching and rising to the atmosphere starting as early as 20 min into the experiment . The ebullition rate in the open-tank experiment was within the range of ebullition rates estimated to be necessary to explain the volume and oxygen concentrations observed in the POP experiments ( estimates indicated by the gray bar in Figure 5A ) . The dissolved oxygen concentration also increased through the course of the incubations ( Figure 5B ) , however , there was no relationship between the rate of bubble production and the dissolved oxygen concentration ( linear regression p=0 . 16 , R2 = 0 . 04 ) or the change in oxygen concentration over time ( linear regression p=0 . 27 , R2 = 0 . 03 ) . The mean bubble diameter produced by the algae Gracialaria sp . was larger than that produced by the algae Chaetomorpha sp . ( Figure 5C and Figure 5—figure supplement 1; 0 . 79 ± 27 mm in diameter and 0 . 64 ± 0 . 23 mm for Gracialaria sp . and Chaetomorpha sp . , respectively , however these differences were not statistically significant ) . The incubation of these same algae species at higher PAR values ( 300 to 650 µmol quanta m−2 s−1 , normally observed on a reef in a sunny day ) yielded different bubble production rates: at 300 µmol quanta m−2 s−1 , the algae Chaetomorpha sp . had the highest ebullition rates ( Figure 5C and Video 2 , 16 . 2 ± 1 . 9 bubbles per min per dm2 ) . The rate decreased to 0 . 9 ± 0 . 04 at 650 µmol quanta m−2 s−1 . The algae Gracilaria sp . also produced fewer bubbles per minute at 650 µmol quanta m−2 s−1 ( 0 . 9 ± 0 . 08 ) . To predict the proportion of the benthic gross oxygen production that is lost through microbial respiration and ebullition , a weighted linear model combined the cell-specific respiration rates from Figure 2 , the microbial abundances sustained by each primary producer from POP experiments , and the rates of oxygen ebullition from Figure 5 . Each of these rates was applied to an idealized reef of 1 m2 benthic surface covered by fleshy green algae and/or scleractinian coral , and the microbial community within 1 m3 of the water column above . Both the microbial respiration and ebullition rates increased with increasing algal cover in this idealized model ( Figure 5D—source data 2 ) . At 0% algal cover , 10 . 4% of the benthic oxygen production is consumed by microbes , compared to 35 . 9% at 50% algal cover , and 47 . 0% at 100% algal cover . The amount of oxygen loss by ebullition is predicted to increase from 1 . 0% at 0% algal cover to 14 . 0% at 50% algal cover and 19 . 7% at 100% algae cover . Combined , microbial respiration and ebullition are predicted to consume 50% of the gross oxygen production at 50% algal cover , and 66 . 7% of the oxygen production at 100% algal cover .
The metagenomic results in Figure 1 show an increase in genes for pathways that shunt carbon to biosynthesis and incomplete oxidation , that is , low O2 consumption ( Russell and Cook , 1995; Haas et al . , 2016 ) . The abundance of genes encoding enzymes involved in the Pentose Phosphate ( PP ) and Entner-Doudoroff ( ED ) pathways , increased with fleshy algae cover ( Figure 1 ) . These pathways are alternatives to the canonical glycolytic route for carbohydrate consumption and produce more NADPH than NADH as intermediate electron acceptors . These two reduced coenzymes have distinct cell fates , with NADH preferentially donating electrons to oxidative phosphorylation that consumes oxygen during ATP production ( Pollak et al . , 2007; Spaans et al . , 2015 ) . Thus , cells utilizing more NADPH pathways consume less O2 relative to carbon compared to cells utilizing more NADH , and store a larger fraction of the consumed carbon into biomass . The tricarboxylic acid cycle ( TCA or Krebs cycle ) , is a canonically oxidative path for complete decarboxylation of pyruvate . But the TCA cycle can work as a biosynthetic pathway by providing intermediates to amino acids , nitrogenous bases and fatty acids syntheses ( Cronan and Laporte , 2013; Sauer and Eikmanns , 2005 ) . TCA cycle stoichiometry is maintained through the resupply of these intermediates by anaplerotic reactions that use pyruvate ( or phosphoenolpyruvate ) ( Owen et al . , 2002 ) . In the reefs analyzed here , the abundance of genes encoding anaplerotic enzymes increased , and genes encoding an oxidative decarboxylation step in the Krebs cycle decreased with increasing algal cover . These results indicate a shift towards the use of the Krebs cycle as an anabolic route ( Bott , 2007; Sauer and Eikmanns , 2005 ) . Anaplerotic activity increases in rapidly growing bacteria with high rates of amino-acid synthesis ( Bott , 2007 ) . These routes shunt organic carbon into biomass accumulation , as opposed to oxidation with electron transfer to O2 ( Russell and Cook , 1995 ) . This overflow metabolism is analogous to the Warburg effect , where cells undergoing fast metabolism are limited by enzymes' catalytic rates , as opposed to substrate concentrations ( Vander Heiden et al . , 2009 ) . This type of metabolism is classically described in cancer cells and fast-growing yeast , and consumes large amounts of organic carbon without complete oxidation to CO2 , even in the presence of O2 . In bacteria , the Warburg effect is a physiological response to changing proteomic demands that optimizes energy biogenesis and biomass synthesis under high energy supply ( Basan et al . , 2015 ) . This biochemical transition occurs because the proteome cost of energy biogenesis by respiration exceeds that of fermentation . A switch to overflow metabolism would benefit heterotrophic bacteria living in a biogeochemical environment of high total DOC but low relative O2 that is observed in algae exudates . The overflow metabolism predicted by metagenomic analysis was observed in experimental incubations of microbes . When growing on algal exudates , microbes increased cell volume and DOC consumption per cell ( Figures 2A and 3 ) . Yet , these cells displayed the same cell-specific O2 consumption compared to small cells growing on coral exudates ( Figure 2B ) . These results indicated that microbes growing on coral exudates fully oxidize the organic carbon consumed , channeling metabolic energy to maintenance costs through NADH- and ATP-producing pathways ( Russell and Cook , 1995 ) . On algal exudates , bacteria had higher DOC consumption , with a greater fraction of the organic carbon being only partially oxidized and shunted to NADPH and biosynthesis . The deviation of the excess carbon to these pathways caused less O2 consumption ( relative to carbon ) and increased community biomass ( Figures 2B and 3 ) . We considered and rejected alternative hypotheses for DOC and O2 consumption patterns observed in our experiments , such as broken TCA cycles and reactive oxygen species ( ROS ) detoxification ( Mailloux et al . , 2007; Steinhauser et al . , 2012 ) . There was no evidence for an increase in genes encoding these pathways in our dataset ( Figure 1—source data 1 ) . Likewise , the metagenomic transitions observed in situ could not be explained by the rise or disappearance of specific taxonomic groups that changed in abundance with increasing algae cover ( Dinsdale et al . , 2008; Kelly et al . , 2014 ) . The disconnection between functional and taxonomic profiles is common in microbial communities and has been previously observed in coral reefs ( Kelly et al . , 2014 ) . This discordance is due to strain-level variability in functional genes a result of genomic adaptation to multiple environment selective pressures ( Klingner et al . , 2015; Martiny et al . , 2006; Martiny et al . , 2009 ) . The cell volume and abundance changes observed in our experiments helps to explain the higher microbial biomass observed in algae-dominated reefs across the Pacific , Caribbean and Indian Oceans ( Haas et al . , 2016 ) . In the Pacific , an increase in total bacterial biomass caused the total bacterial energetic demands to surpass that of macrobes in degraded reefs ( McDole et al . , 2012; McDole Somera et al . , 2016 ) . The increase in overflow metabolism observed in our dataset provides a mechanism for the increase in biomass and energetic demand observed in degraded reefs , affecting reef-scale trophic interactions ( De'ath and Fabricius , 2010; Haas et al . , 2016; Russell and Cook , 1995; Silveira et al . , 2015; Wilson et al . , 2003 ) . The changes in biomass observed in situ cannot be attributed to changes in abundance of a single clade . The only clade with an increase relative abundance that could potentially explain the change in cell sizes was Prochlorococcus , which has cell volumes of ~0 . 1 µm3 ( Kirchman , 2016 ) . However , this is the same average cell size observed in the offshore water used as inoculum in our incubation experiments , and the consumption experiments were run in the dark , selecting for heterotrophic microbes ( Figure 3B ) . The other groups with significant relationships with fleshy algae cover contributed to only 0 . 9% to 1 . 4% of the community ( Figure 1—source data 1 ) , and their average cell sizes are not consistent with the observed increase in cell size in algae: Thermotoga has large cells , but their abundance decreased with algae , Leuconostoc , Rhodopirellula , and Methanobacterium also decreased in abundance with increasing algae . Therefore , none of these clades can explain the change in size observed in this study . Photosynthesis and respiration have a theoretical 1:1 molar ratio of organic carbon and O2 produced and consumed ( Williams et al . , 1979 ) . In holobionts with different proportions of heterotrophic and autotrophic components , such as corals and algae , respiration can consume different fractions of the oxygen and organic carbon produced in photosynthesis ( Tremblay et al . , 2016 ) . The cost of calcification in corals and calcifying algae can also increase the rate of oxygen consumption due to its energetic costs ( Muscatine et al . , 1981 ) . As a result , the DOC:O2 ratios in exudates from calcifying holobionts are predicted to be higher than in fleshy algal exudates . This pattern contradicts experimental data showing higher DOC:O2 ratios in microbe-free algae incubations ( Haas et al . , 2011 ) . Our results suggest that O2 ebullition is the likely cause of this observation . The organic carbon released by algae can be fully solubilized , while a fraction of the O2 nucleates and escapes , leaving behind a high DOC:dissolved O2 ratio . The lack of a relationship between dissolved oxygen concentrations and ebullition rates suggests that the bubbles are formed by heterogenous nucleation on the algae surface ( Freeman et al . , 2018 ) . Differential carbon allocation into biomass can also alter these ratios , and the quantitative analysis of carbon incorporation using isotope probing in combination with dissolved and gaseous O2 dynamics is the next step to resolve these budgets . O2 ebullition and bubble injection by hydrodynamics are recognized as a source of error when estimating production in shallow water ecosystems , yet the extent to which ebullition affects these estimates is rarely quantified due to methodological challenges ( Cheng et al . , 2014; Kraines et al . , 1996 ) . O2 ebullition observed in this study corresponded to 5–37% of net community production . Previous studies quantified that ebullition accounted for the loss of up to 21% and 37% of the O2 production in a salt marsh and a lake , respectively ( Howard et al . , 2018; Koschorreck et al . , 2017 ) . If in situ bubble nucleation and rise rates are comparable to those in incubations , our results imply that coral reef gross primary production has been significantly underestimated , especially in algae-dominated states ( Howard et al . , 2018 ) . Previous studies reported bubbles on the surface of turf algae , on sediments , and inside coral skeletons colonized by endolithic algae ( Bellamy and Risk , 1982; Clavier et al . , 2008; Freeman et al . , 2018; Odum and Odum , 1955 ) . The heterogeneous distribution of bubble nucleation over primary producers entails that benthic community composition determines the magnitude of O2 ebullition on reef-level O2 dynamics . Our model based on the combined results from microbial and primary producer incubations predicts that microbial respiration and ebullition together contribute to the loss of 11 . 5% to 66 . 7% of the gross primary production as reef algae cover increases from 0% to 100% . A combination of future in situ incubations and gas exchange studies is required to test these relationships . Our study suggests that O2 depletion observed in algae-dominated reefs is a result of high bacterial densities and oxygen ebullition . Ebullition causes a decoupling between photosynthetic fixed carbon and O2 , fundamentally changing the biogeochemical environment: O2 loss as bubbles causes an increase in dissolved DOC:O2 ratios , stimulating overflow metabolism in the microbial community ( Figure 6 ) . These biophysical and physiological changes cause a depletion in oxygen standing stocks that may negatively affect corals and other reef animals .
Microbial metagenomes were sequenced from water samples collected on coral reefs in the Pacific and in the Caribbean . The Pacific samples were collected during NOAA RAMP cruises from 2012 to 2014 to the Hawaiian Islands , Line Islands , American Samoa , and Phoenix Islands . Caribbean samples were collected around the island of Curaçao during the Waitt Institute Blue Halo expedition in 2015 . Geographic coordinates for each sampling site are provided in source data 1 for Figure 1 , along with benthic cover data . At each sampling site , SCUBA divers collected water from within 30 cm of the reef surface using Hatay-Niskin bottles ( Haas et al . , 2014a ) . One metagenome was generated from each sampling site , and 1 to 5 sites were sampled around each island in the Pacific , and 22 samples from around the island of Curaçao , in the Caribbean . Samples were brought to the ship , filtered through a 0 . 22 μm cylindrical filter within 4 hr , and stored at −20°C until laboratory processing . DNA was extracted from the filters using Nucleospin Tissue Extraction kits ( Macherey Nagel , Germany ) and sequenced on an Illumina HiSeq platform ( Illumina , USA ) . Pacific microbial metagenomes described here were previously analyzed for the presence of CRISPR elements , competence genes and Shannon diversity in Knowles et al . ( 2016 ) . All fastq files from Pacific and Curaçao metagenomes were quality filtered in BBTools with quality score >20 , duplicate removal , minimum length of 60 bases , and entropy 0 . 7 ( Bushnell , 2014 ) . The 87 metagenomes are available on the NCBI Short Read Archive ( PRJNA494971 ) . The number of samples analyzed here surpassed that estimated by power analysis based on the work of Haas et al . ( 2016 ) ( estimated n = 67 for α = 0 . 05 and β = 0 . 2 based on the relationship between microbial abundance and algae cover shown in Figure 2B of Haas et al . , 2016 ) . Relative abundances of genes encoding rate-limiting enzymes in central carbon metabolism pathways were utilized as proxy for the representation of that pathway in the community . The list of pathways and enzymes analyzed here is provided in Supplementary Table 1 . Enzyme-specific databases were built using amino acid sequences from the NCBI bacterial RefSeq . BLASTx searches were performed using metagenomic reads against each enzyme gene database , using a minimum alignment length of 40 , minimum identity of 60% and e-value <10−5 . Reads mapping to the database were normalized by total high-quality reads resulting in the percent abundance of each enzyme gene . Relative gene abundances are provided as source data one for Figure 1 . The relationship between fleshy algae cover and enzyme gene relative abundances was analyzed by the statistical learning method of supervised random forests regression ( Breiman , 2001 ) . In the random forest procedure , each tree in the ensemble is grown using a different bootstrap sample of the original data . For the final prediction , the mean prediction regression from all the trees is used . Since each tree is grown with a bootstrap sample of the data , there is approximately one-third of the data that is ‘out of sample’ for each tree . This out of sample data acts as an internal validation dataset and can be used to determine variable importance for each predictor variable . This is done for each tree by randomly permuting each variable in the out of sample data and recording the prediction . For the ensemble of trees , the permuted predictions are compared with the unpermuted predictions . The magnitude of the mean decrease in accuracy indicates the importance of that variable . Significance of the importance measures is estimated using a permutation test . The random forest easily handles different types of variables and does not require the data to be rescaled or transformed . The random forest consists of an ensemble of decision trees for regression . Predictions , prediction error , and variable importance are estimated from the ensemble of trees . The random forest was grown using 1000 trees and four tree splitting variables . The mean squared error diagnostic plot of the random forest indicated it settled down and therefore enough trees have been used ( Figure 1—figure supplement 1 ) . The significance of the random forests variable importance was determined by a permutation method , which has an implicit null hypothesis test , and was implemented using the R package rfPermute . The permutation test p-values were obtained using 100 permutation replicates . Variable importance was selected based on the increasing mean squared error , a measure of the importance of a given feature for the correct prediction of a response variable , with alpha cutoff of 0 . 05 . No outliers were removed from any analyses . The changes in bacterial community composition across the gradient in algae cover were compared to changes in relative abundances of metabolic genes to access the relationship between taxonomy and function . This analysis can indicate whether specific taxa with differences in gene copy number and genome sizes would drive the functional variation . First , the abundance of bacterial genera in metagenomes was generated through k-mer profiles using FOCUS2 ( Silva et al . , 2016 ) . FOCUS2 computes the optimal set of organism abundances using non-negative least squares ( NNLS ) to match the metagenome k-mer composition to organisms in a reference database ( Silva et al . , 2014 ) . For annotation at species level , sequences were mapped against a database of bacterial genomes from RefSeq using Smalt at 96% sequence identity ( Ponstingl and Ning , 2010 ) . The abundance of each organism was computed using the FRAP normalization ( Fragment Recruitment , Assembly , Purification ) , which calculates the fractional abundance of each bacteria in the metagenome accounting for the probability of a read mapping to a genome given the length of both the read and the genomes in the database ( Cobián Güemes et al . , 2016 ) . Taxa with median relative abundance across samples equal to zero were removed from the analysis . The relative abundances of genera and species were separately analyzed through permutation regression random forests with percent algae cover as response variable in both cases . The random forest was grown using 1000 trees and four tree splitting variables . The mean squared error diagnostic plot of the random forest indicated it settled down and therefore enough trees have been used ( Figure 1—figure supplement 1 ) . Variable importance was selected based on the increasing mean squared error , a measure of the importance of a given feature for the correct prediction of a response variable , with alpha cutoff of 0 . 05 . The complete genomes of the taxa varying in abundance with algae cover deposited in the NCBI RefSeq were manually checked for copy number variation in the genes identified by the functional analysis according the annotations associated with each genome deposited . This dark incubation experiment was performed in Mo’orea , French Polynesia . Briefly , exudates from corals and algae released during a light incubation period were 0 . 2 μm-filtered and inoculated with freshly collected unfiltered forereef seawater to add a compositionally-representative ambient microbial community to the exudate samples . The forereef water column sample contains a microbial community that is most representative of the offshore community that approaches the reef environment in the incoming current ( Haas et al . , 2011; Nelson et al . , 2013 ) . Inoculation in filtered seawater was utilized as control . All bottles were kept in the dark for 48 hr . DOC , dissolved O2 and microbial abundances were measured at the beginning and at the end of each dark incubation as described above . DOC samples were filtered through pre-combusted GF/F filters ( Whatman , 0 . 7 μm nominal particle retention ) and transferred to acid-washed HDPE vials . Samples were kept frozen until analysis according to Carlson et al . ( 2010 ) . DOC and O2 consumption rates normalized by the bacterial cell yield at the end of each experiment resulted in cell-specific carbon and O2 demands . The fleshy organisms used in this experiment were turf algae and the macroalgae Turbinaria ornata and Amansia rhodantha , and the calcifying organisms were CCA and the coral Porites lobata . Five replicate incubations of each organism were performed ( Power analysis estimate n = 4 for α = 0 . 05 and β = 0 . 2 based on the incubations of Haas et al . , 2011 and Haas et al . , 2013b ) . Changes in bacterial abundance , cell size and total biomass in response to primary producers derived from two independent experiments , described below: Experiment 1: Microbial communities from the reef off the CARMABI Research Station , in the island of Curacao were incubated with four different primary producers: the scleractinian coral Orbicella faveolata , CCA , the fleshy macroalgae Chaetomorpha sp . , and turf algae . O . faveolata colonies were collected at 12 m depth from the Water Factory site ( 12°10’91’ N , 68°95’49’ W ) and cut into ~10 cm2 fragments . Coral fragments were kept for two weeks at the CARMABI Research Station flow-through tank system . The tank was subject to natural diel light cycles with light intensities comparable to 10 m depth , as measured using HOBO Pendant UA-002–64 . CCA , turf , and macroalgae were collected off CARMABI immediately prior to the experiment . Five replicate incubations for each organism were conducted , and five control incubations had no primary producer ( Power analysis estimate n = 4 for α = 0 . 05 and β = 0 . 2 based on the incubations of Haas et al . , 2011 and Haas et al . , 2013a ) . Surface area of organisms is provided in Figure 3—source data 1 . Incubations lasted for 4 days at 24°C with natural diel light cycles . Bacterial communities at the start and end of the incubation were analyzed by fluorescence microscopy according to McDole et al . ( 2012 ) . Briefly , cell volume was calculated by considering all cells to be cylinders with hemispherical caps and individual microbial cell volumes were converted to mass in wet and dry weight using previously established size-dependent relationships for marine microbial communities ( Simon and Azam , 1989 ) . Differences in cell abundance , cell volume , and total microbial biomass were calculated by the difference between final and initial values , and were normalized by the area of the primary producer in the incubation . Experiment 2: Five specimens of the coral Favia sp . were obtained from a long-term holding tank at the Hawaiian Institute of Marine Biology ( HIMB ) and placed in independent 5 L polycarbonate containers filled with 0 . 2 μm-filtered seawater . Five specimens of the fleshy macroalgae Gracilaria sp . were collected off HIMB and placed in independent 5 L containers . Additional control containers were filled with filtered seawater . Primary producers were incubated in natural light conditions for 8 hr to release exudates . At the end of the incubation , 2 L of seawater containing exudates were 0 . 2 μm-filtered and inoculated with 1 L of unfiltered offshore seawater containing water column reef microbial communities . All bottles were incubated for 24 hr in the dark . For microbial abundance and biomass determination , 1 mL samples were collected and analyzed as described above . Differences in cell abundance , cell volume and total heterotrophic microbial biomass were calculated by the difference between final and initial values . The rates of dissolved and gaseous O2 production by benthic primary producers were measured in two tank experiments ( POP Experiments 1 and 2 ) . In both experiments , organisms were incubated in custom-made chambers named peripheral oxygen production ( POP ) bottles . POP bottles are bell-shaped Polyethylene Terephthalate ( PET ) bottles with a removable base and two sampling ports , one for dissolved analyte sampling , and one at the top for gas sampling . Primary producers were placed at the bottom and bubbles released from their surfaces during incubation accumulated at the top . Primary producers were placed on the base of the POP bottles , inside a larger tank filled with reef water . The bell-shaped container was then placed over the base , and the bottle was sealed . At the end of the incubation , the gas accumulated at the top of the bottles was collected in a syringe and the volume of gas was recorded . The gas was transferred to a wide-mouth container and O2 partial pressure was measured using a polarographic probe ( Extech 407510 ) immediately upon collection . The POP chambers and containers used for measurements were located underwater inside a larger holding tank , so no gas was lost during the quantification procedures . Dissolved O2 was determined using a Hatch-Lange HQ40 DO probe . POP Experiment 1: The scleractinian coral Montipora sp . and the fleshy macroalgae Chaetomorpha sp . were collected from coral and macroalgae long-term holding tanks maintained at SDSU . Specimens were collected from the tanks immediately prior to the experiment and placed inside POP bottles . Three specimens of each organism were individually incubated for 2 days with artificial seawater from the coral tank , along with three control bottles under cycles of 12 hr light ( 150 μmol quanta m−2 s−1 ) and 12 hr dark at 27°C . POP Experiment 2: Four benthic primary producers were analyzed: the scleractinian coral Orbicella faveolata , CCA , the fleshy macroalgae Chaetomorpha sp . , and turf algae . O . faveolata colonies were collected at 12 m depth from the Water Factory site in Curaçao ( 12°10’91’ N , 68°95’49’ W ) and cut into ~10 cm2 fragments . Coral fragments were kept for two weeks at the CARMABI Research Station flow-through tank system . The tank was subject to natural diel light cycles with light intensities comparable to 10 m depth , as measured using HOBO Pendant UA-002–64 . CCA , turf , and macroalgae were collected off CARMABI immediately prior to the experiment . Five individual incubations for each organism were conducted . Surface area of organisms is provided in Figure 3—source data 1 . Incubations lasted for 4 days at 24°C with natural diel light cycles ( PAR varied between 150 and 500 μmol quanta m−2 s−1 ) . The dissolved O2 in macroalgae and CCA bottles were above the accuracy range of the oxygen probe used . In these cases , we used the highest value within the accuracy range of the probe for all following calculations . Therefore , the results shown are lower bounds of the actual O2 concentrations . Figure 3A indicates the bottles where supersaturation was observed with gray-filled symbols , and the estimated fraction of oxygen lost by ebullition in Figure 3C that were calculated using these lower bound values . Two species of algae ( Chaetomorpha sp . and Gracilaria sp . ) and two species of coral ( Montipora sp . and Pocillopora sp . ) from the SDSU long-term holding tanks were transferred to incubation tanks immediately prior to each experiment . The surface area of each organism , light intensity ( 150 to 630 µmol quanta m−2 s−1 ) and incubation times are provided in Figure 5—source data 1 . The experiment started in the morning after a 12 hr period of darkness . Each organism was transferred to a 15 cm x 20 cm x 30 cm glass tank containing seawater from the holding tank . The tank was open on the top to allow free gas exchange with atmospheric air . A GoPro camera was positioned inside the tank facing the organism , and videos were recorded at 4K 30 fps ( Videos 1 and 2 are provided as representative videos of the experiments with Chaetomorpha sp . at 150 and 300 PAR , respectively ) . Additional pictures were taken with a Canon RebelT3 with 18–55 mm lens to obtain bubble sizes . Dissolved O2 was measured throughout the experiments using a Hatch-Lange HQ40 DO probe . The number of bubbles produced per minute was determined by manual counts of bubbles rising from the algae in the video . Bubble sizes were determined from a total of 160 bubbles ( at least 40 bubbles from each experiment ) from the still images in Adobe Illustrator . The internal pressure , pi , of a bubble was estimated using the Laplace’s law using a standard surface tension model , pi=pe+2σw/Rb ( Bowers et al . , 1995 ) . Here , pe was the external pressure ( assumed at atmospheric pressure , pe=1 atm ) , σw the surface tension of water ( 71 . 99 ± 0 . 05 mN/m at 25°C , Pallas and Harrison , 1990 ) , and Rb the radius of the bubble . The moles of oxygen in a bubble , nb , were estimated using the ideal gas law , nb= ( piVb ) / ( RT ) , which is a good approximation for oxygen at normal conditions ( Christensen et al . , 1992 ) . Here , R was the universal gas constant ( 8 . 314 J K–1 mol–1 ) , T the temperature ( 25°C or 298 . 15 K ) , and Vb was the volume of the bubble , which was approximated as spherical , Vb= ( 4π/3 ) Rb3 . The rate of bubble production necessary to generate the moles of oxygen in the gas phase of the POP experiment , nPOP , was obtained by assuming a constant bubble production: rb=Nb/tlight . Here Nb=nPOP/nb was the number of bubbles needed , and tlight was the total time of light exposure in the POP experiments ( 2 to 4 days of 12 hours of light per day ) . The rate was normalized by the covered area ( Ap ) of the primary producer , ρb=rb/Ap . The estimated ebullition rate was obtained for macroalgae in the POP experiment using the surface values ( Ap ) and moles ( nb ) values reported in Figure 4—source data 1 . The bubble diameter utilized was derived from the open tank ebullition experiment . The mean , first quartile ( 25% ) and third quartile ( 75% ) of bubble sizes were used to obtain the range of ebullition rates necessary to recover the gas in headspace of the POP experiments , indicated by the shaded area in Figure 5A . All statistical analyses were performed using the software R . The response variables from the incubation experiments were tested for normality using the Shapiro–Wilk test . Due to lack of normality ( Shapiro–Wilk , p<0 . 05 ) , the non-parametric Kruskal-Wallis test was used to test if there were differences in treatments followed by the post-hoc Wilcoxon test with the False Discovery Rate ( FDR ) multiple-comparison correction with a significance cutoff of p<0 . 05 . Because gaseous O2 production and fraction of total O2 in the form of gas were not significantly different between POP Experiments 1 and 2 ( Kruskal-Wallis , p>0 . 05 ) , the results of both experiments were combined in subsequent tests . Statistical tests comparing calcification functional groups were performed by combining corals with CCA as calcifying and turf algae with fleshy macroalgae as fleshy organisms , based solely on the presence of absence of calcification . No outliers were removed . All measurements from incubations had an associated error originating from probe accuracy , organism surface area measurement software , volume in incubation chambers , incubation time and microscope resolution . The uncertainty quantification of rates calculated using combination of measurements was performed based on error propagation ( Taylor , 1997 ) . For error generated from two or more variables , the derivative of the source errors were applied ( Rice , 2006 ) . All uncertainty propagated from the device measurements were at least five times smaller than the statistical standard deviations from the experiments , except for control incubations where the statistical variation is in the same order as the uncertainty , as expected ( Figures 2 , 3 and 4 source datasets ) . Therefore , uncertainty was not incorporated in statistical comparisons between treatments . The relative contribution of microbial respiration and ebullition to the removal of oxygen in an idealized reef was performed using a weighted linear model . The per-cell respiration rates from the dark experiment ( Figure 2—source data 1 ) were applied to the bacterial abundances sustained by one decimeter square ( dm2 ) of primary producer observed in POP experiment 2 ( Figure 3—source data 1 ) . These rates were scaled up to a 1 m3 reef by normalizing the rate by the percent cover of coral and fleshy algae on the benthos ( Figure 5—source data 2 ) . The model focused on these two primary producer groups due to the availability of data on their ebullition rates . The derivation of the model is provided below . The production of O2 in the gas phase per m2 was calculated from the POP experiments for corals , ρcoralgas , and fleshy algae , ρcoralgas . The net production of dissolved O2 per area ( m2 ) was also obtained from the POP experiments for coral , ρcoralnet , and fleshy algae , ρalganet . The microbial respiration rate of dissolved O2 per cell was obtained from the dark incubation experiment ( Figure 2 ) . The microbial concentration from the POP experiment at the end point was assumed as the reference value to obtain the rate of microbial respiration per volume ( m3 ) in the POP experiment . This was obtained for coral , Ωcoralmicrob , and fleshy algae , Ωalgamicrob , incubations . To estimate the gross production rate of dissolved oxygen in a reef , a boundary layer of a meter was considered , and above this layer the influence of the primary producers on the oxygen content was neglected ( Barott and Rohwer , 2012; Shashar et al . , 1996 ) . The gross production of dissolved O2 per meter square of primary producers was , thus , estimated from the coral incubations , Ocoraldiss=ρcoralnet×1 m2+ Ωcoralmicrob×1 m3 , and the fleshy algae incubations , Oalgadiss=ρalganet×1 m2+ Ωalgamicrob×1 m3 . The gross production rate of O2 was obtained by combining the dissolved oxygen and gas oxygen obtained from the coral , Ocoralgross=Ocoraldiss+ ρcoralgas×1 m2 , and algae , Oalgagross=Oalgadiss+ ρalgagas×1 m2 , experiments . The benthos of an ideal reef was assumed to be covered by a combination of coral and algae , C+A=1 , where C was the fraction of coral coverage and A the fraction of algal coverage . The gross production rate of O2 in a reef was estimated by the linear weighted model Oreefgross= C∙Ocoralgross+A∙Oalgaegross . The production rate of O2 gas , Oreefgas , and microbial consumption rate of O2 , Oreefmicrob , in the reef were calculated analogously . The coral coverage fraction was expressed in terms of algal coverage fraction , CA=1-A , to investigate the impact of algal coverage in the oxygen budget of the reef . The fraction of oxygen loss in ebullition was defined as E=Oreefgas/Oreefgross . In terms of the algal coverage , the ebullition fraction was given by the shifted Hill function of order oneEA=ΔOgasΔOgrossA+OcoralgasΔOgross OcoralgrossΔOgross+A Here , Δ ∆Ogross=Oalgagross-Ocoralgross and Δ ∆Ogas=Oalgagas-Ocoralgas . A similar expression was obtained for the fraction of oxygen rate consumed by microbes as a function of the fraction of algal coverage , M ( A ) , where the gas terms above become microbial terms . | Rising water temperatures , pollution and other factors are increasingly threatening corals and the entire reef ecosystems they build . The potential for corals to resist and recover from the stress these factors cause ultimately depends on their ability to compete against fast-growing fleshy algae that can rapidly take over the reefs . Living on the fleshy algae , the coral and in the surrounding water are communities of bacteria and other microbes that help maintain the health of the coral reef . Both corals and algae modify the chemical and physical environment of the reef to alter the composition of the microbial communities for their own benefit . Algae , for instance , release large amounts of sugars and other molecules of organic carbon into the water . These carbon molecules are then taken up by the bacteria , along with oxygen , to produce chemical energy via a process called respiration . This could cause the levels of oxygen in the water to decrease , potentially damaging the corals and creating more open space for the algae . Previous studies have revealed how communities of microbes on coral reefs use organic carbon , but it remains unclear how they affect the levels of oxygen in the reefs . To address this question , Silveira et al . used an approach called metagenomics to analyze the bacteria in samples of water from 87 reefs across the Pacific and the Caribbean , and also performed experiments with reef bacteria grown in the laboratory . The experiments showed that bacteria growing in the presence of fleshy algae became larger and more abundant than bacteria growing near corals , resulting in the water containing lower levels of oxygen . Furthermore , the fleshy algae produced bubbles of oxygen that were released from the water . Silveira et al . developed a mathematical model that predicted that these bubbles , combined with the respiration of bacteria that live near algae , caused the loss of 67% of the oxygen in the water surrounding the reef . These findings represent a fundamental step towards understanding how changes in the levels of oxygen in water affect the ability of coral reefs to resist and recover from stress . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2019 | Biophysical and physiological processes causing oxygen loss from coral reefs |
When microbes acquire new abilities through horizontal gene transfer , the genes and pathways must function under conditions with which they did not coevolve . If newly-acquired genes burden the host , their utility will depend on further evolutionary refinement of the recombinant strain . We used laboratory evolution to recapitulate this process of transfer and refinement , demonstrating that effective use of an introduced dichloromethane degradation pathway required one of several mutations to the bacterial host that are predicted to increase chloride efflux . We then used this knowledge to identify parallel , beneficial mutations that independently evolved in two natural dichloromethane-degrading strains . Finally , we constructed a synthetic mobile genetic element carrying both the degradation pathway and a chloride exporter , which preempted the adaptive process and directly enabled effective dichloromethane degradation across diverse Methylobacterium environmental isolates . Our results demonstrate the importance of post–transfer refinement in horizontal gene transfer , with potential applications in bioremediation and synthetic biology .
Microbes frequently acquire genetic material from organisms that are distant evolutionary relatives but close ecological neighbors ( Ochman et al . , 2000; Gogarten et al . , 2002; Smillie et al . , 2011 ) . These horizontally-transferred genes and pathways play important roles in processes ranging from the spread of antibiotic resistance to ecological differentiation ( Forsberg et al . , 2012; Shapiro et al . , 2012 ) . When , for example , a new environmental niche opens up through the introduction of a xenobiotic compound , horizontal gene transfer ( HGT ) can speed the assembly and dissemination of a corresponding catabolic pathway ( Springael and Top , 2004 ) . However , newly-acquired abilities may be costly for the host until they are carefully integrated into existing metabolic and regulatory networks ( Kim and Copley , 2012; Yadid et al . , 2013 ) . The challenges associated with such adaptations are highlighted by many examples from metabolic engineering , where productively transferring genes and pathways into new hosts often requires significant post–transfer refinement ( Chang et al . , 2007; Michener et al . , 2012 ) . Observing this process of transfer and refinement in nature is challenging , but we can recreate the same process in the laboratory using a combination of genetic engineering and experimental evolution . Laboratory evolution has been used extensively to study natural evolutionary processes . Recent experiments have highlighted the mechanisms by which evolution can optimize existing traits ( Elena and Lenski , 2003 ) , select for the emergence of a novel ability ( Blount et al . , 2012 ) , and refine a rudimentary pathway ( Quandt et al . , 2014 ) . Laboratory studies focusing on genetic exchanges between microbes have described the recombination of mutations within an evolving population ( Winkler and Kao , 2012 ) or the replacement of an endogenous metabolic pathway with an alternative route ( Chou et al . , 2011 ) . Meanwhile , metabolic engineers have used adaptive evolution as an engineering tool , including the evolution of highly modified strains , but under conditions that do not reproduce HGT ( Fong et al . , 2005; Trinh and Srienc , 2009; Lee and Palsson , 2010; Lee et al . , 2012 ) . In this work , we combined these approaches to study how microbes evolve following the acquisition through horizontal gene transfer of a novel metabolic ability by deliberately transferring a mobile metabolic pathway into a new host and then using experimental evolution to select for its efficient use . We investigated this process of HGT followed by evolutionary refinement using dichloromethane ( DCM ) catabolism in Methylobacterium strains as an example . Dichloromethane became an important industrial solvent during World War II and is still the most prevalent chlorinated solvent today ( www . eurochlor . org ) . Following widespread environmental contamination by DCM , several microbial strains have been isolated based on their ability to grow on DCM as the sole source of carbon and energy . Most notably , Methylobacterium extorquens DM4 , a methylotrophic Alphaproteobacterium , has served as a reference model to elucidate the details of bacterial growth with DCM ( Gälli and Leisinger , 1985; Muller et al . , 2011a ) . M . extorquens DM4 expresses a dehalogenase , DcmA , whose gene shows clear signs of horizontal transfer ( Schmid-Appert et al . , 1997; Vuilleumier et al . , 2009 ) and whose product , formaldehyde , feeds directly into central methylotrophic metabolism ( Chistoserdova , 2011 ) . Past efforts to understand the genetics of growth on DCM used gene knockouts to uncover DCM-specific genes ( Figure 1A ) ( Muller et al . , 2011b ) . We complemented these efforts using a synthetic approach involving conjugal transfer of the dcmA gene into several naïve Methylobacterium strains ( Figure 1B ) ( Kayser et al . , 2000; Michener et al . , 2014 ) . These dcmA transconjugants showed DCM dehalogenase activity but grew poorly on DCM , and we previously investigated factors that might explain their limited growth ( Michener et al . , 2014 ) . Growing on DCM using the dcmA catabolic pathway is very challenging for the host , as its products and intermediates include formaldehyde , hydrochloric acid , and the alkylating agent S-chloromethylglutathione ( Figure 1F ) ( Kayser et al . , 2000; Kayser and Vuilleumier , 2001 ) . As DCM dehalogenase activity was generally high in transconjugant strains ( Michener et al . , 2014 ) , we hypothesized that one or more of these stresses was limiting growth in the transconjugants and thus required post–transfer evolutionary refinement of the host to produce highly active DCM-degrading strains . 10 . 7554/eLife . 04279 . 003Figure 1 . Experimental evolution recapitulates post–transfer optimization of a challenging catabolic pathway . The natural isolate , DM4 , grows well on dichloromethane . ( A ) Past efforts to explain the genetics of growth on DCM relied on gene knockouts ( yellow ) to identify important genes ( Muller et al . , 2011b ) . ( B ) We deliberately transferred the dcmA gene ( red ) into naïve recipient Methylobacterium strains , which then grew poorly on DCM ( Michener et al . , 2014 ) . ( C ) Serial propagation on DCM selected for mutants with increased fitness on DCM . ( D ) Whole genome resequencing allowed us to identify these mutations ( blue ) , and reconstructing the individual mutations in wild type cells verified that they were causal . We then worked backwards from the mutations to identify the stress that the mutations overcame . ( E ) Introducing a plasmid containing both the pathway ( red ) and a solution to the most common limiting stress ( blue ) allowed efficient growth on DCM without chromosomal modifications to the host . ( F ) The biochemistry of DCM dehalogenation produces several challenging compounds ( red ) that potentially limit growth on DCM . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 003 In this work , we used experimental evolution to identify the factors that were limiting growth in dcmA-containing transconjugants . Beginning from these ancestral transconjugants , we evolved replicate populations with DCM as the sole source of carbon and energy and obtained evolved isolates with substantially higher fitness on DCM ( Figure 1C ) . We sequenced the genomes of these evolved strains to identify the mutations that increased fitness on DCM and then reconstructed these mutations in isogenic backgrounds to measure their effect on fitness ( Figure 1D ) . Based on these mutations , we could infer both the primary limiting stress that the pathway placed on its host as well as the biochemical mechanisms that the cells used to overcome this stress . Finally , we used this knowledge to uncover naturally-occurring refining mutations in two DCM-degrading environmental isolates and to design an improved gene cassette that enhances DCM bioremediation by preempting the need for post–transfer refinement ( Figure 1E ) .
We previously transferred the DCM dehalogenase dcmA into three strains of M . extorquens as well as three other species of Methylobacterium ( Michener et al . , 2014 ) . Each of the transconjugants was significantly less fit than the natural isolate , M . extorquens DM4 ( hereafter referred to as ‘DM4’ ) . In this work , replicate populations of five of these transconjugants were serially propagated with DCM as the sole carbon and energy source to select for increased fitness on DCM . One of these five transconjugants , M . extorquens AM1 , was initially unable to grow on DCM alone and instead was propagated on a mixture of DCM and methanol . After a total of 150 generations of growth , individual clones were isolated from each replicate population . We measured the fitness of each clone in direct competition with the reference strain , DM4 , and selected the most-fit clone from each population for further analysis . For four out of the five ancestral transconjugant strains , the fitness of isolates from each evolved population was significantly higher than the fitness of the ancestor during growth on DCM ( Figure 2 ) , increasing by 1 . 6- to 13-fold relative to the respective ancestral transconjugant . Methylobacterium radiotolerans was the only transconjugant that did not yield improved isolates after 150 generations , and these non-improved populations were not further characterized . 10 . 7554/eLife . 04279 . 004Figure 2 . Evolved isolates from four of the five transconjugants have improved fitness relative to the ancestor . A single clonal isolate was selected from each replicate population after 150 generations of growth on DCM . Each isolate was mixed with the reference strain , DM4ΔdcmA+pJM10 , and grown on DCM to determine competitive fitness . Error bars show ±1 standard deviation , calculated from three biological replicates . The horizontal dashed line indicates equal fitness to the reference isolate , DM4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 004 The genomes of all ancestral strains had previously been sequenced ( Vuilleumier et al . , 2009; Marx et al . , 2012 ) , so we used a combination of whole-genome resequencing and targeted Sanger sequencing to identify putative causal mutations in the evolved isolates . In the seven isolates that we resequenced , we repeatedly found mutations to the genes for the protein translocase secY , the chloride/proton antiporter clcA , and a hypothetical protein with a domain of unknown function ( DUF599 , which we renamed edgA for Evolved Dichloromethane Growth ) , as well as a single mutation to besA , a homolog of the eukaryal bestrophin family of chloride channels ( Figure 3A ) . Sanger sequencing confirmed these mutations in the resequenced isolates and targeted sequencing of these loci demonstrated that each remaining isolate had mutations at one of these four loci . We note that additional mutations were identified in each resequenced isolate ( Figure 3—source data 1 ) . 10 . 7554/eLife . 04279 . 005Figure 3 . Each evolved isolate has a mutation in at least one of four loci . ( A ) A combination of whole genome resequencing and targeted Sanger sequencing allowed the identification of mutations at four loci , encoding the protein translocase secY , the chloride/proton antiporter clcA , a hypothetical protein that we named edgA , and the bestrophin-homolog chloride channel besA . A simplified phylogenetic tree is shown at left ( Michener et al . , 2014 ) . Mutation names indicate the genetic locus followed by the isolate name in which it has been detected . ( B ) Reconstruction of these single mutations in AM1 and PA1 demonstrated that each individual mutation was beneficial , even in a host in which the mutation was not observed ( secY in PA1 and clcA in AM1 ) . ( C ) In PA1 , the double mutant besAE2/clcAE2 has similar fitness to the besAE2 single mutant when grown on DCM . During the evolution of population E2 , the clcAE2 mutation fixed first , followed by the besAE2 mutation . Each of the reconstructed strains expresses dcmA from plasmid pJM10 . Error bars show ±1 standard deviation , calculated from three biological replicates . N . D . : not determined . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 00510 . 7554/eLife . 04279 . 006Figure 3—source data 1 . Mutations identified during experimental evolution . N . D . : not determined . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 00610 . 7554/eLife . 04279 . 007Figure 3—figure supplement 1 . Mutations have relatively small fitness effects during growth on succinate compared to the benefits of threefold or more on DCM . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 00710 . 7554/eLife . 04279 . 008Figure 3—figure supplement 2 . Comparison of competitive fitness of single and double mutants in M . extorquens PA1 explains evolutionary patterns . All tested strains contain dcmA in pJM10 . ( A ) The secYA1/clcAE1 PA1 double mutant is less fit than the clcAE1 single mutant during growth on DCM . ( B ) The edgAE3/clcAE1 PA1 double mutant is less fit than the clcAE1 single mutant during growth on DCM . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 00810 . 7554/eLife . 04279 . 009Figure 3—figure supplement 3 . Allele frequency dynamics for M . nodulans populations . For each of the four M . nodulans populations , the yield ( as measured by OD600 ) was determined at the end of each serial culture . The near fixation of the indicated mutations , to secY and edgA , correspond to dramatic transitions in the culture yield . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 00910 . 7554/eLife . 04279 . 010Figure 3—figure supplement 4 . In population E2 , the clcAE2 mutation fixed first , followed by besAE2 . As shown in Figure 3C , the besAE2 mutation was moderately beneficial in the clcAE2 background , while the reverse was not true . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 010 Repeated mutations to the same loci in independently evolved replicates strongly suggested that the mutations were causal . However , to conclusively demonstrate causality we introduced several of these mutations into the ancestral strains M . extorquens AM1 and PA1 ( hereafter referred to as ‘AM1’ and ‘PA1’ ) . We focused here upon AM1 and PA1 , as these two strains incurred mutations to all four loci described above and are genetically tractable . Since the secY mutation was not identified in PA1 nor the clcA mutation in AM1 , we also constructed hybrid allele exchange vectors to move these mutations into the hosts in which they were not observed . We then measured the fitness of the reconstructed mutants against the natural isolate , DM4 . Each of the mutations was highly beneficial , regardless of whether the mutation arose in that host during evolution ( Figure 3B ) . In contrast , the mutations had only minor fitness effects during growth on an alternate carbon source , succinate , indicating that their beneficial effects were specific to the challenge of growing on DCM and were not broadly beneficial during laboratory growth ( Figure 3—figure supplement 1 ) . While each of the single mutations was beneficial , the fitness effects were not additive . For example , isolate E2 contained mutations in both clcA and besA . Each of these mutations was beneficial alone , but the double mutant was no more fit than the besAE2 single mutant ( Figure 3C ) . Similarly , when we constructed clcAE1/secYA1 and clcAE1/edgAE3 double mutants , we found that the double mutants were less fit than the clcAE1 single mutant ( Figure 3—figure supplement 2 ) . The protein translocase gene secY was mutated in seven of thirteen evolved isolates , totaling six amino acid mutations and one in-frame deletion . In the crystal structure of the SecY homolog from Methanocaldococcus jannaschii ( Van den Berg et al . , 2004 ) , these seven mutations cluster to the same region of the protein surface , at the interface of the channel pore and its blocking plug ( Figure 4A ) . Several of these mutations have been identified and characterized in the Escherichia coli SecY homolog ( with 58% amino acid identity to M . extorquens DM4 SecY ) based on their ability to secrete proteins with mistranslated sequences ( Emr et al . , 1981 ) . These mutations are predicted to disrupt interactions between the plug and pore , leading to a leaky channel ( Smith et al . , 2005 ) , and detailed biochemical investigation of one such mutant in E . coli demonstrated that this mutation allows the facilitated diffusion of small anions , most notably chloride ( Dalal and Duong , 2009 ) . 10 . 7554/eLife . 04279 . 011Figure 4 . Mutations in secY and clcA are predicted to increase chloride export . ( A ) When mapped to the crystal structure of SecY from Methanocaldococcus jannaschii ( Van den Berg et al . , 2004 ) , the seven mutations to secY cluster in a small region of the protein structure ( blue and cyan ) at the interface between the channel and plug ( red ) that is associated with leakage of small anions . ( B ) Mutations to the clcA promoter in isolates E1 and E2 lead to significantly increased transcription of clcA , as measured by qRT-PCR ( one tailed Welch's t-test with two degrees of freedom ) . Error bars show ±1 standard deviation , calculated from three biological replicates . ( C ) Adding DCM to strains that overexpress clcA leads to a larger decrease in internal pH . Strains contain pJM40 , expressing both dcmA and a fluorescent pH biosensor . The strain with a mutated clcA promoter ( blue ) overexpresses ClcA relative to the wild type strain ( green ) . At t = 0 , DCM was added to a final concentration of 10 mM . After an approximate delay of 20 s , the culture was continuously sampled with a flow cytometer to determine the dynamics of the internal pH . Both strains showed a transient decrease in pH , but the decrease is larger for the clcA overexpression strain ( p = 0 . 036 , one-tailed Welch's t-test with two degrees of freedom ) . ( D ) Mutations to secY , but not to clcA , increase chloride sensitivity . Growth rates of three PA1 strains ( wild type , secYA1 , and clcAE1 ) were measured using an automated assay system . The osmolarity of the medium was varied through the addition of sucrose or sodium chloride . The secYA1 mutant did not grow at the highest chloride concentration . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 01110 . 7554/eLife . 04279 . 012Figure 4—figure supplement 1 . Relative clcA expression is independent of the growth substrate . All five strains were grown to mid-log phase in M-PIPES containing either 5 mM DCM plus 0 . 075 mM methanol ( green ) or 3 . 5 mM succinate ( blue ) . For each growth substrate , clcA expression was measured by qRT-PCR and compared to wild type PA1 under the same conditions . Relative clcA expression across different strains was similar for both growth conditions . Error bars show ±1 standard deviation , calculated from three biological replicates . p values are calculated using a one-tailed Welch's t-test with two degrees of freedom . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 012 Similarly , the promoter of the chloride/proton antiporter clcA was mutated in four of the thirteen evolved isolates . The two evolved PA1 isolates with clcA promoter mutations showed increased levels of clcA mRNA compared to either the ancestor or the PA1 isolate with a wild type clcA promoter ( Figure 4B and Figure 4—figure supplement 1 ) . However , the E . coli ClcA antiporter ( with 33% amino acid identity to DM4 ) is capable of importing or exporting chloride , while necessarily transporting protons in the opposite direction ( Accardi and Miller , 2004 ) . To determine the directionality of ClcA-mediated transport in our strains , we used a pH biosensor to determine the effect of clcA overexpression on proton transport during growth with DCM . We observed a transient decrease in intracellular pH upon addition of DCM that was significantly larger in the strain overexpressing clcA ( Figure 4C ) , consistent with ClcA importing protons and exporting chloride . Next , we tested the effect of secY and clcA mutations on the chloride sensitivity of the mutant strains , since mutations that benefit the cell by facilitating chloride diffusion during growth on DCM may also result in detrimental levels of chloride import in high salt media . We measured the growth rate of strains on succinate with varying external chloride concentrations , using sucrose as a control for the effects of osmolarity . We were unable to detect a change in the sensitivity of AM1 mutants because the strain was already highly chloride sensitive ( Michener et al . , 2014 ) . In PA1 , the clcAE1 mutant was no more sensitive to NaCl or sucrose than the wild type strain ( Figure 4D ) . In contrast , the PA1 secYA1 mutant was highly sensitive to NaCl , but only slightly more sensitive to sucrose ( Figure 4D ) . We next asked whether our laboratory evolution experiments had recapitulated the natural process of adaptation to growth on DCM , thereby allowing the identification of refining mutations in environmental isolates . The sequences of secY , besA , and edgA in the reference isolate , DM4 , did not contain any of the mutations identified in the laboratory-evolved strains . However , we found that DM4 expressed significantly higher levels of clcA mRNA than the other three ancestral strains of M . extorquens tested ( Figure 5A ) . The sequences upstream of clcA differ between all four strains , making it difficult to identify the causal mutations in DM4 . Consequently , we exchanged a 140-bp region upstream of clcA ( containing a total of five nucleotide changes ) between DM4 and PA1 . The resulting DM4 strain with the PA1 clcA promoter had decreased clcA expression and minimal growth with DCM as the sole carbon and energy source , while the PA1 strain with the DM4 clcA promoter had increased clcA expression and was significantly more fit than the wild type PA1 strain ( Figure 5B ) . However , complementing the DM4 clcAE0 mutant strain with a clcA/dcmA expression plasmid restored growth on DCM ( Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 04279 . 013Figure 5 . DM4 requires clcA overexpression to grow on DCM . ( A ) The environmental DCM-degrading isolates , DM4 and DM17 , overexpress clcA relative to three other strains of M . extorquens ( one tailed Welch's t-test with two degrees of freedom ) . Error bars show ±1 standard deviation , calculated from three biological replicates . ( B ) Growth on DCM in DM4 and PA1 depends on the clcA promoter but not the strain background . A 140-bp region upstream of clcA ( the ‘clcA promoter’ ) was swapped between PA1 and DM4 . Regardless of the genetic background , strains with the DM4 clcA promoter overexpressed clcA and were more fit on DCM compared to strains with the PA1 clcA promoter . All strains expressed dcmA from plasmid pJM10 . Error bars show ±1 standard deviation , calculated from three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 013 Having identified necessary refining mutations in DM4 , we examined the other known DCM-degrading Methylobacterium isolate , M . extorquens DM17 ( Firsova et al . , 2010 ) , to determine whether a parallel event had occurred in its evolutionary history . When we sequenced the secY , besA , and edgA loci in this strain , the nucleotide sequences were exactly identical to M . extorquens AM1 . Compared to M . extorquens AM1 , however , the clcA locus of DM17 contained an insertion sequence 52 bp upstream of the clcA promoter , and we found that clcA expression was even higher in DM17 than in DM4 ( Figure 5A ) . Having demonstrated that clcA overexpression increases the fitness in several strains of M . extorquens during growth on DCM , we tested whether this effect would also hold in other environmental isolates across the Methylobacterium genus . We constructed a plasmid that expressed dcmA from the native DM4 promoter as well as clcA from a constitutive promoter . We then mated either our original dcmA plasmid ( pJM10 ) or this new dcmA/clcA plasmid ( pJM83 ) into a collection of natural Methylobacterium isolates . We found that every strain showed higher fitness on DCM with the dual-expression dcmA/clcA plasmid than with the single-expression dcmA plasmid ( Figure 6 ) . 10 . 7554/eLife . 04279 . 015Figure 6 . A mobile genetic element containing both dcmA and clcA allows a diverse array of Methylobacterium isolates to grow efficiently on DCM . Plasmids containing either dcmA ( x-axis ) or dcmA and clcA ( y-axis ) were introduced into 16 different Methylobacterium strains , including the five strains used to found the evolved populations ( black ) and 11 that were new ( blue ) . Each transconjugant was tested against the reference , DM4ΔdcmA+pJM10 , to determine its competitive fitness . Note that negative fitness values indicate net death over the course of the competition . The dashed line indicates equal fitness of the two transconjugants , as would be expected if clcA expression were neutral . Error bars show ±1 standard deviation , calculated from three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 01510 . 7554/eLife . 04279 . 016Figure 6—source data 1 . Environmental Methylobacterium strains characterized with the dual expression plasmid . Closest relative was determined based on the nearest BLAST hit to the 16S rRNA sequence of the isolate . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 01610 . 7554/eLife . 04279 . 014Figure 6—figure supplement 1 . Overexpression of clcA , either by the DM4 clcA promoter or the pJM83 clcA/dcmA dual expression plasmid , confers high fitness on DCM . Replacing the clcA promoter in DM4 ΔdcmA with the clcA promoter from PA1 severely reduces fitness on DCM . However , this fitness defect can largely be compensated through clcA overexpression by the introduction of the pJM83 dcmA/clcA dual expression plasmid . Error bars show ±1 standard deviation , calculated from three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 04279 . 014
Previous studies in experimental evolution generally focused on a single wild type microbial strain . In contrast , we evolved five different Methylobacterium strains , each of which carried the same horizontally-transferred metabolic pathway for DCM catabolism . By sequencing the evolved isolates , we identified mutations across diverse backgrounds that improve the ability to grow on DCM . Working backward from these mutations allowed us to determine that chloride accumulation appears to be the most critical factor limiting growth , and to identify multiple mechanisms that alleviate this stress . We found parallel responses at both the genetic and physiological level , indicating that the different strains faced similar challenges when forced to use their new metabolic capability and overcame these challenges in similar ways . Horizontal transfer of dcmA is not sufficient for efficient growth on DCM and , in both the laboratory and the environment , microbes must also acquire refining mutations to optimize the newly-acquired pathway . Several lines of evidence suggest that chloride export was the key limiting factor during growth on DCM . First , the observed mutations in secY are predicted to turn the protein translocase into a leaky chloride channel , as demonstrated previously in E . coli ( Dalal and Duong , 2009 ) . Our growth rate measurements confirm that secY mutants are more sensitive to extracellular chloride , as we would expect if the mutant SecY facilitates chloride diffusion across the cell membrane . Second , mutations to the promoter of clcA increase clcA expression . The in vivo pH measurements upon DCM addition establish that ClcA imports protons , with a necessary concomitant export of chloride , and that increased ClcA expression leads to increased proton import and chloride export . Similarly , the E . coli response to extreme acid stress uses ClcA to export chloride , with other mechanisms used for proton export ( Accardi and Miller , 2004 ) . Third , BesA is predicted to be a chloride channel ( Sun et al . , 2002 ) and while we cannot yet explain the mechanism by which the identified mutation would affect its function , we hypothesize that the mutation may also allow increased chloride export . Finally , we observed that double mutants are no more fit than the best single mutant , suggesting that all the mutations we detected are addressing the same physiological limitation . Previous experiments found that mutations affecting the same pathway were less beneficial when combined ( Chou et al . , 2011 ) , in some cases to the point of being collectively deleterious ( Kvitek and Sherlock , 2011; Rokyta et al . , 2011; Chou et al . , 2014 ) . Due to the antagonism of edgAE3 with clcAE1 , it is likely that EdgA is also involved in chloride export . While chloride export is not an unexpected stress during growth on DCM , there was no a priori reason to expect that chloride export was the major limiting stress across all of the strains that we tested . Multiple other stresses could have been dominant , including sensitivity to internal acidification , the solvent effects of DCM , changes in formaldehyde fluxes perturbing the native host metabolism , or the mutagenic effects of glutathione-conjugated DCM ( Kayser and Vuilleumier , 2001 ) . Indeed , our previous efforts to find a single physiological difference that explained the varied fitness of the ancestral dcmA transconjugants on DCM did not identify chloride sensitivity as a predictive phenotype for fitness during growth on DCM ( Michener et al . , 2014 ) . There was no reason to expect that chloride was more toxic than protons , but using ClcA to export the chloride , even at the cost of importing yet more protons , was highly beneficial . It appears that excess intracellular chloride is either more toxic or less effectively managed than excess intracellular protons . The apparent linkage between besA and chloride export may offer insights into a human disease . Mutations to human bestrophin homologs can lead to vitelliform macular dystrophy , but little is known about these proteins aside from their roles as chloride channels ( Sun et al . , 2002; Tsunenari et al . , 2003 ) . Bacterial homologs of the bestrophin family are similarly uncharacterized . Select bacterial homologs have been shown to affect swarming ( Inoue et al . , 2007 ) and sporulation ( Rahn-Lee et al . , 2009 ) , but these systems have not explicitly replicated the role of the human homolog as a chloride channel . However , our Methylobacterium system may offer an opportunity to characterize the chloride transporter function of this protein family in a tractable host with a strong linkage between BesA function and fitness . By design , our analysis has focused on mutations with very large fitness effects . Mutations that alleviate stresses imposed during growth on DCM had the potential to dramatically increase fitness , and restricting our evolution experiment to 150 generations predominantly selected these mutations . Our populations undoubtedly sampled mutations that would be generally beneficial during growth in our media and laboratory conditions , as has been seen in earlier experiments with M . extorquens AM1 ( Lee et al . , 2009 ) , but the fitness effects of these mutations appear to have been minor compared to the DCM-specific mutations . For example , isolate N1 lost an entire 458 kbp plasmid , which was likely due to selection but with a small fitness benefit as compared to acquiring a secY mutation ( Lee and Marx , 2012 ) . The rarity of mutations generically beneficial during laboratory growth greatly simplified our analysis of the evolved isolates . We expect that multiple lineages within a population independently arrived at solutions to increase chloride efflux . However , the size of the fitness effects produced evolutionary dynamics that resembled periodic selection ( Atwood et al . , 1951 ) , in which the first large effect mutation to arise in the population rapidly fixed . Combinations of mutations were generally no more beneficial than the best single mutation , so we predominantly observed the result of a single selective sweep ( Figure 3—figure supplement 3 ) . Population E2 is the exception , where the clcAE2 mutation fixed first , followed by the besAE2 mutation ( Figure 3—figure supplement 4 ) . The clcAE2 mutant has a fitness of 0 . 6 as compared to 0 . 8 for besAE2 or the besAE2/clcAE2 double mutant ( Figure 3C ) . Thus , while the double mutant is no more fit than the besAE2 single mutant , the besAE2 mutation is still highly beneficial in the clcAE2 background . Experimental populations of M . radiotolerans did not increase in fitness over 150 generations of growth on DCM . When we tested this strain with the dual-expression clcA/dcmA plasmid , the fitness increased but the increase was smaller than for other strains ( Figure 6—source data 1 ) . We hypothesize that our inability to find improved isolates of M . radiotolerans stems from a combination of fewer accessible beneficial mutations , smaller fitness effects of the beneficial mutations , and a decreased sensitivity to the mutagenic effects of growing on DCM ( Perez-Pantoja et al . , 2013 ) . Experimental evolution is increasingly used to simulate natural evolutionary processes ( Kawecki et al . , 2012 ) ; however , there are few examples where mutations that arise during laboratory evolution can also be identified in natural populations ( Wong et al . , 2012; Traverse et al . , 2013 ) . We have shown that mutations to clcA can be highly beneficial during laboratory growth , and that parallel mutations to clcA have arisen in two independent DCM-degrading environmental isolates of M . extorquens . In the reference environmental isolate M . extorquens DM4 , we can directly link nucleotide mutations in the clcA promoter to clcA overexpression and increased organismal fitness during growth on DCM . Demonstrating linkages between evolutionary processes observed in the laboratory and in nature will be increasingly important for the further development and application of experimental evolution . Having identified clcA overexpression as a beneficial evolutionary refinement following acquisition of dcmA , we used this knowledge to develop a novel bioremediation strategy and designed new expression cassettes for bioremediation that are less dependent on subsequent adaptation for effective function . Rather than introducing an exogenous microbe into a contaminated site , we propose to introduce a genetic cassette to the indigenous microflora , an approach known as genetic bioaugmentation ( Top et al . , 2002; Ikuma and Gunsch , 2013 ) . Our experiments clearly demonstrated that simply introducing the degradation pathway for a contaminant can be inefficient , as the recipients may be unprepared for the stresses produced by the pathway ( Kayser et al . , 2002; Michener et al . , 2014 ) . Instead , we propose to identify these stresses in the laboratory and then provide both the catabolic pathway and solutions to the most common limitations . In our example of DCM degradation , transconjugants that received both the catabolic gene dcmA and the chloride exporter gene clcA were more efficient at degrading DCM than those that received dcmA alone ( Figure 6 ) . In summary , our results highlight the important role of evolutionary refinement in the horizontal transfer of a challenging catabolic pathway . Dichloromethane degradation led to chloride accumulation and required modifications to the recipient to export the chloride . Based on recent examples from metabolic engineering , we expect that these types of host–pathway interactions are common ( Kizer et al . , 2008; Ro et al . , 2008; Verwaal et al . , 2010; Michener et al . , 2012; Bernhardt and Urlacher , 2014 ) and , consequently , that post–transfer refinement of microbial hosts is also widespread in nature . Our approach of deliberate horizontal transfer in the laboratory followed by experimental evolution offers an opportunity to identify both the limiting stresses and the mutations that can overcome these stresses . As screens and selections for complex microbial phenotypes are further developed , we expect that this approach will find many applications , both in the analysis of natural microbial isolates and in the optimization of genetically engineered microbes and mobile genetic elements .
E . coli cultures were grown in LB containing 10 g/l NaCl except for cultures containing the dcmA/clcA dual expression plasmid ( pJM83 ) , which were salt-sensitive and instead were grown in LB containing 0 . 5 g/l NaCl . Unless otherwise noted , Methylobacterium strains were grown in 10 ml cultures in a 50 ml flask at 30°C and 220 rpm in M-PIPES medium supplemented with 3 . 5 mM succinate or 5 mM DCM ( Delaney et al . , 2013a ) . Antibiotics were added to final concentrations of 10 µg/ml streptomycin , 12 . 5 µg/ml tetracycline , or 50 µg/ml kanamycin . All DCM cultures were grown in gas-tight 50 ml screw-top flasks sealed with Mininert valves ( Supelco , Bellefonte , PA ) and teflon tape . Before use , valves were sterilized with ethanol and dried in a laminar flow hood . Unless otherwise noted , chemicals were purchased from Sigma–Aldrich ( St . Louis , MO ) and enzymes from New England Biolabs ( Ipswich , MA ) . The additional Methylobacterium strains presented here were either isolated from the University of Washington campus ( L1 , D21 , D23 , D24; CJM , unpublished ) or from the surroundings of Woods Hole , MA ( J-4-1 , C-7-2 , C-7-1 , C-2-3 , G-1-1 , M-1-1; NF Delaney , unpublished ) . Strains and plasmids are listed in Supplementary file 1 . Transconjugant strains containing pJM10 were streaked to single colonies on M-PIPES + succinate + kanamycin plates . Liquid cultures were inoculated from single colonies into M-PIPES + succinate + kanamycin and grown to saturation . Cultures were then diluted 100× into M-PIPES + DCM to initiate the evolution experiment , starting 3–4 replicate flasks for each transconjugant . Every 3 . 5 days , cultures were diluted into fresh M-PIPES + DCM . Initial dilution factors were 16× , rising to 64× by the end of the experiment . Aliquots were taken at 12 , 24 , 36 , 48 , 60 , 90 , 120 , and 150 generations and frozen at −80°C in 8% DMSO . After 150 generations , the cultures were plated on M-PIPES + succinate plates to isolate single colonies . Individual colonies were restreaked onto M-PIPES + succinate plates , then inoculated into liquid cultures in M-PIPES + succinate . Liquid cultures were diluted 100× into M-PIPES + DCM to test for DCM growth . For each replicate population , the isolate with the highest yield on DCM as assayed by OD600 was selected for further analysis . The ancestral AM1 transconjugant was unable to grow on DCM alone . Consequently , the initial stages of evolution were performed in DCM supplemented with 0 . 75 mM methanol and kanamycin . For two of the populations , A2 and A3 , the populations were able to grow on DCM alone after 60 generations of evolution , and the remaining 90 generations were conducted on DCM alone . For the final population , A1 , the entire evolution experiment was performed with DCM + methanol + kanamycin . Competition experiments were conducted to compare the fitness of the tested strain against M . extorquens DM4 ΔdcmA hptA::Venus + pJM10 , as described previously ( Michener et al . , 2014 ) . Briefly , strains were grown to saturation in M-PIPES + succinate + kanamycin , and then diluted 100× into M-PIPES + DCM . After 3 days , cultures were mixed at defined ratios in fresh M-PIPES + DCM and allowed to grow for 3 more days . A flow cytometer was used to measure the population ratios before and after the final round of growth on DCM . The fitness of the isolate was then calculated based on the change in population ratios and the fold-growth of the mixed population . To prepare genomic DNA for sequencing , evolved isolates were streaked to single colonies on M-PIPES + succinate , and then grown to saturation in M-PIPES + succinate . Saturated cultures were centrifuged at 4500×g for 10 min , washed in 1 ml of water , and centrifuged at 8000×g for 3 min . After discarding the remaining supernatant , cell pellets were stored at −20°C overnight . Pellets were resuspended in 570 µl of TET ( 10 mM Tris pH 7 . 5 , 1 mM EDTA , and 1% Triton X-100 ) , heated to 90°C for 1 hr , then cooled to RT . Lysozyme was added to a final concentration of 2 mg/ml , and the suspension was incubated at 37°C for 30 min . Next , 30 µl of 10% SDS and 0 . 2 mg/ml of proteinase K were added , and the suspension was incubated for a further 1 hr at 37°C . After addition of 0 . 5 mg/ml RNAse A , the suspension was incubated at 37°C for 1 hr , heated to 90°C for 10 min , and 100 µl 5 M NaCl and 80 µl CTAB/NaCl ( 10% CTAB in 0 . 7 M NaCl ) were added . After a further 10 min incubation at 90°C , the mixture was extracted twice with phenol/chloroform , precipitated with isopropanol , air dried , and resuspended in 10 mM Tris pH 7 . 5 . DNA sequencing was performed at the Microarray and Genomic Analysis Core Facility of the University of Utah and the IBEST Genomics Resources Core of the University of Idaho using a HiSeq 2000 and yielding 90×–200× coverage . Genomic DNA libraries were constructed using a TruSeq DNA Sample Prep LT kit ( Illumina , San Diego , CA ) following the manufacturer's instructions . Genome sequences were analyzed using breseq-0 . 24 ( Deatherage and Barrick , 2014 ) . Putative mutations were confirmed by PCR amplification from the chromosome and Sanger sequencing . To determine the first appearance of a given mutation , the chromosomal loci were amplified by PCR from the mixed population aliquots that had been frozen at intermediate time points during evolution . The fraction of the population with a given allele was determined by Sanger sequencing of these mixed PCR products , with an estimated detection limit of ∼5% of the population . To construct plasmid pJM83 , plasmid pJM40 was amplified in an around-the-horn PCR to remove the pHluorin/mCherry coding frame . The clcA gene was amplified from PA1 gDNA with primers that added homology to the pJM40 expression construct ( promoter and terminator ) . The clcA insert was then cloned into the pJM40 PCR product using Gibson assembly ( Gibson et al . , 2009 ) . To construct the allele exchange vectors , plasmid pPS04 was digested with XbaI and SacI . The appropriate chromosomal locus , including ∼500 bp on either side of the desired mutation , was amplified by PCR using primers that added homology to the pPS04 vector backbone . The chromosomal amplicon was then cloned into the pPS04 backbone using Gibson assembly ( Gibson et al . , 2009 ) . In the case of plasmids pJM66 , pJM67 , pJM75 , pJM76 , pJM88 , and pJM89 , the mutant allele was moved into a strain in which it did not occur . Rather than cloning a single ∼1000 bp chromosomal locus , these plasmids were constructed from three separate amplicons . These amplicons included a ∼500 bp upstream homology region amplified from the recipient chromosome , a small region containing the desired mutation amplified from the donor chromosome , and a second ∼500 bp downstream homology region amplified from the recipient chromosome . The PCR primers used added the appropriate homology to combine all three amplicons into the pPS04 backbone in a single Gibson assembly reaction . Plasmids were mated into recipient Methylobacterium strains using triparental matings as described previously ( Fulton et al . , 1984 ) . Allele exchanges were performed as described ( Marx , 2008 ) . Putative allele exchange mutants were confirmed by Sanger sequencing a PCR amplification of the modified chromosomal locus . In vivo pH measurements were conducted as described previously ( Michener et al . , 2014 ) . Briefly , an expression plasmid containing both dcmA and a pHluorin-mCherry translational fusion ( pJM40 ) was mated into the appropriate strain . The transconjugant was grown to saturation in M-PIPES + succinate + tetracycline , then diluted 100× into M-PIPES + DCM , and grown for three days . Cultures were diluted to a final optical density of 0 . 01 before analysis on an LSRII flow cytometer ( BD , Franklin Lakes , NJ ) . pHluorin was excited at 488 nm and measured at 530/30 nm . mCherry was excited at 561 nm and measured at 620/40 nm . Samples were gated for forward scatter , side scatter , and mCherry fluorescence . After determining the population fluorescence , DCM was added to a final concentration of 10 mM . The culture was quickly vortexed and returned to the flow cytometer , leading to an approximately 20 s delay between DCM addition and consistent fluorescence measurements . After DCM addition , the population fluorescence was monitored for a further 10 min . For each strain , a standard curve was constructed by diluting the DCM-grown culture to a final OD600 of 0 . 01 in a solution composed of 30 mM buffer , 50 mM NaCl , 3 mM KCl , 10 µM valinomycin , and 10 µM nigericin . Buffers used were MES pH 5 . 1 , MES pH 5 . 3 , MES pH 5 . 5 , MES pH 5 . 7 , MES pH 6 . 1 , PIPES pH 6 . 5 , PIPES pH 6 . 9 , and PIPES pH 7 . 3 . The population mean fluorescence ratio ( pHluorin fluorescence divided by mCherry fluorescence ) was measured for each combination of strain and pH , then fit to a modified Henderson–Hasselbalch equation . The internal pH of the experimental samples was calculated by finding the fluorescence ratio of each cell , dividing the timecourse into 5 s intervals , calculating the population mean of the fluorescence ratio for each interval , and comparing that ratio to the standard curve to calculate the internal pH . For qRT-PCR measurements , cultures were grown to saturation in M-PIPES plus succinate , and then diluted into the indicated media conditions . Upon reaching mid-log phase ( typically at half the optical density of a saturated culture ) , cultures were centrifuged for 10 min at 4500×g and 4°C , washed once with 1 ml of water , and pelleted again at 8000×g for 3 min . Cell pellets were flash frozen in liquid nitrogen and stored at −80°C overnight . RNA was extracted using an RNeasy Kit ( Qiagen , Germantown , MD ) and RNAse-free DNAse ( Qiagen ) according to the manufacturer's instructions . Total RNA was reverse transcribed using SuperScript III ( Life Technologies , Carlsbad , CA ) , approximately 1 µg of RNA , and gene specific primers ( rpsB FWD 5′-ACCAACTGGAAGACCATCTC-3′; rpsB REV 5′-CTTCTCGAGCTTGTCCTTCTCAC-3′; clcA FWD 5′-ATCGTCACCGAGATGACCCAG-3′; clcA REV 5′-CCAAGGTGTGATAGAGGCCG-3′ ) according to the manufacturer's directions . cDNA was quantified by qPCR , using a CFX-96 qPCR machine ( Bio-Rad , Hercules , CA ) and EvaGreen qPCR mix ( Biotium , Hayward , CA ) . Three technical replicates were performed for each biological replicate . For each replicate , the clcA concentration was normalized by the rpsB concentration . For each strain , the average normalized clcA concentration was compared to the reference strain , typically PA1 . Growth rates were measured using an automated system as described previously ( Delaney et al . , 2013b ) . Briefly , strains were grown to saturation in M-PIPES + succinate and then diluted 64× into 640 µl of M-PIPES + succinate with the appropriate concentration of osmolyte ( NaCl or sucrose ) in a 48-well plate . After the cultures reached saturation , they were again diluted 64× into 640 µl of the appropriate media , with three replicate wells per condition . The optical density of these assay plates was monitored every 45 min until the cultures again reached saturation . The growth rate of the culture was calculated using CurveFitter ( Delaney et al . , 2013a ) . | Many microbes can rapidly evolve to adapt to new or extreme habitats . Most often the characteristics that develop via evolution result from individuals inheriting new combinations of genes from their parents . However , many species can also acquire new genes through a process called horizontal gene transfer , where organisms that share an environment exchange potentially useful genes . The ability of bacteria to acquire genes by horizontal gene transfer is thought to be the main reason for the rapid evolution of antibiotic-resistant bacteria such as MRSA . A horizontally transferred gene rarely works efficiently when first transferred , and may even cause problems for its new host . The host organism must therefore evolve further after receiving a new gene , but it is difficult to trace how this occurs . A toxic chemical called dichloromethane ( or DCM ) has been used in many industries since World War II , and has caused widespread contamination to the environment . Strains from several microbial genera have been identified that have adapted to break down DCM and use it as their sole energy source . Moreover , a gene responsible for breaking down DCM appears to have been shared between different species by horizontal gene transfer . In work that was presented earlier in 2014 , researchers introduced this gene into several bacterial strains from the genus Methylobacterium that had not been previously exposed to DCM . However , the resulting new strains of bacteria still had difficulties growing on DCM . There were diverse ways that the bacteria could have been prevented from growing—for example , multiple by-products generated when DCM is broken down are highly toxic . Now Michener et al . —including the researchers that performed the 2014 work—have experimentally evolved five of the modified strains of Methylobacterium to discover how they adapt over time to living on DCM . Examining the DNA sequences of these bacteria showed several mutations that both improved the ability of the bacteria to survive on DCM and helped the bacteria to more efficiently remove chloride ions from their systems . Chloride ions accumulate as DCM is broken down; therefore , Michener et al . suggest that being exposed to too much chloride prevented the bacteria in previous experiments from growing well . This idea is further supported by mutations found in previously isolated bacteria that live on DCM . These mutations increase the bacteria's ability to excrete chloride ions from their cells , and when Michener et al . removed these mutations , the bacteria no longer grew well on DCM . Understanding how bacteria adapt once they have acquired the DCM-degrading gene could make it easier to help bacteria to mop up DCM and other contaminants in the environment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2014 | Effective use of a horizontally-transferred pathway for dichloromethane catabolism requires post–transfer refinement |
Bruton’s tyrosine kinase ( BTK ) is targeted in the treatment of B-cell disorders including leukemias and lymphomas . Currently approved BTK inhibitors , including Ibrutinib , a first-in-class covalent inhibitor of BTK , bind directly to the kinase active site . While effective at blocking the catalytic activity of BTK , consequences of drug binding on the global conformation of full-length BTK are unknown . Here , we uncover a range of conformational effects in full-length BTK induced by a panel of active site inhibitors , including large-scale shifts in the conformational equilibria of the regulatory domains . Additionally , we find that a remote Ibrutinib resistance mutation , T316A in the BTK SH2 domain , drives spurious BTK activity by destabilizing the compact autoinhibitory conformation of full-length BTK , shifting the conformational ensemble away from the autoinhibited form . Future development of BTK inhibitors will need to consider long-range allosteric consequences of inhibitor binding , including the emerging application of these BTK inhibitors in treating COVID-19 .
Bruton’s tyrosine kinase ( BTK ) is a non-receptor tyrosine kinase belonging to the TEC family ( Rawlings and Witte , 1995; Berg et al . , 2005 ) . BTK is expressed primarily in B cells and myeloid cells , where it functions downstream of receptors including the B-cell receptor , Fc receptors , and toll-like receptors ( Kurosaki , 2011; López-Herrera et al . , 2014; Xu et al . , 2012; Hartkamp et al . , 2015; Ellmeier et al . , 2011 ) . B-cell receptor engagement leads to BTK activation ( via phosphorylation on Y551 in the activation loop ) , which results in the direct phosphorylation and activation of phospholipase C gamma 2 ( PLCγ2 ) triggering calcium flux and subsequent regulation of gene transcription ( Kurosaki , 2011 ) . In addition to the clear role of the BTK catalytic domain in activating PLCγ2 , the N-terminal regulatory domains of BTK , which include the Pleckstrin homology and Tec homology ( PHTH ) domain , a proline-rich region ( PRR ) , a SRC homology 3 ( SH3 ) domain , and a SRC homology 2 ( SH2 ) domain ( Figure 1a ) , mediate autoinhibitory and activating interactions that control BTK function . Indeed , both kinase-dependent and independent activities of BTK have been shown to contribute to proper signaling downstream of immune receptors ( Saito et al . , 2003; Halcomb et al . , 2007 ) . Previous work has established that the regulatory SH3 and SH2 domains of BTK assemble into a compact autoinhibitory conformation on the ‘distal’ surface of the kinase domain , similar to the autoinhibited SRC kinases ( Figure 1b; Joseph et al . , 2017; Wang et al . , 2015 ) . Solution data support an additional autoinhibitory interface on the activation loop face of the kinase domain occupied by the PHTH domain ( Figure 1b; Amatya et al . , 2019; Devkota et al . , 2017 ) . Upon receptor-mediated activation , the regulatory domains of BTK bind their activating ligands , which releases the catalytic domain from its intramolecular interactions , thereby permitting activation loop phosphorylation and a conformational shift of the catalytic machinery from the inactive to active state . BTK has become an important pharmacological target ( Bond et al . , 2019; Feng et al . , 2019; Kim , 2019 ) due to its essential role downstream of immune receptors; for example , increased receptor signaling increases BTK activity driving multiple B-cell malignancies , autoimmune diseases , and allergies ( Pal Singh et al . , 2018; Smith , 2017; Mohamed et al . , 2009 ) . Ibrutinib is an irreversible ATP-competitive inhibitor that binds covalently to BTK C481 within the kinase active site ( Figure 1c , d ) . This small molecule is the first BTK-specific inhibitor to be approved by the FDA and has become the frontline drug in efforts to control a variety of B-cell disorders . Crystallographic analysis reveals that Ibrutinib-bound BTK kinase domain adopts an inactive conformation ( Bender et al . , 2017 ) ; the activation loop is collapsed into the active site , the αC-helix is out , the regulatory spine is not assembled , and the conserved Lys/Glu salt bridge is broken ( Figure 1d ) . Ibrutinib is currently approved for the treatment of chronic lymphocytic leukemia ( CLL ) , Mantle cell lymphoma ( MCL ) , Waldenstrom’s macroglobulinemia , Marginal zone lymphoma ( MZL ) , and chronic graft versus host disease and is in various stages of clinical trials for the treatment of other B-cell disorders ( Bond et al . , 2019; Zi et al . , 2019; Davids and Brown , 2014; Molica et al . , 2020; Treon et al . , 2015; Aw and Brown , 2017; Wang et al . , 2013 ) . More selective , second-generation BTK inhibitors , acalabrutinib and zanubrutinib ( Barf et al . , 2017; Guo et al . , 2019; Wu et al . , 2016 ) , have since received FDA approval bringing the total number of approved therapies that target BTK to three . Recent findings demonstrate a possible role for the BTK inhibitor acalabrutinib in treating excessive inflammation in the context of severe COVID-19 ( Roschewski et al . , 2020 ) . Off-label use of acalabrutinib for hospitalized COVID-19 patients showed normalization of inflammatory markers and decreased oxygen requirements ( Roschewski et al . , 2020 ) . Development of reversible BTK inhibitors has also progressed with a number of candidates in clinical development ( Brown et al . , 2016 ) . To date , all BTK inhibitors currently being used or in clinal trials are active site inhibitors . Despite the success of Ibrutinib in the clinic , drug resistance is a major ongoing challenge . Based on previous kinase-targeted therapies , it has become clear that acquisition of drug resistance occurs via a variety of mechanisms including loss of drug binding , activation of the kinase , or the activation of downstream substrates ( Lovly and Shaw , 2014; Sherbenou et al . , 2010; Yun et al . , 2008 ) . Of the Ibrutinib resistance mutations that have been identified in BTK , all except one ( T316A ) cluster to the BTK active site ( Puła et al . , 2019 ) . The most prevalent mutation occurs at BTK C481 , the residue in the kinase active site that covalently binds to Ibrutinib ( Figure 1d; Lampson and Brown , 2018 ) . The BTK C481S mutation and other kinase active site Ibrutinib resistance mutations are thought to interfere with stable Ibrutinib binding ( Furman et al . , 2014; Woyach et al . , 2014 ) . In contrast , the BTK T316A mutation occurs in the BTK SH2 domain ( Figure 1b ) and is the only Ibrutinib resistance mutation that has been identified to date outside of the BTK kinase domain ( Sharma et al . , 2016; Kadri et al . , 2017 ) . The mechanism by which the remote T316A mutation confers Ibrutinib resistance is currently unknown . Although an abundance of data including pharmacology , associated toxicities , metabolic processing , clearance , potency , and efficacy are available for the full-length BTK protein ( Bond et al . , 2019; Kim , 2019; Molica et al . , 2020; Owen et al . , 2019 ) , the major molecular-level insights provided by high-resolution BTK:drug complex crystal structures are limited to the kinase domain fragment of BTK with drug bound to the active site cavity ( Kuglstatter et al . , 2011; Xing and Huang , 2014 ) . Structural consequences of drug binding to the full-length BTK protein have remained unexplored . Importantly , work from other kinase systems have shown that drug binding to the kinase active site can block catalytic activity as expected but can also have unanticipated and important long-range allosteric effects that impact protein/ligand interactions ( Sonti et al . , 2018; Leonard et al . , 2014 ) . To expand our mechanistic understanding of drug resistance and the effect of different inhibitors on the full-length BTK protein , we developed solution-based Nuclear Magnetic Resonance ( NMR ) and Hydrogen Deuterium Exchange Mass Spectrometry ( HDX-MS ) tools to probe local and long-range conformational consequences of resistance mutation and drug occupancy in the BTK active site . A panel of five BTK inhibitors , Ibrutinib , GDC-0853 ( Fenebrutinib ) , CGI1746 , CC-292 ( Spebrutinib , AVL-292 ) , and Dasatinib , was assembled based on availability of crystallographic data and commercial availability of each drug . Additionally , drug choices were prioritized to represent a range of kinase structures ( activation loop ‘in’ versus ‘out’ , αC-helix ‘in’ versus ‘out’ ) and different inhibitor binding modes ( covalent versus non-covalent ) . All inhibitors in this study , except Dasatinib , have been developed as BTK-specific inhibitors . Our findings provide the first molecular insights into the conformational response of full-length BTK toward a spectrum of small-molecule active-site inhibitors . The inhibitors fall into two categories , differentiated by the presence or absence of an inhibitor-induced effect on the regulatory regions distant from the active site . We find that Ibrutinib and Dasatinib have profound effects on the entire BTK protein that extend beyond the active site . In contrast , GDC-0853 , CGI1746 , and CC-292 do not alter the conformational state of the regulatory domains . Additionally , we explore the mechanism of action of the T316A resistance mutation . This single remote amino acid mutation destabilizes the autoinhibited conformation of full-length BTK . We propose that an increase in the population of active enzyme contributes to the decreased Ibrutinib sensitivity observed in vivo . Taken together , these data provide a framework for developing new strategies to target BTK in the face of acquired drug resistance and reveal significant differences in the overall conformational preferences of full-length BTK bound to distinct small molecule inhibitors . We anticipate that further development of BTK inhibitors will rely on such information to fully understand , and ultimately control , the consequences of drug interactions within the full-length BTK protein .
A conserved structural switch in the αC-helix of protein kinases plays a crucial role in controlling catalytic activity ( Taylor et al . , 2015 ) . The conformational equilibrium between an inactive and active kinase domain involves a switch in the αC-helix between an ‘αC-out’ and ‘αC-in’ conformation , respectively , which in turn either breaks or creates a salt bridge between a conserved Lys/Glu pair ( Figure 2a ) . Crystal structures of active and inactive BTK kinase domain show that the conformational shift in αC is accompanied by a change in the rotamer conformation of the BTK W395 sidechain ( Figure 2a ) . We have established use of the BTK W395 side chain indole NH resonance to monitor the conformational preference of the BTK kinase domain by solution NMR methods ( Joseph et al . , 2017 ) . The upfield W395 resonance was previously assigned ( Joseph et al . , 2017 ) to the inactive , αC-out conformation and in the full-length apo BTK protein this resonance is in slow exchange with the downfield resonance corresponding to the active , αC-in kinase domain conformation ( Figure 2a ) . This W395 resonance signature is consistent with previously established resonance frequencies of tryptophan indoles that are hydrogen bonded to water ( Fenwick et al . , 2018 ) . The ~0 . 2 ppm upfield shift observed for the W395 indole NH proton likely reflects weaker hydrogen bonding in the more nonpolar environment of the W395 indole NH in the αC-out conformation compared to αC-in rather than a complete removal of the bound water , which is characterized by a > 2 ppm upfield shift ( Fenwick et al . , 2018 ) . To test whether Ibrutinib binding to the BTK kinase domain in solution stabilizes the αC-out conformation observed in the co-crystal structure ( Bender et al . , 2017 ) , we monitored Ibrutinib binding to the linker-kinase domain fragment of BTK ( Figure 1a , LKD , residues 381–659 ) focusing on the tryptophan indole region of 1H-15N TROSY HSQC spectra ( Figure 2b ) . The W395 indole NH resonance in the apo BTK LKD is in the downfield position ( Figure 2b , top spectrum ) showing that the LKD fragment , in the absence of Ibrutinib , adopts the active or αC-in conformation in solution consistent with activity assay results reported previously ( Joseph et al . , 2007 ) . Titration of Ibrutinib into BTK LKD shows an increase in the intensity of the W395 peak in the upfield position , and a corresponding decrease in the intensity of the W395 peak in the downfield position ( Figure 2b ) . This is consistent with a conformational change in solution from the αC-in to the αC-out state observed in the crystal structure of the kinase:Ibrutinib complex ( Figure 2b ) . In fact , at a 1:1 molar ratio of protein to drug , the BTK LKD almost exclusively populates the inactive conformation consistent with a single binding site for Ibrutinib ( Figure 2b ) . That the upfield resonance corresponds to W395 in Ibrutinib-bound BTK LKD was confirmed by acquisition of the same spectrum of the BTK W395A mutant bound to Ibrutinib ( Figure 2—figure supplement 1 ) . The intact mass analysis of the full-length BTK:Ibrutinib sample also supports the single drug binding site ( Figure 2c ) . Next , we acquired 1H-15N TROSY HSQC spectra for full-length apo BTK and full-length BTK bound to Ibrutinib ( Figure 2d , iii , iv ) . As previously reported ( Joseph et al . , 2017 ) , full-length apo BTK ( as compared to the BTK LKD ) gives rise to more than one W395 resonance indicating exchange between αC-in and αC-out kinase domain conformations likely reflecting the increased conformational heterogeneity in the full-length protein due to the presence of the regulatory domains ( Figure 2d , iii ) . Ibrutinib binding to full-length BTK shifts the conformational equilibrium to the αC-out conformation as observed for Ibrutinib-bound LKD ( Figure 2d , iv ) . The greater W395 upfield shift induced by Ibrutinib compared to apo full-length BTK ( red versus gray dashed line ) may be due to local chemical shift changes in the presence of Ibrutinib and/or further stabilization of the αC-out conformer . Ibrutinib binding to full-length BTK also results in appearance of a new peak corresponding to the indole NH resonance of a different tryptophan , W251 in the SH3 domain ( Figure 2d , iv , peak denoted with asterisk ) . This resonance is broadened beyond detection in the apo full-length BTK spectrum due to intermediate exchange and so its appearance upon addition of Ibrutinib suggests that dynamics in the SH3 domain region have been altered . We have previously found that the W251 indole resonance is also visible in spectra of BTK Y223A ( Joseph et al . , 2017 ) , a mutation that disrupts the autoinhibited conformation of BTK shifting the SH3 domain toward an unengaged state more similar to the isolated SH3 domain ( Figure 2—figure supplement 2 ) . Thus , the appearance of the W251 resonance in the spectrum of Ibrutinib-bound full-length BTK ( Figure 2d , iv ) suggests that drug binding to the kinase active site shifts the conformational preference of the regulatory SH3 domain away from its autoinhibitory contacts on the distal side of the kinase domain . This observation prompted further exploration of the consequences of other BTK active site inhibitors on full-length BTK to determine whether this is a shared feature of active site inhibition of BTK . Crystal structures of the BTK kinase domain have been solved in complex with different active site inhibitors ( Bender et al . , 2017; Crawford et al . , 2018; Marcotte et al . , 2010; Di Paolo et al . , 2011 ) . A comparison of the available structures shows that the majority of inhibitors ( e . g . Ibrutinib [irreversible , covalent] , GDC-0853 [reversible] , CGI1746 [reversible] , and CC-292 [irreversible , covalent] ) stabilize the inactive or αC-out conformation ( Figure 3a ) . In addition to the αC-out conformation , a key structural feature of the inactive kinase domain includes the activation loop collapsed toward the active site burying the Y551 phosphorylation site . In most of the BTK:drug complex structures , the activation loop is resolved but electron density is missing for part of the activation loop in the CC-292-BTK structure ( Figure 3a ) . Superposition of these four complexes shows that the overall conformation of the kinase domain is similar regardless of bound inhibitor ( Figure 3b ) . The Dasatinib-bound structure is different; the kinase domain in this complex adopts the active conformation with the αC-helix moved in toward the active site ( αC-in ) and the activation loop is not visible in the electron density ( Figure 3c ) . Superposition of the Dasatinib- and Ibrutinib-bound BTK reveals the conformational change in the αC-helix , which is accompanied by the rotamer shift in the side chain of W395 ( Figure 3d ) . We next compared the tryptophan indole region of the 1H-15N TROSY HSQC spectra of apo BTK LKD to BTK LKD bound to the five different active site inhibitors ( Figure 4a ) and find that binding of GDC-0853 and CGI1746 shifts the conformational equilibrium to the inactive ( αC-out ) conformation in a manner similar to Ibrutinib ( Figure 4a ) . In the inhibitor-bound spectra for GDC-0853 and CGI1746 , the W395 indole NH resonance is entirely in the upfield position ( Figure 4a ) and , like the ibrutinib-bound protein , there is no visible peak in the downfield position . These data recapitulate the crystallography; BTK LKD bound to Ibrutinib , GDC-0853 , and CGI1746 all stabilize the inactive ( αC-out ) kinase domain conformation ( Figures 3a and 4a ) . Interestingly , in the GDC-0853-bound BTK LKD spectrum , there is a weak W395 peak adjacent to the major W395 peak ( Figure 4a ) , indicating that the BTK LKD additionally adopts a minor drug-bound conformation . The crystal structure of BTK linker-kinase bound to CC-292 reveals clear features associated with the inactive kinase ( Bender et al . , 2017 ) ; the αC-helix is positioned out and away from the active site , and the W395 side chain adopts the rotamer conformation consistent with inactive state ( Figure 3a ) . However , unlike the Ibrutinib , GDC-0853 , and CGI1746 complexes , the activation loop in the CC-292 crystal structure is not resolved , suggesting that this segment of the protein remains dynamic when bound to CC-292 . Comparison of the tryptophan resonances for apo BTK LKD to BTK LKD bound to CC-292 shows the W395 indole NH resonance remains in the downfield position indicating that binding of CC-292 to the active site does not shift the conformational preference of the kinase domain in solution toward the αC-out state ( Figure 4a ) . Chemical shift perturbations evident elsewhere in the spectrum confirm binding of CC-292 to BTK LKD ( Figure 4a ) . Our results suggest that the features of the kinase domain that are trapped by crystallography ( Figure 3a ) do not accurately represent the conformational ensemble of the CC-292-bound kinase domain in solution . Finally , the crystal structure of Dasatinib-bound BTK ( Marcotte et al . , 2010 ) is distinct from the other inhibitor-bound structures ( Figure 3b–d ) and the behavior of the W395 resonance is consistent with the crystal structure . Crystallography and NMR both indicate Dasatinib-bound BTK adopts the active ( αC-in ) conformation ( Figure 4a ) . Dasatinib binding causes a further downfield shift in the 1H resonance frequency of W395 suggesting that either binding of this drug increases the population of the αC-in conformation compared to apo BTK LKD or may be reflective of local chemical shift changes due to drug binding to the active site . Acquisition of separate 1H-15N TROSY HSQC spectra for Ibrutinib , GDC-0853 , CGI1746 , CC-292 , and Dasatinib bound to the BTK linker-kinase ( W395A ) mutant confirms the W395 assignments ( Figure 2—figure supplement 1 ) . To test whether the conformational states of the BTK LKD complexed to the different drugs are similar for full-length BTK , we analyzed the corresponding NMR spectra of full-length BTK bound to each of the drug molecules . Binding of Ibrutinib , GDC-0853 , and CGI1746 all shift the conformational equilibrium of the kinase domain in the full-length protein to the αC-out conformation in a manner that is nearly identical to drug binding to the shorter LKD fragment ( Figure 4b ) . Although the magnitude and direction of the chemical shift changes are the same for the fragment and full-length BTK , the inhibitor-bound W395 indole NH resonances in these full-length BTK complexes are more complex with multiple overlapping peaks likely reflecting conformational heterogeneity around the αC-out state of the multi-domain , full-length protein ( Figure 4a and b ) . Despite its covalent binding to the active site , CC-292 bound to full-length BTK highlights the dynamic nature of the bound complex that is partially evident in the BTK:CC-292 crystal structure ( Figure 3a ) . The 1H-15N TROSY HSQC spectrum of full-length BTK bound to CC-292 gives rise to a range of W395 indole NH resonances corresponding to both the αC-in and αC-out states ( Figure 4b ) . It is likely that the crystal structure of BTK LKD bound to CC-292 captured the lowest energy , inactive conformation , whereas solution NMR methods reveal the extent of conformational sampling of the CC-292/BTK complex . And last , Dasatinib binding to full-length BTK leads to extensive line-broadening throughout the spectrum . The W395 resonance is no longer detected but other resonances remain visible ( Figure 4b ) . The observed line-broadening likely reflects increased dynamics of the conformational ensemble upon Dasatinib binding to the active site . While distinctly different from one another , the BTK:Dasatinib and BTK:CC-292 complexes both exhibit increased dynamic motions in the kinase domain compared to BTK bound to Ibrutinib , GDC-0853 , or CGI1746 . To complement the NMR data ( which reveals the conformational status of just the kinase domain ) , we next sought further details about the effects of drug binding on the regulatory domains within the full-length protein by subjecting the BTK:drug complexes to hydrogen-deuterium exchange mass spectrometry ( HDX-MS ) . We have previously established that HDX-MS can be used to assess the conformational status of each domain in full-length BTK as the conformational equilibrium shifts between the ‘closed’ autoinhibited and the ‘open’ active states ( Joseph et al . , 2017 ) . In solution , full-length BTK exists predominantly in the ‘closed’ autoinhibited conformation ( Figure 1b ) . BTK activating mutations that destabilize the autoinhibitory conformation result in an increase in deuterium uptake , compared to wild-type BTK , in peptides derived from the SH3 , SH2 , and SH2-kinase linker region ( Joseph et al . , 2017 ) . Building on this approach , full-length BTK was mixed separately with either a twofold molar excess of Ibrutinib or a 10-fold molar excess of GDC-0853 , CGI1746 , CC-292 , or Dasatinib and each sample was subjected to HDX-MS analysis . We compared the effects of each inhibitor on full-length BTK using a set of peptic peptides common among all six forms ( apo and five inhibitor-bound states , the identities of this subset of peptides are listed in Source data 1 ) which report on the conformational status of the regulatory domains and the kinase domain . Inhibitor binding to full-length BTK causes clear effects on conformation and dynamics as measured by HDX-MS , and there are stark differences between the different inhibitors ( Figures 5 and 6 ) . Comparison of deuterium uptake for Ibrutinib conjugated to full-length BTK with that of apo BTK ( Figures 5a and 6 ) shows that Ibrutinib labeling of C481 of BTK FL alters deuterium exchange in several different regions of the protein . As expected , the β2 and β3 strands in the kinase N-lobe as well as the β7 , β8 strands and the N-terminus of the kinase activation segment ( DFG motif ) show reduced deuterium incorporation that is consistent with the location of Ibrutinib binding in the kinase active site ( Figure 6 ) . Importantly , despite Ibrutinib stabilizing an inactive ( αC-out ) conformation in the kinase domain , increases in deuterium exchange are observed in both SH3 and SH2 domains as well as in the SH2-kinase linker , consistent with destabilization of the autoinhibitory interface ( Figures 5a and 6 ) . Similar increases in deuterium uptake in BTK regulatory elements have been observed for mutations that shift the equilibrium away from the autoinhibited BTK conformation and increase activity ( Joseph et al . , 2017 ) . Together , the NMR and HDX-MS data indicate that Ibrutinib binding to the active site favors a ‘hybrid’ conformational state of full-length BTK: an inactive ( αC-out ) conformation is stabilized in the kinase domain , while the regulatory domains are at least partially released from the distal surface of the kinase domain to populate an open ‘active’ conformation . The allosteric effects of Ibrutinib binding to the BTK active site are quite distinct from the effects induced by GDC-0853 and CGI1746 ( Figures 5a , b , c and 6 ) . Despite the strong similarities between the W395 resonance frequencies ( Figure 4 ) and the crystal structures of Ibrutinib , GDC-0853 , and CGI1746 bound to the BTK kinase domain ( Figure 3a , b ) , the HDX-MS data show that GDC-0853 and CGI1746 lead to pronounced protection from exchange in the kinase domain but they do not affect the regulatory domains ( Figures 5b , c and 6 ) . Protection induced by GDC-0853 and CGI1746 in the regions surrounding the active site is similar to that observed for Ibrutinib albeit greater in magnitude . Specifically , upon binding of GDC-0853 and CGI1746 the activation loop ( e . g . Figure 5b , c , peptide iv ) is more strongly protected than when bound to Ibrutinib while parts of the αF and αG helices are strongly protected with GDC-0853 and CGI1746 but not protected at all by Ibrutinib ( Figures 5a , b , c and 6 ) . It is also important to note that peptides that cover β1–3 in the N-lobe of the kinase domain ( e . g . Figure 5 , peptide iii ) are both sensitive to the presence of the inhibitor ( protection from exchange in the inhibitor-bound forms compared to the apo-form ) and see an attenuation in the degree of that protection ( difference between BTK:Ibrutinib and BTK:GDC-0853/CGI1746 ) as the regulatory domains are released from the autoinhibited conformation . While the HDX-MS changes within the kinase domain are consistent with the binding mode and contacts being made by the drug within the BTK active site in the BTK:drug co-crystals ( Figure 3b ) , HDX-MS of full-length BTK reveals that there are clear conformational differences in the BTK regulatory domains between the drug-bound forms in solution . An extension of this type of drug-dependent allostery has been recently reported in the context of a dual-liganded enzyme ( Ghode et al . , 2020 ) . The kinase domain conformation of the BTK:CC-292 complex shares structural similarity with the Ibrutinib , GDC-0853 , and CGI1746 BTK complex crystal structures but the solution behavior of the CC-292 complex points to differences . While CC-292 is covalently attached to BTK C481 and contacts the hinge region of the kinase , it does not extend further into the kinase active site setting it apart from the other BTK inhibitors ( Figure 3 ) . Consistent with this , HDX-MS reveals only limited protection upon CC-292 binding localized primarily to the N-lobe surrounding the active site ( Figure 5d , peptide iii , 6 ) . The lack of protection from deuterium exchange in the activation loop of BTK upon binding to CC-292 is consistent with high flexibility and the absence of electron density for this region in the crystal structure of CC-292 bound to BTK LKD ( Figure 3a ) . The smaller and more confined HDX differences between apo- and CC-292-bound full-length BTK are consistent with the similar conformational modulation evident in the W395 resonance for apo BTK and BTK:CC-292 ( Figure 4b ) . Finally , the BTK:CC-292 complex , like the GDC-0853 and CGI1746 complexes , is characterized by a lack of changes in the regulatory domains . As full-length BTK is predominantly in the autoinhibited conformation in solution ( Joseph et al . , 2017 ) , this result indicates that binding of GDC-0853 , CGI1746 , and CC-292 do not shift the equilibrium away from the inactive , autoinhibited conformation of full-length BTK . HDX-MS of the BTK:Dasatinib complex shows that this complex exhibits the largest differences compared to apo full-length BTK ( Figures 5e and 6 ) . In contrast to the other BTK:drug complexes , Dasatinib binding causes dramatic increased deuterium uptake in a large part of the kinase domain C-lobe and activation loop ( Figure 5e , peptide iv , 6 ) , a result suggesting that these regions of the protein become less structured and more solvent exposed upon binding to Dasatinib . Such changes are consistent with the extensive NMR line broadening ( Figure 4b ) and the absence of electron density for the activation loop in the crystal structure of Dasatinib bound to BTK ( Figure 3c ) . In addition , the SH3 and SH2 domains and the SH2-kinase linker region all exhibit increased deuteration upon Dasatinib binding to the BTK active site ( Figures 5e and 6 ) and the extent of deuterium uptake is greater than that observed for Ibrutinib binding suggesting Dasatinib shifts the population away from the autoinhibited form to a greater extent than Ibrutinib ( Figure 5a , e , peptide i and 6 ) . Indeed , the W395 resonance suggests that the kinase domain in the Dasatinib:BTK complex populates the αC-in conformation rather than the αC-out conformation stabilized by Ibrutinib providing a possible explanation for how Dasatinib induces greater exposure in the regulatory domains . Taking the results for all the inhibitors together , NMR and HDX-MS data provide unique insights into the extent to which different BTK inhibitors affect the conformational preferences and dynamics of both the kinase domain itself and the regulatory domains within full-length BTK . Ibrutinib and Dasatinib promote changes in the conformational equilibria of full-length BTK , whereas the other inhibitors , GDC-0853 , CGI1746 , and CC-292 , cause significantly more localized conformational adjustments limited to the kinase domain . Refocusing on the clinically important BTK inhibitor , Ibrutinib , we turn next to evaluating the cause of Ibrutinib resistance upon mutation of BTK T316 to alanine . Unlike the majority of Ibrutinib resistance mutations , including the most common C481S mutation in the kinase domain active site ( Furman et al . , 2014; Woyach et al . , 2014; Young and Staudt , 2014; Cheng et al . , 2015; Zhang et al . , 2015 ) , BTK T316A is located outside of the kinase domain in the BTK SH2 domain ( Figure 1b; Sharma et al . , 2016; Kadri et al . , 2017 ) . The location of T316 suggests that mutation at this position might destabilize autoinhibitory contacts between the SH2 domain and the kinase domain C-lobe . T316 is adjacent to R307 , a key residue in the autoinhibitory salt bridge interaction between the SH2 domain and D656 in the C-terminus of the kinase domain ( Figure 7a , b; Joseph et al . , 2017 ) . Based on the crystal structure of this region of autoinhibited BTK ( Wang et al . , 2015 ) , T316 is predicted to stabilize the productive autoinhibitory conformation of the R307 sidechain via a network of hydrogen bonds , hydrophobic , and cation-pi contacts involving R288 , R307 , S309 , H333 , D656 , and E658 ( Figure 7a , b ) . Loss of even the small threonine side chain could be deleterious to this local structure and lead to a conformational change that is propagated across the autoinhibitory interface involving the entire SH3-SH2-linker region , thereby activating BTK . To more fully understand the consequences of the T316A mutation on BTK and to gain insight into how a mutation that is remote from the active site confers resistance to Ibrutinib , we applied NMR and HDX-MS approaches to the full-length BTK T316A mutant . Comparing the extent of deuterium uptake for wild-type BTK versus the T316A mutant shows small increases in exchange in the non-catalytic regulatory SH3-SH2 region upon mutation of T316 to alanine ( Figure 7c , d ) . Mapping the peptides that show increased deuterium uptake for the BTK T316A mutant onto the autoinhibited structure shows that the regions affected by the mutation in the SH2 domain localize to the SH3-SH2 domains and SH2-kinase linker that together create autoinhibitory contacts with the distal side of the kinase domain ( Figure 7e ) . It should be noted that these small 0 . 5–1 . 0 Da changes cannot be specifically localized within the regions indicated and so the peptides shown encompass a region larger than the actual change in deuteration . As well , the changes observed for the T316A mutation are consistent with a more transiently populated ‘open’ conformation compared to the more drastic changes in deuterium uptake observed upon mutation of D656 in previous work ( Joseph et al . , 2017 ) . Loss of the D656 sidechain abolishes the autoinhibitory salt bridge with R307 ( Figure 7b ) and leads to activation of BTK ( Joseph et al . , 2017 ) . In contrast to the changes in deuterium exchange observed for the BTK T316A mutant , the BTK C481S active site resistance mutant shows no change in exchange behavior compared to the wild-type protein ( Figure 7d , Source data 1 ) . NMR analysis confirms that the T316A mutation in the BTK SH2 domain shifts the conformational ensemble of BTK toward the active kinase ( αC-in ) population . Comparing the 1H-15N TROSY HSQC spectrum of the full-length BTK T316A to wild-type BTK shows a change in the conformational equilibrium upon mutation ( Figure 8a ) . Peak intensities for the W395 resonance in the upfield and downfield positions indicate that 72% of full-length wild-type BTK adopts a conformation consistent with inactive kinase ( αC-out ) and mutation of T316 to alanine results in only 60% of the full-length kinase in an inactive conformation . Accordingly , the population of active BTK as measured by W395 peak intensity increases ~1 . 5-fold , from 28% for wild-type BTK to 40% for the BTK T316A mutant ( Figure 8a ) . Consistent with the HDX-MS data ( Figure 7c ) , the partial shift in population to the αC-in state in the T316A mutant is small when compared to the the BTK D656K activating mutant , which shows an almost complete shift in the conformational ensemble to the active ( αC-in ) state ( Joseph et al . , 2017 ) . To directly test kinase activity , we compared autophosphorylation of BTK wild-type and T316A mutant under identical kinase assay conditions . Phosphorylation on Y551 in the BTK T316A activation loop is twice that of wild-type BTK consistent with the shift in the conformational ensemble measured by NMR ( Figure 8b , c ) . BTK C481S mutant is included the kinase assay and , as predicted by the identical deuterium exchange behavior of this mutant compared to wild-type , no difference in kinase activity is observed between wild-type BTK and the C481S mutant ( Figure 8b , c ) . Thus , the remote T316A resistance mutation in the BTK SH2 domain leads to an increase in the population of active enzyme that is reflected by an increase in activity . The shift in the conformational ensemble of BTK T316A away from the autoinhibited conformation could give rise to a conformational state that disfavors Ibrutinib binding , thereby contributing to Ibrutinib resistance . We examined Ibrutinib conjugation to the BTK T316A mutant by following the W395 resonance and the HDX-MS signatures . Both experimental methods require protein concentrations in the micromolar range and so it is noted that the experiments are carried out well above the nanomolar IC50 values previously reported for Ibrutinib ( Furman et al . , 2014; Woyach et al . , 2014; Hamasy et al . , 2017 ) . W395 reports a shift toward the inactive ( αC-out ) kinase domain conformation upon Ibrutinib binding to the BTK T316A mutant that mirrors the results of Ibrutinib binding to wild-type BTK ( Figure 8d ) . Ibrutinib binding to the T316A mutant results in increased deuterium exchange in the region spanning the SH3 , SH2 domains and SH2-kinase linker in a manner that is nearly identical to wild-type protein ( Figure 8e , Figure 8—figure supplement 1 ) . Thus , the Ibrutinib-induced remote conformational changes in the regulatory region of BTK appear to be a general consequence of Ibrutinib binding that occurs regardless of the T316A mutation in the BTK SH2 domain . The covalent nature of Ibrutinib binding to the BTK active site prevents standard measurement of affinity constants . To assess whether the T316A mutation in the SH2 domain affects drug binding to the active site , we instead performed an HDX-MS titration series to compare Ibrutinib binding to wild-type and T316A BTK . No meaningful differences ( >0 . 5 Da ) in deuterium uptake are observed between wild-type BTK and the BTK T316A mutant over a range of Ibrutinib concentrations ( Figure 8—figure supplement 2 ) . Thus , our results indicate that the BTK T316A mutation likely causes resistance to Ibrutinib ( Sharma et al . , 2016 ) by shifting the conformational ensemble of full-length BTK to favor the active state . At concentrations of Ibrutinib that exceed the IC50 , there is no evidence that the T316A mutation in the SH2 domain limits binding of drug to the BTK active site .
The essential role that kinases play in cell signaling make them attractive pharmacological targets ( Zhao and Bourne , 2018; Müller et al . , 2015 ) . The vast majority of available kinase inhibitors target the kinase active site and our work shows that the occupancy of the kinase active site by different inhibitors can have profoundly distinct effects on the distant BTK regulatory domains , effects that are missed in crystal structures of drug-bound kinase domain alone . Here , we provide the first report of the impact of binding of the irreversible FDA-approved drug Ibrutinib on the conformation of full-length BTK . Approval of an increasing number of covalent inhibitors demonstrate the effectiveness of irreversible drugs compared to reversible inhibitors in the targeted therapies of kinases ( Zhao and Bourne , 2018; Bauer , 2015 ) . A major advantage of irreversible inhibitors is that once covalently bound to the kinase , they permanently trap the protein , thereby sequestering its catalytic function until fresh protein is produced by translation ( Barf and Kaptein , 2012 ) . Consequently , covalent inhibitors are particularly effective in the treatment of kinases that have a long cellular half-life ( Bauer , 2015 ) . Additionally , covalent inhibitors are selected for rapid clearance from the serum to prevent toxicity effects arising from off-target modification ( Barf and Kaptein , 2012 ) . Ibrutinib levels in the serum peak 1–2 hr post-ingestion and return to basal levels after 3–4 hr ( Advani et al . , 2013 ) . The BTK C481S resistance mutation prevents the covalent attachment of Ibrutinib to the active site during the period when Ibrutinib is available in the serum but Ibrutinib still binds noncovalently ( Furman et al . , 2014 ) . This reversible drug binding coupled with rapid serum clearance implies that the BTK C481S mutant is not ‘trapped’ by Ibrutinib . Once serum Ibrutinib levels diminish , BTK signaling by the BTK C481S mutant occurs resulting in drug resistance . The BTK T316A resistance mutant retains the ability to be covalently modified by Ibrutinib . Thus , when the BTK T316A mutant encounters Ibrutinib it is likely ‘trapped’ by the covalent drug in a manner similar to wild-type BTK . However , the BTK T316A mutation is itself activating; loss of the T316 sidechain destabilizes a key down-regulatory interaction within BTK and shifts the conformational equilibrium away from the autoinhibited state ( Figures 7c and 8a ) . And so , newly synthesized BTK T316A could escape the narrow inhibition time window created by rapid Ibrutinib clearance . The unconjugated and active T316A mutant could then promote BTK signaling , unlike properly autoinhibited wild-type BTK . The idea that resistance mutations can counteract the autoinhibitory state has been elegantly demonstrated in recently published work on the ABL kinase ( Xie et al . , 2020 ) . It should be noted that we have not characterized binding of Ibrutinib to the phosphorylated form of BTK and so it is possible that phosphorylated BTK T316A could exhibit diminished affinity for Ibrutinib thereby also contributing to the observed resistance . Our findings now show that mutation is not the only mechanism by which the BTK conformational ensemble shifts away from the ‘closed’ autoinhibited state . Binding of either Ibrutinib or Dasatinib triggers the same shift in the conformational equilibrium of the regulatory domains away from the autoinhibited conformation creating the opportunity for BTK regulatory domain interactions with exogenous ligands despite the inhibited kinase domain . This conformational shift could promote kinase-independent signaling events ( Saito et al . , 2003; Middendorp et al . , 2003; Middendorp et al . , 2005 ) and/or create a dominant negative . Varied effects of inhibitor binding on the non-catalytic functions of kinases have been observed previously and range from alterations in interactions with upstream or downstream regulatory kinases and phosphatases to changes in ligand affinity ( Sonti et al . , 2018; Leonard et al . , 2014; Tong et al . , 2017; Skora et al . , 2013 ) . For BTK specifically , recent evidence suggests that non-catalytic functions of Ibrutinib-bound BTK activate CLL-specific PLCγ variants ( Wist et al . , 2020 ) . Thus , selection of BTK inhibitor type to treat specific disease states as well as design of the next generation of BTK inhibitors need to carefully consider the impact of the inhibitor on BTK regulatory domain conformation . Of the five active site inhibitors studied here only CGI1746 , CC-292 , and GDC-0853 inhibit kinase activity without disrupting the autoinhibitory conformation of the full-length kinase . Ibrutinib and Dasatinib alter the conformational equilibrium of the multidomain kinase decreasing the population of the autoinhibited kinase even though the drugs block and inactivate the kinase domain . The effect of an active site inhibitor on the overall conformational preference of a multidomain kinase must originate in the kinase domain . Multiple mobile elements including the αC-helix , DFG motif , and the activation loop switch the kinase domain between its inactive and active conformations ( Taylor et al . , 2015; Taylor and Kornev , 2011; Fabbro et al . , 2015 ) . A number of studies are emerging that focus on the long-range conformational effects of active site inhibitor binding to the SRC module kinase families: SRC , ABL , and TEC . There is an overriding effort to define a unifying metric that can predict the outcome of inhibitor binding on the conformation of the regulatory domains and different studies have endeavored to link disassembly of the regulatory domains to different mobile elements within the kinase domain . For the SRC family , the αC-in configuration has been correlated with release of the regulatory domains from the autoinhibited state ( Leonard et al . , 2014; Tsutsui et al . , 2016 ) . Other studies suggest that the αC-helix conformation alone can be used to predict the regulatory domain conformation in the SRC , ABL , and TEC kinases; inhibitors that stabilize the αC-in conformation disrupt the regulatory domains from their autoinhibitory conformation , whereas αC-out stabilizing inhibitors do not ( Fang , 2020 ) . However , ABL focused studies show that Dasatinib stabilizes αC-in but does not disrupt the regulatory domains from their compact , autoinhibitory state ( Sonti et al . , 2018 ) . For ABL , a strict correlation has been observed instead between activation loop conformation and regulatory domain assembly ( Sonti et al . , 2018; Skora et al . , 2013 ) . Inhibitors that stabilize an activation loop ‘in’ conformation disrupt the ABL SH3 and SH2 domains from their autoinhibited conformation , whereas inhibitors that stabilize the activation loop ‘out’ conformation do not . Our results suggest simple αC-helix or activation loop ‘in’ versus ‘out’ metrics are not sufficient to explain BTK inhibitor effects . Just one example from the data presented here shows that both Ibrutinib and CGI1746 favor the activation loop ‘in’ configuration , but the regulatory domains are displaced upon Ibrutinib binding and remain in the autoinhibited conformation on binding CGI1746 . To understand why Ibrutinib and Dasatinib disrupt the autoinhibited conformation of full-length BTK while GDC-0853 , CGI1746 , and CC-292 do not , we turn back to the crystal structures of BTK LKD bound to each inhibitor ( Figure 3 ) . Probing for features that are common to the Ibrutinib- and Dasatinib-bound structures but distinct from the other inhibitor-bound structures , we observed two distinct drug orientations among the five BTK:drug complexes ( Figure 9a , b ) . Interestingly , the orientations of Ibrutinib and Dasatinib ( the two inhibitors that disrupt the full-length BTK autoinhibitory structure ) , differ significantly from that of GDC-0853 , CGI1746 , and CC-292 ( Figure 9a , b ) . Ibrutinib and Dasatinib are oriented toward the ‘back pocket’ of the kinase active site ( also called the back cleft or hydrophobic pocket II ( HPII ) ) ( Roskoski , 2016 ) , whereas GDC-0853 , CGI1764 , and CC-292 bypass the back pocket and extend into the front pocket toward the activation loop ( Figure 9a , b ) . All five compounds maintain contacts with the kinase hinge region ( Figure 9a ) . In accessing the back pocket of the active site , Ibrutinib and Dasatinib alter the back pocket around the regulatory ( R- ) spine ( Taylor et al . , 2015; Kornev and Taylor , 2010; Kornev et al . , 2008 ) residue L460 ( Figure 9c ) . Ibrutinib directly contacts L460 , while Dasatinib binding induces the inward movement of the αC-helix bringing M449 in direct contact with L460 ( Figure 9c ) . In contrast , the compounds that extend into the front pocket , such as GDC-0853 ( Figure 9c ) , avoid all contact , direct or indirect with L460 . L460 is adjacent to Y461 , a component of the conserved ‘hydrophobic stack’ . The ‘hydrophobic stack’ consists of two residues on the distal surface of the kinase domain ( Y461 and W421 in BTK ) and a single hydrophobic residue from the linker spanning the SH2 and kinase domains ( L390 in BTK ) ( Figure 9a ) . Regulatory domain release from the distal side of the kinase domain leads to disruption of the hydrophobic stack and is accompanied by exchange of ADP for ATP in the active site , thus linking the conformational preference of the regulatory domains and the occupancy of the active site ( von Raußendorf et al . , 2017 ) . Given the impact of Ibrutinib and Dasatinib on the R-spine residue L460 and the fact that the hydrophobic stack residue Y461 is adjacent to L460 , we can speculate that small molecules that fill ( directly or indirectly ) the back pocket of the active site , may allosterically trigger conformational adjustments in the hydrophobic stack region that result in displacement of the regulatory domains from their autoinhibited pose . This is a new metric for predicting global conformational effects of specific BTK inhibitors that does not depend on strict conformational requirements being met in the αC helix and/or activation loop . Structural studies that combine data from X-ray crystallography , solution NMR and HDX-MS provide a detailed and more complete picture to identify key regulatory interactions at work in protein kinases . In addition , use of the full-length kinase reveals unexpected differences between BTK inhibitors that are missed in crystallographic analyses of fragments . For BTK , this work has demonstrated that the clinically important inhibitor Ibrutinib imposes conformational adjustments on the distant non-catalytic domains of BTK , which may be functionally relevant in vivo . BTK inhibitors that do not induce a conformational shift toward the ‘open’ form of BTK may improve the efficacy of disease treatment where the kinase-independent functions of BTK play an important role . Additionally , targeting downstream BTK substrates such as PLCγ2 or combination therapies of Ibrutinib together with reversible active site/allosteric inhibitors that have a prolonged serum half-life may reduce the type of resistance embodied by the T316A mutation . The sustained presence of reversible inhibitors in the serum may block the activity of newly synthesized BTK T316A and prevent escape from Ibrutinib . Recognizing the conformational impact that mutations and inhibitors such as Ibrutinib have on full-length BTK should inform future strategies to effectively target this kinase in the context of disease .
The bacterial expression constructs for murine BTK LKD , SH3-SH2-Linker-kinase domain ( 32LKD ) and full-length ( FL ) have been described previously ( Joseph et al . , 2017 ) . All BTK constructs carry the solubilizing Y617P mutation ( Joseph et al . , 2017 ) . All mutations were made using the site-directed mutagenesis kit ( Agilent ) , and the sequences of all constructs were confirmed by sequencing at the Iowa State University DNA facility . Ibrutinib , Dasatinib , CC-292 and CGI1746 were purchased from Selleckchem . GDC0853 was purchased from MedKoo Biosciences . Full-length , kinase-active BTK , 32LKD , and BTK LKD were produced by co-expressing BTK with YopH in BL21 ( DE3 ) ( Millipore Sigma ) or BL21-Gold ( DE3 ) cells ( Agilent Technologies ) as described previously ( Joseph et al . , 2017 ) . Briefly , the culture was grown at 37°C to an O . D . 600 nm of 0 . 6 to 0 . 8 . The temperature of the culture was lowered to 18°C and then induced with either 1 . 0 mM IPTG for BTK LKD construct or 0 . 1 mM IPTG for BTK full-length and 32LKD . The culture was harvested 24 hr after induction and the pellets were resuspended in lysis buffer ( 50 mM KH2PO4 , pH 8 . 0 , 150 mM NaCl , 20 mM imidazole and 0 . 5 mg/ml lysozyme ) and stored at −80°C . Cells were lysed by thawing and the action of lysozyme , and 3000 U DNAse I ( Sigma ) and 1 mM PMSF were added to the lysate , incubated at RT for 20 min and then spun at 16 , 000 rpm for 1 hr at 4°C . Glycerol was added to the supernatant to a final concentration of 10% and was then incubated with Ni-NTA resin ( QIAGEN ) for 2 hr , washed with Tris pH 8 . 0 , 75 mM NaCl , 40 mM imidazole , and eluted in 20 mM Tris pH 8 . 0 , 150 mM NaCl , 250 mM Imidazole , and 10% glycerol . Eluted protein was flash frozen in liquid nitrogen and stored at −80°C . The proteins were concentrated and dialyzed into the final NMR buffer which consists of 20 mM Tris pH 8 . 0 , 150 mM Sodium chloride , 0 . 02% Sodium azide , and 10% glycerol . For HDX-MS analysis , the proteins were further purified by size exclusion chromatography ( Hiload Superdex 26/60 200 pg or Hiload Superdex 26/60 75 pg , GE Healthcare ) . The fractions containing pure protein were pooled , concentrated , snap frozen and stored at −80° C . The final buffer consists of 20 mM Tris pH 8 . 0 , 150 mM Sodium chloride , 0 . 02% Sodium azide and 10% glycerol . Phosphorylation level on Y551 of all purified BTK proteins used in this study is below western immuno-detection . Uniformly 15N-labeled BTK samples were produced as described earlier by growth in modified M9 minimal media containing 15N ammonium chloride ( 1 g/L , Cambridge Isotope Laboratories , Inc ) as the sole source of nitrogen ( Joseph et al . , 2017 ) . For the expression of BTK full-length , the M9 minimal media was supplemented with 0 . 1 mM Zinc chloride . The final NMR sample buffer consists of 20 mM Tris , 150 mM Sodium chloride , 10% glycerol , pH 8 . 0% , and 0 . 02% Sodium azide . All NMR spectra were collected at 298 K on a Bruker AVIII HD 800 spectrometer equipped with a 5 mm HCN z-gradient cryoprobe operating at a 1H frequency of 800 . 41 . NMR samples with inhibitors consisted of 150 μM 15N-labeled BTK , mixed with 200 μM inhibitor in 2% DMSO . All data were analyzed using NMRViewJ ( Johnson and Blevins , 1994 ) . Peak intensities were quantified using the integrate function within Bruker TopSpin . Prior to intact mass analysis , purified BTK full-length wild-type ( 20 μM ) and Ibrutinib ( 40 μM ) , both in 20 mM Tris pH 8 . 0 , 150 mM NaCl , 10% glycerol , 2% DMSO , were allowed to interact at 21°C for 1 hr . Both the free wild-type BTK , as a control , and Ibrutinib-labeled proteins ( 70 picomoles ) were injected into a Waters nanoACQUITY with HDX technology set-up for intact mass analysis at 20°C . The proteins were injected into the sample loop and desalted for 3 min at 100 μL/min using water ( 0 . 1% formic acid ) on an in-house packed POROS 20 R-2 trap . Proteins were eluted into the mass spectrometer using a 15–70% ACN ( 0 . 1% formic acid ) gradient in 10 min at a flow rate of 100 μL/min . Mass spectra were acquired using a Waters Synapt HDMSE mass spectrometer operated in TOF only mode with a standard electrospray source , capillary voltage of 3200 V and a cone voltage of 40 V with a mass range of 50–2000 m/z . Intact mass values for both free and ibrutinib-labeled wild-type BTK were calculated from the raw m/z spectra using MaxEnt1 within MassLynx 4 . 1 ( Waters ) with a resolution of 0 . 10 Da and an output mass range of 65 , 000–90 , 000 Da . General procedures for HDX-MS of BTK have been described in detail previously ( Joseph et al . , 2017 ) . Details specific to experiments conducted here are provided in Source data 1 in the format recommended ( Masson et al . , 2019 ) for HDX-MS experimental descriptions . All HDX-MS data have been deposited to the ProteomeXchange Consortium via the PRIDE ( Vizcaíno et al . , 2016 ) partner repository with the dataset identifier PXD020029 . Briefly , prior to continuous labeling HDX experiments , purified BTK full-length wild-type , T316A , and C481S ( 20 μM ) and Ibrutinib 40 μM , ( 20 mM Tris pH 8 . 0 , 150 mM NaCl , 10% glycerol , 2% DMSO ) were allowed to interact at 21°C for 1 hr ( Note: Ibrutinib was only added to wild-type and T316A BTK ) . For the Ibrutinib titration experiment , labeling was performed for 10 min for BTK:Ibrutinib at molar ratios of 1:0 . 2 , 1:0 . 5 , 1:1 , and 1:2 . After the binding reactions , both the free kinase and kinase bound to Ibrutinib were placed on ice prior to deuterium labeling . Deuterium labeling proceeded for the times described using labeling buffer , and labeling was stopped with an equal volume of quench buffer at 0°C ( details in Source data 1 ) . Full-length wild-type BTK was independently mixed with CGI-1746 , CC-292 , GDC-0853 , or Dasatinib to final concentrations of BTK 20 µM , inhibitor 200 µM and these mixtures were incubated for 1 hr at 21°C . Unbound kinase ( as control ) or kinase bound to each of the small molecules was deuterated for the times described using labeling buffer , and labeling was stopped with an equal volume of quench buffer at 0°C ( details in Source data 1 ) . Quenched samples were immediately analyzed using a Waters nanoACQUITY with HDX technology using online pepsin digestion with a Waters Enzymate immobilized pepsin column and UPLC separation of the resulting peptic peptides . Mass spectra were acquired using a Waters Synapt HDMSE mass spectrometer . Peptides generated from online pepsin digestion were identified with Waters Protein Lynx Global Server 3 . 0 using separate unlabeled protein that was prepared in the same manner as protein labeled with deuterium . Deuterium incorporation was quantified using Waters DynamX 3 . 0 . Deuterium levels for each peptide were calculated by subtracting the average mass of the undeuterated control sample from that of the deuterium-labeled sample; the data were not corrected for back exchange and are therefore reported as relative . Vertical difference maps in Figures 5 , 7 and 8 do not represent a linear sequence of non-overlapping peptides . All coincident and overlapping peptides for comparisons in each figure are provided in figure identified tabs of Source data 1 . In vitro kinase assays were performed by incubating 1 μM BTK FL , BTK FL C481S or BTK FL T316A proteins in a kinase assay buffer ( 50 mM Hepes pH 7 . 0 , 10 mM MgCl2 , 1 mM DTT , 5% glycerol , 1 mM Pefabloc , and 200 μM ATP ) at room temperature for 10 min . The reactions were stopped by the addition of SDS-PAGE loading buffer , and the samples were boiled , separated by SDS−PAGE , and western blotted with the anti-BTK pY551 antibody ( BD Pharmingen ) or anti-His antibody ( EMD Millipore ) as described previously ( Joseph et al . , 2007 ) . The bands were quantified using the ChemiDoc ( Biorad ) gel imaging system . The phosphorylation levels ( Anti-BTK pY551 blot ) were normalized to the total protein level ( Anti-His blot ) . The BTK FL value was set to one and compared to BTK FL C481S or BTK FL T316A . Initial phosphorylation levels of BTK WT , C481S , and T316A on Y551 , prior to the start of the activity assay were undetectable by western immuno-detection . | Treatments for blood cancers , such as leukemia and lymphoma , rely heavily on chemotherapy , using drugs that target a vulnerable aspect of the cancer cells . B-cells , a type of white blood cell that produces antibodies , require a protein called Bruton’s tyrosine kinase , or BTK for short , to survive . The drug ibrutinib ( Imbruvica ) is used to treat B-cell cancers by blocking BTK . The BTK protein consists of several regions . One of them , known as the kinase domain , is responsible for its activity as an enzyme ( which allows it to modify other proteins by adding a ‘tag’ known as a phosphate group ) . The other regions of BTK , known as regulatory modules , control this activity . In BTK’s inactive form , the regulatory modules attach to the kinase domain , blocking the regulatory modules from interacting with other proteins . When BTK is activated , it changes its conformation so the regulatory regions detach and become available for interactions with other proteins , at the same time exposing the active kinase domain . Ibrutinib and other BTK drugs in development bind to the kinase domain to block its activity . However , it is not known how this binding affects the regulatory modules . Previous efforts to study how drugs bind to BTK have used a version of the protein that only had the kinase domain , instead of the full-length protein . Now , Joseph et al . have studied full-length BTK and how it binds to five different drugs . The results reveal that ibrutinib and another drug called dasatinib both indirectly disrupt the normal position of the regulatory domains pushing BTK toward a conformation that resembles the activated state . By contrast , the three other compounds studied do not affect the inactive structure . Joseph et al . also examined a mutation in BTK that confers resistance against ibrutinib . This mutation increases the activity of BTK by disrupting the inactive structure , leading to B cells surviving better . Understanding how drug resistance mechanisms can work will lead to better drug treatment strategies for cancer . BTK is also a target in other diseases such as allergies or asthma and even COVID-19 . If interactions between partner proteins and the regulatory domain are important in these diseases , then they may be better treated with drugs that maintain the regulatory modules in their inactive state . This research will help to design drugs that are better able to control BTK activity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2020 | Differential impact of BTK active site inhibitors on the conformational state of full-length BTK |
The auditory pathway faithfully encodes and relays auditory information to the brain with remarkable speed and precision . The inner hair cells ( IHCs ) are the primary sensory receptors adapted for rapid auditory signaling , but they are not thought to be intrinsically tuned to encode particular sound frequencies . Here I found that under experimental conditions mimicking those in vivo , mammalian IHCs are intrinsically specialized . Low-frequency gerbil IHCs ( ~0 . 3 kHz ) have significantly more depolarized resting membrane potentials , faster kinetics , and shorter membrane time constants than high-frequency cells ( ~30 kHz ) . The faster kinetics of low-frequency IHCs allow them to follow the phasic component of sound ( frequency-following ) , which is not required for high-frequency cells that are instead optimally configured to encode sustained , graded responses ( intensity-following ) . The intrinsic membrane filtering of IHCs ensures accurate encoding of the phasic or sustained components of the cell’s in vivo receptor potential , crucial for sound localization and ultimately survival .
Sound pressure waves are used by many animal species to gain sensory input from their environment that is vital for their survival and communication . The sensory receptors of the auditory system , the hair cells , are responsible for the transduction of sound waves into electrical signals that travel along the auditory pathway . The mechanosensory apparatus of hair cells is the stereociliary hair bundle that protrudes from their apical surface in the form of a staircase-like structure ( Schwander et al . , 2010 ) . Mechano-sensitive ion channels are positioned near the tip of the shorter rows of stereocilia ( Beurg et al . , 2009 ) , the opening of which initiates a receptor potential that is shaped by voltage and ion-gated ion channels in the hair cell’s basolateral membrane . The receptor potential activates the Ca2+ channels in the basal pole of the cell that trigger the release of neurotransmitter onto afferent terminals ( Glowatzki and Fuchs , 2002 ) , so that sound information is conveyed to the higher auditory brain areas . Auditory hair cells in lower vertebrates such as the turtle and bullfrog ( Fettiplace and Fuchs , 1999 ) , where the hearing sensitivity spans a generally low-frequency range ( from around 100 Hz to 1 kHz ) , show intrinsic membrane tuning such that the membrane potential oscillates at the cells’ characteristic frequency ( CF ) . In these animals , the interplay between the Ca2+ and Ca2+-activated K+ channels , which are differentially expressed along the auditory organ , are thought to be the main mechanism determining hair cell frequency selectivity ( Hudspeth and Lewis 1988; Wu et al . , 1995 ) . The mammalian auditory system has evolved to be able to encode a much wider dynamic range of sound frequencies and intensities . In order to do this , mammals have developed an elaborate auditory organ where much of the frequency tuning is carried out by the mechanics of the basilar membrane ( Robles and Ruggero , 2001 ) and electromotility of the outer hair cells ( OHCs ) ( Brownell et al . , 1985; Ashmore , 1987 ) driven by the motor protein prestin ( Liberman et al . , 2002; Dallos et al . , 2006 ) . These morphological and functional differences along the mammalian cochlea produce a tonotopic place-frequency map that is preserved throughout the auditory pathway . Thus the main sensory receptors , the inner hair cells ( IHCs ) , are believed not to show intrinsic tuning in mammals ( Marcotti et al . , 2003 ) , although recent studies have shown tonotopic differences in the Ca2+ currents ( Johnson and Marcotti , 2008 ) , the size of Ca2+ hotspots at ribbon synapses ( Meyer et al . , 2009 ) , and Ca2+ dependence of exocytosis ( Johnson et al . , 2008 ) along the cochlea . Moreover , the in vivo receptor potentials of low and high-frequency mammalian IHCs differ because of the relative filtering properties of the cells’ basolateral membrane . Low-frequency IHCs ( up to a few kilohertz ) have a predominantly phasic ( AC ) component that follows the stimulation frequency and is graded in size to the intensity ( Dallos , 1985 ) . The receptor potentials of high-frequency cells cannot follow the sound stimulus and as such do not have a large phasic component . Instead , they have a predominantly sustained ( DC ) component that is the average depolarization resulting when there is an asymmetric transduction current and it oscillates too fast for the cell membrane potential to keep up . The DC component is graded in size to sound intensity ( Russell and Sellick , 1978 ) . In the present study , I have investigated the biophysical properties of mature gerbil IHCs under near physiological recording conditions in low-frequency ( apical turn: ~300 Hz ) and high-frequency ( basal turn: ~30 kHz ) regions of the cochlea . The results show that the size of the resting mechanotransducer ( MT ) current and the composition of basolateral membrane currents are optimized to confer low- and high-frequency IHCs with different resting membrane potentials ( Vm ) and kinetics in vivo . These biophysical differences allow apical IHCs to follow precisely the phasic component of sound , while basal cells are able to accurately encode the sustained and graded component of high-frequency stimuli , providing direct evidence for the existence of intrinsic filtering in mammalian cochlear IHCs .
MT currents from apical ( ~300 Hz ) and basal ( ~30 kHz ) coil IHCs of the mature gerbil ( P17–P25 ) were elicited by displacing their hair bundles along their axis of mechanosensitivity with a 50 Hz sinusoidal stimulus using a piezoelectric fluid jet stimulator . The fluid jet produces a uniform deflection of the hair bundle ( Corns et al . , 2014 ) and for IHCs , which are stimulated in vivo by fluid motion within the cochlear partition ( Fettiplace and Kim , 2014 ) , it provides a near-physiological stimulation without affecting their resting position . To mimic in vivo conditions as closely as possible , recordings were obtained from IHCs maintained at body temperature and their hair bundles were perfused with an extracellular solution containing an endolymphatic low-Ca2+ concentration ( 40 µM: see Materials and methods ) . The same low-Ca2+ solution was also used inside the fluid jet . A large inward MT current was elicited in both apical ( 1 . 1 ± 0 . 1 nA , n = 5 , P20–P25: Figure 1A ) and basal coil IHCs ( 1 . 0 ± 0 . 2 nA , n = 6 , P17–P25: Figure 1B ) upon moving their bundle in the excitatory direction ( i . e . , towards the taller stereocilia using positive driver voltages ) . Apical cells had a significantly larger resting MT current ( 464 ± 29 pA: Figure 1A , P<0 . 0001 ) , which shut off during inhibitory phases of the stimulus , than that recorded from basal cells ( 70 ± 17 pA: Figure 1B ) . This difference was also observed with FM1-43 , a styryl dye that permeates the open transducer channels and accumulates inside the hair cells ( Gale et al . , 2001; Furness et al . , 2013 ) . Bath application of FM1-43 onto P18–P22 gerbil cochleae ( n = 6 ) labeled IHCs strongly in the apical coil but only weakly in the base ( Figure 1C and D , respectively ) . The opposite was observed in the OHCs , consistent with the fact they have larger resting MT currents in the basal region ( Johnson et al . , 2011 ) . 10 . 7554/eLife . 08177 . 003Figure 1 . The resting mechanotransducer ( MT ) current in apical and basal inner hair cells ( IHCs ) from adult gerbils . ( A and B ) Saturating MT currents recorded from an apical ( A ) and a basal IHC ( B ) by applying 50 Hz sinusoidal force stimuli to the hair bundles at the membrane potential of –84 mV . Recordings were carried out in the presence of a low-Ca2+ ( 40 µM ) extracellular solution . The driver voltage ( DV ) signal of up to ± 20 V to the fluid jet is shown above the traces ( positive deflections are excitatory ) . The dashed lines indicate the closure of the transducer channels and disappearance of the resting current during inhibitory bundle displacements . ( C and D ) FM1-43 fluorescence images of apical ( C ) and basal ( D ) gerbil cochlear sections ( P18 ) with the differential interference contrast ( DIC ) image superimposed . Apical IHCs , and to a much less extent outer hair cells ( OHCs ) , were labeled by the dye whereas the opposite pattern was seen in basal cells . Scale bar represents 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 003 The likely resting potential of mature gerbil IHCs in vivo was determined by performing whole-cell current clamp experiments with endolymphatic low-Ca2+ solution bathing the stereociliary hair bundle . Current clamp recordings were performed on mature apical and basal coil gerbil IHCs ( P17–P22 ) in response to current injection steps of up to 1 nA from their resting potential ( Figure 2A–C ) . In the presence of 1 . 3 mM extracellular Ca2+ , which is normally used when studying basolateral membrane currents , the resting potentials of apical ( –68 . 6 ± 1 . 3 mV , n = 12 ) and basal ( –71 . 7 ± 0 . 6 mV , n = 11 ) IHCs were very similar ( Figure 2A and D ) , and within the same range previously found in mature IHCs from mice ( Marcotti et al . , 2004 ) . The low ( 40 µM ) extracellular Ca2+ solution caused apical IHCs to depolarize ( –56 . 1 ± 1 . 5 mV , n = 12 ) significantly more than basal IHCs ( P<0 . 0001; –65 . 8 ± 1 . 3 mV , n = 11; Figure 2B and D ) . The voltage responses of apical IHCs showed an initial decline and compression for the larger depolarizing current steps , whereas basal were larger , more sustained , and graded to the stimulus , especially in low Ca2+ ( Figure 2B and E ) . In the presence of the MT channel blocker dihydrostreptomycin ( DHS ) ( 100 µM ) the voltage responses returned to control levels ( 40 µM Ca2+ + DHS: Figure 2C and D ) , indicating that the shift in the resting membrane potential in low-Ca2+ was due to the increased MT channel open probability . These findings demonstrate that the resting membrane potential of apical IHCs is likely to be substantially more depolarized than that of basal cells in vivo , which would speed up their membrane time constant . The time constant of the voltage rise , following 100 pA depolarizing current injections , were significantly more rapid in apical than in basal IHCs in both control 1 . 3 mM Ca2+ ( P<0 . 001; apical: 1 . 03 ± 0 . 11 ms , n = 12; basal: 1 . 89 ± 0 . 19 ms , n = 11 ) and low-Ca2+ conditions ( P<0 . 0005; apical: 0 . 44 ± 0 . 08 ms , n = 12; basal: 1 . 07 ± 0 . 13 ms , n = 11 ) ( Figure 2F ) . The different voltage responses between apical and basal IHCs indicates the existence of cellular specializations in the underlying basolateral membrane currents . 10 . 7554/eLife . 08177 . 004Figure 2 . The resting membrane potential in apical and basal inner hair cells ( IHCs ) from adult gerbils . ( A–C ) Current clamp responses of apical ( top panels ) and basal ( lower panels ) IHCs from adult gerbils in response to 100 pA current steps from -100 pA to 900 pA from the cells’ resting potential ( Vm ) . From left to right are responses in 1 . 3 mM extracellular Ca2+ ( A ) , low-Ca2+ ( 40 µM ) endolymph-like solution ( B ) , and low-Ca2+ solution with 100 µM dihydrostreptomycin ( DHS ) ( C ) . Traces are averages from all IHCs ( apical: n = 12; basal: n = 11 ) . ( D ) Average Vm values measured before the current steps in the different extracellular conditions as in ( A–C ) , including the washout with 1 . 3 mM Ca2+ , for apical ( red ) and basal ( blue ) IHCs . ( E ) Average peak ( open symbols ) and steady-state ( closed symbols ) Vm from apical and basal IHCs measured at different current injection levels from the current clamp recordings in low-Ca2+ in ( B ) . The arrowheads/arrows in ( B ) indicate where the peak and steady state Vm were measured . Note that the voltage responses of basal cells to a sustained current step , similar to a high-frequency tone , are more clearly graded to the stimulus amplitude than those in apical cells in terms of the dynamic range of voltages covered . ( F ) Average onset of the initial Vm response to 100 pA current injection in 1 . 3 mM Ca2+ ( left ) and low-Ca2+ ( right ) for apical IHCs ( top panels , n = 12 ) and basal IHCs ( bottom panels , n = 11 ) . The initial rise to peak was fit with a single exponential function . The average time constant ( τ ) , obtained from fitting the individual cells , was ( apical IHCs ) 1 . 03 ± 0 . 11 ms in 1 . 3 mM Ca2+ and 0 . 44 ± 0 . 08 ms in low-Ca2+; ( basal IHCs ) 1 . 89 ± 0 . 19 ms ( 1 . 3 mM Ca2+ ) and 1 . 07 ± 0 . 13 ( low-Ca2+ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 004 Basolateral membrane currents were recorded from gerbil IHCs at body temperature in response to 10 mV voltage steps from the holding potential of –64 mV ( Figure 3A and B ) . Both apical ( n = 26 ) and basal IHCs ( n = 27 ) showed the currents characteristic of mature IHCs ( Marcotti et al . , 2004; Kros et al . , 1998 ) , namely , the rapidly activating large conductance ( BK ) , Ca2+-activated K+ current ( IK , f ) , and the negatively activating delayed rectifier current ( IK , n: carried by KCNQ4 channels; potassium channel , voltage gated KQT-like subfamily Q , member 4 ) . The overall size of the peak and steady-state inward and outward currents appeared comparable between apical and basal IHCs ( Figure 3C ) . The peak current measured at 2 ms gives a reliable and accurate measurement of the size of IK , f in isolation due to its very rapid activation kinetics ( Marcotti et al . , 2004 ) , which was similar between apical and basal gerbil IHCs . The overall slope conductance , measured at around the likely in vivo membrane potential ( Ac: –54 mV; Bc: –64 mV: Figure 2D ) from the steady-state current-voltage ( I-V ) curves , was found to be significantly larger in apical ( 56 . 3 ± 3 . 0 nS , n = 26 , P<0 . 0001 ) than in basal IHCs ( 10 . 0 ± 0 . 4 nS , n = 27 ) . Profound differences were observed in the current activation time-course and size of the tail currents between apical and basal IHCs ( apical: Figure 3D , F , and H; basal: Figure 3E , G , and H ) . All IHCs investigated showed a very negative current activation ( Figure 3H , inset ) . Pharmacological analysis revealed a differential expression of K+ currents between apical and basal IHCs ( Figure 4A–E ) . The main difference between the cells was the outward delayed-rectifier K+ current that was mainly a linopirdine-insensitive IK , s in basal IHCs ( Figure 3D ) while in apical cells it was predominantly a linopirdine-sensitive current ( Figure 3E ) reminiscent of the outward IK , n found in adult OHCs ( Marcotti and Kros , 1999 ) . The deactivating tail currents of the linopirdine-sensitive current ( Figure 4F ) indicate that there is more than one linopirdine-sensitive current component . The negatively activating component is larger in basal cells ( Figure 4G ) and is similar in size and voltage-dependence to the IK , n previously shown in apical-coil mouse IHCs that are also comparatively high-frequency receptors ( Marcotti et al . , 2003; Oliver et al . , 2003 ) . The tonotopic difference in this negatively activating IK , n is consistent with the expression gradient of KCNQ4 channels observed in the rat ( Beisel et al . , 2005 ) . The more positively activating linopirdine-sensitive component was much larger in apical IHCs and could result from the expression of a different KCNQ channel subtype . 10 . 7554/eLife . 08177 . 005Figure 3 . Differences in the basolateral membrane currents of apical and basal adult gerbil inner hair cells ( IHCs ) . ( A and B ) Basolateral membrane currents elicited from apical ( A , red ) and basal ( B , blue ) IHCs in response to depolarizing voltage steps in nominal 10 mV increments from –144 mV ( holding potential: –64 mV ) . Tail currents were measured upon returning to –124 mV ( protocol shown above the current traces ) . Traces are averages from 26 apical and 27 basal IHCs . ( C ) Average current-voltage ( I-V ) curves for apical and basal IHCs . Peak currents were measured at 2 ms from the onset of the test step and steady-state values were obtained at around 160 ms . ( D and E ) Average current traces as in ( A ) and ( B ) , but on an expanded y-scale to show more clearly the time course of the currents at more negative membrane potentials . ( F and G ) Average tail currents from apical and basal IHCs , respectively , from the different test potentials shown by some of the traces ( amplified from [A] and [B] ) . ( H ) Average activation curves for apical and basal IHCs made by plotting the instantaneous tail current amplitude against the preceding voltage step . The inset shows the activation of the K+ currents on an expanded scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 00510 . 7554/eLife . 08177 . 006Figure 4 . Differences in the K+ current components in apical and basal gerbil inner hair cells ( IHCs ) . ( A–E ) average basolateral currents elicited as in Figure 3 from apical ( red ) and basal ( blue ) IHCs in the presence of different K+ channel blockers to isolate the underlying current components . The control traces ( A ) are the same as those in Figure 3A and B . ( B ) The perfusion of linopirdine ( 80 µM ) blocked the inward IK , n in both apical and basal IHCs leaving IK , s and IK , f ( apical n = 5 , basal n = 7 ) . Linopirdine also removed the outward delayed component in apical cells but not in basal cells . ( C ) The perfusion of iberiotoxin ( IbTx; 60 nM ) was used to block the rapidly activating IK , f leaving a current composed of IK , s and IK , n ( apical n = 4 , basal n = 4 ) . ( D ) The combination of linopirdine and iberiotoxin ( Linop + IbTx ) was used to isolate IK , s ( apical n = 4 , basal n = 4 ) , which was small in apical cells but of a similar size to that in C for basal cells . ( E ) The isolated IK , n ( Linop sensitive ) was obtained by subtracting the current in the presence of both linopirdine and IbTx ( D ) from the current obtained in the presence of IbTx ( C ) . ( F ) Average tail currents at –124 mV from the linopirdine sensitive current in E for apical and basal IHCs . Only the most negative traces are shown to emphasize the negatively activating component of the linopirdine-sensitive current . ( G ) Average activation curves for the negatively activating component of the linopirdine-sensitive current from the instantaneous tail current amplitudes in F for apical and basal cells . The inset shows the activation curves up to more positive voltages as in Figure 3H . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 006 Possible differences in the endogenous Ca2+ buffering in IHCs along the cochlea ( gerbil: Pack and Slepecky , 1995 ) could affect Ca2+ loading and lead to differences in the basolateral membrane current properties or MT channel resting open probability . However , recent electrophysiological observations have indicated that the buffering strength in gerbil IHCs along the cochlea is likely to be similar ( Johnson et al . , 2008 ) . In order to verify this observation , I performed recordings from mature gerbil IHCs ( P29–P36 ) under the perforated patch configuration , a condition that prevents the dialysis of the intracellular Ca2+ binding proteins with the Ca2+ buffer in the pipette solution and therefore preserves any tonotopic gradient that could be present . The basolateral currents recorded using perforated patch ( Figure 5A–C ) were similar to those obtained with whole cell recording ( Figure 3A–C ) . The overall current size was similar between apical and basal IHCs ( Figure 5C ) , but the tail currents were much larger in apical cells ( Figure 5A and B ) as shown in Figure 3 . The size of the resting transducer current , which contributes to the cells holding current in voltage clamp , was measured by perfusing IHCs with low-Ca2+ extracellular solution ( Figure 5D and E ) . The addition of DHS to the low-Ca2+ solution largely prevented the increase in holding current magnitude , indicating that the transducer channel component was absent . The change in holding current in perforated patch ( apical IHCs –764 ± 28 pA , n = 3; basal IHCs –99 ± 21 pA , n = 4 ) was similar to that obtained with whole cell recordings ( apical IHCs –549 ± 81 pA , n = 6; basal IHCs –69 ± 22 pA , n = 4 ) and also similar to the resting transducer current values obtained from direct MT current measurements in low-Ca2+ ( Figure 1A and B ) . 10 . 7554/eLife . 08177 . 007Figure 5 . Endogenous Ca2+ buffering does not affect differences in basolateral membrane currents or the resting mechanotransducer ( MT ) current in apical and basal adult gerbil inner hair cells ( IHCs ) . ( A and B ) Basolateral membrane currents elicited from apical ( A , red ) and basal ( B , blue ) IHCs as in Figure 3A and B . ( C ) Average peak and steady-state current-voltage ( I-V ) curves for apical ( n = 3 ) and basal ( n = 4 ) IHCs . ( D and E ) Membrane currents covering a physiological range of IHC membrane potentials elicited from the holding potential of around -70 mV . The holding current is indicated in the first panel as IH . Recordings were made in standard 1 . 3 mM extracellular Ca2+ ( left panels ) , 40 µM Ca2+ ( middle panels ) , and 40 µM Ca2+ with 100 µM DHS ( right panels ) . Low Ca2+ ( 40 µM ) elicited a large inward holding current only in apical IHCs ( asterisk ) that was blocked by DHS . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 007 These findings show that the biophysics of low-frequency and high-frequency IHCs , including resting MT current size , K+ channel composition , and voltage responses , differ between low-frequency and high-frequency IHCs , with apical IHCs having a more depolarized resting Vm and faster membrane time constant in vivo . The functional implication of the biophysical differences in the resting MT current and voltage responses between low-frequency and high-frequency IHCs ( Figures 1–5 ) was investigated by using a sound-like sinusoidal current stimulus applied to IHCs maintained at their predicted in vivo resting potential ( Figure 6 ) . The sinewave stimulus was superimposed on a sustained holding current , equivalent in value to the resting MT current size measured in endolymphatic low-Ca2+ ( apical: 500 pA; basal: 50 pA; Figure 1 ) . 10 . 7554/eLife . 08177 . 008Figure 6 . Sound-like stimulation of apical and basal adult gerbil inner hair cells ( IHCs ) . ( A and B ) Current clamp protocols used to mimic the in vivo sound-induced stimulation of apical ( A ) and basal ( B ) IHCs . A 300 Hz sinewave of 1 nA peak-to-peak amplitude was superimposed on a step current injection ( apical: 500 pA; basal: 50 pA ) , which was used to depolarize lHCs to their predicted in vivo resting potential ( see Results ) . Expanded versions of the sinewave stimulus ( top ) are shown below . ( C and D ) Average voltage responses from apical ( n = 7 ) and basal ( n = 6 ) IHCs in response to the whole stimulus shown in ( A ) and ( B ) ( top ) . ( E and F ) Average voltage responses of apical ( E ) and basal ( F ) IHCs to a 300 Hz ( top ) , 2 kHz ( middle ) , and a 10 kHz ( bottom ) 1 nA sinewave . Only the sinewave portion of the responses is shown and the holding current levels were as in ( A ) and ( B ) , respectively . An expanded portion of the 2 kHz and 10 kHz voltage responses is shown below the main trace in order to show some cycles of the sine wave . For the basal response at 2 kHz the predominantly phasic ( AC ) and predominantly sustained ( DC ) components are indicated . Apical cells showed only the AC component for 1 nA stimuli . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 008 The sinewave , 1 nA in amplitude and of varying frequency , was used to mimic the transducer current during the sound induced deflection of the hair bundle that would open and close the MT channels from their resting state ( Figure 6A and B ) . The averaged voltage responses of apical ( n = 7 ) and basal ( n = 6 ) IHCs to the full protocol , with a 300 Hz sinewave , are shown in Figure 6C and D , respectively . Apical cells responded to the different frequencies used ( 300 Hz , 2 kHz , and 10 kHz sinewave ) with a phasic ( AC ) component around the resting potential that became progressively smaller with increasing frequency ( Figure 6E ) . In basal IHCs the phasic responses to all three frequencies ( Figure 6F ) were superimposed on a sustained shift in membrane potential ( DC component ) . Note that the voltage recordings obtained using sinewave frequencies of 300 Hz ( Figure 6E , top ) and 10 kHz ( Figure 6F , bottom ) are likely to be the closest approximation to in vivo responses for apical and basal IHCs , respectively . In addition to frequency , IHC responses are also determined by differences in sound intensity , which was mimicked by varying the amplitude of the sinewave ( Figure 7 ) . Both apical and basal IHCs responded with very small changes in membrane potential when small amplitude sinewaves were applied from the estimated in vivo resting Vm ( 100 pA: Figure 7A and C ) . However , the more depolarized Vm of apical IHCs , and the consequently larger resting membrane conductance , made positively going responses to stimulation generally smaller than in basal cells ( Figure 7A and B ) . The hyperpolarizing phase of apical IHCs was more pronounced than in basal cells ( Figure 7B ) . For larger stimuli , at frequencies approaching ( 2 kHz ) or exceeding ( 10 kHz ) the maximum limit for phase locking in vivo ( Palmer and Russell , 1986 ) , the first phase of the responses in apical IHCs reached its maximum value immediately after the onset of stimulation . By contrast , the responses of basal cells built up to reach a peak after a few cycles ( Figure 7D ) because they are superimposed on a rising DC response that grows with timing set by the membrane time constant . While apical IHCs appear to be suited for phasic signaling , basal cells are more adept at signaling graded responses to high-frequency stimuli ( Figure 7E and F ) since their receptor potential builds up to represent stimulus intensity . 10 . 7554/eLife . 08177 . 009Figure 7 . Voltage responses of apical and basal inner hair cells ( IHCs ) to sound-like stimulation of different intensity . ( A ) Voltage responses of apical and basal IHCs to 100 pA , 300 Hz sinewave stimulation . Dashed vertical lines are spaced at 3 . 33 ms to represent one cycle at 300 Hz with the first being aligned to the initial peak of the apical response . Basal responses show a delay compared to the peaks in apical cells . ( B ) Voltage responses as in ( A ) but using a 1 nA , 300 Hz sinewave ( only the initial four cycles are shown ) . Dashed vertical lines are aligned as in ( A ) . Note the leftward leaning of the basal upper peaks following the initial one . ( C ) Responses to 100 pA , 2 kHz stimulation showing the initial few milliseconds . The dashed vertical lines are as in ( A ) but spaced to show cycles at 2 kHz . The responses of apical and basal IHCs are reduced in magnitude compared to those at 300 Hz ( A ) , but the delay in the basal peaks was maintained . ( D ) Responses to 1 nA , 2 kHz stimulation with vertical lines as in ( C ) . Note that the depolarization builds up in basal responses and that the peaks are delayed compared to those for apical cells . ( E and F ) The initial few ms of 10 kHz responses to three different amplitude stimuli , shown next to the traces for apical ( E , n = 6 ) and basal cells ( F , n = 6 ) . While both apical and basal IHCs have a graded phasic ( AC ) response to the stimulus , basal cells show a pronounced graded ( DC ) response to the stimulus intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 009 The timing and accuracy of phasic responses in apical and basal IHCs were compared by measuring the time difference of the peaks and troughs of the first few cycles to sound-like stimulation ( Figure 8A ) . The time differences are also shown as phase angle shifts , calculated based on the size of the time difference and the stimulus frequency . At 300 Hz the timing was generally more advanced in apical IHCs for the lower stimulus levels ( 100 pA: closed symbols in Figure 8B; see also Figure 7A ) . The same also occurred at 1 nA for the first peak , after which positive peaks in basal IHCs occurred sooner ( Figure 7B ) , but this made the half-cycle width much less accurate in basal cells ( Figure 8C ) . At 300 Hz the half-cycle error was generally lower at each stimulus level in apical IHCs . At 2 kHz , where the timing and accuracy of the phasic component would be vital in vivo , the peak responses of apical IHCs again occurred generally more rapidly than in basal cells ( Figure 8D ) . In contrast to 300 Hz the timing difference at 2 kHz declined with progressive peaks to a stable level , but remained with an apical advance . This stable apical lead seems to be due to the underlying basolateral current differences between the two cell types ( Figures 3–5 ) , because when apical and basal cells were stimulated with the same apical-like 2 kHz 1 nA protocol a similar stable apical lead was evident ( Figure 9A–C ) . Additionally , when the outward linopirdine-sensitive K+ current , present only in apical cells ( Figure 4E ) , was blocked , the voltage responses became more basal-like ( Figure 9D–F ) . The 2 kHz responses of apical cells were very accurate at each stimulus level ( Figure 8E ) , whereas basal cells began with a large half-cycle error that improved progressively , as seen for the timing differences . The decline in basal cell half-cycle error is likely to reflect the decrease in cell membrane time constant , to a fairly steady value , as depolarization activates the K+ currents . Therefore , the intrinsic characteristics of apical IHCs seem to be best suited for the accurate representation of the phasic component of the receptor potential . 10 . 7554/eLife . 08177 . 010Figure 8 . The timing of phasic responses to sound-like stimulation is faster and more accurate in apical inner hair cells ( IHCs ) . ( A ) IHC responses to a 1 nA , 2 kHz stimulus showing the relative timing shift between apical and basal cells ( above ) to the sinewave stimuli ( below; apical black and basal grey ) . Dashed vertical lines delineate each half cycle of the 2 kHz stimulus with the half-cycle number shown . The numbers on the dashed vertical lines indicate the maxima and minima peak number used for subsequent analysis . Note that the responses of apical and basal IHCs are both delayed compared to the stimulus but the corresponding peaks and half cycles can be seen . The difference in timing between apical and basal traces is shown and was measured as the time delay between the occurrences of the equivalent peaks . The half widths were measured as the time between maximal and minimal peaks for each response half cycle . ( B ) The average timing difference between the arrival of the different peaks between apical and basal responses for 300 Hz stimuli ( the symbols represent two stimulus amplitudes ) . Data points were obtained by subtracting the average apical value from that of basal cells . Therefore positive values ( light red area ) indicate an apical lead whereas negative values ( light blue area ) indicate a basal lead . The right-hand scale shows the corresponding shift in phase angle for the equivalent timing differences , calculated based on the size of the time difference and the stimulus frequency . ( C ) Average half-cycle width errors for apical and basal responses to 300 Hz stimuli of different amplitudes . Errors were calculated for each cell as the width of each half cycle in milliseconds divided by the actual half-cycle width of the stimulus ( 1 . 67 ms for 300 Hz ) and converted to percentage . ( D and E ) Timing differences and half width errors , respectively , calculated as in ( B ) and ( C ) but for 2 kHz stimuli ( half-cycle width of 0 . 25 ms ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 01010 . 7554/eLife . 08177 . 011Figure 9 . Response timing is determined by the intrinsic membrane properties of apical and basal inner hair cells ( IHCs ) . ( A ) Average voltage responses of apical ( n = 7 ) and basal ( n = 6 ) IHCs to a 2 kHz 1 nA current sinewave using a stimulus with a holding current of 500 pA for both apical and basal IHCs . The dashed vertical lines are spaced to the 2 kHz half cycle and aligned to the initial apical peak . Even though both IHC types were stimulated with the same protocol , basal cells maintained a delayed response compared to apical cells . ( B ) The average timing difference between the arrival of the different peaks between apical and basal voltage responses shown in ( A ) . Data points were obtained by subtracting the average apical value from that of basal cells ( as in Figure 8B ) . All positive values indicate an apical lead . ( C ) Average half width errors , calculated as in Figure 8C , for the apical and basal voltage responses in A . Blocking IK , n in apical IHCs makes their response timing and accuracy more basal-like . ( D ) Average voltage responses of apical IHCs in the presence of 80 µM linopirdine ( Ac+Linop , n = 6 ) to a 1 nA , 300 Hz current sinewave using a holding current of 500 pA for apical cells . Voltage responses in apical ( Ac ) and basal ( Bc ) IHCs are as in Figure 7B ( holding current of 500 pA for apical and 50 pA for basal cells ) . The dashed vertical lines are spaced to the 300 Hz half cycle and aligned to the initial apical peak ( red trace ) . Note that when IK , n in apical cells was blocked by linopirdine their responses closely resembled those from basal cells with a similar leftward lean in the upper peaks following the initial one . Between –65 mV and –20 mV the IK , n represents a large component only in apical cells ( Figure 3D and E and Figure 4E ) . ( E ) The average timing difference between the arrival of the different peaks between the voltage responses of apical and either basal or apical in linopirdine , shown in ( D ) . ( F ) Average half width errors for the apical , apical in linopirdine and basal voltage responses in ( D ) . The error values for apical cells in linopirdine were shifted towards those from basal values . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 011 The maximal size of the MT current has been predicted to reach up to around 6 nA in vivo ( Kennedy et al . , 2003 ) due to the presence of a large endocochlear potential that drives ions through the MT channels , which is difficult to achieve in vitro . Therefore , a 6 nA sinewave stimulus ( 300 Hz ) was applied to apical and basal IHCs ( Figure 10A and B ) , in which the holding current was also increased to reflect the larger standing in vivo MT current . The responses of apical IHCs showed a slight depolarizing DC shift reaching a maximum of –25 . 7 ± 1 . 3 mV . Basal cells responded with larger changes in membrane potential , with very little or no hyperpolarization from rest and depolarizing up to –13 . 9 ± 1 . 6 mV , which corresponds to the peak of the calcium current ( Johnson and Marcotti , 2008 ) . The DC shift is likely to be slightly underestimated because a maximal MT current would show a greater saturation than that provided by the sinewave . 10 . 7554/eLife . 08177 . 012Figure 10 . Maximal sound-like stimulation of apical and basal gerbil inner hair cells ( IHCs ) and predicted synaptic vesicle release probability suit phase-locking in apical cells . ( A and B ) Current clamp protocols ( top panels ) applied to apical ( A ) and basal ( B ) IHCs as in Figure 6 but the holding currents and sinewave amplitudes have been increased to account for the contribution of the endocochlear potential ( see Results ) . Average voltage responses from apical ( n = 4 ) and basal ( n = 3 ) IHCs are shown below with the sinewave region amplified in the lower panels . ( C and D ) Normalized average △Cm values plotted against membrane potential for apical ( A ) and basal ( B ) gerbil IHCs obtained from the synaptic transfer functions shown in the insets ( from Johnson et al . , 2008 ) . The normalized △Cm approximates to a vesicle release probability ( PRelease ) and data points were fit with a Boltzmann curve with values of: apical IHCs Vhalf –25 . 8 mV and S 5 . 5 mV-1; basal IHCs Vhalf –31 . 4 mV and S 7 . 7 mV-1 . ( E ) The average IHC voltage responses to maximal sound-like stimulation in ( A ) and ( B ) were transformed into PRelease using the functions shown in ( C ) and ( D ) for apical and basal cells , respectively . ( F ) Same traces as in ( E ) but on an expanded time scale to emphasize the cycle-by-cycle differences in PRelease between apical and basal cells , showing pronounced ‘on’ and ‘off’ responses to positive and negative phases of the sinusoidal currents , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 08177 . 012 In order to provide an understanding on how these maximal voltage responses would drive neurotransmitter release at IHC synapses , Vm values were transformed into a probability of release ( PRelease ) using synaptic transfer functions previously obtained from adult apical and basal gerbil IHCs ( Johnson et al . , 2008 ) . These synaptic transfer functions represent a relatively steady-state relation of IHC exocytosis ( obtained using 100 ms voltage steps ) . I have used them to infer an overall PRelease from all IHC synapses in this more dynamic situation based on the fact that IHC exocytosis is very fast ( Beutner et al . , 2001 ) and the release sites are within a nanodomain of the Ca2+ channels ( Johnson et al . , 2008 ) . However , there could be subtle differences in exocytosis kinetics for apical and basal IHCs especially amongst individual synapses that are likely to have different thresholds and operating ranges ( Ohlemiller et al . , 1991 ) . The high-order and linear exocytotic Ca2+ dependence relations of apical ( Figure 10C , inset ) and basal cells ( Figure 10D , inset ) were used to express the normalized △Cm values as a function of cell Vm ( Figure 10C and D ) and to obtain an approximated release probability relation . The fits to the data were used to transform the Vm traces in Figure 10A and B into PRelease ( Figure 10E and F ) . PRelease reached higher values in basal IHCs ( 0 . 98 ± 0 . 02 , n = 3 ) than in apical cells ( 0 . 56 ± 0 . 07 , n = 4 ) . At the cell resting potential , PRelease was not zero in either apical or basal IHCs ( apical: 0 . 021 ± 0 . 002 , n = 4; basal: 0 . 064 ± 0 . 022 , n = 3; not significantly different ) , consistent with measured spontaneous activity in afferent fibers ( Ohlemiller et al . , 1991; Ohlemiller and Siegel , 1994 ) . In apical IHCs PRelease went below the resting level to reach almost zero during pronounced ‘off’ periods ( responses to negative phases of the sinusoidal currents ) that were longer lasting than those in basal cells where PRelease did not reduce significantly below resting levels ( apical minimal PRelease: 0 . 003 ± 0 . 001 , n = 4; basal minimal PRelease: 0 . 038 ± 0 . 006 , n = 3; P<0 . 001 ) . The emphasized ‘off’ periods in apical IHCs ( Figure 10F ) would provide greater contrast for accurate phase-locking .
The sound induced receptor potentials of low-frequency ( below a few kilohertz ) and high-frequency IHCs are very different . While low-frequency cells have a predominant sound frequency-following , or phase-locked , AC component ( Dallos , 1985 , 1986; Cheatham and Dallos , 1993 ) , the main component of high-frequency cells is a graded and sustained DC shift in potential in proportion to sound intensity ( Russell and Sellick , 1978 ) . Mammals use two main strategies to localize sounds of different frequency . Low frequencies are localized by principal cells in the medial superior olive ( MSO ) that compare the inter-aural timing differences ( ITDs ) of phase-locked activity in auditory afferents from the two ears ( Chirila et al . , 2007 ) . High-frequency sounds are localized by principal cells in the lateral superior olive ( LSO ) that compares inter-aural level differences ( ILDs ) arising from the graded responses of IHCs of each ear ( Caird and Klinke , 1983 ) . Therefore , the biophysical properties of low-frequency and high-frequency IHCs should be best suited to maximize the accuracy of response-timing in apical cells and response-level in basal cells to allow the detection of ITDs down to 10 µs and ILDs of only 1–2 dB ( Grothe et al . , 2010 ) . In the present study I found that low-frequency IHCs have a significantly more depolarized resting Vm than basal cells arising from the large resting MT current , which increases their resting conductance and speeds up their voltage responses . In response to simulated sound stimuli , apical IHCs were faster and preserved the phase width of the stimulus better than basal cells up to around the maximum frequency for phase-locking . However , the superior speed and timing of apical cells is not just down to their more depolarized Vm since basal cells remained slower even when depolarized at similar values to those of apical cells ( Figure 9A–C ) . I found that the differential expression of the K+ current IK , n in apical and basal IHCs was a major determinant for response speed and timing at low frequencies ( Figure 9D–F ) . The depolarized resting Vm of apical cells ( about –55 mV ) caused their ‘off’ response to sound-like stimulation to be larger than their ‘on’ component ( responses to positive phases of the sinusoidal currents ) for moderate stimuli ( Figure 5B ) . This is comparable to the visual system where photons of light cause a hyperpolarization of the photoreceptor membrane ( Baylor et al . , 1979 ) and a reduction in the release of glutamate . This ‘backwards’ mechanism introduces much less noise into the signal and enhances contrast , which would be favorable for phase-locking to auditory stimuli ( Palmer and Russell , 1986 ) . The high-order Ca2+ dependence of synaptic vesicle release in apical gerbil IHCs ( Johnson et al . , 2008 ) enhances these ‘off’ periods when Vm is transformed into PRelease ( Figure 10 ) , which would further increase the signal contrast as it is relayed onto afferent fibers and push the boundary for phase-locking up to a higher frequency . The smaller overall PRelease of apical IHCs to the maximal predicted stimulation would be compensated by multivesicular ( Glowatzki and Fuchs , 2002 ) or uniquantal ( Chapochnikov et al . , 2014 ) release at individual ribbon synapses , generating large excitatory postsynaptic potentials ( EPSPs ) in the afferent terminals , the size of which has been shown to be Ca2+ independent and larger ones show more accurate phase-locking ( Glowatzki and Fuchs , 2002; Grant and Glowatzki , 2010; Li et al . , 2014 ) . Recent findings have shown that a more depolarized membrane potential in hair cells leads to the facilitation of glutamate release ( Cho et al . , 2011; Goutman and Glowatzki , 2011 ) , which would have important consequences for IHC neurotransmitter release and short-term plasticity , and spike adaptation during a sound stimulus . However , a generally smaller exocytotic response , as calculated for the maximal PRelease in apical IHCs ( Figure 10E and F ) , would again be favorable for maximizing response speed , at the expense of accurately signaling intensity . This could explain why sound intensity discrimination is better towards the higher frequencies ( 1–6 kHz; Fletcher and Munson , 1933 ) . While apical cells are adapted for response speed , basal cells are specialized for intensity coding . Their more hyperpolarized Vm ( about –65 mV ) , due to the much smaller resting MT current , and consequently greater membrane time constant , means that the responses of basal cells to even low level sounds summate into a DC membrane potential shift that is proportional to stimulus intensity . By contrast , apical IHCs only showed signs of a DC component for large stimuli ( Figure 10A ) . The lower resting conductance of basal cells also means that their response amplitudes are bigger , although slower , than apical cells and are more spread out giving better discrimination between stimulus levels ( Figure 2E ) . The overall linear Ca2+ dependence of neurotransmitter release in adult gerbil basal IHCs ( Johnson et al . , 2008 ) would extend the dynamic range , enabling cells to accurately relay information from both low and high intensity sounds into afferent activity . The differences in mature gerbil IHCs seen here are opposite to what we previously observed in mature OHCs ( Johnson et al . , 2011 ) . In OHCs it is the high-frequency basal cells that have the fastest membrane properties and largest currents . This is important because they follow sound up to the highest frequencies in order to drive the prestin motors on a cycle-by-cycle basis for cochlear amplification ( Dallos et al . , 2006; Fettiplace and Hackney , 2006; Ashmore , 2008 ) . In order for IHCs to be fast enough to encode the phasic component of high-frequency sounds , the size of the receptor potential would be so small that the graded component would be lost and the response would be too fast to be encoded by the release of synaptic vesicles . Classical in vivo sharp electrode recordings of IHC receptor potential responses to sound reveal similar distinctions between apical and basal IHCs ( Palmer and Russell , 1986; Cheatham and Dallos , 1993 ) . Low-frequency guinea pig IHCs respond to sound with a predominant AC component and a large hyperpolarization from rest ( Dallos , 1986 ) . By contrast , the responses in high-frequency cells are dominated by the DC component ( Russell and Sellick , 1978 ) , although when stimulated with low-frequency sound they also show an AC component but with little hyperpolarization from rest ( Palmer and Russell , 1986 ) , indicative of a smaller resting MT current in basal cells . In vivo studies on IHCs and afferent fibers have independently suggested that there are intrinsic differences between apical and basal regions , which generate a resting bias for apical cells towards the ‘on’ or depolarized condition and basal cells to the ‘off’ hyperpolarized condition ( Palmer and Russell , 1986; Ohlemiller and Siegel , 1994; Cheatham and Dallos , 1993 ) . A different resting bias would result from differences in the standing current through the IHCs ( i . e . , resting MT current ) , such that the larger current in apical cells would reduce the DC component to favor AC responses , whereas the smaller current in basal cells would enhance the DC Vm shifts . A different resting bias would also explain why afferent fibers in the gerbil generally have a higher spontaneous firing rate ( SR ) at the apex than those at base ( Ohlemiller et al . , 1991; Ohlemiller and Siegel , 1994 ) . Therefore , both low-frequency and high-frequency IHCs are specialized for their respective tasks , showing a preference for either timing or intensity coding , respectively . While low-frequency cells are able to respond rapidly and accurately to phasic stimuli at around the phase-locking limit , high-frequency cells show clearly defined graded shifts in membrane potential over an extended dynamic range . Evidence in support of a specialization of apical cells for phase-locking comes from in vivo recordings from chinchilla auditory nerve fibers showing that the strength of phase-locking in high-CF fibers starts to deteriorate at lower stimulus frequencies than that in low-CF fibers ( Temchin and Ruggero , 2010 ) . The findings presented here show that the primary sensory IHCs play a more active role in auditory information processing than previously thought .
Gerbil cochlear IHCs ( n = 115 ) were studied in acutely dissected organs of Corti from postnatal day 17 ( P17 ) to P40 , where the day of birth is P0 . Animals of either sex were killed by cervical dislocation , under schedule 1 in accordance with UK Home Office regulations . IHCs were positioned at a frequency range of 250–420 Hz in apical and 20–37 kHz in basal cells ( Müller , 1996 ) . Cochleae were dissected as previously described ( Johnson et al . , 2008 , 2012 ) in normal extracellular solution ( in mM ) : 135 NaCl , 5 . 8 KCl , 1 . 3 CaCl2 , 0 . 9 MgCl2 , 0 . 7 NaH2PO4 , 5 . 6 D-glucose , 10 Hepes-NaOH . Sodium pyruvate ( 2 mM ) , MEM amino acids solution ( 50X , without L-Glutamine ) , and MEM vitamins solution ( 100X ) were added from concentrates ( Fisher Scientific , UK ) . The pH was adjusted to 7 . 5 ( osmolality ~308 mmol/kg ) . The dissected organs of Corti were transferred to a microscope chamber , immobilized using a nylon mesh fixed to a stainless steel ring , and continuously perfused with the above extracellular solution . The organs of Corti were observed with an upright microscope ( Nikon , Japan ) with Nomarski differential interference contrast optics ( X60 water immersion objective and X15 eyepieces ) . Whole-cell patch clamp recordings were performed at body temperature ( 34–37ºC ) using an Optopatch ( Cairn Research Ltd , UK ) amplifier . Patch pipettes ( 2–3 MΩ ) were coated with surf wax ( Mr . Zogs SexWax , USA ) to minimize the fast capacitance transient of the patch pipette . The pipette intracellular solution contained ( in mM ) : 131 KCl , 3 MgCl2 , 1 EGTA-KOH , 5 Na2ATP , 5 Hepes-CsOH , 10 Na2-phosphocreatine ( pH 7 . 3; osmolality ~296 mmol/kg ) . The perforated-patch technique ( Rae et al . , 1991 ) was used on a few mature gerbil IHCs ( n = 7 ) in order to see whether endogenous Ca2+ buffering affected the resting MT current and the other basolateral membrane currents . For these experiments the pipette filling solution contained ( mM ) : 21 KCl , 110 potassium aspartate , 3 MgCl2 , 5 Na2ATP , 1 EGTA–KOH , 5 Hepes–KOH , 10 sodium phosphocreatine ( pH 7 . 3 , 295 mmol/kg ) . The antibiotic amphotericin B ( Calbiochem , UK ) was dissolved in dry dimethylsulfoxide ( DMSO ) prior to its dilution into the above intracellular solution to a final concentration of 120–240 μg/ml . The patch pipette was tip-filled with the above normal potassium aspartate intracellular solution before back-filling with the amphotericin B containing solution to prevent leakage of the antibiotic onto the IHC prior to sealing onto the membrane . Data acquisition was controlled by pClamp software using a Digidata 1440A ( Molecular Devices , USA ) . Voltage-clamp recordings were low-pass filtered at 2 . 5 kHz ( 8-pole Bessel ) and sampled at 5 kHz . Current clamp steps were recorded at 5 kHz and filtered at 2 . 5 kHz and the sound-like stimulation protocols were recorded at 100 kHz and filtered at 20 or 50 kHz . Data analysis was performed using Origin software ( OriginLab , USA ) . The residual series resistance ( Rs ) after compensation ( 60–80% ) was 1 . 16 ± 0 . 08 MΩ ( n = 60 ) for apical IHCs and 1 . 15 ± 0 . 09 MΩ ( n = 55 ) for basal cells . The average voltage-clamp time constant ( product of Rs and membrane capacitance Cm; apical: 12 . 4 ± 0 . 1 pF , n = 60 , basal: 11 . 7 ± 0 . 2 pF , n = 55 ) was 14 . 5 ± 1 . 0 µs for apical IHCs and 13 . 2 ± 0 . 8 µs for basal cells . Membrane potentials were corrected for a liquid junction potential measured between electrode and bath solutions of –4 mV for the KCl intracellular solution or –10 mV for the K-aspartate solution . In some voltage and current clamp experiments an extracellular solution containing low-Ca2+ ( 40 µM Ca2+: obtained by buffering 3 . 7 mM Ca2+ with 4 mM ( 2-hydroxyethyl ) ethylenediaminetriacetic acid ) was used to mimic the endolymphic Ca2+ concentration ( 20–40 µM: Bosher and Warren , 1978; Salt et al . , 1989 ) . To verify that the effects of the low-Ca2+ solution were mediated by the transducer channel , the transducer channel blocker dihydrostreptomycin ( DHS; 100 µM , Marcotti et al . , 2005 ) was added to the low-Ca2+ solution . In order to isolate and identify the different K+ currents expressed in IHCs ( IK , s , IK , n , IK , f: Marcotti et al . , 2003 , 2004 ) , different ion channel blockers were added to the extracellular solution: the KCNQ channel blocker linopridine ( 80 µM ) , the BK Ca2+-activated K+ channel blocker iberiotoxin ( 60 nM ) , both from Tocris Bioscience ( UK ) , and the inward rectifier channel blocker Cs+ ( 5 mM ) . Extracellular solutions containing different ion concentrations or channel blockers were topically perfused onto IHCs using a multi-barreled pipette positioned close to the cells , but far enough away as to not cause any hair bundle movement . Statistical comparisons of means were made by Student’s two-tailed t-test or , for multiple comparisons , analysis of variance , usually one-way ANOVA followed by the Tukey test . Mean values are quoted ± s . e . m . where P<0 . 05 indicates statistical significance . MT currents were elicited by stimulating the hair bundles of apical and basal coil mature gerbil cochlear IHCs ( P17–P25 ) using a fluid jet from a pipette ( tip diameter 10–12 µm ) driven by a piezoelectric disc ( Corns et al . , 2014 , Kros et al . , 1992 ) . The pipette tip of the fluid jet was positioned near to the bundles to elicit a maximal transducer current . Mechanical stimuli were applied as 50 Hz sinusoids ( filtered at 0 . 5 kHz , 8-pole Bessel ) with driving voltages of up to ± 20 V . Positive driver voltage ( fluid flowing out of the jet ) caused hair bundles to move towards the taller stereocilia and produced excitatory responses . The pressure in the fluid jet was zeroed before approaching the IHC to ensure hair bundles were not displaced from their resting position . All recordings were carried out at body temperature and in the presence of the endolymph-like low-Ca2+ solution ( 40 µM ) . FM1-43 experiments were performed on acutely dissected cochleae as previously described ( Gale et al . , 2001 ) . Briefly , the apical and basal cochlear coils from the mature gerbil ( P18–P22 ) were dissected in normal extracellular solution and then placed next to each other in the recording chamber and held down using a nylon mesh . At this stage the normal extracellular solution containing 1 . 3 mM Ca2+ was changed to the low-Ca2+ ( 40 µM ) solution . The cochlear coils were then bath exposed to a low-Ca2+ solution containing 3 μM FM1-43 for 8 s and immediately washed several times with normal 1 . 3 mM Ca2+ extracellular solution . The cochleae were then viewed with an upright microscope ( Olympus , Japan ) equipped with epifluorescence optics and fluorescein isothiocyanate ( FITC ) filters ( excitation , 488 nm; emission , 520 nm ) using a X63 water immersion objective . Images were captured using a charge-coupled device ( CCD ) camera and were taken within 15 min after exposure to FM1-43 . Fluorescence images were taken with 4 s exposure time . Stock solutions of 3 mM FM1-43 were prepared in water . A total number of six cochleae from three gerbils were used and the tectorial membranes were left intact . These experiments were performed at room temperature ( 22–25°C ) . | Many animals’ survival depends on them accurately and quickly identifying sounds in their environment . In animals with backbones , cells with hair-like projections ( called hair cells ) inside the ear convert information collected from sound waves into electrical signals . These signals are then transmitted to the brain , which processes the information further . Animals like bullfrogs are adapted to hearing low frequency sounds , like their own mating calls . These frog’s hair cells are individually tuned so that they can capture sounds in this low frequency range . Mammals , on the other hand , have evolved to hear a much wider range of sounds from loud and low frequency sounds , such as thunder , to soft and high frequency sounds , like the cries of their young . In mammals , the part of inner ear involved in hearing ( called the cochlea ) has an elaborate spiral-like shape . The structure of the cochlea results in different frequencies of sound being transformed by the hair cells into electrical signals at different points around the spiral . Because of this , most researchers didn’t think that hair cells in mammals were individually tuned like those in bullfrogs . Now , Stuart Johnson demonstrates that hair cells in different parts of the gerbil’s cochlea are specialized for encoding sounds of specific frequencies . In conditions that mimic the environment inside the ear , a very precise jet of fluid was used to stimulate single hair cells in a similar way to a sound wave . The experiments then compared how hair cells from the upper and lower parts of the cochlea’s spiral responded . Johnson found that hair cells from the upper portion of the gerbils’ cochlea are specialized to capture low frequency sounds . They have electrical properties that allow them to quickly transmit information to the brain about low frequency sounds . In the lower portion of the cochlea , hair cells are specialized to capture high frequency sounds . That is , their electrical properties make it easier for these hair cells to transmit detailed information to the brain about the volume of high frequency sounds . Together , these findings help explain how these animals are able to localize sounds , which requires capturing both the timing and intensity of different types of sounds . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Membrane properties specialize mammalian inner hair cells for frequency or intensity encoding |
Cell division is essential to expand , shape , and replenish epithelia . In the adult small intestine , cells from a common progenitor intermix with other lineages , whereas cell progeny in many other epithelia form contiguous patches . The mechanisms that generate these distinct patterns of progeny are poorly understood . Using light sheet and confocal imaging of intestinal organoids , we show that lineages intersperse during cytokinesis , when elongated interphase cells insert between apically displaced daughters . Reducing the cellular aspect ratio to minimize the height difference between interphase and mitotic cells disrupts interspersion , producing contiguous patches . Cellular aspect ratio is similarly a key parameter for division-coupled interspersion in the early mouse embryo , suggesting that this physical mechanism for patterning progeny may pertain to many mammalian epithelia . Our results reveal that the process of cytokinesis in elongated mammalian epithelia allows lineages to intermix and that cellular aspect ratio is a critical modulator of the progeny pattern .
Epithelia are sheets of polarized cells that function as barriers between compartments of multicellular organisms and between the organism and the external environment . In addition to providing a physical barrier , specialized epithelial cell types provide functions including sensation , absorption and secretion , and contribute to the identities of nearby cells through cell-cell signaling . Proper epithelial function requires that these diverse cell types are positioned appropriately within the tissue and that this distribution is maintained as new cells are added through cell division . The adult mammalian small intestine is a prime example of an epithelium that contains many cell types and maintains a high degree of spatial organization during rapid turnover ( Barker , 2014 ) . In the small intestine , divisions of stem cells in the crypts of Lieberkühn replenish the stem cell pool and generate absorptive and secretory progenitor cells in the crypt , which in turn produce differentiated cells that carry out the absorptive and protective functions of the gut ( Gracz and Magness , 2014 ) . Throughout the epithelium , cells derived from a given progenitor intersperse with other cells ( Carroll et al . , 2017 ) . In particular , lineage tracing in fixed tissues has established that cells derived from secretory progenitors intermix with cells derived from absorptive progenitors along the crypt and villus length ( Yang et al . , 2001 ) . At the crypt base , stem cells are interspersed with Paneth cells ( Farin et al . , 2016 ) . Interspersion of cell lineages plays important roles in determining local signaling environments required for intestinal homeostasis . For example , intestinal stem cells receive signals critical to their identity from neighboring Paneth cells ( Sato et al . , 2011 ) . Indeed , direct contact between stem and Paneth cells supports stem cell maintenance ( Farin et al . , 2016 ) . However , the molecular mechanisms that underlie the intermixing of lineages are poorly understood . Here , we use light sheet and confocal imaging of live murine small intestinal organoids to define the mechanisms of cell interspersion . We find that rearrangements of the actin cytoskeleton displace mitotic cells along the apical-basal axis , such that cell division occurs at the apical surface . Interspersion arises when elongated interphase neighboring cells wedge between apically dividing daughters during cytokinesis . We find that the propensity to intersperse during division requires an elongated shape of cells in the epithelium; reducing the cellular aspect ratio ( height: width ) in organoids disrupts interspersion , resulting in outgrowth of lineage patches . Consistent with our data indicating that the physical parameters of the tissue are a critical determinant of interspersion during division , we demonstrate that the elongated epiblast/primitive ectoderm of post-implantation ( E7 . 5 ) mouse embryos , but not the short visceral endoderm , also undergoes division-coupled cell interspersion . Thus , tissues of distinct developmental context from the adult small intestine exhibit similar mechanisms for patterning cellular progeny according to cellular dimensions . Together , our data indicate that cell shape differences between interphase and mitotic cells in elongated mammalian epithelia can allow a neighboring cell to insert between nascent daughter cells during cytokinesis and drive interspersion of cellular progeny .
To identify the basis for cell interspersion , we performed time-lapse imaging of adult murine small intestinal organoids ( Kretzschmar and Clevers , 2016; Sato et al . , 2009 ) by confocal and light sheet microscopy ( single plane illumination microscopy - SPIM ) ( Wu et al . , 2013 ) ( Figure 1A ) . To visualize cell lineages , we first used organoids in which the cytoplasm of cells of the secretory lineage was labeled with RFP ( Atoh1CreER; R26RFP ) . Strikingly , we observed that daughter cells separated from one another in approximately half of divisions ( 31/50 divisions , Figure 1A and Video 1; also see [Carroll et al . , 2017] ) . We observed that Atoh1-expressing secretory daughters along the crypt length separated from one another , mixing with unlabeled cells ( Figure 1A and Video 1 ) . 3D SPIM data confirmed that cells were fully separated on their basal surface , although they maintained a minimal contact on the apical surface , creating a V-shaped geometry ( Figure 1C , Figure 1—figure supplement 1E , Figure 1—video 1 , 9/16 daughter pairs ) . When daughters did not separate during the division ( Figure 1D , top panels , 7/16 daughter pairs ) , these daughters either became separated at later time points by division of a neighboring cell ( Figure 1D , bottom panels and Figure 1—video 2 ) , or remained as neighbors for the duration of imaging . These data indicate that separation of nascent daughter cells during cell division makes substantial contributions to the relative positioning of cell types within the intestinal epithelium . We next tested whether daughter cell separation was a common feature of cell lineages in the intestinal epithelium . Notch1-expressing cells ( from Notch1CreERT2; R26RFP organoids ) , which comprise all non-secretory cells including stem cells and absorptive cells , also interspersed during division ( Figure 1—figure supplement 1A ) . Finally , dividing stem cells ( labeled with Lgr5DTR-GFP ) at the crypt base also separated , with secretory ( Paneth ) cells ( labeled with Atoh1CreER; R26RFP ) inserting between them ( Figure 1B , Figure 1—figure supplement 1B and Video 2 ) . Altering cell fates , for example by inhibiting Notch signaling to cause an expansion of secretory cells , did not alter the frequency of this process ( Figure 1—figure supplement 1C , D ) . Thus , cells intersperse during a subset of divisions in all cell lineages of the crypt epithelium . We next sought to determine whether the interspersion of cellular progeny observed in organoids also occurred in the intestine in vivo . To this end , we labeled a subset of cells in the intestines of adult mice with different fluorophores by induction of the stochastic multicolor reporter allele , R26Brainbow2 . 1 ( Vil1CreERT2; R26Brainbow2 . 1 ) . After three days of Cre induction , which is sufficient for most crypt epithelial cells to divide at least once ( Snippert et al . , 2010 ) , the intestines were fixed and the positions of progeny analyzed in thick sections . Consistent with our organoid imaging , we observed that a subset of progeny ( 18/40 progeny pairs , n = 3 mice ) were interspersed with unlabeled cells or differently labeled cells in the intact intestine ( Figure 1E ) . Thus , progeny intersperse with neighboring cells in intestinal organoids and in the intestinal epithelium in vivo . We next sought to characterize the cell behaviors that give rise to interspersion during cell division in the intestinal epithelium . We observed that mixing occurred as cells underwent cytokinesis on the apical surface of the epithelium , during which neighboring cells intruded within the ingressing cytokinetic furrow ( Figure 1B , Video 2 ) . First , mitotic cells displaced to the apical surface of the epithelium , and the dramatic reduction in their basal footprint caused neighboring cells to reposition and occupy the position above ( basal to ) the mitotic cell ( Figure 1B , Figure 1—figure supplement 1B ) . Cells progressed through a polarized ( non-concentric ) cytokinesis ( Figure 2A , Video 2 , Figure 2—videos 1 , 2 and 3 ) ( also see [Fleming et al . , 2007] ) , in which the cleavage furrow initiated from the basal surface and then progressed to the apical surface . As cytokinesis continued , a minimal daughter-daughter contact remained on the apical surface ( Figure 1—figure supplement 1E ) . We note that this minimal vertex contact is consistent with other reports of daughter cell geometry during vertebrate cytokinesis ( Higashi et al . , 2016 ) , but contrasts with the long daughter-daughter interface generated during cytokinesis in Drosophila epithelia ( Gibson et al . , 2006; Herszterg et al . , 2013; Pinheiro et al . , 2017 ) , as we will return to in the Discussion . The minimal contact between daughters generated by cytokinesis allowed a neighboring interphase cell to wedge between the daughters ( Video 2 ) . Finally , as the division completed , the daughter cells elongated on either side of the invading neighbor cell to occupy the full apical-basal axis in interphase ( Figure 1 , Video 2 ) . In contrast to the dramatic shape changes on the basal surface of dividing cells , the apical surface remained unperturbed: the apical footprint of the mitotic cell was similar to its interphase neighbors ( Figure 2—figure supplement 1A–C ) , and a cytokinetic furrow was absent from the apical surface as in many metazoan epithelia ( Fleming et al . , 2007; Guillot and Lecuit , 2013; Herszterg et al . , 2013; Founounou et al . , 2013 ) . Previous studies showed that cell-cell junctions on the apical surface of the intestine persist throughout mitosis ( Jinguji and Ishikawa , 1992 ) and staining with junctional markers indicated that the same is true for intestinal organoids ( Figure 2—figure supplement 1A ) . To test the possibility that persistent cell-cell contacts oppose mitotic shape changes on the apical surface , we dissociated organoids into single cells or pairs of cells and performed time-lapse imaging of mitotic exit . In contrast to the polarized cytokinesis that occurs in the tissue , cytokinesis occurred symmetrically in dissociated cells ( Figure 2B , Figure 2—video 4 ) , suggesting that tissue architecture plays a crucial role in this polarization . Together , these data indicate that mixing arises during cytokinesis as part of a suite of mitotic cell shape changes that are confined to the basolateral surface within the context of the tissue . Our observations suggested that a critical initiating step during cell interspersion was the positioning of the dividing cell on the apical surface of the epithelium . We therefore sought to determine the mechanism that gives rise to this apical displacement . Apical displacement initiated concurrently with mitotic entry ( Figure 2C , Figure 2—figure supplement 1D , Video 3 and Figure 2—video 5 ) , indicating that it was distinct from interkinetic nuclear migration , a process in which the nucleus is moved apically during interphase ( interkinesis ) ( Sauer , 1936 ) by actin or microtubule-based forces ( reviewed in [Norden , 2017] ) . Apical displacement occurred as cells adopted the rounded geometry classically associated with mitosis ( reviewed in [Théry and Bornens , 2008] ) ( Figure 2—figure supplement 1D , Video 1 , Figure 2—video 5 , Figure 2—video 6 ) ; at metaphase and anaphase , only fine membranous processes tethered the cell to the basal surface ( Figure 2—figure supplement 1E–F ) , consistent with previous observations ( Carroll et al . , 2017; Fleming et al . , 2007; Jinguji and Ishikawa , 1992; Trier , 1963 ) . Mitotic rounding also contributes to late stages of interkinetic nuclear migration in some systems ( Meyer et al . , 2011; Spear and Erickson , 2012 ) . Therefore , we tested the importance of actin-driven mitotic rounding for apical displacement . Treatment with the actin depolymerizing small molecule Latrunculin A disrupted rounding and apical displacement ( Figure 2D , E , Figure 2—video 7 ) ; in contrast , cells treated with the microtubule depolymerizing drug nocodazole rounded onto the apical surface similarly to control cells ( Figure 2D and E , Figure 2—video 8 , Figure 2—video 9 ) . As Latrunculin-treated cells entered anaphase , the chromosome masses were positioned orthogonally to the plane of the epithelium , in contrast to the planar divisions observed in control cells ( Figure 2F , Figure 2—figure supplement 1G , Figure 2—videos 7 , 9 ) . This suggests that cell rounding is crucial for the normal planar orientation of the spindle in the intestine , as in some Drosophila epithelia ( Chanet et al . , 2017; Nakajima et al . , 2013 ) . Collectively , our data suggest that actin-based cell rounding displaces mitotic cells apically and is required for planar spindle orientation . We next assessed the mechanisms that restore the basal footprint and the basal position of the nuclei after division . After division , we observed that the basal edge of nascent daughters extended a protrusive front that resembled the leading edge of migrating cells ( Figure 2G; Figure 2—video 10 ) . Therefore , we tested the contributions of the actin cytoskeleton for basal reinsertion . As actin disruption blocks the initial displacement of mitotic cells to the apical surface ( Figure 2D and E ) , determining the requirements for actin in basal reinsertion required that mitotic cells be positioned on the apical surface before disrupting actin . To achieve this , we first blocked cells on the apical surface by arresting them in mitosis with the mitotic kinesin ( Eg5 ) inhibitor S-trityl-L-cysteine ( STLC ) . Cells arrested in mitosis did not reinsert unless mitotic exit was induced by inhibition of the spindle assembly checkpoint ( SAC; Mps1 inhibitor AZ3146 ) or cyclin-dependent kinase ( CDK; RO-3306 ) ( Figure 2—figure supplement 1H , Figure 2—video 11 ) . Thus , mitotic exit and reversal of CDK phosphorylation are sufficient for basal reinsertion , even in the absence of chromosome segregation . Using this mitotic arrest and exit protocol , we tested the requirements for the actin and microtubule cytoskeletons for basal reinsertion ( Figure 2—figure supplement 1I ) . When we disrupted the actin cytoskeleton and induced mitotic exit , the nucleus reformed its interphase morphology on the apical surface and the cell boundary did not protrude toward the basal surface ( Figure 2H , I , Figure 2—video 12 ) . In contrast , depolymerizing microtubules with nocodazole and inducing mitotic exit did not interfere with the ability of nuclei or the cell boundary to reach the basal surface ( Figure 2H , I , Figure 2—video 13 ) . Although actin also plays a critical role in cytokinesis , nuclei reinserted normally following inhibition of cytokinesis using the Polo-like kinase one inhibitor , BI2536 ( Lénárt et al . , 2007; Steegmaier et al . , 2007 ) ( Figure 2—figure supplement 1J ) , indicating that cytokinesis is dispensable for basal movement . Collectively , these data indicate that actin-driven cell elongation after mitotic exit re-establishes the interphase architecture of daughter cells . Our data indicate that the displacement of cells along the elongated apical-basal axis over the course of cell division plays a role in cell interspersion . To test the importance of an elongated apical-basal axis for cell interspersion , we imaged cell behavior in spherical organoids derived from fetal intestine ( Fordham et al . , 2013; Mustata et al . , 2013 ) , in which cells are very short in the apical-basal dimension , and are instead elongated along the sphere circumference ( Figure 3A , Figure 3—video 1 ) . Fetal spheroids did not exhibit apical-basal mitotic movements and the daughters did not intersperse with other cells during division ( Figure 3A , B , Figure 3—video 1 , Figure 3—video 2 ) ( 50/50 divisions ) . We also induced a subset of adult intestinal organoids to adopt a spherical geometry and short apical-basal axis by addition of exogenous Wnt to the medium ( Sato et al . , 2011 ) ( Figure 3—figure supplement 1A ) . These adult spheroids also failed to exhibit apical-basal mitotic movements and the daughters did not intermix with other cells ( 50/50 divisions ) ( Figure 3B , Figure 3—figure supplement 1A ) . Consistent with the lack of interspersion , these spheroids contained patches of cellular progeny ( Figure 3C ) , in contrast to the interspersed pattern of cell lineages observed in normal adult organoids ( Figure 1A ) . As an internal control , a subset of organoids cultured in high Wnt conditions retained their budded morphology and elongated apical-basal cell shape; these organoids continued to exhibit apical displacement and the interspersed pattern of cell lineages ( Figure 3C ) . This experiment , as well as our observations of adjacent progeny in the fetal spheroids , which exhibit very low expression of the Wnt reporter gene Axin2 ( Mustata et al . , 2013 ) , indicate that the effect of cell shape on interspersion is separable from hyperactive Wnt signaling , in contrast with previous work ( Carroll et al . , 2017 ) . Together , these data indicate that an elongated apical-basal axis is critical for apical mitosis and cell interspersion during division . Based on our data suggesting a crucial role for the cellular aspect ratio in interspersion in the organoids ( Figure 3D ) , we next examined whether the mechanisms that we defined in the intestine may be relevant to other tissues with similar physical parameters . Pioneering work by Gardner and Cockroft ( 1998 ) revealed that cells injected into mouse blastocysts to generate chimeras become dispersed throughout the epiblast and primitive ectoderm of the post-implantation embryo . The authors proposed that this pattern might arise as a consequence of cell division , which they and others have observed occurs on the apical surface of the tissue ( Gardner and Cockroft , 1998; Ichikawa et al . , 2013 ) . Therefore , we tested this prediction by performing time-lapse SPIM imaging of E7 . 5 ( late streak-early bud ) mouse embryos ( Figure 3E ) , in which the epiblast/primitive ectoderm was mosaically labeled ( CAGGSCreER; R26Brainbow2 . 1 ) . We imaged cell divisions in these embryos for at least 3 hr and observed that divisions proceeded in a similar manner to the intestinal epithelium , with mitotic cells displacing to the apical surface as they rounded ( Figure 3F–G ) . Daughter cells then separated from one another and interspersed with unlabeled cells during cytokinesis ( Figure 3H , Video 4 ) ( 8/10 divisions , n = 3 embryos from three pregnancies ) . Thus , daughter cells positioned on the apical surface intersperse with other cells during cytokinesis in the elongated epiblast/primitive ectoderm of the embryo , as in the adult small intestine . In contrast , the cells of the visceral endoderm ( the low aspect ratio cells that surround the epiblast ) did not exhibit apical displacement and daughters remain adjacent ( Figure 3—figure supplement 1B , Figure 3—video 3 , 12/12 divisions , n = 3 embryos from three pregnancies ) , consistent with classical experiments reporting outgrowth of contiguous clones in this tissue ( Gardner , 1984; 1985; Lawson et al . , 1991 ) . Thus , cell division generates distinct progeny patterns in the two layers of the early post-implantation mouse embryo , consistent with a central role for cellular aspect ratio in determining the spatial patterning of cell progeny .
The functions of epithelial organs rely on the concerted action of multiple cell types . As these cell types are replenished as the organ renews , they must be positioned appropriately within the tissue . In some mammalian epithelia , such as the small intestine , daughter cells derived from a common progenitor disperse throughout the tissue and intermingle with cells of other lineages , a process that plays an important role in determining local signaling environments . Previous studies have reported that intermingling of cells can occur during cell division ( Carroll et al . , 2017; Firmino et al . , 2016; Gardner and Cockroft , 1998; Higashi et al . , 2016; Lau et al . , 2015; Packard et al . , 2013 ) but the mechanism by which this occurs has not been clear . Here , we show that intermixing arises when a neighboring cell inserts between apically displaced daughter cells during cytokinesis . The process of intermixing requires that the neighboring cell and dividing cell are positioned in such a way that the neighbor can occupy the wedge between the daughters generated by the ingressing furrow . Our data support a model in which the neighboring cell can become opportunely positioned for invasion into the cytokinetic furrow as a consequence of the cell shape changes associated with vertebrate mitosis in tissues comprised of cells with a high aspect ratio ( Figure 4 ) . In cells with a high aspect ratio , the actin-driven cell shape changes required for mitosis ( rounding and subsequent elongation ) displace the dividing cell along the apical-basal axis ( Figure 4 ) . As a result , an elongated interphase neighboring cell can surround the dividing cell both basally and laterally , allowing it to follow the path of the ingressing furrow between the daughters . Consistent with a key role for cell aspect ratio in interspersion behavior , reducing the aspect ratio in organoids generates patches . Live imaging of cell division in the two epithelial layers of the peri-gastrulation mouse embryo further supports a model in which cell aspect ratio is a critical parameter for determining whether cellular progeny intersperse , raising the intriguing possibility that the patterning principles that we define in the intestine may be a common feature of many mammalian epithelia . Several lines of evidence support a model in which interspersion arises as a mechanical consequence of executing planar cell division in elongated cells , rather than being determined by developmental signaling or differential adhesion between cells . First , daughter separation is observed throughout the intestinal crypt for all progenitor cell identities: stem cells , Notch-expressing absorptive progenitors and Atoh1-expressing secretory progenitors ( Figure 1 , Figure 1—figure supplement 1A ) . However , importantly , daughter separation is a frequent but not universal event , occurring in approximately half of the divisions observed , including when observing cells of a specific lineage ( Figure 1 , Figure 3D ) . Additionally , altering cell fates , for example by inhibiting Notch signaling to cause an expansion of secretory cells , does not alter the frequency of this process ( Figure 1—figure supplement 1C , D ) . In contrast , altering epithelial geometry in culture disrupts interspersion ( Figure 3 ) . Since our data indicate that interspersion can arise from the execution of planar cell division coupled with the physical parameters of the tissue , it raises the possibility that the mechanisms of interspersion that we define for the intestinal epithelium may be generalizable to other vertebrate tissues with similar physical parameters . Consistent with this notion , we observed similar interspersion in the high aspect ratio epithelium of the early mouse embryo , while the surrounding low aspect ratio epithelium did not exhibit division-coupled interspersion ( Figure 3 ) . Several tissues across vertebrates with a high aspect ratio have also been reported to exhibit division-coupled interspersion ( Carroll et al . , 2017; Firmino et al . , 2016; Gardner and Cockroft , 1998; Higashi et al . , 2016; Packard et al . , 2013 ) . In contrast , in numerous tissues in which cells have a low aspect ratio , progeny remain adjacent and form contiguous patches , including the interfollicular epidermis ( Ouspenskaia et al . , 2016; Rompolas et al . , 2016 ) , MDCK cells ( Reinsch and Karsenti , 1994 ) , and alveolar epithelial cells ( Desai et al . , 2014 ) . Our model raises the possibility that isolated reports of division-coupled interspersion in diverse vertebrates including frog , chick and mouse may be unified by a common physical mechanism arising from the aspect ratio of the tissue and the mechanics of cell division . While our data indicate that cellular aspect ratio is an important parameter for interspersion , the mechanics and geometry of cytokinesis also appear to play a central role . In vertebrates , the mechanism of furrow ingression minimizes the contact between the daughters and progresses until a single apex physically connects the two cells ( Higashi et al . , 2016 ) ( Figure 1—figure supplement 1E , Figure 4—figure supplement 1 ) . An important component of our model is that the development of the furrow creates a position , both basally and laterally , for neighboring cells to invade and occupy . However , in contrast , during cytokinesis in Drosophila , the two daughters form a long adhesive contact between them ( Gibson et al . , 2006 ) ( Figure 4—figure supplement 1 ) , dependent on myosin II accumulation in the neighboring cells ( Herszterg et al . , 2013; Pinheiro et al . , 2017 ) . In this regard , it is interesting to note that Drosophila epithelia exhibit a high aspect ratio , apical mitosis and non-concentric cytokinesis , yet do not exhibit cell interspersion and form contiguous patches of progeny ( Bryant , 1970; Bryant and Schneiderman , 1969; Founounou et al . , 2013; Gibson et al . , 2006; Guillot and Lecuit , 2013; Herszterg et al . , 2013; Meyer et al . , 2011; Morais-de-Sá and Sunkel , 2013 ) . We speculate that the extended cell-cell contact formed between daughter cells in Drosophila would oppose the invasion of a neighboring cell . In the future , it will be interesting to attempt to modify the extent of interactions between daughter cells either in Drosophila or vertebrate epithelia and determine the effects on progeny patterning . Broadly , since our data suggest that cell interspersion requires a set of criteria that are satisfied by many vertebrate epithelia , it is unlikely to be unique to those tissues in which it has been reported . Although our work has focused on the columnar epithelium of the small intestine , in which mitotic cell shape changes are sufficient to displace dividing cells relative to their neighbors , the numerous elongated pseudostratified epithelia that undergo apical mitosis due to interkinetic nuclear migration ( reviewed in [Norden , 2017] ) are particularly attractive candidates for division-coupled interspersion . Together , our model suggests that interspersion during cell division may be widespread across elongated vertebrate epithelia .
Adult mice of the following lines were used to generate organoids . R26mTmG/mTmG ( Muzumdar et al . , 2007 ) ( female ) Vil1Cre-ERT2/+ ( el Marjou et al . , 2004 ) ; R26mTmG/+ ( male ) Atoh1CreERT/+ ( Chow et al . , 2006 ) ; R26RFP/+ ( Madisen et al . , 2010 ) ; Lgr5DTR-GFP/+ ( Tian et al . , 2011 ) ( female ) Notch1CreERT2 ( SAT ) /+ ( Fre et al . , 2011 ) ; R26RFP/ RFP ( Madisen et al . , 2010 ) ( female ) Fetal organoids were generated from E13 . 5 C57BL/6J embryos . For imaging of cell interspersion in the intact intestine , adult Vil1Cre-ERT2/+ ( el Marjou et al . , 2004 ) ; R26Brainbow2 . 1/+ ( Snippert et al . , 2010 ) mice were used . Recombination was induced by oral gavage with one dose of 2 . 5 mg tamoxifen in corn oil 3 days before analysis . Brainbow embryos were generated by crossing CAGGSCreER/+ males ( Hayashi and McMahon , 2002 ) to R26Brainbow2 . 1/Brainbow2 . 1 ( Snippert et al . , 2010 ) females . Plugged females were injected intraperitoneally with 2 . 5 mg tamoxifen in corn oil at E5 . 5 . H2B-GFP embryos were generated by crossing H2B-GFP males ( Hadjantonakis and Papaioannou , 2004 ) to C57BL/6J females . Embryos were dissected at E7 . 5 and staged according to ( Delling et al . , 2016; Downs and Davies , 1993 ) . The strains of these mice were the same as previously described in their respective references at the time of acquisition but were subsequently maintained on mixed backgrounds after breeding between different lines . All experiments involving mice were approved by the Institutional Animal Care and Use Committee of the University of California , San Francisco ( protocol #AN151723 ) . Small intestinal crypts were isolated from adult mice or E13 . 5 embryos and cultured in medium supplemented with human recombinant EGF , human recombinant Noggin and R-Spondin conditioned medium ( ENR medium ) as described ( Sato et al . , 2009 ) . Catalog numbers for culture medium components are described in ( Mahe et al . , 2013 ) . R-spondin and Wnt3a conditioned medium were used where indicated . Lentiviral transduction of adult organoids was performed as described ( Koo et al . , 2011 ) . Fetal organoids were transduced according to the same protocol , but without the addition of exogenous Wnt3a to the medium at any step . For propagation , organoids were grown in 24-well plastic plates . For spinning disc imaging and immunofluorescence , organoids were grown in 96-well glass bottom dishes ( Matriplate , Brooks ) . For SPIM , organoids were grown on glass coverslips which were then transferred to the SPIM imaging chamber ( see below ) . For immunofluorescence , organoids were fixed in 4% PFA in PBS for 1 hr before blocking in 3% BSA , TBS , 0 . 1% Triton X-100 . Primary antibody was incubated overnight at four degrees and secondary antibody was incubated for >2 hr at RT . Reagents used for immunofluorescence were as follows: rabbit anti-ZO-1 antibody ( Thermo Fisher ) , Alexa488-Phalloidin ( Thermo Fisher # A12379 ) , Hoechst 33342 ( Molecular Probes H3570 ) . For organoid dissociation , organoids in one well of a 24 well plate were washed once in PBS before Matrigel was manually disrupted by pipetting in TrypLE Select ( Life Technologies ) in the well . The plate was then incubated at 37°C for 7–8 min before additional disruption with a P200 pipette . The cell suspension was centrifuged in medium +5% fetal bovine serum at 1000 x g for 5 min . The pellet was resuspended in Matrigel , allowed to polymerize for 10 min and covered with ENR medium and immediately transferred to the microscope for imaging for 45 min – 1 hr . Animals were anesthetized by intraperitoneal ( i . p . ) injection of 250 mg/kg of body weight avertin ( 2 , 2 , 2-tribromoethanol ) and transcardially perfused with 4% paraformaldehyde ( PFA ) in 0 . 1 M phosphate-buffered saline ( PBS ) . Dissected tissues were post-fixed in 4% PFA for 3 hr at 4°C and cryoprotected in 30% sucrose in 1 × PBS overnight at 4°C . For whole mount tissue , the external smooth muscle and fat of the most proximal 3 cm of the small intestine was removed and epithelial tissue was coverslipped with ProLong Gold Antifade ( P36930 , Thermo Fisher Scientific ) . For tissue sections , tissue was embedded in OCT compound ( 4583 , Sakura ) , frozen and stored at −80°C . Small intestine swiss rolls were cryosectioned at 50 µm and coverslipped with ProLong Gold Antifade . Whole mount tissue and sections were counterstained with DAPI ( 1:10000; D9542 , Sigma ) for 45 min or 15 min , respectively . For spinning disc confocal imaging , images were acquired on a Yokogawa CSU-X1 spinning disk confocal attached to an inverted Nikon TI microscope , an Andor iXon Ultra 897 EM-CCD camera , using Micro-Manager software ( Edelstein et al . , 2010 ) . Imaging of 12 × 1 µm z-stacks was performed either at 4 min time intervals with a 40 × 1 . 30 NA Plan Fluor oil objective or a 20 × 0 . 75 NA objective , or at 20 s time intervals with a 60XA 1 . 20 NA Plan Apo water immersion objective . Maximum intensity projections of 1–5 Z-stacks are shown unless otherwise noted . Point-scanning confocal imaging of intact intestines was performed using a Leica TCS SP8 X confocal microscope , with HyD and LAS X software . 0 . 76 μm optical sections were acquired sequentially with a 63 × 1 . 40 HC PL APO CS2 oil objective . 4-dimensional imaging was performed on an ASI diSPIM microscope equipped with 40 × 0 . 80W NA NIR-Apo water dipping objectives , Hamamatsu Flash 4 . 0 cameras , and 488 nm and 561 nm solid state lasers from Vortran , using a nightly build of the Micro-Manager software . The structure of the environmental control chamber is described in detail at https://valelab4 . ucsf . edu/~nstuurman/protocols/diSPIMIncubator/ . Temperature was maintained using 3 × 50 ohm resistors attached to the stainless steel incubation chamber holding the coverslip and medium , a 10 kOhm thermistor inserted in the medium and a temperature controller ( TE Technology , Inc . TC-48–20 ) . O2 and CO2 tensions in the medium were kept constant by flowing humidified gas underneath the sample chamber . To allow gas exchange , the sample was placed on a sandwich of 2 × 24 ×50 mm coverslip glasses in which 2 ~ 12×12 mm windows had been laser-cut and between which a piece of ~37 . 5 µm thick Teflon AF-2400 ( a gift from BioGeneral , Inc . ) was placed . Evaporation was minimized by layering mineral oil ( Howard ) over the sample . Organoids were imaged in ENR medium; embryos were imaged in DMEM +25% rat serum ( Rockland , Inc . ) . 3D reconstructions were generated using a Micro-Manager plugin ( https://github . com/nicost/MMClearVolumePlugin ) that uses the ClearVolume library ( Royer et al . , 2015 ) . 3D reconstructions are scaled with gamma adjustment . All imaging experiments were performed at 37°C , 5% CO2 , 20% O2 . Small molecule concentrations are described in Table 1 . All stock solutions were prepared in DMSO . All pharmacological experiments were performed in the presence of 10 µM Verapamil to inhibit drug efflux . Details of statistical tests are provided in the figure legends . A statistical method of sample size calculation was not used during study design . Data were pooled from at least three biological replicates . When the observations presented were observed in less than 100% of cases , their frequency is noted in the figure , figure legend and/or text . | The body has an impressive ability to renew itself by replacing old and damaged cells with new ones . This can happen rapidly; for example , the lining of the intestine renews itself approximately every five days . The lining contains many different cell types , which exchange important signals with their neighbors . This means that the new cells need to occupy similar positions to the ones they are replacing to keep the intestine working . New cells form when existing cells double their contents and divide . In many tissues the resulting cells sit side-by-side . But when cells in the intestine divide , the new cells often separate , ending up on either side of a cell that did not divide . To investigate how this happens , McKinley et al . used live microscopy techniques to watch in real time as new cells divide and position themselves in mouse intestinal organoids – miniature versions of organs that can be grown outside the body . This revealed that the shape of intestinal cells explains why the newly formed cells become separated . Intestinal cells are taller than they are wide , and divide near their top edge . This enables a neighboring cell to squeeze between the new cells as they divide . Further experiments showed that tall cells in other mouse tissues also become separated after division . The process of new cells interspersing with their neighbors due to their height is therefore not unique to the intestine . It may also be common in other mammalian tissues . There is great potential for investigating this further because labs can now grow many types of organoids , representing different organs . Using live microscopy to examine them could reveal more about how various tissues grow . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2018 | Cellular aspect ratio and cell division mechanics underlie the patterning of cell progeny in diverse mammalian epithelia |
Many epithelial cancers show cell cycle dysfunction tightly correlated with the overexpression of the serine/threonine kinase Aurora A ( AURKA ) . Its role in mitotic progression has been extensively characterised , and evidence for new AURKA functions emerges . Here , we reveal that AURKA is located and imported in mitochondria in several human cancer cell lines . Mitochondrial AURKA impacts on two organelle functions: mitochondrial dynamics and energy production . When AURKA is expressed at endogenous levels during interphase , it induces mitochondrial fragmentation independently from RALA . Conversely , AURKA enhances mitochondrial fusion and ATP production when it is over-expressed . We demonstrate that AURKA directly regulates mitochondrial functions and that AURKA over-expression promotes metabolic reprogramming by increasing mitochondrial interconnectivity . Our work paves the way to anti-cancer therapeutics based on the simultaneous targeting of mitochondrial functions and AURKA inhibition .
The mitotic kinase AURKA controls centrosomal maturation , the timing of mitotic entry , bipolar and central spindle assembly , and cytokinesis ( Nikonova et al . , 2013 ) . AURKA is over-expressed in epithelial cancers , and it is believed to act as an oncogene through the induction of genomic instability . AURKA alters the number of centrosomes , the properties of the mitotic spindles , it induces aneuploidy and a defective cell division ( Bischoff et al . , 1998; Zhang et al . , 2004; Nikonova et al . , 2013 ) . To perform these functions , AURKA interacts with multiple proteins at centrosomes and on the mitotic spindle , in a kinase-dependent or independent manner ( Nikonova et al . , 2013 ) . In mice , the over-expression of AURKA suffices to induce the appearance of mammary tumours similar to human breast cancers ( Wang et al . , 2006; Treekitkarnmongkol et al . , 2016 ) . Thus , AURKA has been of high interest for pharmaceutical companies as a drug target , despite the currently available drugs show only modest beneficial effects in patients . Although still at a preclinical level , the only promising strategy appears to be the combination of AURKA inhibitors with agents simultaneously targeting multiple cancer-relevant AURKA partners and functions ( Nikonova et al . , 2013 ) . Epithelial cancers were found to be dependent on mitochondrial ATP , produced through oxidative phosphorylation , as a source of energy ( Whitaker-Menezes et al . , 2011 ) . Particularly , epithelial cancer cells were found to carry a specific molecular signature constituted by 38 genes regulating the functionality of the mitochondrial respiratory chain , the synthesis of mitochondrial ribosomes and the import of proteins into mitochondria . It is known that mitochondria adapt their ATP production rate by modulating their morphology from fragmented organelles to an interconnected network ( Mishra and Chan , 2016 ) . In particular , mitochondrial fusion was shown to increase ATP production in several paradigms ( Mitra et al . , 2009; Tondera et al . , 2009 ) , granting protection against apoptosis ( Lee et al . , 2004 ) and increasing cell proliferation ( Mitra , 2013 ) . However , how these effects are established by increasing mitochondrial connectivity remains to be mechanistically elucidated . In mammals , fusion is mediated by the Outer Mitochondrial Membrane GTPases MFN1 and 2 and the Inner Mitochondrial Membrane OPA1 , which undergoes a proteolytic cleavage from a long ( L-OPA1 ) to a short ( S-OPA1 ) isoforms . The balance between L- and S-OPA1 is controlled by multiple mitochondrial proteases to ensure mitochondrial fusion both in normal and in stress conditions ( Mishra and Chan , 2016 ) . Conversely , mitochondrial fission is regulated by the cytosolic GTPase DNM1L through its interaction with several mitochondrial receptors ( MFF , MIEF1 and 2 , FIS1 ) , but also through the regulation of its post-translational modifications ( phosphorylation , SUMOylation , acetylation and S-nitrosilation ) ( Mishra and Chan , 2016 ) . Mitochondrial fission is also used to separate dysfunctional mitochondria from the healthy network prior to their degradation by mitophagy , a selective type of autophagy ( Youle and Narendra , 2011 ) . Given that AURKA is a hallmark of epithelial cancers , understanding the complexity of the AURKA interactome is mandatory to optimise therapeutic strategies in patients with epithelial cancers derived from AURKA over-expression . Despite often considered a mitotic protein , recent evidence showed that AURKA is active at interphase as well ( Mori et al . , 2009; Bertolin et al . , 2016 ) , although its roles beyond mitosis are still largely unexplored . We here show that interphasic AURKA localises to mitochondria , where it is imported and processed . While exploring the potential roles of mitochondrial AURKA , we observed that it increases mitochondrial fusion through a direct interaction with proteins regulating mitochondrial dynamics . The modulation of mitochondrial fusion and fission mechanisms when AURKA is over-expressed increases ATP production via the mitochondrial respiratory chain .
While exploring the localisation of AURKA at interphase , we observed that AURKA co-localises with the mitochondrial processing peptidase PMPCB in human MCF7 cell lines ( Figure 1A ) . The fluorescence signal of AURKA observed at mitochondria is specific , as it disappeared after AURKA knockdown by siRNA-mediated gene silencing ( Figure 1A compare the two left panels and histograms ) . AURKA depletion also leads to profound changes in the organisation of the mitochondrial network , strongly suggesting a functional role of AURKA at mitochondria ( Figure 1A compare the two middle panels ) . In addition , AURKA localises to mitochondria regardless of the cell cycle phase and of its relative abundance ( Figure 1—figure supplement 1A ) . We then explored how AURKA is imported into mitochondria . The vast majority of mitochondrial proteins undergo one or sequential proteolytic cleavages when imported into these organelles ( Chacinska et al . , 2009 ) . First , the mitochondrial matrix peptidase PMPCB cuts the Mitochondrial Targeting Sequence ( MTS ) off the mitochondrial precursor protein ( Chacinska et al . , 2009 ) . Then , a second or multiple mitochondrial proteases can further cleave the pre-protein and allow it to reach the mitochondrial sub-compartment of destination . We therefore searched for mitochondrial AURKA isoforms representing one or more cleavage products . In immunoblots of total cell and mitochondrial lysates from HEK293 cells , AURKA could be detected as three isoforms: a predominant full-length isoform of ~46 kDa ( AURKA46 ) , an intermediate isoform of ~43 kDa ( AURKA43 ) compatible with a first proteolytic cleavage by PMPCB ( Chacinska et al . , 2009 ) , and a short isoform of ~38 kDa ( AURKA38 ) that presumably represents the mature mitochondrial isoform and which is detectable with a monoclonal ( Figure 1B ) and a polyclonal anti-AURKA antibody ( Figure 1—figure supplement 1B ) . These results also corroborate the mitochondrial localisation of AURKA in a second cell line . To confirm that AURKA43 and AURKA38 correspond to two AURKA isoforms processed inside mitochondria , we depleted 80% of the mitochondrial protease PMPCB . This almost completely blocked the intra-mitochondrial cleavage of AURKA ( Figure 1C ) , strongly suggesting that AURKA43 and AURKA38 are mitochondrial cleavage products issued by PMPCB-related import pathways in the matrix . To identify the cleavage sites from where AURKA43 and AURKA38 originate we used nanoLC-ESI MS/MS to search for semi-tryptic peptides corresponding to the N-terminally processed isoforms AURKA43 and AURKA38 . Compared to canonical peptides generated by the proteolytic action of trypsin , semi-tryptic peptides are cut non-canonically and they can therefore be generated by the action of mitochondrial proteases ( Vögtle et al . , 2009 ) . We retrieved two semi-tryptic peptides starting at residues 33 and between residues 80–82; the size of these peptides was consistent with the molecular weights of AURKA43 and AURKA38 , respectively ( Figure 1—figure supplement 1C and Supplementary file 1 ) . We then investigated the exact sub-mitochondrial localisation of AURKA in HEK293 cells using transmission electron microscopy ( TEM ) . We first detected ectopic AURKA fused to GFP in the mitochondrial matrix and in contact with mitochondrial cristae ( Figure 1D ) . This localisation was confirmed by employing different antibody combinations ( Figure 1—figure supplement 1D and E ) and by illustrating that AURKA localises to the mitochondrial matrix similarly to the mitochondrial matrix protein SOD2 ( Figure 1—figure supplement 1F ) . To determine whether AURKA is a soluble protein or it is strongly attached to mitochondrial membranes , protein extraction with sodium carbonate was carried out on mitochondrial fractions of HEK293 cells ( Figure 1E ) . AURKA was retrieved exclusively in the soluble fraction , reinforcing the conclusion that AURKA is mainly localised in the matrix as observed in TEM analyses . To further investigate the import of AURKA , we digested mitochondrial fractions of HEK293 cells with trypsin to eliminate mitochondrial protein precursors and peripheral Outer Mitochondrial Membrane ( OMM ) proteins , and we further combined trypsin with detergents to access the matrix . AURKA38 showed a degradation pattern similar to the one of the matrix PMPCB protease ( Figure 1F ) . Together , these data demonstrate for the first time that AURKA is localised to mitochondria at interphase and that it is imported and processed in the mitochondrial matrix . Positively charged motifs that are located at the amino termini of mitochondrial pre-proteins often target the pre-proteins to mitochondria ( Chacinska et al . , 2009 ) . Once there , pre-proteins interact primarily with the Translocase of Outer Mitochondrial Membrane ( TOMM ) complex , a multi-subunit machinery regulating the recognition and the entry of pre-proteins inside mitochondria ( Chacinska et al . , 2009 ) . Given that AURKA enters mitochondria and it is processed in the matrix , we analysed whether it follows a canonical import route through TOMM . In this light , we searched for the physical proximity of AURKA with the subunits of the TOMM machinery by FRET/FLIM ( Padilla-Parra and Tramier , 2012 ) . Decreases in donor fluorescence lifetime in FRET/FLIM analyses indicated physical proximity compatible with protein-protein interactions between exogenously expressed AURKA and all the major TOMM subunits , TOMM 20 , 22 , 40 and 70 ( Figure 1—figure supplement 2A ) . We therefore searched the N-terminus of AURKA for a Mitochondrial Targeting Sequence ( MTS ) signature . The first 30 or 100 amino acids of AURKA fused to GFP were not detected in mitochondria by immunofluorescence microscopy , or in mitochondrial fractions of HEK293 cells subjected to western blotting ( Figure 1—figure supplement 2B and C ) . This indicates that the N-terminal of AURKA is not sufficient to shuttle GFP to mitochondria and it suggests the absence of a canonical MTS in these regions of the protein . However , an AURKA truncation mutant in which the first 30 amino acids were removed ( AURKA ΔNter ) did not localise to mitochondria by confocal microscopy and western blotting ( Figure 1—figure supplement 3A and B ) and did not interact with TOMM ( Figure 1—figure supplement 3C ) , although catalytically active in vitro towards two AURKA substrates , histone H3 and RALA ( Lim et al . , 2010; Kashatus et al . , 2011; Bertolin et al . , 2016 ) ( Figure 1—figure supplement 3D ) . It has been described that AURKA is needed for RALA localisation to mitochondria , and that the two proteins participate in a common pathway to regulate mitochondrial fission at mitosis ( Kashatus et al . , 2011 ) . We therefore explored whether the import of AURKA inside mitochondria depends on the presence of RALA . We observed that the depletion of RALA does not inhibit the entry of AURKA into mitochondria . However , it impacts the abundance of AURKA to the same extent in total fractions and inside mitochondria . Therefore , the mitochondrial import of AURKA is RALA-independent ( Figure 1—figure supplement 3E ) . Intriguingly , AURKA38 was also detected in cytosolic fractions of HEK293 cells , indicating that the kinase is exported from mitochondria after processing ( Figure 2A ) . To establish whether this feature is intrinsic to the AURKA MTS , we replaced the first 30 amino acids of AURKA with a well characterised , strong MTS derived from cytochrome c oxidase , which is known to be recognised by TOMM and cleaved by PMPCB ( Chacinska et al . , 2009 ) . This yields an exclusively mitochondrial AURKA ( mitoAURKA ) , which was not exported back to the cytosol ( Figure 2B and Supplementary file 3C ) . In addition , mitoAURKA did not localise at the centrosome when compared to normal AURKA ( Figure 2C ) . The export of AURKA to the cytosol is further supported by the observation that ectopically expressed AURKA also shows abundant cytosolic staining ( Figure 2B and C and Figure 1—figure supplement 2A ) , which could include the fraction of AURKA38 exported from mitochondria . Taken together , our data show for the first time that AURKA bears an atypical MTS that is necessary but not sufficient for its intracellular transport to and from mitochondria . We next explored whether the kinase activity of AURKA is involved in its transport to , and function within , mitochondria . To this end , we used a previously published AURKA FRET biosensor , which allows to track the activation of AURKA via its autophosphorylation on Thr288 by FRET/FLIM ( Bertolin et al . , 2016 ) . The biosensor consists of a full-length AURKA carrying the donor FRET fluorophore EGFP at the N-terminus and the acceptor FRET fluorophore mCherry at the C-terminus , under the control of the minimal transcriptional regulatory region of AURKA to ensure the physiological expression of the kinase ( Bertolin et al . , 2016 ) . The AURKA biosensor localises at mitochondria as endogenous AURKA in MCF7 cells does ( Figure 2D , E and see Figure 1A for comparison ) , and the decrease in the lifetime of GFP revealed that it is autophosphorylated on Thr288 and activated in mitochondria . This activation is abolished by the AURKA inhibitor Alisertib ( MLN8237 ) ( Figure 2D ) , an ATP-analogue currently under clinical trials ( Görgün et al . , 2010 ) . Intriguingly , AURKA cleavage inside mitochondria is also supported by the presence of a double band in western blots from mitochondrial fractions of MCF7 cells expressing GFP-AURKA-mCherry ( Figure 2E ) . In these fractions we detected a band at ~110 kDa and corresponding to the full GFP-AURKA-mCherry biosensor construct , and a second band at ~70 kDa potentially representing the cleaved ( imported ) biosensor without the GFP moiety . Given that AURKA undergoes a two-step proteolytic cleavage when entering mitochondria , its activation is likely to occur prior to the cleavage of the MTS by PMPCB as this step removes the entire N-terminus , which comprises the MTS and the FRET donor . In this light , we assessed whether a kinase-dead mutant of AURKA ( AURKA Lys162Met ) could shuttle to mitochondria . Indeed , AURKA Lys162Met was not retrieved in mitochondrial fractions ( Figure 2F ) , indicating that the kinase activity of AURKA is required for its mitochondrial localisation . The enzymatic activity of AURKA at mitochondria suggests that the kinase might be involved in the regulation of mitochondrial functions . To understand the potential role played by AURKA at this compartment , we explored how the over-expression of the kinase may act on two interlinked mitochondrial functions: mitochondrial dynamics and energy production . The mitochondrial dynamics balance has been shown to be crucial in cancer progression , as the mitochondrial network reshapes to meet the increasing energy requirements of cancer cells ( Vyas et al . , 2016; Wai and Langer , 2016 ) . As we detected a role of over-expressed AURKA in the control of mitochondrial energy production , we next evaluated whether AURKA plays a role in mitochondrial dynamics . First , we analysed mitochondrial morphology by TEM in HEK293 cells . Knockdown of AURKA led to mitochondrial elongation: the organelles appeared swollen , but they showed intact cristae and no apparent loss of intramitochondrial content ( Figure 3A ) . Analyses of mitochondrial length and branching ( Koopman et al . , 2005 ) revealed that the silencing of AURKA increases the length of the whole mitochondrial network ( Figure 3—figure supplement 4A ) , confirming previous results obtained with a kinase-dead version of AURKA ( Kashatus et al . , 2011 ) . Intriguingly , TEM analyses showed that mitochondria interconnectivity increases also when AURKA is over-expressed , again with no apparent signs of intramitochondrial content loss ( Figure 3A ) . Under these conditions , analyses of mitochondrial network morphology showed that mitochondria are interconnected and they pack into mitochondrial clusters ( Figure 3—figure supplement 4A ) . We then compared whether the degree of mitochondrial interconnectivity when AURKA is silenced or over-expressed . To this end , we used the diffusion of a photoconvertible Dendra2 targeted to the mitochondrial matrix ( mitoDendra2 ) . This fluorescent protein is photoconverted from green to red with a 405 nm laser , and the diffusion of the red species throughout the network is achieved only if mitochondria are organised in an electrochemical continuum . In parallel , we knocked-out AURKA in a multicellular organism as the fruit fly . In contrast to tumorigenic cells having heterogeneous genotypes and where the complete depletion of AURKA is not achievable , the fruit fly allows to compare the effects of the physiological abundance of AURKA on mitochondria to the ones observed after complete knock-out or overexpression of the kinase . In this model , we measured mitochondrial connectivity in the notum , a monolayer of epithelial cells that display three-dimensional and dynamic spatial organisation of the mitochondrial network . AURKA loss-of-function mutants or those harbouring an AURKA-targeted RNAi and gain-of-function mutants ( overexpression of Drosophila AURKA ) showed more interconnected mitochondria than did controls ( Figure 3B , Figure 3—figure supplement 1B and C , red panels ) . By using the diffusion of red mitoDendra2 , we confirmed that mitochondria are interconnected also in MCF7 cells both when AURKA is silenced or over-expressed ( Figure 3—figure supplement 1D ) . We observed that mitochondrial interconnectivity in the presence of over-expressed AURKA depends on its capacity to be imported/exported , and on the kinase activity of AURKA itself , as the over-expression of AURKA ΔNter , mitoAURKA or the kinase-dead AURKA Lys162Met have no effect on mitochondrial elongation ( Figure 3—figure supplement 2A and B ) . Therefore , the role of AURKA in the regulation of mitochondrial morphology is conserved in flies and humans with no differences in mitochondrial connectivity due to the silencing or the overexpression of the kinase . Given that an increase in mitochondrial connectivity is observed both when AURKA is up- or downregulated , we sought to define the respective molecular mechanisms involved . We first analysed the abundance of the proteins involved in mitochondrial fusion and fission when AURKA was silenced or over-expressed . When AURKA was downregulated , the abundance of MFN1 and OPA1 increased while the level of DNM1L decreased ( Figure 3C and Figure 3—figure supplement 3A ) . We then analysed the phosphorylation state of DNM1L on Ser637 , since DNM1L localises in the cytosol when phosphorylated on this residue and to mitochondrial when dephosphorylated ( Kashatus et al . , 2011; Wai and Langer , 2016 ) . Downregulation of AURKA lead to an increase of Ser637 phosphorylation , corresponding to an increased cytosolic localisation of DNM1L ( Figure 3—figure supplement 3B ) . Under these experimental conditions , increased mitochondrial connectivity at interphase could be reverted by normal AURKA expressed at physiological levels . On the contrary , the cytosolic-only AURKA ΔNter did not rescue mitochondrial elongation ( Figure 3—figure supplement 3C ) , although this protein retains its catalytic activity and its capacity to phosphorylate RALA on Ser194 ( Figure 3—figure supplement 3D ) . The phosphorylation of DNM1L on Ser616 was previously shown to play an active role in promoting organelle fission at mitosis ( Kashatus et al . , 2011 ) . On the contrary , the phosphorylation of DNM1L on Ser637 observed at interphase and after silencing of AURKA indicates a lack of fission , which then results in increased mitochondrial connectivity . In physiological conditions , two distinct pathways appear to regulate mitochondrial fragmentation according to the cell cycle phase . At mitosis , fission mechanisms require translocation of RALA and its effector RALBP1 to the mitochondrion ( Kashatus et al . , 2011 ) , whereas this step is dispensable at interphase . When AURKA was over-expressed , we observed the increase of the short isoform of OPA1 ( S-OPA1 ) , a phenomenon previously reported to be directly caused by an increased mitochondrial respiratory chain activity ( Mishra et al . , 2014 ) . In addition , we also observed a small but significant decrease of the levels of DNM1L and its receptor MFF ( Figure 3D and Figure 3—figure supplement 3D ) . Under these conditions , the phosphorylation of DNM1L on Ser637 remained globally unaltered ( Figure 3—figure supplement 3E ) . While analysing mitochondria in tumorigenic cells , we observed that nearly 60% of cells with ectopic AURKA showed interconnected mitochondria clustered in the perinuclear region ( Figure 3—figure supplement 3F ) . These structures are similar to insoluble mito-aggresomes , aggregates of mitochondria which cannot be degraded ( Driscoll and Chowdhury , 2012 ) . As mito-aggresomes , we found these AURKA-positive mitochondrial aggregates to be SDS-insoluble as indicated by dot-blot filter retardation assays ( Figure 3—figure supplement 3G ) . In vivo , the over-expression of the fly homologue of DNM1L – Drp1 – in flies over-expressing AURKA rescued mitochondrial interconnectivity analysed with mitoDendra2 ( Figure 3—figure supplement 3H ) . Furthermore , we retrieved a direct interaction between AURKA , DNM1L and MFF by localised decrease of the lifetime of AURKA-GFP on mitochondria in FRET/FLIM analyses ( Figure 3E ) . A similar decrease in GFP lifetime was not observed between over-expressed AURKA and the fusion protein MFN2 or the mitochondrial protein SNPJN2 not involved in mitochondrial dynamics ( Figure 3—figure supplement 3H ) . This further corroborates the specificity of the interaction between AURKA , DNM1L and DMFF analysed by FRET/FLIM . Accordingly , the interaction between AURKA , DNM1L and MFF in the AURKA interactome was also detected by nanoLC-ESI MS/MS ( Supplementary file 2 ) . Together , over-expressed AURKA directly interacts with the fission proteins DNM1L and MFF and drives mitochondrial elongation . To validate our data in a cancer cell context , we examined the morphology of the mitochondrial network and its correlation to the levels of AURKA expression in four breast cancer cell lines . Hs578T and MDA-MB-231 cells show low expression levels of AURKA , whereas MDA-MB-468 and T47D express AURKA at higher levels ( Figure 4A ) . In all these cell lines , we also retrieved an AURKA-positive signal at mitochondria by immunoblotting . When looking at the morphology of the mitochondrial network , mitochondria appeared more fragmented in Hs578T and MDA-MB-231 cells , and more elongated in MDA-MB-468 and T47D ( Figure 4B ) . To correlate this phenotype with the abundance of AURKA , we inhibited the kinase with MLN8237 . MLN8237 had no effect on mitochondrial length in MDA-MB-468 and T47D cells , which express high levels of AURKA and where the mitochondrial network is already dramatically interconnected per se . Conversely , MLN8237 increased mitochondrial length and branching in Hs578T and MDA-MB-231 cells , where the abundance of AURKA is low and mitochondria appear fragmented in basal conditions ( Figure 4B ) . This is consistent with what observed in MCF7 cells upon the depletion of AURKA by siRNA . Together , AURKA plays two opposing functions in mitochondrial dynamics according to its abundance in the cell: contributing to organelle fission when expressed under physiological conditions , and directly increasing mitochondrial fusion when over-expressed . These roles of AURKA are conserved in four carcinoma cell lines , further increasing the relevance of these results . We then analysed the mitochondrial energy capacity in HEK293 and MCF7 cells , in the presence and absence of over-expressed AURKA . We evaluated key parameters as the abundance of mitochondrial respiratory chain complexes , oxygen consumption for ATP production , the mitochondrial membrane potential and possible sources of mitochondrial stress as the activation of autophagy and of the Ubiquitin-Proteasome System ( UPS ) . Among the levels of steady-state respiratory complexes , western blotting analysis revealed that the levels of the respiratory complex IV subunits increased in the presence of over-expressed AURKA ( Figure 5A ) . It has been reported that increased abundance and activity of the respiratory complex IV are part of a mitochondrial-specific signature of epithelial cancers , which mainly rely on oxidative phosphorylation for ATP production ( Whitaker-Menezes et al . , 2011; Vyas et al . , 2016 ) . In this light , we observed that the oxygen consumption rate ( OCR ) – a measure of mitochondrial respiration – was increased when AURKA was over-expressed in HEK293 cells ( Figure 5B ) ( Whitaker-Menezes et al . , 2011; Vyas et al . , 2016 ) . Although only the steady-state levels of complex IV increased significantly in the presence of ectopic AURKA , the analysis of the interactome of AURKA by proteomics showed that AURKA directly interacts with multiple subunits of all respiratory complexes in HEK293 cells ( Supplementary file 2; Figure 5—figure supplement 1 ) . This reinforces the conclusion that over-expressed AURKA globally acts on the mitochondrial respiratory chain to increase ATP production . We then analysed mitochondria-related stress levels in the presence and absence of AURKA by flow cytometry . The over-expression of AURKA increased the proportion of cells exhibiting ongoing autophagy ( Figure 5C ) . On the contrary , the downregulation of AURKA increased the activity of the ubiquitin-proteasome system ( Figure 5D ) , which has been proposed to be a complementary system of autophagy for the degradation of selective mitochondrial proteins ( Zhu et al . , 2010 ) . To specifically evaluate the turnover of mitochondria , we calculated the red/green ratio of MitoTimer ( Ferree et al . , 2013 ) . Increasing levels of red mitoTimer were observed in cells transfected with an AURKA-shRNA ( Figure 5E ) , indicating that mitochondrial turnover was attenuated under these conditions . The increased red mitoTimer was specifically due to mitochondrial turnover , as we did not detect significant reactive oxygen species variations when downregulating or over-expressing AURKA ( Data not shown ) . Mitochondrial membrane potential ( ΔΨ ) – another indicator of mitochondrial functionality – measured with JC-1 ( Figure 5F ) or with Tetramethylrhodamine Methyl Ester ( TMRM ) ( Figure 5G ) decreased upon AURKA knockdown , further confirming that mitochondria are defective in the absence of AURKA . Although mitochondria are depolarised after AURKA knockdown , no global effect on the mitochondrial oxygen consumption rate was observed under these conditions ( Figure 5H ) , despite both AURKA knockdown and over-expression have an impact on cell viability as previously published ( Figure 5I ) ( Zhang et al . , 2004; Bavetsias and Linardopoulos , 2015 ) . In addition , the mitochondrial ATP production did not differ from that of control cells when AURKA was inhibited with MLN8237 , corroborating the finding that both silencing and inhibition of AURKA do not alter mitochondrial ATP production ( Figure 5L ) . In conclusion , our results indicate that AURKA maintains mitochondrial fission when expressed at physiological levels and that mitochondrial interconnectivity in the absence of AURKA is a consequence of a lack of fission . This results in the mere accumulation of elongated mitochondria without any increase in the energetic capabilities of the mitochondrial network . On the contrary , over-expressed AURKA actively enhances ATP production by promoting mitochondrial interconnectivity . These data reveal a novel role of AURKA in the control of mitochondrial bioenergetics , by acting on the mitochondrial respiratory chain and on mitochondrial functionality ( Figure 5M ) .
Over-expression of AURKA is observed in many epithelial cancers . Increased copy number of the AURKA gene region is generally associated with an aggressive disease and poor patient survival . The AURKA gene region is located on chromosome 20 , and its amplification includes the enhanced expression of additional genes ( e . g . genes regulating cell cycle progression , and the most well-described AURKA interactor TPX2 ) ( Belt et al . , 2012; Sillars-Hardebol et al . , 2012 ) . In addition , the overexpression of AURKA has been linked with chromosomal instability ( Baba et al . , 2009 ) . These events are common in different cancer types as in ovarian , pancreatic , lung and colon cancers and lead to bad prognosis . For instance , the increased copy number of AURKA is associated with the evolution of colorectal polyp into carcinoma ( Carvalho et al . , 2012 ) . In breast cancer , the overexpression of AURKA is also linked to poor survival and it is associated with the overexpression of the human growth factor receptor 2 ( HER2 ) and progesterone receptor ( Nadler et al . , 2008 ) . Although epithelial cancers are non-glycolytic tumours and use the OXPHOS chain to produce ATP ( Whitaker-Menezes et al . , 2011 ) , none of the above-mentioned studies took into account mitochondrial dysfunctions caused by or appearing in the presence overexpressed AURKA . Our study is thus pioneer in correlating for the first time this multifaceted kinase and mitochondrial physiology . In addition to its well-characterised roles in mitosis , new functions of AURKA during interphase are regularly discovered ( Mori et al . , 2009; Bertolin et al . , 2016; Zheng et al . , 2016 ) . We here demonstrated that AURKA is imported in the mitochondrial matrix . To reach this compartment , AURKA physically interacts with the TOMM complex , the major entry gate for mitochondrial proteins . Once it enters mitochondria , AURKA is cleaved in a two-step process to become a fully mature mitochondrial protein , potentially capable of interacting with multiple mitochondrial partners as the mitochondrial respiratory chain subunits . We discovered that the signal required for the import of AURKA into mitochondria is located within the first 36 amino acids of the kinase . Conventionally , MTS are incapable of shuttling to mitochondria when fused at the C-terminus of a fluorophore ( Chacinska et al . , 2009 ) . AURKA MTS is indeed atypical , as its mitochondrial import is not blocked by the presence of a GFP at its N-terminus . In addition , the inability of this MTS to shuttle a generic GFP to mitochondria further suggest that the AURKA MTS may belong to a new class of weak mitochondrial targeting signals ( Matthews et al . , 2010 ) , previously reported to require a specific folding conformation or post-translational modification to shuttle to mitochondria ( Karniely and Pines , 2005 ) . The hypothesis that centrosomal proteins play additional roles at mitochondria has already been raised ( Moore and Golden , 2009 ) . It has been shown that the mitochondrial protein SUCLA2 , which catalyses the conversion of succinyl CoA into succinate inside mitochondria , has a mitochondrial and centrosomal double localisation in Drosophila ( Hughes et al . , 2008 ) . It was shown that centrosomal SUCLA2 regulates the number and the stability of centrosomes , and this raises the fascinating hypothesis that mitochondrial proteins could in turn play roles at the centrosome under certain conditions ( e . g . the cell cycle phase ) . Given that AURKA is preferentially a centrosomal protein now shown to directly regulate mitochondrial functions , it is tempting to speculate that this mitochondria-to-centrosome crosstalk could also work in a retrograde manner from the centrosome to mitochondria , with centrosomal proteins as AURKA also regulating mitochondrial functions . In this light , it will be interesting to explore whether the centrosomal and mitochondrial pools of AURKA are spatiotemporally connected . To this end , further studies are required to establish the exact molecular mechanisms that allow the first 36 amino acids of AURKA to act as a MTS only when bound to the rest of the protein . We then searched for potential roles of AURKA at mitochondria and for impairments of mitochondrial functionality when AURKA is over-expressed . We discovered that the over-expression of AURKA enhances mitochondrial ATP production . In exploring the mitochondrial interactome of AURKA during interphase , we discovered that one out of five interactors of AURKA is a mitochondrial protein as revealed by nanoLC-ESI MS/MS analyses ( Supplementary file 2 , Figure 5—figure supplement 1 ) . Proteins regulating energy metabolism in the cell , including multiple subunits of the mitochondrial respiratory chain , were found to significantly interact with AURKA at interphase ( Supplementary file 2 , Figure 5—figure supplement 1 ) . This strongly supports a role of AURKA in the control of the mitochondrial respiratory chain functionality . It is known that an interconnected mitochondrial network favours ATP production through mixing of the intramitochondrial content , which also counteracts the effects of deleterious mtDNA mutations in vivo ( Nakada et al . , 2001 ) . Interconnected mitochondrial networks have been proposed to act as energy-transmitting cables , delivering energy to parts of the cell in which oxygen for mitochondrial respiration is low ( Westermann , 2012 ) . Mitochondrial fusion is also stimulated in selected stress conditions as starvation , helping the cell to cope with increasing energy demands ( Tondera et al . , 2009 ) . The over-expression of AURKA represents a mitotic stress paradigm with centrosome abnormalities , chromosome misalignment , aberrant DNA inheritance at cell division and apoptosis ( Zhang et al . , 2004; Nikonova et al . , 2013 ) . Therefore , it is not surprising that mitochondria under these conditions modify their functionality beyond mitosis as well , adapting to stress by increasing their connectivity and the production of ATP during interphase . On the contrary , the increased mitochondrial connectivity observed in the absence of AURKA or when the kinase is pharmacologically inhibited does not lead to an increased ATP production . However , the connectivity of the mitochondrial network under these conditions resembles the one observed when AURKA is overexpressed . As AURKA drives mitochondrial fission when expressed at physiological levels , the absence of AURKA or the inhibition of its catalytic activity lead to the appearance of interconnected mitochondrial networks . In the absence of an active AURKA , the paradigms regulating mitochondrial fission are therefore limited and fusion remains the only mechanism left to regulate mitochondrial morphology , as previously proposed in conditions of fission inhibition ( Hoitzing et al . , 2015 ) . In this light , we indeed demonstrate that the molecular mechanisms used by AURKA to regulate mitochondrial dynamics are distinct according to the expression levels of the kinase . When AURKA is silenced , fission is inhibited . In conditions where AURKA is overexpressed , fusion is enhanced and so is the energetic capability of the entire mitochondrial network . In conclusion , the modifications induced by AURKA to mitochondrial morphology are multifaceted and , in this context , the simple interconnectivity of the mitochondrial network is not a direct readout of the energetic capabilities of mitochondria . The AURKA interactome and FRET by FLIM analyses show that AURKA directly interacts with DNM1L and MFF . A direct interaction of cytosolic AURKA with proteins regulating mitochondrial dynamics is conceivable , as they are mainly located on the OMM . Nevertheless , the physical interaction of the kinase with multiple components of the mitochondrial respiratory chain , located on the Inner Mitochondrial Membrane , would require the kinase to be imported into mitochondria to ultimately promote ATP production . This is also in agreement with an increased abundance of the Inner Mitochondrial Membrane protein S-OPA1 induced by AURKA overexpression . Of note , increased S-OPA1 is a previously characterised hallmark of augmented mitochondrial respiratory chain activity ( Mishra et al . , 2014 ) , which further reinforces our conclusion that over-expressed AURKA directly potentiates the functionality of the mitochondrial respiratory chain . However , further studies are required to understand how AURKA spatiotemporally interacts with its multiple partners and if different sub-mitochondrial pools of AURKA are capable of driving changes in mitochondrial morphology and in energy production . Recent findings showed that mitotic AURKA promotes mitochondrial fission through the phosphorylation of RALA on Ser194 in the cytosol ( Kashatus et al . , 2011 ) . Through this modification , RALA can shuttle to mitochondria to ensure the correct segregation of these organelles to daughter cells . We therefore sought to understand whether the role played by AURKA at mitochondria during interphase falls within the interplay between AURKA and RALA . First , the kinase is imported into mitochondria regardless of RALA . Second , the MTS of AURKA is the only portion of the protein strictly required for the pro-fission role of AURKA in physiological conditions . Indeed , a version of AURKA devoid of this fragment is unable to induce organelle fragmentation although it is fully capable of phosphorylating RALA on Ser194 . Therefore , our study demonstrates that AURKA can impact on mitochondrial functions following two parallel pathways: a RALA-independent pathway during interphase , and by interacting with RALA at mitosis ( Kashatus et al . , 2011 ) . Although we demonstrated the existence of two different pathways taken by AURKA , further studies are necessary to better characterise their molecular players and their potential interplay . In conclusion , we propose the mitochondrial pool of AURKA regulates the fusion of interconnected organelles and , by doing so , it controls ATP production ( Figure 6 ) . Mitochondria with high metabolic capacity might escape turnover through fusion mechanisms and thus sustain the high metabolic needs of cancer cells , potentially representing a selective advantage for epithelial cancer progression . Targeting mitochondrial hyperfunctionality together with AURKA inhibition might therefore represent an innovative approach in the development of anti-cancer treatments .
Unless purchased from Addgene , DNA constructs were generated using Gibson Assembly Master Mix ( New England Biolabs ) and T4 DNA ligase ( Thermo Fisher Scientific ) . All restriction enzymes were purchased from Thermo Fisher Scientific . All cloning reactions were verified on a 3130 XL sequencer ( Applied Biosystems ) . All site-directed mutagenesis reactions were performed by QuickChange site-directed mutagenesis ( Stratagene ) . Vectors carrying AURKA ΔNter were constructed by removing the first 30 aminoacids of AURKA; mitoAURKA was constructed by adding the MTS of cytochrome c oxidase to AURKA ΔNter . The complete list of plasmid used in the study is reported in Supplementary file 3 . Mycoplasma-free MCF7 ( HTB-22 ) and HEK293 ( CRL-1573 ) , cells were purchased from the American Type Culture Collection and grown in Dulbecco’s modified Eagle’s medium ( DMEM , Sigma-Aldrich ) supplemented with 10% foetal bovine serum ( GE Healthcare ) , 1% L-glutamine ( GE Healthcare ) and 1% penicillin–streptomycin ( GE Healthcare ) . Hs578T , MDA-MB-231 , MDA-MB-468 and T47D cells were a kind gift of P . Legembre ( CLCC Eugène Marquis , Rennes ) and were grown in Dulbecco’s modified Eagle’s medium ( DMEM , Sigma-Aldrich ) supplemented with 10% foetal bovine serum ( GE Healthcare ) , 1% L-glutamine ( GE Healthcare ) and 1% penicillin–streptomycin ( GE Healthcare ) . For live microscopy experiments , cells were incubated in phenol red-free Leibovitz’s L-15 medium ( Thermo Fisher Scientific ) supplemented with 20% foetal bovine serum , 1% L-glutamine and 1% penicillin–streptomycin . All live microscopy experiments were performed at 37°C in Nunc Lab-Tek II Chamber slides ( Thermo Fisher Scientific ) . Validated siRNA against PMPCB was purchased from Dharmacon ( L-004747- 00–0005 ) , AllStars negative control ( SI03650318 ) and validated siRNA against RALA ( SI02662835 ) were purchased from Qiagen; the siRNA against AURKA was synthesised as previously described ( Bertolin et al . , 2016 ) ( sequence: 5’-AUGCCCUGUCUUACUGUCA-3’ ) and purchased from Eurogentec . The AURKA-specific ( SHCLNG-NM_003600 ) and non-targeting control ( SHC002 ) shRNAs were purchased from Sigma-Aldrich . Plasmids , siRNAs and shRNAs were transfected by the calcium phosphate method or with Lipofectamine 2000 ( Thermo Fisher Scientific ) according to the manufacturer’s instructions . Cells were plated at 70% confluence in 96-well cell plates for plate reads , in 24-well cell plates for immunocytochemistry , or on 10 or 15 cm2 petri dishes for total cell lysates and subcellular fractionation . Cells were harvested , fixed or imaged 48 hr after transfection unless otherwise indicated . MLN8237/Alisertib was purchased from Selleck Chemicals and used at a final concentration of 250 nM for 10 min or of 100 nM for 3 hr before imaging , cell fixation or harvesting . Tetramethylrhodamine methyl ester perchlorate ( TMRM , Thermo Fisher Scientific , 50 nM ) was incubated with the cells for 30 min at 37°C in phenol-free medium before imaging . AurA2G and aurA3A indels were generated using the approach described in ( Kondo and Ueda , 2013 ) . Briefly , gRNAs targeting exon 2 ( GGCGCTTTGATCAGGAAGCCAGG ) or exon 3 ( GGAAAAGGAATCCCAGTTCGTGG ) of AurA were cloned into the pBFv-U6 . 2 vector . Following molecular validation by sequencing , the pBFv-U6 . 2 . exon2 gRNA or pBFv-U6 . 2 . exon3 gRNA plasmids were injected into y1 v1 P{nos-PhiC31C\int . NLS}X; P{caryP}attP40 ( stock BDSC25709 ) by Rainbow Transgenic Flies , Inc . The resulting male transformants y1 v1 P{nos-PhiC31C\int . NLS}X; P{U6 . 2-exon2 ( or exon 3 ) gRNA} attP40 were balanced over CyO using y2 cho2 v1; Sp/CyO stock . Male y2 cho2 v1; P{U6 . 2-exon2 ( or exon 3 ) gRNA}attP40/CyO flies were next crossed with female y2 cho2 v1; P{nos-cas9}attP40/CyO flies . The resulting male y2 cho2 v1; P{nos-cas9}attP40/P{U6 . 2-exon2 ( or exon 3 ) gRNA}attP40; aurA ( indel ? ) flies were crossed with w; krIf/CyO; MKRS/TM6 , Tb , Hu females . Single males of the genotype w; P{nos-cas9}attP40/CyO; AurA ( indel ? ) /TM6 , Tb , Hu were crossed with w; krIf/CyO; MKRS/TM6 , Tb , Hu females . The resulting w; rIf/CyO; AurA ( indel ? ) /TM6 , Tb , Hu stock was established and characterised by sequencing . AurA2G is a 17 bp deletion that induces a frame-shift and eventually a STOP codon , encoding a 100-aa protein containing the first 61 aa of AurA . AurA3A is a 7 bp deletion encoding a 180-aa protein containing the first 177 aa of AurA . Both alleles fail to complement aurAST null allele ( Moon and Matsuzaki , 2013 ) , and no AurA protein was detected by western blot . UAS-mitoDendra2 flies were established by adding the cytochrome c oxidase MTS to the Dendra2 fluorophore . The remaining strains used in this study are listed in Supplementary file 4 . Drosophila melanogaster crossings were set up and grown at 25°C . All crossings and the corresponding Fig . panels are listed in Supplementary file 5; w1118 pupae were used as wild-type controls for all experiments . Pupae were collected as white pupae , aged for 16 hr at 25°C and mounted on glass slides prior to imaging . All images collected in this study were acquired from epithelial cells of the dorsal thorax ( notum ) at room temperature . Total protein fractions were obtained by lysing cells in 50 mM Tris-HCl ( pH 7 . 5 ) , 150 mM NaCl , 1 . 5 mM MgCl2 , 1% Triton X-100 , and 0 . 5 mM dithiothreitol ( DTT ) supplemented with 0 . 2 mM Na3VO4 , 4 mg/ml NaF , 5 . 4 mg/ml β-glycerophosphate and protease inhibitors ( Complete Cocktail , Roche ) followed by centrifugation at 13 , 000 g for 20 min at 4°C . Isolated mitochondrial fractions were obtained by differential centrifugation as previously described ( Bertolin et al . , 2013 ) and were digested with 1 μg of trypsin ( Sigma-Aldrich ) per 10 µg of protein at 37°C for the indicated incubation times ( Figure 1F ) . Digestion of the inner mitochondrial membrane was performed by adding 2 µg/µl digitonin or 3% Triton X-100 to trypsin-digested fractions . Insoluble/soluble mitochondrial protein fractions were obtained by alkaline extraction . Briefly , mitochondrial fractions were incubated on ice with 200 mM Na2CO3 or NaCl as a control followed by incubation of the pellet and soluble fractions in 600 mM sorbitol ( Sigma-Aldrich ) and 20 mM HEPES-KOH ( Sigma-Aldrich , pH 7 . 4 ) . Proteins were precipitated in 12% trichloroacetic acid ( Sigma-Aldrich ) , washed three times with acetone , dried and resuspended in Laemmli sample buffer . Cytosolic fractions were obtained from mitochondria-free protein fractions using Amicon Ultra 4 ml filters for protein purification and concentration ( 10 kDa cutoff , Merck Millipore ) according to the manufacturer’s instructions . Protein purification and in vitro kinase assays were performed as described in ( Bertolin et al . , 2016 ) . All protein fractions were assayed using the Bradford reagent ( Bio-Rad ) and then boiled in Laemmli sample buffer , resolved by SDS–PAGE , transferred onto a nitrocellulose membrane ( GE Healthcare ) and analysed by western blotting . Dot-blot filter retardation assays were performed in a 96-well BioDot microfiltration unit ( Bio-Rad ) using a 0 . 22 µm cellulose acetate membrane ( Dutscher ) . After treatment , the samples were resuspended in 2% SDS , loaded onto the membrane , filtered and washed twice with 0 . 1% SDS . The list of primary antibodies is in Supplementary file 6 . Secondary horseradish-peroxidase-conjugated antibodies ( anti-mouse and anti-rabbit ) were purchased from Jackson ImmunoResearch Laboratories; anti-rat antibodies were purchased from Bethyl Laboratories . The membranes were incubated with commercially available ( Pierce ) or homemade enhanced chemiluminescence substrate as described in ( Bertolin et al . , 2016 ) . Chemiluminescence signals were captured on film ( CP-BU new , Agfa Healthcare ) , developed using CURIX 60 developer ( Agfa Healthcare ) and quantified with ImageJ software ( NIH ) . The relative abundance of specific bands of interest was calculated by normalising it towards the abundance of loading controls and indicated in each graph . Cells were trypsinised , resuspended in growth medium , and placed in the respiratory chamber of an Oroboros Oxygraph-2k ( WGT ) . Cellular respiration was determined under basal conditions in the presence of oligomycin ( 1 µg/ml , Sigma-Aldrich ) to estimate leakage and/or in the presence of increasing amounts ( 2 . 5–5 µM ) of CCCP ( Sigma-Aldrich ) to obtain maximal respiration . The respiration reserve capacity was calculated by subtracting the basal respiration from the maximal respiration . The mitochondrial respiratory control corresponded to the basal/leak ratio . ‘ATP’ indicates the O2 consumption used for ATP synthesis . Mitochondrial respiration was inhibited by the addition of 1 mM potassium cyanide ( KCN , Sigma-Aldrich ) . AURKA-GFP isolated by affinity-purification in whole cell extracts or mitochondrial fractions were resolved by SDS-PAGE using 4–12% Criterion XT Bis-Tris gradient gel ( Bio-Rad ) and stained with Sypro Ruby ( Bio-Rad ) . Bands corresponding to AURKA46 , AURKA43 and AURKA38 were extracted from gels and processed for in-gel digestion . Alternatively , AURKA-GFP affinity-purification extracts were subjected to in-solution digestion . Peptide samples were separated by online reversed-phase ( RP ) nanoscale capillary liquid chromatography ( nanoLC ) and analyzed by electrospray mass spectrometry ( ESI MS/MS ) . The experiments were performed with a Dionex UltiMate 3000 nanoRSLC chromatography system ( Thermo Fisher Scientific/Dionex Softron GmbH ) connected to a Orbitrap Fusion Tribrid mass spectrometer ( Thermo Fisher Scientific ) equipped with a nanoelectrospray ion source . Mass spectra data generated by the Orbitrap Fusion Tribrid instrument ( * . raw files ) were analyzed with Byonic version 2 . 12 . 0 ( Protein Metrics , San Carlos , USA ) using the optimal search parameters ( mass tolerances and post-translational modifications ) generated by Preview version 2 . 12 . 0 ( Protein Metrics , San Carlos , USA ) . Precursor mass tolerance was set to 2 ppm and the fragment mass tolerance was set to 0 . 2 Da . Both precursor and fragments m/z measurements were recalibrated based on Preview calculations . The following mass additions were used as variable modifications: oxidation of methionine , histidine and tryptophan [+15 . 9949 Da] , dioxidation of methionine and tryptophan [+31 . 9898 Da] , deamidation of asparagine and glutamine [+0 . 9840 Da] , N-terminal protein acetylation [+42 . 0105 Da] , phosphorylation of serine and threonine [+79 . 9663 Da] , glutamine conversion to pyroglutamate [−17 . 0265] and glutamate conversion to pyruglutamate [−18 . 0105 Da] . Carbamidomethylation of cysteines [+57 . 0214 Da] was set as a fixed modification . A semi-specific trypsin digestion setting allowing for N- or C-ragged peptides was specified . Byonic was used to search the MS/MS data against the Uniprot human reference proteome ( 71 657 entries as of May 23 , 2017 ) complemented with a list of common contaminants maintained by Protein Metrics and concatenated with the reversed version of all sequences ( decoy mode ) . The Byonic automatic score cutoff option was specified to maintain the false discovery rate ( FDR ) for peptide spectrum matches ( PSMs ) in the range of 0–5% . The FDR for protein identifications was set to 1% . The cellular localizations and biological functions of identified proteins were further analyzed based on information available from the Gene Ontology ( GO ) classification tool available in DAVID Bioinformatics Resources ( https://david . ncifcrf . gov/ ) ( Huang et al . , 2009a; Huang et al . , 2009b ) . All peptide spectrum matches corresponding to AURKA peptides identified by Byonic were extracted from the peptide dataset for the identification of putative mitochondrial processing peptidases cleavage sites into AURKA . AURKA peptides containing N- and C-ragged termini were collected and classified relative to AURKA amino acid sequence . Only high quality peptides were considered by applying a stringent p value cutoff of 0 . 001 ( i . e . a -log ( p value ) of 3 . 0 ) . The number of occurrences of each non-tryptic cleavage sites was calculated and plotted relative to AURKA amino acid sequence . For the CRAPome filtering , protein datasets of affinity-purified AURKA-GFP-interacting proteins generated by LC-MS/MS analysis were grouped together to establish a repertoire of putative AURKA interactors ( Supplementary Material ) . The protein list was filtered against CRAPome ( http://crapome . org/ ) , a contaminant repository database that scores high-confidence interaction data from AP-MS experiments . A group of 30 control AP-MS experiments based on the isolation of GFP-tagged proteins by anti-GFP antibodies coupled to magnetic Dynabeads was used as a reference database to score AURKA-GFP interactions . The CRAPome primary scores ( FC-A ) distribution was plotted to perform statistical analysis . The upper quartile , a value that cuts off 75% of protein identifications with the lowest FC-A scores was used as a cutoff criteria . Only the top-scoring 25% ( 641 out of 2601 protein IDs ) proteins were accepted as putative AURKA-interacting proteins and used for Gene Ontology analysis . Cells were fixed in 4% paraformaldehyde ( Sigma-Aldrich ) , stained using standard immunocytochemical procedures and mounted in ProLong Gold Antifade reagent ( Thermo Fisher Scientific ) . The antibodies used were: primary monoclonal mouse anti-AURKA , 1:20 ( Cremet et al . , 2003 ) ; anti-GFP , 1:1000 ( Sigma-Aldrich , 11814460001 ) ; polyclonal rabbit anti-PMPCB , 1:500 ( Proteintech , 16064–1-AP ) ; and secondary anti-mouse or anti-rabbit antibodies conjugated to Alexa 674 at a 1:500 dilution , Alexa 555 or 488 both at a 1:5000 dilution ( Thermo Fisher Scientific ) . Cells displaying mitochondrial clusters were scored after viewing in a DMRXA2 microscope ( Leica ) equipped with a 63X oil-immersion objective ( numerical aperture ( NA ) 1 . 32 ) and driven by MetaVue software ( Molecular Devices ) . Multicolour images of cultured cells were acquired with a Leica SP8 inverted confocal microscope ( Leica ) and a 63X oil-immersion objective ( NA 1 . 4 ) driven by LAS software or alternatively with a BX61WI FV-1000 confocal microscope ( Olympus ) driven by Olympus FV-1000 software and equipped with a 60X oil- immersion objective ( NA 1 . 35 ) . Multicolour images of Drosophila pupae were acquired with an SPE DM 5500 microscope ( Leica ) and a 63X oil-immersion objective ( NA 1 . 4 ) . The excitation and emission wavelengths for GFP/Alexa 488 were 488 and 525/50 nm , respectively; for mCherry/Alexa 555 , they were 561 and 605/70 nm . GFP was used as a FRET donor in all experiments , and its decrease was measured by FLIM microscopy as in ( Bertolin et al . , 2016 ) . MitoDendra2 photoconversion was performed on a region of interest ( ROI ) with a 405 nm laser at 0 . 25% power for 5 msec on an inverted Leica SP8 confocal microscope . Images of the green ( λex: 490 nm; λem: 507 nm ) and red ( λex: 553 nm; λem: 573 nm ) species of mitoDendra2 were then acquired using a 63X oil-immersion objective ( N . A . 1 . 4 ) and a hybrid detector , driven by the LAS software . The total number of red objects present 120 s after photoconversion was normalised to the number of red objects in the ROI in the first image obtained after the photoconversion procedure ( 5 s ) . Fluorescence co-localisation was calculated with the JaCoP plugin ( Bolte and Cordelières , 2006 ) of the ImageJ software after applying an automatic threshold mask to the confocal images; AURKA-positive mitochondria in each cell cycle phase were calculated by normalising AURKA and PMPCB-co-localising objects to the total number of AURKA-positive objects . Mitochondrial aspect ratio and form factor were calculated from confocal images as in ( Koopman et al . , 2005 ) . TMRM and mitoTimer fluorescence were acquired in a FluoSTAR OMEGA plate reader ( BMG Labtech ) equipped with 485/520 490 nm and 540/615 nm excitation/emission filters . When overexpressing AURKA or one of its variants for confocal microscopy , non-fluorescent 6xHis-tagged AURKA was used instead of GFP-AURKA where indicated . For conventional electron microscopy , the cells were rinsed with 0 . 15 M sodium cacodylate and fixed by adding 2 . 5% glutaraldehyde for 1 hr . After fixation , the cells were rinsed several times with 0 . 15 M sodium cacodylate and post-fixed with 1 . 5% osmium tetroxide for 1 hr . After further rinsing , the samples were dehydrated in increasing concentrations of ethanol ( 50 , 70 , 90% and 100% v/v ) . The cells were gradually infiltrated with increasing concentrations of epoxy resin ( 30 , 50 , 70% v/v in ethanol ) for a minimum of 3 hr per step . The samples were then incubated overnight in pure epoxy resin before continuing the infiltration procedure with a two-step incubation in 2 , 4 , 6-Tris ( dimethylaminomethyl ) phenol ( DMP30 , Sigma-Aldrich ) -epoxy resin , first for 3 hr and then at 60°C for 24 hr to polymerise the samples en bloc . Ultra-thin sections of 80 nm were then cut from the blocks using a UCT ultramicrotome ( Leica ) , placed on grids , and post-stained with uranyl acetate for 30 min and with lead citrate for 20 min . For immunoelectron microscopy , cells were centrifuged for 5 min at 800xg , recovered and rapidly fixed in 4% paraformaldehyde and 0 . 1% glutaraldehyde in 0 . 1 M phosphate buffer ( PB ) for 4 hr as previously described ( Slot and Geuze , 2007 ) . The cells were rinsed in PB and suspended in gelatin ( 12% wt/vol ) at 37°C for 10 min . After solidification on ice , the cell blocks were cut and immersed in 2 . 3 M sucrose at 4°C overnight . The blocks were then mounted on a pin holder and placed in a UC7 cryo-ultramicrotome ( Leica ) . Rapid trimming was performed using a 90°C trim tool ( DTB20 , Diatome AG ) at −80°C to determine a region of interest . Ultrathin cryosections ( 70–90 nm ) were cut at −120°C using a dry diamond knife ( DCIMM 3520 , Diatome AG ) , picked up with a mixture ( 1:1 vol/vol ) of 2 . 3 M sucrose and 2% wt/vol methylcellulose and transferred to formvar-coated copper or nickel grids . The grids were subjected to standard immunolabelling procedures ( Slot and Geuze , 2007; Griffiths et al . , 1984; Nicolle et al . , 2015 ) before a final contrast on ice with a mix of 2% wt/vol methylcellulose and 4% wt/vol uranyl acetate in a ratio of 8:2 . The combinations of primary and secondary antibodies used are listed in Supplementary file 7 . The grids used for electron microscopy were examined at 120 kV with a JEOL 1400 ( Peabody ) transmission electron microscope equipped with an SC 1000 camera ( Gatan Orius ) . Mitochondrial length , lysosomal abundance and number of gold beads were scored using the ImageJ software . Analyses of autophagy , apoptosis , mitochondrial membrane potential and proteasome peptidase activity were performed on a BD Accuri C6 flow cytometer ( BD Biosciences ) . Annexin V-FITC/PI apoptosis detection kit was used as described by the manufacturer ( Thermo Fisher Scientific ) . MAP1LC3A activation was measured using the FlowCellect Autophagy LC3 Antibody-based Assay Kit ( Merck Millipore ) . Mitochondrial inner membrane potential was measured with the JC-1 probe ( Thermo Fisher Scientific ) as previously described ( Agier et al . , 2012 ) . The peptidase activity of proteasomes was monitored using the fluorogenic peptide succinyl-Leu-Leu-Val-Tyr-7-amido-4-methylcoumarin , LLVY-AMC ( Sigma-Aldrich ) ( Bulteau et al . , 2006 ) . Two-way ANOVA tests were employed to compare two variables among multiple conditions , and one-way ANOVA when just one variables needed to be tested among multiple conditions . Student’s t-test was employed to compare two conditions . Statistical tests were performed after testing data for normality . Two-way ANOVA and the Holm-Sidak method were used to compare the the effect of siRNAs and AURKA isoforms on the relative mitochondrial abundance of AURKA ( Figure 1C , Figure 1—figure supplement 3E ) , the effect of the pharmacological treatment and the mitochondrial respiratory parameter on mitochondrial respiration ( Figure 5L ) the effect of pharmacological treatment and the fluorescence protein on lifetime ( Figure 2D ) , the effect of pharmacological treatment and transfection conditions on TMRM fluorescence ( Figure 5G ) , the effect of time and transfection conditions or Drosophila genotypes on the number of mitoDendra2 red objects ( Figure 3B , Figure 3—figure supplements 4B–D and 5 ) and the effect of the pharmacological treatment and the cell line on mitochondrial aspect ratio and form factor ( Figure 4B ) . One-way ANOVA and the Holm-Sidak method were used to compare the relative mitochondrial abundance of AURKA isoforms ( Figure 1B ) , the number of AURKA-positive mitochondria ( Figure 1—figure supplement 1A ) , the effect of acceptors on FRET efficiencies for given donor-acceptor pairs ( Figures 3E and 4F , Figure 1—figure supplements 2A , 3C and 6I ) , Mander’s co-localisation coefficients ( Figure 2B and Figure 1—figure supplement 2B ) , the relative total or mitochondrial abundance of AURKA with normal or kinase-dead AURKA ( Figure 2F ) , the relative abundance of each oxidative phosphosphorylation complex ( Figure 5A ) , the proportion of autophagic cells ( Figure 5C ) , the percentage of cells showing mitochondrial aggregates ( Figure 3—figure supplement 6E ) , the abundance of mitochondrial fusion and fission proteins ( Figure 3D ) , proteasomal activity ( Figure 5D ) , MitoTimer red/green ratio ( Figure 5E ) , JC-1 red fluorescence ( Figure 5F ) and the percentage of live , dead or apoptotic cells ( Figure 5I ) . One-way ANOVA on ranks and Dunn’s method were used to compare mitochondrial length ( Figure 3A ) . One-way ANOVA on ranks and the Kruskal-Wallis method were used to compare mitochondrial aspect ratio and form factor ( Figure 3—figure supplement 4A ) and the abundance of phosphorylated DNM1L forms and their ratios to total DNM1L ( Figure 3—figure supplement 6E ) . Student’s t-test was used to compare Mander’s co-localisation coefficients ( Figure 1A ) , the relative total or mitochondrial abundance of AURKA isoforms ( Figure 1B , Figure 3—figure supplement 3E ) , the relative O2 flux ( Figure 5B and H ) , PMPCB downregulation efficiency ( Figure 1C ) , mitochondrial aspect ratio and form factor ( Figure 3—figure supplement 6C ) , the abundance of mitochondrial fusion and fission proteins ( Figure 3C ) , AURKA downregulation efficiency ( Figure 3C ) and the abundance of phosphorylated DNM1L forms and the ratios of these forms to total DNM1L ( Figure 3—figure supplement 6B ) . Alpha for statistical tests used in this study was equal to 0 . 05 . | Structures called mitochondria power cells by turning oxygen and sugar into chemical energy . Each cell can have thousands of mitochondria , which work together to supply changing energy demands . They can fuse together or break apart , forming networks that change size and produce different amounts of energy . Getting the balance right is crucial; if energy levels are too low , the cell will not be able to grow and divide . If energy levels are too high , the cell can grow at a faster rate , which can contribute to the cell becoming cancerous . Although we know that mitochondria provide energy , it is not clear how they communicate to fine-tune the supply . Some clues come from cancer cells that seem dependent on their mitochondria for survival . In these cells , levels of a protein called AURKA are higher than normal . AURKA helps cells to divide , and it interacts with many different proteins . This complexity makes it difficult to work out exactly what AURKA does , but it is possible that it plays a role in energy supply . Bertolin et al . have now investigated whether mitochondria use AURKA to communicate inside human breast cancer cells . Tagging AURKA proteins with a fluorescent marker revealed that it accumulates inside mitochondria . Once it gets there , AURKA changes the shape of the mitochondria , which has dramatic effects on their capacity to produce energy . At normal levels , AURKA causes the mitochondria to fragment , breaking apart into smaller pieces . This maintains their energy output at a normal level . If AURKA levels are too high , the mitochondria fuse together and produce more energy . This means AURKA could help to fuel fast-growing cancer cells . Current drugs that aim to treat cancer by blocking the activity of AURKA show poor results . This is partly due to the fact that the protein has so many different roles in the cell . Finding that AURKA affects mitochondria is the first step in understanding one of its unknown roles . It also suggests the possibility of developing new drugs to change how mitochondria make energy in cancer cells that contain high levels of AURKA . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"cancer",
"biology"
] | 2018 | Aurora kinase A localises to mitochondria to control organelle dynamics and energy production |
Malaria transmission relies on the production of gametes following ingestion by a mosquito . Here , we show that Ca2+-dependent protein kinase 4 controls three processes essential to progress from a single haploid microgametocyte to the release of eight flagellated microgametes in Plasmodium berghei . A myristoylated isoform is activated by Ca2+ to initiate a first genome replication within twenty seconds of activation . This role is mediated by a protein of the SAPS-domain family involved in S-phase entry . At the same time , CDPK4 is required for the assembly of the subsequent mitotic spindle and to phosphorylate a microtubule-associated protein important for mitotic spindle formation . Finally , a non-myristoylated isoform is essential to complete cytokinesis by activating motility of the male flagellum . This role has been linked to phosphorylation of an uncharacterised flagellar protein . Altogether , this study reveals how a kinase integrates and transduces multiple signals to control key cell-cycle transitions during Plasmodium gametogenesis .
Malaria is caused by vector-borne protozoan parasites of the genus Plasmodium that cycle between mosquitoes and vertebrate hosts . Malaria pathology is linked to the proliferation of asexual blood stage parasites , whereas transmission to the mosquito is mediated by an obligatory sexual life cycle phase . Differentiation from asexually replicating stages into non-dividing male and female gametocytes takes place inside red blood cells . Following a period of maturation , the sexual precursors are available to initiate transmission when ingested by a mosquito . Gametocytes resume their development by responding to environmental signals including a small mosquito molecule , xanthurenic acid ( XA ) , and a simultaneous drop in temperature ( Billker et al . , 1998 ) . Upon ingestion by a mosquito vector , Plasmodium male gametocyte undergoes explosive development . Within 10 min , it completes three rounds of genome replication followed by endomitosis within a single nucleus , assembles the component parts of eight axonemes , and escapes the red blood cell in a process called exflagellation . Circulating microgametocytes are arrested at a G0-like stage of the cell cycle at the haploid level . After 15 s of induction by XA , eight basal bodies are assembled from a single microtubule organising centre ( Sinden et al . , 1976 ) . After 1 min , the first genome replication is completed and the spindle of mitosis I is formed ( Billker et al . , 2002 ) . At the same time each basal body nucleates one of the eight axonemes from large quantities of tubulin contained in the cytoplasm . By six minutes , the four spindles of mitosis III have formed and chromatin condensation only sets in at the end of mitosis III . In parallel gametocytes escape their host cell following the exocytosis of specialised secretory vesicles containing proteins with membranolytic activities . At the onset of exflagellation , axonemes become motile and swim out of the residual gametocyte body . As each basal body remains attached to a mitotic spindle pole , they drag a haploid genome that is incorporated into the exflagellating gamete . Malaria parasites are highly divergent from model organisms and significant differences in the composition and properties of cell cycle regulators have been reported ( Gerald et al . , 2011 ) . As a consequence little is known about how progression through the cell cycle is regulated in these parasites . Gametocyte stimulation by XA is followed by Ca2+ mobilisation from internal stores after a lag phase of ~10 s ( Billker et al . , 2004 ) which requires active cGMP-dependent protein kinase G , PKG ( Brochet and Billker , 2016; Brochet et al . , 2014 ) . In activated microgametocytes , the plant-like Ca2+-dependent protein kinase 4 ( CDPK4 ) , which belongs to a family absent from the human genome , is required to enter S-phase in the rodent parasite Plasmodium berghei ( Billker et al . , 2004 ) . Selective inhibitors of CDPK4 were shown to block exflagellation of P . berghei in vivo and of the human parasite P . falciparum in vitro placing CDPK4 as a promising drug target to reduce transmission of malaria ( Ojo et al . , 2014 , 2012 ) . Despite the importance of CDPK4 for transmission to the mosquito vector , its molecular functions remain unknown and none of its substrates have been identified . In this study , we took advantage of the highly synchronised nature of P . berghei gametogenesis to exactly identify when CDPK4 activity is required . By combining reverse and chemical genetics with molecular and cellular phenotyping , we found that CDPK4 plays at least three distinct roles during male gametogenesis and we identified three effectors mediating each of these roles .
Small bumped-kinase inhibitors targeting Plasmodium CDPK4 were recently developed by capitalising on a small serine gatekeeper residue in the active site of the enzyme ( Ojo et al . , 2014 , 2012 ) . One of these , compound 1294 , was found to inhibit P . falciparum exflagellation through CDPK4 ( Ojo et al . , 2014 ) . To ascertain for 1294 specificity in P . berghei , we set out to replace cdpk4 with a drug-resistant allele , cdpk4S147M-3xHA to prevent binding of 1294 to the ATP binding pocket ( Figure 1—figure supplement 1 parts 1 to 5 ) . We however discovered that introduction of both a C-terminal 3xHA epitope tag and a S147M substitution imposed a fitness cost on CDPK4 function preventing exflagellation . Investigation of the effect of each single modification revealed exflagellation is delayed in CDPK4-3xHA but not in CDPK4S147M parasites . This indicates that a S147M substitution only imposes a minor cost that , on its own , does not impact on exflagellation . Compared to the parental wild-type strain , CDPK4S147M transgenic parasites showed a ~ 50 fold decrease in exflagellation susceptibility to 1294 . Importantly , 1 µM 1294 blocked exflagellation by specifically targeting CDPK4 and this concentration was further used this study ( Figure 1A ) . 10 . 7554/eLife . 26524 . 003Figure 1 . A chemical genetic approach reveals CDPK4 activity is required to initiate the first round of DNA replication , assemble the first mitotic spindle and initiate axoneme motility . ( A ) A line expressing a CDPK4S147M allele is 50 times more resistant to compound 1294 . Error bars show standard deviations , n = 3 . ( B ) Effect of 1 μM 1294 addition at multiple time points after XA activation on microgametocytes ploidy . CDPK4 activity is required between 10 and 20 s to initiate the first round of DNA replication and the second genome replication . CDPK4 activity is however not required for DNA replication per se . Error bars represent standard deviations , n = 2 . ( C ) Immunofluorescence assays showing the effect of 1294 at different time points after activation . Addition of 1 μM 1294 at the time of activation prevents the formation of mitotic spindles as observed 1 min after activation in presence of DMSO . Addition of 1294 at 30 s or at 9 min post-activation does not inhibit axoneme formation but blocks initiation of axoneme motility and condensation of chromatin . Absence of 1294 treatment leads to the exflagellation of male gametes 10 min after activation . Scale bars = 1 µm . ( D ) Quantification of macrogametocytes showing mitotic spindles at 1 min post-activation in DMSO or 1294 treated parasites . Error bars show standard deviations , n = 2 , *** Student's T-test , p≤0 . 001 . ( E ) Quantification of macrogametocytes showing axonemes formed at 10 min after activation . Error bars show standard deviations , n = 2 , *** Student's T-test , p≤0 . 001 . ( F ) Quantification of exflagellation events in the WT and CDPK4S147M lines when DMSO or 1294 were added 9 min after activation . Error bars show standard deviations n = 2 , *** Student's T-test , p≤0 . 001 . ( G ) When added 15 min post-XA activation , 1294 does not inhibit motility of active microgametes , n = 3 , *** Student's T-test , p≤0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 00310 . 7554/eLife . 26524 . 004Figure 1—figure supplement 1 . CDPK4 is required to initiate the first round of DNA replication but not for ookinete development or motility . ( 1–4 ) Genetic modification vectors and genotyping data for CDPK4 targeting constructs . Oligonucleotides used for PCR genotyping are indicated and agarose gels for corresponding PCR products from genotyping reactions are shown . ( 1 ) We first attempted to generate CDPK4S147M-3xHA parasites but we repeatedly obtained pure parasite populations in which the 3xHA tag and the selection cassette were chromosomally integrated while only a minor population also incorporated the S147M substitution which was rapidly lost after one parasite passage . The resulting CDPK4-3xHA line was further cloned . ( 2 ) To select for the S147M substitution , we first generated a marker-free CDPK4-KO clonal line . ( 3 ) This CDPK4-KO line was then complemented with a CDPK4S147M-3xHA allele and the line was cloned . ( 4 ) To test for the respective fitness cost of S147M substitution and 3xHA tagging we generated a clonal parasite line expressing CDPK4S147M . ( 5 ) Effect of S147M substitution or 3xHA tagging on microgametocyte exflagellation 10 or 20 min after stimulation with XA . S147M substitution does not affect exflagellation while C-terminal 3xHA tagging of CDPK4 delays exflagellation . Combination of S147M substitution with a 3xHA tag fusion completely blocks male gametogenesis indicating that both mutations impose a fitness cost on CDPK4 activity but that S147M substitution only imposes a minor cost . The range of whisker plots indicates the 2 . 5–97 . 5% percentile , the box includes 50% of all values and the horizontal line shows median values . Data are from two biological replicates . ( 6 ) Identification of populations of RFP positive female gametocytes and GFP positive males by FACS in blood infected with parasites of line 820cl1 . ( 7 ) FACS analysis of Vybrant dye cycle violet fluorescence intensity of micro-gametocytes showing evolution from haploid to octoploid cells in less than 6 min post XA stimulation . ( 8 ) Relative proportion of 1N , 2N , 4N and 8N microgametocytes from 0 to 10 min post XA stimulation . Error bars represent standard deviations from two independent biological replicates . ( 9 ) Addition of 1 μM 1294 one hour after XA stimulation does not block ookinete development . Error bars are standard deviations from three biological replicates . ( 10 ) Addition of 1 μM 1294 on mature ookinetes does not affect their motility as assessed in Matrigel . The range of whisker plots indicates the 2 . 5–97 . 5% percentile , the box includes 50% of all values and the horizontal line shows median values . Data are representative of three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 004 To decipher how CDPK4 controls entry into S-phase during male gametogenesis , we first determined the kinetics of DNA replication ( Figure 1—figure supplement 1 parts 6 and 7 ) . In non-activated parasites , around 85% of male gametocytes were haploid . A majority of diploid parasites could be observed between 30 s and 1 min after XA activation , while tetraploid microgametocytes were mainly detected between 2 and 5 min . The three rounds of genome replication seemed to be completed by 6 min when 75% of male parasites were octoploid ( Figure 1—figure supplement 1 part 8 ) . We then investigated the requirement for CDPK4 activity to control genome replications by adding 1294 at different time points during gametogenesis ( Figure 1B ) . Addition of 1294 prior to activation dramatically reduced the ability of male gametocytes to replicate their genome as previously described for CDPK4-KO parasites ( Billker et al . , 2004 ) . We were however able to detect 25% of microgametocytes progressing to the diploid state but not further to the tetra- or octoploid levels . When 1294 was added 20 s after activation , no inhibition of DNA replication was observed . Altogether , this reveals that CDPK4 activity between 0 and 20 s post-activation is important for the 1N/2N and subsequent 2N/4N transitions but not for DNA replication per se . Interestingly , almost no diploid parasites reached the tetraploid state suggesting that CDPK4 plays a second role between the first and second rounds of genome replication . To test this hypothesis , parasites treated with 1294 or DMSO were imaged 1 min after activation , when the first mitotic spindle is observable ( Figure 1C ) . In DMSO treated parasites , 64% of male gametocytes showed spindle-like structures while this proportion decreased to 7% in parasites treated with 1294 ( Figure 1D ) indicating that CDPK4 activity is also required for mitotic spindle formation . We further took advantage of the versatility of the chemical approach to test whether CDPK4 is required later during male gametogenesis . To do so , we added 1294 30 s after activation , a time point at which CDPK4 is not required to initiate DNA replication anymore . With such a treatment , parasites normally developed to the 8N state and assembled axonemes . However , DNA did not condense , axonemes remained around the nucleus and no exflagellation was observed ( Figure 1C and E ) . A similar phenotype was observed when 1294 was added only seconds prior to the initial exflagellation events at 9 min post-activation ( Figure 1C and E–F ) . However , when added at 15 min post-activation , 1294 did not inhibit flagellar beating ( Figure 1G ) or ookinete formation and motility ( Figure 1—figure supplement 1 parts 9 and 10 ) . Altogether , this shows that CDPK4 activity seems to be redundant between 30 s and 9 min after activation but is further required at a late stage immediately preceding the onset of axoneme motility , DNA condensation and cytokinesis . To gain further insights into how CDPK4 regulates these transitions , we first aimed at identifying its interacting proteins in non-activated gametocytes . A total of 150 proteins were immunoprecipitated in both CDPK4-3xHA and CDPK4-2xmyc lysates after cross-linking but not in a WT control ( Figure 2—figure supplement 1 parts 1 to 3 and Supplementary file 1 ) . MCM2-7/Cdt1 and ORC proteins that are part of the pre-replicative complex represented the most enriched molecular components ( Figure 2A ) . Proteins of the microtubule cytoskeleton and of the replisome were also enriched including polymerases α , δ , and ε , the proliferating nuclear antigen , replication factor C complexes and replication factor A . 10 . 7554/eLife . 26524 . 005Figure 2 . CDPK4 activity facilitates the assembly of the pre-replicative complex during microgametogenesis . ( A ) GO term enrichment analysis of proteins co-immunoprecipitated with CDPK4-3xHA in non-activated gametocytes reveals the kinase is interacting with proteins of the pre-replicative complex . Bonferroni corrected p-values are indicated . ( B ) 178 proteins are immunoprecipitated with both CDPK4-3xHA and MCM5-3xHA including all components of the ORC and the MCM complex . The relative abundance of proteins was determined as the number of spectral counts for each protein divided by the number of total spectral counts in the respective immunoprecipitate . ( C ) The interaction between CDPK4-3xHA and MCM2-7/Cdt1 proteins is stable during the first 15 s following XA stimulation while interaction with ORC1-5/Cdc6 proteins is increased when the kinase is activated . Data are representative of two independent biological replicates . ( D ) MCM5-3xHA is enriched in chromatin-enriched NaCl fractions at 15 s post-activation in gametocytes but not when 1294 is added; α-tubulin was used as a soluble control . Error bars show standard deviations , n = 3 , ** Student's T-test , p≤0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 00510 . 7554/eLife . 26524 . 006Figure 2—figure supplement 1 . CDPK4 activity is required to assemble the pre-replicative complex . ( 1 ) Complementation of a CDPK4-KO line with a CDPK4-2xmyc allele . ( 2 ) A transgenic line expressing CDPK4-2xmyc does not show exflagellation defects . Error bars are standard deviations from two biological replicates . ( 3 ) Western blot and silver staining analysis of CDPK4-3xHA immunoprecipitation . Eluates show specific isolation of the 3xHA fusion protein . Molecular weights are indicated in kDa . hc = heavy chain , lc = light chain . ( 4 ) Generation and genotyping of a non-clonal line expressing MCM5-3xHA . ( 5 ) Western blot analysis of MCM5-3xHA lysates produced from non-activated gametocytes . ( 6 ) CDPK4-3xHA is incorporated in the exflagellated male gamete . Scale bar is 1 μm . ( 7 ) Relative abundance of proteins co-immunoprecipitated with CDPK4-3xHA in non-activated gametocytes compared with 15 second-activated gametocytes in presence or absence of 1 µM 1294 . The six components of the ORC/cdc6 complex show a significant increase in abundance when CDPK4 is active . ( 8 ) Relative abundance of proteins co-immunoprecipitated with CDPK4-2xmyc in non-activated gametocytes compared with 15 second-activated gametocytes in presence or absence of 1 µM 1294 . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 006 As MCM2-7/Cdt1 represented a highly enriched and abundant protein complex immunoprecipitated with CDPK4 in non-activated gametocytes , we first confirmed this interaction by determining the partners of MCM5-3xHA in the same conditions ( Figure 2—figure supplement 1 parts 4 and 5 ) . We found 103 proteins interacting with CDPK4-3xHA , CDPK4-2xmyc and MCM5-3xHA indicating that CDPK4 and MCM5 share an overlapping molecular environment ( Figure 2B ) . Consistently , CDPK4 represented one of the most abundant interactors of MCM5-3xHA . Altogether this suggests that CDPK4 is part of the MCM2-7/Cdt1 complex and may regulate the pre-replicative complex to initiate DNA replication . To look for putative molecular roles explaining the late requirement of CDPK4 , we identified interacting partners of CDPK4-2xmyc and CDPK4-3xHA 10 min after activation ( Supplementary file 1 ) . Cytoskeletal proteins predicted to be components of the microgamete axonemes ( Talman et al . , 2014 ) were the most abundant and significantly enriched cellular components ( p-value=5 . 10−4 ) . It is however important to note that most of this cellular component was also enriched in non-activated gametocytes and that cytoskeletal proteins are abundant in male gametocytes ( Khan et al . , 2005 ) . It is currently unclear whether these proteins represent relevant or specific interactors . Nuclear proteins involved in DNA replication were nevertheless much less abundant compared to earlier time points ( Supplementary file 1 ) . Traces of CDPK4 were also found to be incorporated in exflagellated male gametes ( Figure 2—figure supplement 1 parts 4 and 5 - [Talman et al . , 2014] ) in accordance with a role in the initiation of axoneme activity . Assembly of the pre-replicative complex is a dynamic process ( Bell and Kaguni , 2013 ) . In eukaryotes , ORC1-6 proteins bind to origin sequences and associate with Cdc6 . Cdc6 subsequently recruits the licensing factor Cdt1 and the heterohexameric MCM helicase complex . To investigate the dynamics of the CDPK4/MCM2-7/Cdt1 complex , we compared the CDPK4 interactome in non-activated gametocytes and in 15 second-activated gametocytes in the presence or absence of 1294 . For 215 out of the 221 proteins immunoprecipitated with CDPK4-3xHA under all conditions , no obvious differences were detected between the three conditions ( Figure 1—figure supplement 1 part 8 ) . However six proteins showed up to 8-fold increase in abundance upon gametocytes activation which was not observed when CDPK4 was inhibited ( Figure 2C ) . A similar result was observed when using the CDPK4-2xmyc line ( Supplementary file 1 and Figure 2—figure supplement 1 part 8 ) . Among these proteins were ORC1 , ORC2 , and ORC5 , while the three remaining proteins were not functionally annotated . Sequence analysis indicated that these three proteins likely correspond to homologues of Cdc6 , ORC4 and ORC3 . PBANKA_110290 contained an AAA ATPase domain and a HMM profile analysis indicated that its closest homologue in human and yeast was Cdc6 . Similarly , the closest homologue of PBANKA_134880 was ORC4 in human and yeast . Finally , homologues of PBANKA_051390 were only found in Plasmodium and Babesia but encoded a PF07034 PFAM domain corresponding to the N-terminus of ORC3 . We then determined the relative abundance of MCM5-3xHA in chromatin-enriched NaCl fractions between non-activated and 15 second-activated gametocytes . Western blot analysis showed that 30% of MCM5-3xHA was present in the high salt fraction in non-activated gametocytes . This ratio almost doubled 15 s after activation but remained unchanged in the presence of 1294 ( Figure 2D ) . Altogether , this suggests that loading of the MCM2-7/Cdt1 complex onto ORC1-5/Cdc6 complex happens around 15 s after activation of gametogenesis and requires active CDPK4 . CDPK4 has previously been shown to be myristoylated in P . falciparum ( Wright et al . , 2014 ) . We thus interrogated whether myristoylation could differentially regulate CDPK4 functions . To this aim , we generated a line in which the myristoylated glycine 2 was replaced by an alanine residue preventing myristoylation ( Figure 3—figure supplement 1 part 1 ) . As for CDPK4-KO parasites , CDPK4G2A-2xmyc microgametocytes did not exflagellate while a CDPK4-2xmyc control was not impaired ( Figure 3A ) . Consistently , more than 85% of CDPK4G2A-2xmyc and CDPK4-KO gametocytes remained haploid while 28% of CDPK4-2xmyc gametocytes became octoploid . However 65% showed spindle-like structures showing that CDPK4 myristoylation is important to initiate the first round of DNA replication but nor for mitotic spindle assembly ( Figure 3A ) . 10 . 7554/eLife . 26524 . 007Figure 3 . Myristoylation of CDPK4 is required to initiate the first round of DNA replication while non-myristoylated CDPK4 is important to complete gametogenesis . ( A ) Effect of cdpk4 deletion or G2A substitution on microgametocyte DNA replication , mitotic spindle assembly , and exflagellation . G2A mutation mimics cdpk4 deletion with ca . 70% of microgametocytes remaining at the 1N state and another 20% blocked at the diploid stage ( n = 3 ) . Consistently G2A substitution completely blocks exflagellation ( n = 3 ) but does not prevent formation of mitotic spindles ( n = 2 ) . Error bars show standard deviations , *** Student's T-test , p≤0 . 001 . ( B ) Western blot analysis showing expression of CDPK4-2xmyc and CDPK4G2A-2xmyc proteins in complemented CDPK4-KO lines . Migration of the CDPK4G2A-2xmyc isoform around 63 kDa ( i63 ) is affected , suggesting it is myristoylated while the 55 kDa isoform ( i55 ) migration is not affected suggesting it is not myristoylated . ( C ) CDPK4-2xmyc and CDPK4G2A-2xmyc are both mainly detected in the PBS fraction . A minor population of CDPK4G2A-2xmyc i63 is partly recovered in membrane-associated or NaCl fractions while the CDPK4G2A-2xmyc i55 is fully solubilised in PBS . Antibodies against the ERD2 integral membrane protein and the soluble PRF protein were used as controls . ( D ) M57A substitution leads to expression of only i63 , indicating that i55 originates from a second translation start at methionine 57 . ( E ) A line expressing CDPK4M57A-2xmyc shows a strong defect in exflagellation but is not affected in DNA replication and axoneme assembly indicating that the short non-myristoylable CDPK4 isoform is mainly required to control late gametogenesis but not to initiate the first round of DNA replication . Error bars show standard deviations , n = 3 , *** Student's T-test , p≤0 . 001 . ( F ) The two major CDPK4 isoforms and their identified roles during gametogenesis . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 00710 . 7554/eLife . 26524 . 008Figure 3—figure supplement 1 . CDPK4 myristoylation does not affect the kinase localisation and related protein-protein interactions . ( 1 ) Complementation of a CDPK4-KO line with CDPK4G2A-2xmyc or CDPK4M57A-2xmyc alleles and genotyping . ( 2 ) Western blot analysis of CDPK4-3xHA during gametogenesis indicates that the same migration pattern is observed at 0 , 4 , 8 and 10 min after stimulation with XA . ( 3 ) Acyl-RAC assay suggesting an absence of CDPK4 palmitoylation . Western blot analysis of CDPK4-3xHA gametocytes after acyl-biotin exchange in the absence ( −NH2OH ) or presence ( +NH2OH ) of hydroxylamine . Cleavage of the thioester bond , linking palmitate to a putative cognate cysteine residues , by hydroxylamine should create reactive cysteine residues that can be detected by the thio-reactive resin . No signal is observed for CDPK4-3xHA or profilin . Toxoplasma gondii tachyzoites lysates processed in parallel together with an α-TgGAP45 against the palmitoylated protein TgGAP45 were used as a positive technical control for S-acylated protein enrichment . ( 4 ) Immunofluorescence assays did not allow the detection of any obvious differences in CDPK4G2A-2xmyc localisation at 0 or 15 s post activation suggesting that CDPK4 myristoylation is not a major determinant for the cellular localisation of CDPK4 . Scale bars = 1 μm . ( 5 ) No difference in relative abundance of protein co-immunoprecipitated with CDPK4-2xmyc and CDPK4G2A-2xmyc could be detected suggesting myristoylation does not affect CDPK4 interaction with its partners . Data is from one biological replicate . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 008 Western blots of CDPK4-2xmyc revealed a large band in non-activated gametocytes close to the expected size of 63 . 1 kDa and a second band migrating around 55 kDa ( Figure 3B ) , a pattern that was conserved during gametogenesis ( Figure 3—figure supplement 1 part 2 ) . In CDPK4G2A-2xmyc extracts , the large isoform migrated slower . In some instances myristoylation has been shown to accelerate electrophoretic mobility ( Demetriadou et al . , 2017; Zhao et al . , 2011 ) suggesting that the large isoform is myristoylated . Conversely , the 55 kDa band did not show a different mobility . Proteins can undergo additional post-translational covalent modifications with one or more palmitoyl groups after N-myristoylation . Using an acyl-RAC assay , we were however not able to detect significant evidence of CDPK4 palmitoylation ( Figure 3—figure supplement 1 part 3 ) . To investigate how myristoylation could regulate CDPK4 function , we first determined its importance for protein localisation and protein-protein interactions . Immunofluorescence assays on CDPK4-2xmyc and CDPK4-3xHA , revealed that CDPK4 showed a broad cellular distribution in microgametocytes at 0 and 15 s after activation ( Figure 3—figure supplement 1 part 4 ) . We were however not able to detect any obvious difference in CDPK4G2A-2xmyc localisation at both time points . Similarly , immunoprecipitation of CDPK4-2xmyc and CDPK4G2A-2xmyc did not reveal any significant changes in interacting partners under the tested conditions ( Figure 3—figure supplement 1 parts 5 and Supplementary file 1 ) . Myristoylation is also crucial to promote weak and reversible protein-membrane interactions . The CDPK4-2xmyc 55 kDa isoform was fully solubilised during hypotonic lysis while 90% of the CDPK4-2xmyc large isoform was solubilised during hypotonic lysis and fully released with a subsequent carbonate treatment ( Figure 3C ) . Interestingly , the large isoform in CDPK4G2A-2xmyc lysates was fully solubilised during hypotonic lysis . In parallel we also asked whether CDPK4 could be identified in a chromatin-enriched NaCl fraction . We found a minor population of myristoylated CDPK4-2xmyc in the NaCl fraction which was not observed for CDPK4G2A-2xmyc . Altogether this highlights that a major population of CDPK4 is soluble whereas a minor population of the large CDPK4 isoform is associated with membranes and possibly with chromatin . As the short CDPK4 isoform did not appear to be myristoylated , we hypothesised it may correspond to a shorter protein whose translation could be initiated at a second start codon . The only possible second start codon upstream the catalytic domain was methionine +57 , which we replaced by a codon coding for an alanine residue ( Figure 3—figure supplement 1 part 1 ) . The resulting CDPK4M57A-2xmyc line only expressed the large isoform ( Figure 3D ) confirming that the translation of the 55 kDa isoform is initiated at a second start codon and cannot be myristoylated . Phenotyping of this line revealed that microgametocytes only expressing the large 63 kDa myristoylated isoform , progressed to the octoploid stage , assembled axonemes but showed a dramatic reduction in exflagellation compared to a control line ( Figure 3E ) . Altogether this demonstrates that the large CDPK4 isoform is mainly required to initiate the first round of DNA replication while the short non-myristoylated isoform is essential to complete late gametogenesis ( Figure 3F ) . Interestingly , either short or long CDPK4 isoforms seem to support mitotic spindle assembly . To better understand how CDPK4 regulates these distinct biological processes we set out to identify its substrates using an analogue sensitive kinase ( AS-kinase ) engineered to contain a small gatekeeper residue ( Allen et al . , 2007 ) . Bulky ATP analogues have been used to specifically label the targets of such modified kinases . The γ-phosphate of these artificial ATP-analogues is replaced with a thiophosphate and can then be used to purify the cognate substrates by chemical means . We first attempted to generate a line expressing a CDPK4S147G-3xHA allele . As for CDPK4S147M-3xHA , only a transient population having incorporated the S147G mutation and the 3xHA tag could be detected ( data not shown ) . We were however able to clone a line in which endogenous cdpk4 was replaced by a cdpk4S147G allele ( Figure 4—figure supplement 1 part 1 ) . This line showed a limited reduction in exflagellation indicating that S147G substitution also imposes a slight fitness cost on CDPK4 function ( Figure 4—figure supplement 1 part 2 ) . To identify CDPK4 substrates , we compared incorporation of thiophosphate from N6-phenylethyl ATP-γS in lines expressing cdpk4S147G and cdpk4S147M alleles , respectively . Data processing with MaxQuant and Proteome Discoverer allowed to detect a total of 19 phosphorylated peptides in CDPK4S147G lysates only , ten of which were found to interact with CDPK4 ( Figure 4A and Supplementary file 2 ) . Among these proteins , we decided to name those without predicted function SOC , for substrate of CDPK4 . Ten proteins were indeed conserved Plasmodium proteins of unknown function and one protein , SOC2 , was annotated as cyclin-related protein 2 but it was suggested be unrelated to cyclins ( Roques et al . , 2015 ) . Intriguingly , we also identified GAP40 , a protein expressed in the merozoite stage ( Otto et al . , 2014 ) and important for the inner membrane complex biogenesis in Toxoplasma gondii ( Harding et al . , 2016 ) . 10 . 7554/eLife . 26524 . 009Figure 4 . Identification of CDPK4-dependent phosphorylation during the first seconds of gametocyte activation . ( A ) Analogue sensitive-CDPK4S147G uses N6-phenylethyl ATP-γS ( Red ) to thiophosphorylate its substrates while the CDPK4S147M allele that cannot accommodate N6-phenylethyl ATP-γS is used as a control . Thiophospho-tryptic peptides are captured by SulfoLink Coupling Resin and submitted to LC-MS/MS analysis . Nineteen phosphopeptides were found to be thiophosphorylated by AS-CDPK4S147G . Phosphoproteomic analysis revealed that 70 phosphopeptides on 61 proteins showed a 2-fold increase at 18 s post XA activation in the WT but not in the CDPK4-KO line . Data are from two biological replicates . ( B ) Among these latter , SOC1 , SOC2 and SOC4 are phosphorylated by AS-CDPK4S147G and could represent early effectors of CDPK4 . Conversely , GAP40 and SOC3 were not differentially phosphorylated in the CDPK4-KO line but were robustly identified as phosphorylated by AS-CDPK4S147G and could represent CDPK4 effector to control late gametogenesis . List of corresponding peptides and description of their proteins , MaxQuant intensity/Proteome Discoverer Xcorr , and log2[CDPK4-KO/WT]18sec phosphorylation ratio; sites assigned as phosphorylated by the two search engines are highlighted in grey . ( C ) GAP40 could not be deleted in asexual stages , suggesting an essential role in erythrocytic asexual multiplication . Knocking-out SOC4 or the SOC5 and SOC6 controls did not produce any defect in exflagellation nor DNA replication while SOC1 , 2 and 3 KO lines showed a strong reduction in exflagellation . For SOC1-KO and SOC2-KO lines a defect in DNA replication was observed while SOC3-KO was not impaired in DNA replication and axoneme assembly . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 00910 . 7554/eLife . 26524 . 010Figure 4—figure supplement 1 . Identification and preliminary characterisation of CDPK4 substrates . ( 1 ) Generation of a CDPK4S147G clonal line and genotyping . ( 2 ) A CDPK4S147G clonal line shows a slight decrease in exflagellation at both 10 and 20 min after stimulation by XA . The range of whisker plots indicates the 2 . 5–97 . 5% percentile , the box includes 50% of all values and the horizontal line shows median values . Data are representative of two biological replicates . ( 3–8 ) Generation and genotyping of SOC4/5/6-KO non-clonal lines and of SOC1/2/3-KO clonal lines and their phenotyping for exflagellation ( 9 ) and DNA replication during male gametogenesis ( 10 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 010 Target identification with AS-kinases relies on cell lysates as a source of substrates and would identify potential substrates of CDPK4 implicated in regulation of both early and late gametogenesis . To identify physiologically relevant substrates in early gametogenesis , we took advantage of the high synchronicity of early gametogenesis to profile the phosphoproteome in a CDPK4-KO line and its wild-type counterpart at 0 and 18 s after XA activation , when CDPK4 is required to initiate DNA replication and mitotic spindle assembly . In total we were able to identify 10 , 580 phosphopeptides mapping on 1 , 734 P . berghei proteins ( Supplementary file 3 ) . Among those , we identified 70 phosphopeptides mapping on 61 proteins that showed a 2-fold upregulation in the WT between 0 and 18 s but not in the CDPK4-KO line ( Figure 4A ) . Among the 19 peptides phosphorylated by AS-CDPK4S147G , those mapping on SOC1 , SOC2 and SOC4 showed a 2-fold upregulation or more at 18 s in the WT but not in the CDPK4-KO line . This suggests that these three proteins are likely to represent CDPK4 physiological substrates to regulate early gametogenesis and were retained for further analysis ( Figure 4B ) . Conversely , the remaining substrates identified with the AS-kinase could represent CDPK4 effectors controlling late gametogenesis . Among those , further analysis was focused on SOC3 and GAP40 proteins that were identified with the highest confidence based on both MaxQuant and Proteome Discoverer analyses ( Figure 4B ) . To investigate the roles of SOC1 to 4 and GAP40 during gametogenesis , we attempted to individually knock-out their encoding genes ( Figure 4B and C , Figure 4—figure supplement 1 parts 3 to 6 ) . As controls , we included soc5 and soc6 ( Figure 4—figure supplement 1 parts 7 and 8 ) that were only detected by either MaxQuant or Proteome Discoverer but not differentially phosphorylated in the CDPK4-KO mutant . For gap40 , no transgenic parasites could be obtained and this gene was not retained for further analysis . Viable KO parasites could be detected in mixed asexual stages for six genes and non-clonal populations were first assessed for microgametocyte DNA replication and exflagellation ( Figure 4—figure supplement 1 parts 9 and 10 ) . No defects were observed in the SOC4-KO line as well as in the SOC5-KO and SOC6-KO control lines suggesting that these proteins do not represent crucial CDPK4 effectors in the regulation of male gametogenesis . Defects in DNA replication or exflagellation were observed for SOC1-KO , SOC2-KO , and SOC3-KO lines that were cloned for in depth characterisation . SOC1-KO line showed a dramatic reduction in exflagellation and 50% of microgametocytes remained haploid ( Figure 5A ) . As SOC1 phosphorylation requires CDPK4 activity within the first 20 s it is likely that SOC1 is a CDPK4 effector to control loading of the pre-replicative complex . Consistently co-immunoprecipitation of SOC1-3xHA in non-activated gametocytes identified MCM proteins and CDPK4 as abundant interactors ( Figure 5B ) . As observed for CDPK4-3xHA and CDPK4-2xmyc , SOC1-3xHA did not show a significant enrichment in the nucleus but rather showed a diffuse cytoplasmic localisation ( Figure 5B ) raising the possibility it may also have other functions during male gametogenesis . Interestingly , 40% of SOC1-KO microgametocytes reached the octoploid state suggesting this protein is not essential to initiate the next two rounds of DNA replication . Although no domain could be identified in SOC1 , this protein was found to be conserved in eukaryotes ( Figure 5—figure supplement 1 part 1 ) . Eukaryotic homologues contained SAPS domains that have been found to be involved in G1 to S-phase transitions by associating to SIT4 phosphatase ( Luke et al . , 1996 ) . SOC1 is also expressed in asexual blood stages and its deletion was associated with a slight fitness cost in these stages ( Figure 5—figure supplement 1 part 2 ) suggesting its role is conserved in multiple stages of the Plasmodium life cycle . The SOC2-KO line showed a dramatic reduction in exflagellation . Unlike SOC1-KO , proportions of haploid microgametocytes in 10 min activated microgametocytes were similar in the KO and its WT counterpart indicating that SOC2 is not required for initiation of DNA replication and DNA replication per se ( Figure 5C ) . However , 47% of the parasites remained diploid and 20% and 16% reached the 4N and 8N levels , respectively . SOC2 could thus represent a CDPK4 effector required to complete mitosis in order to enter the subsequent round of replication . This prompted us to image mitotic spindles in the SOC2-KO line one minute after activation . As opposed to the 820cl1 parental line , we were rarely able to observe incorporation of α-tubulin into distinct mitotic spindles at 1 min post-activation , a phenotype similar to what was observed when control parasites were treated with 1294 prior to activation by XA ( Figure 5C ) . At 10 min post-activation , SOC2-KO showed normal axonemes encircling the nucleus ( Figure 5C ) . This 623 kDa protein appeared to be specific to Plasmodium ( Figure 5—figure supplement 1 part 1 ) and is predicted to encode six transmembrane domains and six sialidase penultimate C-terminal domains . C-terminal 3xHA tagging of SOC2 only allowed detection of a c . a . 140 kDa protein , 70% of which was extracted in PBS . Interestingly , this protein contained the peptide phosphorylated by CDPK4 . Immunoprecipitation of SOC2-3xHA recovered peptides mapping on the last 1137 amino acids suggesting SOC2 is processed into at least two different isoforms . The 3xHA tagged C-terminal end of SOC2-3xHA was found to colocalise with mitotic spindles further suggesting a role in their assembly ( Figure 5D ) . Similarly to SOC1 , SOC2 is also expressed in asexual blood stages and its deletion was associated with a strong fitness cost in these stages ( Figure 5—figure supplement 1 part 2 ) suggesting a conserved role in mitotic spindle assembly during the atypical schizogony of Plasmodium parasites . Finally , SOC3-KO clone did not exflagellate but defects neither in DNA replication nor in axoneme assembly could be detected ( Figure 5E ) suggesting it represents a crucial effector of CDPK4 to complete cytokinesis . Concordantly , SOC3-3xHA was found to partially colocalise with axonemal α-tubulin in exflagellating microgametes ( Figure 5F ) . As opposed to SOC1 and 2 , SOC3 is mainly expressed in the male gametocyte ( Otto et al . , 2014 ) and deletion of its coding gene was not associated with any growth defect of asexual blood stages ( Figure 5—figure supplement 1 part 2 ) . Phylogenetic analysis indicated that SOC3 is restricted to Plasmodium species suggesting that the molecular function of SOC3 is highly specific to control the activation of male gametocyte axonemes . 10 . 7554/eLife . 26524 . 011Figure 5 . Functional characterisation of three identified substrates of CDPK4 . ( A ) A SOC1-KO line shows a strong reduction in exflagellation ( n = 3 ) . SOC1 is required only for the 1N/2N transition as parasites reaching the 2N level were able to further progress to the octoploid state ( n = 2 ) . Error bars show standard deviations , *** Student's T-test , p≤0 . 001 . ( B ) SOC1-3xHA shows a diffuse cytoplasmic distribution ( scale bar = 2 µm ) but interacts with CDPK4 and proteins of the MCM complex . ( C ) SOC2-KO line is strongly impaired for exflagellation ( n = 3 ) . SOC2 is not required for the 1N/2N transition but for each following transition indicating it is not involved in the initiation of DNA replication nor for DNA replication per se but most likely during the three successive endomitoses ( n = 2 ) . Consistently , a strong reduction in parasites exhibiting mitotic spindles at 1 min post-activation is observed in the SOC2-KO line ( n = 2 ) . However no defect in axoneme assembly could be detected ( n = 2 ) . Error bars show standard deviations , *** Student's T-test , p≤0 . 001 . ( D ) SOC2 is predicted to be a 645 kDa polytopic protein . Immunoprecipitation of SOC2-3xHA recovers peptides covering the last 1707 a . a . only; corresponding unique spectral counts are indicated in pink . Similarly co-immunoprecipitation of CDPK4 only recovered peptides covering the last 1137 a . a . of SOC2; corresponding unique spectral counts are indicated in blue . The C-terminal fragment colocalises with mitotic spindle α-tubulin but not with peripheral α-tubulin . Scale bar = 2 µm . ( E ) SOC3 is not required for DNA replication ( n = 2 ) and axoneme assembly ( n = 2 ) but is essential for exflagellation ( n = 3 ) . Error bars show standard deviations , *** Student's T-test , p≤0 . 001 . ( F ) SOC3 colocalises with axonemal α-tubulin in exflagellating gametes . Scale bar = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 01110 . 7554/eLife . 26524 . 012Figure 5—figure supplement 1 . Characterisation of three identified substrates of CDPK4 . ( 1 ) Taxonomic distribution of homologues of each characterised substrate of CDPK4 determined by HMMER ( http://hmmer . org/ ) . Numbers represent the number of homologues in each clade . ( 2 ) Effect of gene knock-out of soc1/2/3 on asexual blood stages growth showing that deletion of SOC1 and SOC2 are associated with different fitness costs . Error bars show standard deviations from three infections . ( 3–5 ) Generation and genotyping of SOC1/2/3-3xHA non-clonal lines . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 012
Kinases represent important regulators of Plasmodium development and play crucial roles for the asexual proliferation in erythrocytes , in liver cells and during transmission stages ( Brochet et al . , 2015; Solyakov et al . , 2011; Tewari et al . , 2010 ) . However little is known about how kinases regulate progression of the life cycle of malaria parasites . By taking advantage of the highly synchronised nature of P . berghei gametogenesis we were able to pinpoint when and how CDPK4 controls multiple cell cycle events during this biological process ( Figure 6 ) . 10 . 7554/eLife . 26524 . 013Figure 6 . Model showing the roles of CDPK4 during Plasmodium male gametogenesis . Activation of male gametocytes by a drop in temperature and xanthurenic acid leads to mobilisation of intracellular calcium within a lag phase of ten seconds . Calcium activates a myristoylated isoform of CDPK4 which activity is required during the next ten seconds to facilitate the loading of the MCM2-7/Cdt1 complex onto ORC1-5/Cdc6 complex requiring SOC1 . The subsequent assembly of mitotic spindles also requires ( i ) CDPK4 activity within the 30 fist second of gametogenesis and ( ii ) SOC2 which is a microtubule-associated protein phosphorylated by CDPK4 within the twenty fist seconds of gametogenesis . Between 30 s and 9 min , CDPK4 activity is neither required for the three successive rounds of genome replication and endomitosis nor for the assembly of the eight axonemes . Finally , activity of a non-myristoylable isoform of CDPK4 is required for the initiation of axoneme motility via the CDPK4 substrate SOC3 . DOI: http://dx . doi . org/10 . 7554/eLife . 26524 . 013 In response to ingestion by the mosquito , male gametocytes undergo three rounds of DNA replication alternating with three endomitotic divisions to generate a cell with an 8N genomic complement . In eukaryotes , developmentally programmed polyploidy involves the selective loss of mitotic cyclin-dependent kinases ( CDKs ) activity and bypassing many of the processes of mitosis ( Zielke et al . , 2013 ) . In this study , we have found that CDPK4 represents a pivotal kinase to enter S phase in response to mosquito ingestion by promoting the recruitment of the MCM2-7/Cdt1 complex onto origins of replication . We identified SOC1 as an important effector of CDPK4 in the first twenty seconds of activation for the 1N/2N transition . Although this protein does not contain recognisable domains , conserved eukaryotic homologues of SOC1 contain a SAPS domain ( Sit4-associated proteins ) . In yeast , SAPS proteins function in G1 to promote timely DNA replication by regulating G1 cyclin transcription ( Luke et al . , 1996 ) . Given the time frame of microgametogenesis it is possible that SOC1 regulates initiation of DNA replication through different mechanisms in Plasmodium . Genome replications alternate with three endomitotic divisions . We found that SOC2 , a second protein that is phosphorylated by CDPK4 within 20 s of activation , is not necessary to reach the 2N level but is important to progress to the 4N state , suggesting this early substrate is not involved in DNA replication but is required to complete the first mitosis . Accordingly , SOC2 localised at the mitotic spindles during mitosis and deletion of its coding gene was associated with marked defects in mitotic spindle formation . It is likely that SOC2 represents a Plasmodium-specific microtubule-associated protein required for chromosome segregation by controlling mitotic spindle assembly during the atypical cell cycle of Plasmodium parasites . Interestingly most SOC2-KO microgametocytes did not reach the 4N state suggesting a checkpoint might be operational to ensure that the second genome replication is initiated only after successful completion of the first mitosis . This is in agreement with recent findings showing that down-regulation of the cyclin-related kinase 4 leads to defects in both DNA replication and mitotic spindle formation during schizogony of P . falciparum ( Ganter et al . , 2017 ) . Interestingly , we show here that regulation of DNA replication and mitotic spindle assembly are simultaneously controlled by a single kinase suggesting that , in Plasmodium , some S- and M-phase events may be simultaneously co-regulated . However , previous findings showed that small molecules affecting mitotic spindle stability did not impair DNA replication in asexual erythrocytic stages and in microgametocytes ( Billker et al . , 2002; Sinou et al . , 1998 ) and it remains unclear how cell cycle transitions are regulated in Plasmodium parasites . On the other hand , we were able to observe fully formed axonemes in SOC2-KO gametocytes indicating that once initiated , axoneme assembly proceeds independently from some cell cycle checkpoints . Further studies will be required to investigate how S- and M-phase transitions are regulated during the atypical cell cycle of Plasmodium microgametes . It is only during the final steps of gametogenesis that the eight genomes are incorporated into the eight exflagellating haploid gametes . This is achieved when the forming gamete swims out of the residual gametocyte body and drags a haploid genome that remains attached to a mitotic spindle pole . Here , we have found that CDPK4 activity is also required to initiate axoneme motility and hence , to complete cytokinesis . We have linked this late role to an axoneme associated protein , SOC3 . Its deletion did not impair genome replication nor axoneme assembly but blocked axoneme activation . Protein phosphorylation and dephosphorylation is known to be important to initiate axoneme motility but the cause-and-effect relationship between phosphorylation and motility remains poorly understood ( Wirschell et al . , 2011 ) . From this study , we propose that SOC3 represents a Plasmodium-specific phosphoprotein required to initiate axoneme motility and consequently cytokinesis during microgametogenesis . Deletion of soc1 and soc2 did not block the 1N/2N and 2N/4N transition as drastically as CDPK4 inhibition . This strongly suggests that other identified or non-identified CDPK4 substrates cooperate with SOC1 and SOC2 to load the pre-replicative complex and assemble mitotic spindles . In addition , multiple kinases and phosphatases have been shown to control male gametogenesis ( Guttery et al . , 2012; Tewari et al . , 2010 ) and changes in phosphorylation levels between 0 and 18 s after activation highlighted a broad phospho-dependent response . It is thus likely that a plastic phosphorylation-dependent network underlies regulation of the initiation of DNA replication and mitotic spindle assembly and the exact molecular roles of SOC1 and SOC2 and of their phosphorylation remains to be elucidated . Conversely , deletion of soc3 completely blocked exflagellation suggesting that SOC3 may be a focal component for the activation of axoneme motility . During microgametogenesis , CDPK4 activity is essential only during two short time windows: within 10 to 30 s after activation and seconds prior to exflagellation . This highlights that CDPK4 is likely integrating multiple signals to be active at the right time and place . During the first window , the kinase is activated by a PKG-dependent Ca2+ mobilisation that occurs ten seconds after XA stimulation ( Brochet et al . , 2014 ) . Myristoylation of the large CDPK4 isoform is another signal required to initiate the first round of DNA replication but not to initiate axoneme motility . Interestingly , both large myristoylated and short non-myristoylated isoforms seemed to support mitotic spindle assembly . This indicates that myristoylation per se is not essential for CDPK4 activity in vivo in the absence of the 56 amino acids upstream of methionine +57 . In vitro , full length non-myristoylated recombinant CDPK4 proved to be an active enzyme suggesting myristoylation is important for in vivo regulation ( Ojo et al . , 2012 ) . We were however not able to provide evidence that CDPK4 myristoylation is required for specific cellular targeting nor for protein-protein interactions . Intriguingly , global profiling of protein myristoylation in human revealed that a large part of N-myristoyltransferase substrates are localised to the nucleus and mediate nuclear processes ( Thinon et al . , 2014 ) . Collectively , this suggests that the additional 56 amino acids upstream the second start codon not only serves as an extra sequence to accommodate N-terminal myristoylation but may also represent a regulatory domain to control CDPK4 nuclear functions . CDPK4 is regarded as a potential target for a transmission blocking strategy and proof of principle that chemical inhibition of CDPK4 by bumped kinases inhibitors such as 1294 can block Plasmodium transmission in vivo has been demonstrated ( Ojo et al . , 2012 ) . It is likely that CDPK4 strict essentiality during malaria transmission is explained by its multiple functions identified in this study . Bumped kinase inhibitors such as 1294 appear to have a broad-spectrum anti-protozoal activity by targeting functional orthologues of CDPK4 ( Van Voorhis et al . , 2017 ) and this study provides a first insight into the molecular functions of this atypical Ca2+-dependent protein kinase . Remarkably , CDPK4 has been previously reported to be important for sporozoite motility ( Govindasamy et al . , 2016 ) and oocyst formation ( Billker et al . , 2004 ) , two processes unrelated to male gametogenesis . Similarly , in the related apicomplexan parasite Toxoplasma gondii , the functional orthologue of CDPK4 ( TgCDPK1 ) is essential for host cell egress and invasion ( Lourido et al . , 2010 ) . It thus likely that this kinase not only represents a pleiotropic regulator of the cell cycle during gametogenesis but also a broad spectrum regulator of multiple other biological processes in the Apicomplexa .
P . berghei strain ANKA ( Vincke et al . , 1966 ) derived clone 2 . 34 ( Billker et al . , 2004 ) , clone 820cl1 ( Mair et al . , 2010 ) and derived transgenic lines ( Supplementary file 4 ) were maintained in CD1 outbred mice . The parasitemia of infected animals was determined by methanol fixed and Giemsa stained thin blood smears . Female mice were used for all experiments . CD1 outbred mice were obtained from Charles River laboratories . Mice were specific pathogen free and subjected to regular pathogen monitoring by sentinel screening . They were housed in individually ventilated cages furnished with a cardboard mouse house and Nestlet . Mice were maintained at 21 ± 2°C under a 12 hr light/dark cycle and given commercially prepared autoclaved dry rodent diet and water ad libitum . Mice were used for experimentation at 6–9 weeks of age . All animal experiments were conducted with the authorisation numbers ( GE/82/15 and GE/41/17 ) according to the guidelines and regulations issued by the Swiss Federal Veterinary Office . For gametocyte production , parasites were maintained in mice phenyl hydrazine-treated three days before infection . One day after infection , sulfadiazine ( 20 mg/L ) was added in the drinking water to eliminate asexually replicating parasites . Microgametocyte exflagellation was quantified three or four days after infection of mice by adding 4 µl of blood from a superficial tail vein to 70 µl exflagellation medium ( RPMI 1640 containing 25 mM HEPES , 4 mM sodium bicarbonate , 5% FCS , 100 µM xanthurenic acid , pH 7 . 4 ) . To calculate the number of exflagellation centres per 100 microgametocytes , the percentage of RBCs infected with microgametocytes was assessed on Giemsa-stained smears . For gametocyte purification , parasites were harvested in suspended animation ( SA - RPMI1640 medium containing 25 mM HEPES , 5% FCS , 4 mM sodium bicarbonate , pH 7 . 20 ) and separated from uninfected erythrocytes on a Histodenz cushion made up from 48% of a Histodenz stock ( 27 . 6% w/v Histodenz -Sigma- in 5 . 0 mM Tris-HCl [pH 7 . 20] , 3 . 0 mM KCl , 0 . 3 mM EDTA ) and 52% SA with a final pH of 7 . 2 . Gametocytes were harvested from the interphase . Schizonts for transfection were purified from overnight cultures on a Histodenz cushion made up from 55% of a Histodenz stock and 45% PBS . Purified parasites were harvested from the interphase and centrifuged at 500 g for 3 min , resuspended in 100 μL Amaxa Basic parasite Nucleofector solution ( Lonza , Switzerland ) and added to 10–20 µg of precipitated DNA resuspended in 10 µl of H2O . Cells were electroporated using the U-0033 program of the Amaxa Nucleofector II . Transfected parasites were resupended in 200 µl of fresh red blood cells and injected intraperitoneally into mice . Selection with 0 . 07 mg/mL pyrimethamine ( Sigma-Aldrich , Switzerland ) in drinking water ( pH ~4 . 5 ) was initiated from day one post infection . Negative selection of CDPK4S147M and CDPK4-3xHA parasites expressing yFCU was performed through the administration of 5 fluorocytosine ( 1 mg/mL , Sigma ) via the drinking water . Each mutant was genotyped using three combinations of primers , specific for either the WT or modified locus on both sides of the targeted locus ( experimental designs are shown in Supplemental Figures ) . For allelic replacements , sequences were confirmed by Sanger sequencing using indicated primers . WT DNA controls were included in each genotyping panel . Lines were cloned when indicated . CDPK4 mutants and derivatives and MCM-3xHA parasites were generated in the 2 . 34 background and all other transgenics were generated in the 820cl1 background . 3xHA tagging , knock-out , and allelic replacement constructs were generated using phage recombineering in Escherichia coli TSA bacterial strain with PlasmoGEM vectors ( http://plasmogem . sanger . ac . uk/ ) . The vector used to tentatively knock-out gap40 was PbGEM-335107 . For final targeting vectors not available in the PlasmoGEM repository , generation of knock-out and tagging constructs was performed using sequential recombineering and gateway steps as previously described ( Pfander et al . , 2013 , 2011 ) . A list of oligonucleotides used in this study is available in Supplementary file 5 . For each gene of interest ( goi ) , the Zeocin-resistance/Phe-sensitivity cassette was introduced using oligonucleotides goi HA-F x goi HA-R and goi KO-F x goi KO-R for 3xHA tagging and KO targeting vectors , respectively . Substitutions of CDPK4S147 gatekeeper residue were introduced with a two-step strategy using λ Red-ET recombineering as described in ( Brochet et al . , 2014 ) . The first step involved the insertion by homologous recombination of a Zeocin-resistance/Phe-sensitivity cassette flanked by 5’ and 3’ sequences of the codon of interest , which is amplified using the cdpk4-del147F x cdpk4-del147R primer pair . Recombinant bacteria were then selected on Zeocin . The recombination event was confirmed by PCR and a second round of recombination replaced the Zeocin-resistance/Phe-sensitivity cassette with a PCR product containing the S147M or S147G substitution amplified using cdpk4 S147M mut-F or cdpk4 S147G mut-F with cdpk4 S147 mutR primer pairs , respectively . Bacteria were selected on YEG-Cl kanamycin plates . Mutations were confirmed by sequencing vectors isolated from colonies sensitive to Zeocin with primers cdpk4-seq1F to cdpk4-seq4F . The modified library inserts were then released from the plasmid backbone using NotI . Constructs to generate CDPK4-2xmyc , CDPK4G2A-2xmyc , and CDPK4M57A-2xmyc parasites were derived from plasmid p150 ( Billker et al . , 2004 ) . For CDPK4-2xmyc , CDPK4G2A-2xmyc targeting constructs , cdpk4 open reading frame was amplified from genomic DNA with primers cdpk4 WT-F x cdpk4 WT-R and cdpk4 G2A-F x cdpk4 G2A-R , respectively . For CDPK4M57A-2xmyc targeting construct , cdpk4 open reading frame was amplified from genomic DNA with primers cdpk4 M57A-1F x cdpk4 M57A-1R and M57A-2F x cdpk4 M57A-2R . Fragments were assembled by Gibson cloning into p150 plasmid previously digested with NheI and ApaI . cdpk4 coding sequence was confirmed using primers cdpk4-seq1F to cdpk4-seq4F and GW1 . Resulting sequences did not show any undesired mutations and plasmids were linearized at a unique HpaI site 689 bp upstream of cdpk4 for insertion into the 5′ flanking sequence of the genomic cdpk4 locus by a single crossover . Presence of the gene in transgenic parasites was confirmed by sequencing using primer cdpk4-seq1F to cdpk4-seq4F and GW1 . Gametocytes immunofluorescence assays were performed as previously described ( Volkmann et al . , 2012 ) . For HA , c-myc and α-tubulin staining , purified cells were fixed with 4% paraformaldehyde and 0 . 05% glutaraldehyde in PBS for one hour , permeabilised with 0 . 1% Triton X-100/PBS for 10 min and blocked with 2% BSA/PBS for 2 hr . Primary antibodies were diluted in blocking solution ( rat anti-HA clone 3F10 , 1:1000; polyclonal rabbit anti-c-myc ref C3956 , 1:5000; mouse anti-α-tubulin clone DM1A , 1:1000 , all from Sigma-Aldrich ) . Anti-rat Alexa594 , anti-mouse Alexa488 , anti-rabbit Alexa 488 , Anti-rabbit Alexa594 were used as secondary antibodies together with DAPI ( all from Life technologies , Switzerland ) , all diluted 1:1000 in blocking solution . Confocal images were acquired with a LSM700 or a LSM800 scanning confocal microscope ( Zeiss ) . DNA content of microgametocytes was determined by FACS measurement of fluorescence intensity of cells stained with Vybrant dye cycle violet ( life Technologies ) . Gametocytes were purified and resuspended in 100 μl of SA . Activation was induced by adding 100 μl of modified exflagellation medium ( RPMI 1640 containing 25 mM HEPES , 4 mM sodium bicarbonate , 5% FCS , 200 µM xanthurenic acid , pH 7 . 8 ) . To rapidly block gametogenesis , 800 μl of ice cold PBS was added and cells were stained for 30 min at 4°C with Vybrant dye cycle violet and analysed using a Beckman Coulter Gallios 4 . Microgametocytes were selected on fluorescence by gating on GFP positive microgametocytes when the 820cl1 or its derivatives were used . In this case ploidy was expressed as a percentage of male gametocytes only . When 2 . 34 or its derivatives were analysed , gating was performed on both micro- and macrogametocytes and ploidy was expressed as a percentage of all gametocytes . Per sample , >50 . 000 cells were analysed . For cell fractionation , purified gametocytes were washed and resuspended in PBS , PBS and 1% Triton X-100 , PBS and 1M NaCl , or PBS and 0 . 1 M Na2CO3 [pH 11 . 5] . Cells were lysed by freezing and thawing followed by sonication on ice . Pellet and soluble fractions were separated by centrifugation for 1 hr at 14 , 000 rpm at 4°C . The solubility of ERD2 ( MRA-1 from bei resources ) and PRF from ( Plattner et al . , 2008 ) were also assessed in the different conditions as controls . Co-immunoprecipitation of CDPK4-3xHA , CDPK4-2xmyc , MCM5-3xHA , SOC1-3xHA protein complexes were performed with gametocytes fixed for 10 min with 1% formaldehyde , lysed in RIPA buffer and the supernatant was subjected to affinity purification with magnetics beads conjugated with anti-HA antibody ( Roche ) or anti-c-myc antibody ( Sigma ) . A WT control was included in parallel and proteins for which we recovered peptides in the WT control were not retained for further analysis . In CDPK4-3xHA interaction experiments at 0 and 15 s post-activation , 1 µl/mL of benzonase nuclease was added prior to immunoprecipitation . Magnetic beads were then resuspended in 50 µl of 1x NuPAGE LDS sample loading buffer containing 5 mM TCEP and incubated at 70°C for 10 min . Alkylation was carried out by addition of 2 mM iodoacetamide and incubation at room temperature in the dark for 30 min . Samples were separated by polyacrylamide electrophoresis , stained with colloidal Coomassie and processed for mass spectrometry analysis as previously described ( Pardo et al . , 2010 ) , with each gel lane excised into eight slices . Samples were re-dissolved in 0 . 5% FA before LC -MS/MS analysis . Samples were re-dissolved in 0 . 5% FA before LC -MS/MS analysis on LTQ Orbitrap Velos coupled with Ultimate 3000 RSLCnano system . Peptides were loaded and desalted on a peptide trap ( Acclaim PepMap C18 , 100 μm i . d . x 20 mm , 100 Å , 5 μm ) at 10 µl/min and then separated on a nano-analytical column ( Acclaim PepMap C18 , 75 μm i . d . x 250 mm , 100 Å , 3 µm ) at a linear gradient of 4–32% ACN/0 . 1% FA in 60 min ( cycle time 95 min ) at 300 nl/min . The HPLC , mass spectrometer and columns were all from ThermoFisher ( United Kingdom ) . The Orbitrap mass spectrometer was operated in the standard ‘top 15’ data-dependant acquisition mode while the preview mode was disabled . The MS full scan was set at m/z 380–1600 with the resolution at 30 , 000 at m/z 400 and the lock mass at 445 . 1200 and AGC at 1 × 106 with a maximum injection time at 200 msec . The 10 most abundant multiply-charged precursor ions , with a minimal signal above 3000 counts , were dynamically selected for CID fragmentation ( MS/MS ) in the ion trap , which had an AGC set at 5000 with the maximum injection time at 100 msec . The dynamic exclusion duration time was set for 45 s with ±10 ppm exclusion mass width . Raw data was processed with Proteome Discoverer 1 . 4 and searched with Mascot ( MatrixScience ) against a combined mouse and P . berghei ANKA database ( P . bergheiANKA . proteome , v2 May 2015 ) with the following parameters: trypsin/P as enzyme , maximum of 2 missed cleavages , 10 ppm parent ion mass tolerance , 0 . 5 Da fragment ion mass tolerance , and variable modifications of oxidized M , carbamidomethyl C , deamidated NQ , Gln to pyroGlu ( N-terminal Q ) , N-terminal acetylation and N-terminal formylation . Database search results were refined using Mascot Percolator ( significance threshold <0 . 05 , FDR < 1% ) . High confidence peptides were apportioned to proteins using Mascot Protein Family summary . Protein identification required at least three high-confidence unique peptide ( FDR < 1% ) . For all other interaction experiments , magnetic beads were suspended in 100 μl of 6 M Urea in 50 mM ammonium bicarbonate ( AB ) . To this solution , 2 µl of DTT ( 50 mM in LC-MS grade water ) were added and the reduction was carried out at 37°C for 1 hr . Alkylation was performed by adding 2 µl of iodoacetamide ( 400 mM in distilled water ) during 1 hr at room temperature in the dark . Urea concentration was lowered to 1M with 50 mM AB , and protein digestion was performed overnight at 37°C with 15 µL of trypsin Promega ( 0 . 2 µg/µl ) . After beads removal , the sample was desalted with a C18 microspin column ( Harvard Apparatus ) , dried under speed-vacuum , and re-dissolved in H2O/CH3CN/FA 94 . 9/5/0 . 1 before LC-ESI-MS/MS analysis . LC-ESI-MS/MS was performed on a Q-Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer ( Thermo Fisher Scientific , United Kingdom ) equipped with an Easy nLC 1000 system . Peptides were trapped on an Acclaim pepmap100 , C18 , 3 μm , 75 μm x 20 mm nano trap-column and separated on a 75 μm x 500 mm , C18 , 2 μm Easy-Spray column . The analytical separation was run for 90 min using a gradient of H2O/FA 99 . 9%/0 . 1% ( solvent A ) and CH3CN/FA 99 . 9%/0 . 1% ( solvent B ) . The gradient was run as follows: 0–5 min 95% A and 5% B , then to 65% A and 35% B in 60 min , then to 10% A and 90% B in 10 min , and finally stay at 10% A and 90% B for 15 min . The entire run was at a flow rate of 250 nL/min . ESI was performed in positive mode . For MS survey scans , the resolution was set to 70000 , the ion population was set to 3 × 106 with a maximum injection time of 100 ms and a scan range window from 400 to 2000 m/z . For MS2 data-dependent acquisition , up to fifteen precursor ions were selected for higher-energy collisional dissociation ( HCD ) . The resolution was set to 17500 , the ion population was set to 1 × 105 with a maximum injection time of 50 ms and an isolation width of 1 . 6 m/z units . The normalized collision energies were set to 27% . Peak lists were generated from raw data using the MS Convert conversion tool from ProteoWizard . The peaklist files were searched against the Plasmodium Berghei_ANKA database ( Plasmodium Genomic Resource , release 28 , 5076 entries ) using Mascot search engine ( Matrix Science , London , UK; version 2 . 5 . 1 ) . Trypsin was selected as the enzyme , with one potential missed cleavage . Fragment ion mass tolerance was set to 0 . 020 Da and parent ion tolerance to 10 . 0 ppm . Variable amino acid modification was oxidized methionine and fixed amino acid modification was carbamidomethyl cysteine . The mascot search was validated using Scaffold 4 . 7 . 3 ( Proteome Software ) . Protein identifications were accepted if they could be established at greater than 95 . 0% probability and contained at least three unique identified peptides . Purified CDPK4-3xHA gametocytes were lysed in 20 mM Hepes , 1 mM EDTA , 1 . 5% triton X-100 and protease inhibitor at pH 7 . 4 . S-acylated proteins were purified by Acyl-Resin-Assisted Capture as previously described ( Forrester et al . , 2011 ) . Briefly , following cell lysis , free sulfhydryl groups are blocked with 1 . 5% methyl methanethiosulfonate , and the lysate is then treated with the nucleophile hydroxylamine to cleave thioester bonds , with the subsequent capture of proteins containing free cysteine thiolates on thiopropyl sepharose beads . Samples with and without hydroxylamine ( -HA and +HA ) were loaded on a SDS-PAGE gel for western blot analysis . α-HA antibodies are used to detect CDPK4-3xHA , α-PRF was used as a negative control . Toxoplasma gondii tachyzoites lysates together with a α-TgGAP45 ( Frénal et al . , 2010 ) were used as a positive control for S-acylated protein enrichment . Purified gametocytes were activated with exflagellation medium for 10 s and washed twice with ice-cold PBS . Cell were resuspended in 100 μl lysis buffer ( 20 mM HEPES pH7 . 5 , 150 mM NaCl , 1% NP40 , protease phosphatase inhibitors ) and lysed on ice for 30 min . Lysates were then incubated at 37°C in 1 . 6 ml of kinase buffer ( 20 mM HEPES pH7 . 5 , 10 mM MgCl2 , 500 μM N6-PhEt-ATPγS , 3 mM GTP , 200 μM ATP , protease phosphatase inhibitors and 10 μM CaCl2 ) . After 30 min EDTA was added to a final concentration of 20 mM . Samples were then centrifuged at 20 , 000 g for 10 min at 4°C and the supernatant was kept on ice . The insoluble pellet was resuspended in 500 μl of 8 M urea , 100 mM TEAB , protease/phosphatase inhibitor and sonicated . The remaining lysate was centrifuged at 14 , 000 rpm for 5 min at 4°C . The soluble and urea extracts were combined and 3 . 6 mg of total proteins were precipitated with chloroform/methanol to remove detergents . Pellets were solubilised and denatured in 8 M urea/100 mM TEAB/45 mM TCEP at 37°C for 15 min and further diluted to just under 4 M urea with 100 mM TEAB/9 mM TCEP . Proteins were digested with Lys-C ( 1:300 , Wako ) at 37°C for 2 hr and further diluted to 1 M urea with 100 mM TEAB/9 mM TCEP and digested with trypsin ( 1:30 , Thermo Fisher ) at 37°C for 14 hr . Peptides were desalted on C18 Sep-Pak cartridges ( Waters ) and dried in SpeedVac , then resolubilised in 20 mM HEPES buffer pH 7 . 0/50% ACN/1 mM TCEP . The thiophosphopeptides capture used SulfoLink Coupling resin as described in ( Hertz et al . , 2010 ) . Briefly , the SulfoLink Coupling resin was washed with 20 mM HEPES pH 7 . 0/50% ACN ( binding buffer ) then blocked with BSA in the dark for 10 min . The resin was incubated with peptides overnight at 21°C in the dark , then washed sequentially for ten minutes with 50% ACN , H2O , 5M NaCl , 50% ACN , 5% FA , 10 mM DTT and H2O . Samples were then treated with 1 mg/ml oxone ( Sigma ) for 2 min ( twice ) to elute the bond thiophosphopeptides which were converted to phosphopeptides in this step . The eluted peptides were desalted in SDB-XC tip . Phosphopeptides were enriched on TiO2 tip ( Thermo Fisher ) and sequentially eluted with 1 . 5% NH3 and 5% pyrrolidine ( Sigma ) . Both eluates were treated with 10 mM TCEP for 10 or 15 min , respectively , and acidified before LC-MS/MS analysis on a LTQ Orbitrap Velos coupled with Ultimate 3000 RSLCnano system ( Thermo Fisher ) as for CDPK4-3xHA samples with the following differences: peptides were separated at a linear gradient for 95 min , the mass spectrometer was operated in the standard top 15 data dependent acquisition mode , and ions with a signal above 3000 counts were selected for CID fragmentation . Raw data was processed in Proteome Discoverer 1 . 4 with both SequestHT and Mascot search engines . The dynamic modifications set in Mascot were Acetyl ( Protein N-term ) , Deamidated ( NQ ) , Iodo ( Y ) , Phospho ( STY ) and Oxidation ( M ) with fragment mass tolerance set at 0 . 8 Da , while in SequestHT the selected dynamic modifications were Deamidated ( NQ ) , Iodo ( Y ) , Phospho ( STY ) with fragment ions mass tolerance set at 0 . 5 Da . Precursor mass tolerance was set at 20ppm for both . The peptide list was filtered with Percolator where the q-value was set at 0 . 01 , and the search results were merged . In the AS-CDPK4S147G samples , peptides with an XCorr and an IonScore of more than 3 . 8 or 35 , respectively , were considered as positive . The raw data were also processed by MaxQuant ( version 1 . 5 . 2 . 8 ) with most of the parameters settings at default value except the following parameters: trypsin with maximum two missed cleavages sites; Oxidation ( M ) , Deamidation ( NQ ) and Phospho ( STY ) were set as variable modification while no fixed modifications were set on Carbamidomethyl ( C ) . Cut off of 0 . 75 and 4 . 0 were retained for the localisation probability and score difference . In the AS-CDPK4S147G samples , peptides with an intensity >7 . 5×104 were considered as positive . Activated and no-activated purified parasites were snap frozen in liquid nitrogen . Cells were lysed in 4% SDS , 50 mM NaCl , 100 mM Tris buffer ( pH 7 . 4 ) , 5 mM EDTA , 40 mM TCEP and Halt phosphatase and proteinase inhibitor ( 2x , Thermo Fisher ) , and heated at 95°C for 10 min , then processed by ultrasonic probe for 20 s ( 1s on , 1 s off ) at 40% power . Samples was centrifuged at 14 , 000 rpm for 30 min then supernatant was saved . 300 µg of total protein was taken and alkylated with 20 mM IAA at RT for 45 min and precipitated with MTBE ( methyl tert-butyl ether ) . The pellet was digested with 4 µg trypsin in 100 mM TEAB at 37°C . After two hours of incubation , 4 µg of trypsin were added and samples were incubated for another five hours . 80 µg of the tryptic digest was taken and labelled with TMT10plex ( Life Technologies ) in 300 mM HEPES buffer ( pH 8 . 0 ) . Samples were pooled together and SpeedVac dried . The mixed peptides were fractionated on a 4 . 6 mm i . d . x 250 mm XBridge BEH C18 column ( 130 Å , 3 . 5 µm , Waters ) at pH 10 over a linear gradient of 5–35% ACN/0 . 1% NH3/60 min cycle time . Fractions were collected every 30 s in a 96 well plate and pooled into 12 fractions . Enrichment for phosphopeptides were then performed by immobilised affinity chromatography ( IMAC ) chromatography with PHOS-Select Iron Affinity Gel ( Sigma-Aldrich ) and then TiO2 tips ( Thermo Fisher ) . All procedures followed s manufacturer’s instruction with some modifications . 100 µl of 50% suspension of PHOS-Select Iron Affinity Gel was used for each fraction . The peptides were redissolved in 50% CH3CN/250 mM acetic acid/0 . 1% TFA ( trifluoroacetic acid ) ( binding solution ) and then added to the prewashed beads . After binding at room temperature with end-to-end rotation for 30 min , the beads were washed three times with the binding solution and once with H2O . Phosphopeptides were eluted twice with 100 µl of 1 . 5% NH3/25% ACN and then dried in SpeedVac . The flow through and the first wash of IMAC beads were collected and dried in SpeedVac , and then phosphopeptides were enriched with TiO2 tips . Phosphopeptides were eluted from the tip with 1 . 5% NH4OH and then 5% pyrrolidine . Both eluates were pooled , acidified and desalted on Graphite Spin Columns ( Thermo ) as instructed by the manufacturer’s protocol . IMAC and TiO2 enriched phosphopeptides were redissolved in 0 . 5% FA before LC-MS/MS analysis . Peptides were loaded on a trap column ( Acclaim PepMap C18 , 100 µm i . d . x 20 mm ) then separated on a 75 µm i . d . x 500 mm column ( Acclaim PepMap C18 ) with a linear gradient of 4–32% ACN/0 . 1% FA for 120 min and a total of 150 min per cycle . The Orbitrap Fusion was operated at a Top 15 method . The MS full scan was in Orbitrap ( m/z 380–1500 ) with a lock mass at 445 . 120025 , a resolution at 120 , 000 at m/z 200 , and an AGC at 4 × 105 with a maximum injection time at 50 msec . The 15 most abundant multiply-charged precursor ions ( 2+ to 6+ ) , with a minimal signal above 50 , 000 counts , were dynamically selected for HCD fragmentation ( MS/MS ) and detected in Orbitrap with a resolution at 50 , 000 at m/z 200 , and the AGC at 1 × 105 with the maximum injection time at 105 msec . The isolation width was 1 . 2 Da in quadrupole , and the collision energy was set at 38% . The dynamic exclusion duration time was set for 60 s with ±10 ppm exclusion mass width . Raw data were processed in Proteome Discoverer 2 . 1 with both SequestHT and Mascot search engines against a combined protein database of Plasmodium berghei ( P . bergheiANKA . proteome , v2 May 2015 ) and mouse ( UniprotKB ) . The dynamic modifications set in both Mascot were Acetyl ( N-term ) , Deamidated ( NQ ) , Phospho ( STY ) and Oxidation ( M ) , while in SequestHT Camabidomethyl ( C ) was set as a fixed modification . Settings were: Precursor mass tolerance at 20 ppm , fragment at 0 . 5 Da , and TMTplex as fixed modification . The search results were merged and the peptide list was filtered with Percolator where the q-value was set at 0 . 01 , and the phosphorylation sites were localised by phosphoRS implemented in the PD2 . 1 with site probability at 0 . 75 as cut-off . Both unique and razor peptides were used for protein quantification , and the reporter abundances were based on S/N , then the abundances was normalised on Total Peptide Amount , and the scaled with On Channels Average ( per file ) . The co-isolation threshold was set at 50% to reduce the isolation interference . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD005884 . PFAM domains were searched using SMART ( Letunic et al . , 2015 ) ( http://smart . embl-heidelberg . de/ ) and protein sequence similarity searches were performed using HMMER ( Finn et al . , 2011 ) ( http://hmmer . org/ ) . | Malaria is caused by parasites called Plasmodium , which are carried between humans by infected mosquitoes . The parasite needs to move between its insect and human hosts to complete its life cycle , and efforts to stop the spread of malaria are now focussed on blocking this movement . In the blood of infected humans , the parasite exists in a form known as a gametocyte , which can be male or female . When a mosquito bites and ingests the blood of a person with malaria , the gametocytes experience a sudden drop in temperature , which stimulates them to develop into sex cells . In less than ten minutes , each male gametocyte replicates its genetic material three times , divides this material into eight sex cells , and these new cells make lash-like appendages called flagella . The flagella allow the male sex cells to “swim” around the mosquito’s blood meal in search of female sex cells . Several signalling molecules help to control the development of male gametocytes . One such molecule is a protein known as CDPK4 . Drugs targeting this protein inhibited the spread of a malaria parasite that infects rodents , yet a full understanding of how the CDPK4 protein controls the development of male gametocytes remains a mystery . Here , Fang et al . reveal that in the ten minutes following a parasite entering a mosquito , CDPK4 controls three distinct events required for the development of male gametocytes . CDPK4 is required to stimulate the replication of the parasite DNA and to separate the duplicated genetic material into the new cells . CDPK4 influences both events within 30 seconds of the parasite sensing a drop in temperature , which was unexpected because these events are not usually directly connected in other organisms . Shortly after , CDPK4 is required to stimulate the sex cells’ flagella to move . Further experiments analysing the effect of CDPK4 on over 2 , 000 proteins show that CDPK4 controls the activity of multiple proteins in the parasite . Three of these proteins had not been characterized before and Fang et al . found that they are involved in critical stages leading up to the movement of the flagella . Along with inhibiting the spread of Plasmodium , drugs that target CDPK4 can inhibit the spread of many related parasites , including the parasite that causes a disease called toxoplasmosis . These findings provide a first insight into how this protein works at a molecular level and may aid the development of more effective drugs in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Multiple short windows of calcium-dependent protein kinase 4 activity coordinate distinct cell cycle events during Plasmodium gametogenesis |
Prevention of relapses is a major challenge in treating anxiety disorders . Fear reinstatement can cause relapse in spite of successful fear reduction through extinction-based exposure therapy . By utilising a contextual fear-conditioning task in mice , we found that reinstatement was accompanied by decreased c-Fos expression in the infralimbic cortex ( IL ) with reduction of synaptic input and enhanced c-Fos expression in the medial subdivision of the central nucleus of the amygdala ( CeM ) . Moreover , we found that IL dopamine plays a key role in reinstatement . A reinstatement-inducing reminder shock induced c-Fos expression in the IL-projecting dopaminergic neurons in the ventral tegmental area , and the blocking of IL D1 signalling prevented reduction of synaptic input , CeM c-Fos expression , and fear reinstatement . These findings demonstrate that a dopamine-dependent inactivation of extinction circuits underlies fear reinstatement and may explain the comorbidity of substance use disorders and anxiety disorders .
Anxiety disorders are often treated with cognitive-behavioural interventions such as exposure therapy ( McNally , 2007; Vervliet et al . , 2013 ) . Fear conditioning and extinction are used in animal models of anxiety disorders and their treatment ( Davis , 2002 ) . In extinction , conditioned responses can be reduced by prolonged presentations of conditional stimuli ( CS ) without the associated unconditional stimuli ( US ) ( LeDoux , 2000 ) . Many studies have shown that the infralimbic cortex ( IL ) is a critical brain region for extinction ( Herry et al . , 2010; Sotres-Bayon and Quirk , 2010 ) . Extinction is suppressed by pharmacological inactivation of the IL ( Laurent and Westbrook , 2009; Sierra-Mercado et al . , 2011 ) as well as by local injection of N-methyl-D-aspartate receptor ( NMDAR ) antagonists ( Burgos-Robles et al . , 2007; Sotres-Bayon et al . , 2009 ) or cannabinoid antagonists ( Lin et al . , 2009 ) into the IL . The IL inhibits the medial subdivision of the central nucleus of the amygdala ( CeM ) , a key region for fear expression ( Wilensky et al . , 2006; Ciocchi et al . , 2010 ) , partly through intercalated amygdala neurons ( ITCs ) ( Likhtik et al . , 2008; Amano et al . , 2010 ) , which are also necessary for extinction . Relapse is common in anxiety disorders . About 40% of patients in remission experience a relapse ( Bruce et al . , 2005; Ansell et al . , 2011 ) . While clinical observations have limitations on experimental control , relapse studies in the laboratory provide more information because of the greater potential for experimental manipulation ( Vervliet et al . , 2013 ) . In experimental animals , fear can be reinstated by one or more US-only presentations after successful extinction ( Rescorla and Heth , 1975; Bertotto et al . , 2006 ) . We previously reported that fear reinstatement occurs through NMDAR- and protein synthesis-dependent neural plasticity ( Shen et al . , 2013 ) . It has also been reported that fear reinstatement requires β-adrenergic receptor activation , gamma-aminobutyric acid type A receptor endocytosis , and actin rearrangement in the basolateral amygdala ( BLA ) ( Lin et al . , 2011; Motanis and Maroun , 2012 ) . However , neural circuit mechanisms responsible for fear reinstatement are poorly understood . Interestingly , the BLA–medial prefrontal cortex ( mPFC ) pathway is potentiated following fear extinction and depotentiated following reinstatement ( Vouimba and Maroun , 2011 ) . Therefore , the mPFC could be a key region mediating fear reinstatement . Nevertheless , little is known about activity changes of the mPFC and its downstream brain regions during fear reinstatement , synaptic modifications within the mPFC , and potential molecular regulation of such modifications . To identify brain regions involved in processing fear reinstatement , we mapped the regional expression of the inducible immediate early gene ( c-Fos ) . We focused on the mPFC , amygdala , and hippocampus because they are important in fear modulation and have reciprocal connections ( Quirk and Mueller , 2008 ) . In addition , we used in vitro patch-clamp recording to explore synaptic modifications within the mPFC . Building on our results from the c-Fos and electrophysiology experiments , we hypothesised that prefrontal dopamine plays a key role in reinstatement and tested this hypothesis pharmacologically . Together , these data suggest that a dopamine-dependent inactivation of extinction circuits underlies fear reinstatement .
To examine the neural circuits for fear reinstatement , we utilised a contextual fear-conditioning task , as described previously ( Shen et al . , 2013 ) ( Figure 1A ) . Mice learned an association between CS ( chamber A ) and US ( foot shocks ) on Day 1 . On Day 2 , they received a prolonged CS presentation without any US ( extinction training ) , then their freezing time gradually decreased . On Day 3 , they were re-exposed to CS to confirm retention of extinction ( test 1 ) . To reinstate the conditioned fear , they immediately received a weak US ( reminder shock ) in chamber B on Day 4 , and they were exposed to CS again on Day 5 ( test 2 ) . The mice showed higher freezing time in test 2 than they did in test 1 , suggesting successful reinstatement . Freezing time was comparable between tests 1 and 2 if mice were exposed to chamber B without a reminder shock on Day 4 ( 39 . 5 ± 4 . 2% in test 1 and 32 . 6 ± 5 . 3% in test 2 , n = 4 ) . When we gave the reminder shock to naive mice , the reminder shock alone did not induce high fear responses ( 6 . 0 ± 2 . 1% , n = 5 ) , indicating that the reminder shock-induced increase in freezing was derived from the original conditioned fear , not from new learning . 10 . 7554/eLife . 08274 . 003Figure 1 . Reinstatement is associated with low IL activity . ( A ) A reminder shock reinstated extinguished fear ( n = 10 mice; paired t-test , t ( 9 ) = 3 . 6 , p = 0 . 0059 ) . ( B ) Representative images of the infralimbic cortex ( IL ) , the ventral intercalated amygdala neurons ( ITCv ) , and the central nucleus of the amygdala ( CeM ) in the Fear , Extinction , and Reinstatement groups . ( C ) c-Fos+ cell density decreased in the IL and the ITCv and increased in the CeM with reinstatement ( n = 8–11 mice; F ( 2 , 27 ) = 4 . 3 , p = 0 . 023 [IL]; F ( 2 , 26 ) = 4 . 8 , p = 0 . 0016 [ITCv]; F ( 2 , 26 ) = 6 . 3 , p = 0 . 0058 [CeM]; Tukey's test , PExtinction vs Reinstatement = 0 . 029 [IL] , 0 . 035 [ITCv] , 0 . 013 [CeM] ) . ( D ) IL muscimol infusions resulted in high freezing ( n = 10 mice; t ( 18 ) = 2 . 4 , p = 0 . 030 ) . **p < 0 . 01 , *p < 0 . 05 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 00310 . 7554/eLife . 08274 . 004Figure 1—figure supplement 1 . Freezing behaviour of the mice subjected to c-Fos activity mapping . Mice of the Reinstatement group showed greater freezing behaviour relative to the Extinction group ( F ( 2 , 25 ) = 8 . 7 , p = 0 . 0013; Tukey's test , PFear vs Extinction = 0 . 0023 , PExtinction vs Reinstatement = 0 . 0073 ) . **p < 0 . 01 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 00410 . 7554/eLife . 08274 . 005Figure 1—figure supplement 2 . c-Fos+ cell density in the IL calculated in the analysis using a normal threshold and in an additional analysis using a strict threshold . ( n = 8–11 mice; F ( 3 , 35 ) = 5 . 9 , p = 0 . 0023 [Normal threshold]; F ( 3 , 35 ) = 3 . 0 , p = 0 . 042 [Strict threshold]; Tukey's test , PNaive vs Extinction = 0 . 0017 [Normal threshold] , 0 . 032 [Strict threshold] ) . **p < 0 . 01 , *p < 0 . 05 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 00510 . 7554/eLife . 08274 . 006Figure 1—figure supplement 3 . The density of c-Fos+ cells in the PL , LA , BA , CeL , ITCd , CA1 , CA2 , CA3 , and DG was comparable between the Extinction and Reinstatement groups ( PL , F ( 2 , 28 ) = 3 . 6 , p = 0 . 041; LA , F ( 2 , 26 ) = 1 . 4 , p = 0 . 27; BA , F ( 2 , 26 ) = 0 . 068 , p = 0 . 93; CeL , F ( 2 , 26 ) = 4 . 8 , p = 0 . 017; ITCd , F ( 2 , 27 ) = 0 . 97 , p = 0 . 39; CA1 , F ( 2 , 22 ) = 1 . 3 , p = 0 . 29; CA2 , F ( 2 , 22 ) = 0 . 29 , p = 0 . 75; CA3 , F ( 2 , 22 ) = 1 . 0 , p = 0 . 38; DG , F ( 2 , 22 ) = 0 . 46 , p = 0 . 64; Tukey's test , CeL: PFear vs Extinction = 0 . 015 ) . PL: prelimbic cortex , LA: lateral amygdala , BA: basal amygdala , CeL: lateral subdivision of central nucleus of the amygdala , ITCd: dorsal intercalated amygdala neurons . *p < 0 . 05 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 00610 . 7554/eLife . 08274 . 007Figure 1—figure supplement 4 . Histological verification of cannula placements in the experiment with muscimol infusions into the IL . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 007 To identify the brain regions involved in processing reinstatement , we employed activity mapping with c-Fos immunohistochemistry . Mice were exposed to the CS one day after reminder shock ( Reinstatement group ) . The Fear and Extinction groups were exposed to the CS one day after conditioning and one day after extinction training , respectively . The freezing time of the Reinstatement group was higher than it was in the Extinction group and comparable to that of the Fear group ( Figure 1—figure supplement 1 ) . Brains were removed 90 min later and subjected to c-Fos immunohistochemistry ( Figure 1B ) . The density of c-Fos+ cells in the IL in the Reinstatement group was lower than it was in the Extinction group and comparable to that in the Fear group ( Figure 1C ) , which was not affected by thresholding ( Figure 1—figure supplement 2 ) . Given that the IL inhibits the CeM partly through the ITC , the reduced IL activity could result in low ITC and high CeM activities . Consistent with this idea , the density of c-Fos+ cells in the ventral ITC and CeM decreased and increased , respectively , in the Reinstatement group compared to the Extinction group ( Figure 1C ) . There were no significant differences between the Extinction and Reinstatement groups in other sub-regions of the mPFC , amygdala , or hippocampus ( Figure 1—figure supplement 3 ) . These results suggest that low IL activity disinhibits the CeM during fear reinstatement . Next , we tested whether inactivation of the IL would lead to high fear responses . Mice underwent conditioning and extinction training . Muscimol , a gamma-aminobutyric acid type A receptor agonist , or a vehicle was infused into the IL 30 min before 5 min of re-exposure to the CS ( Figure 1—figure supplement 4 ) . Mice infused with muscimol showed higher freezing compared with those infused with a vehicle ( Figure 1D ) , which is consistent with previous works in rats ( Quirk et al . , 2000; Laurent and Westbrook , 2009 ) . These data suggest that inactivation of the IL is sufficient to enhance fear responses . To examine the cellular basis of lowered IL activity , we prepared brain slices 1 hr after the last test and obtained whole-cell recordings from pyramidal neurons in layer 5 of the IL . Frequency of miniature excitatory postsynaptic current ( mEPSC ) was lower in the Reinstatement group than it was in the Extinction group ( Figure 2A , B ) , while mEPSC amplitude was comparable across groups ( Figure 2C ) . Thus , excitatory synaptic inputs to the IL were decreased with reinstatement . To probe release probability , we measured paired-pulse ratio ( PPR ) by layer 2/3 stimulation . PPR was higher in the Reinstatement group than it was in the Extinction group ( Figure 2E , F ) , indicating decreased transmitter release to IL neurons . Moreover , in the Reinstatement group , increased freezing time between tests 1 and 2 was negatively and positively correlated with mEPSC frequency and PPR , respectively ( Figure 2D , G ) . These results suggest that presynaptic depression in the IL is associated with reinstatement . 10 . 7554/eLife . 08274 . 008Figure 2 . Reinstatement is associated with presynaptic depression in the IL . ( A ) Representative miniature excitatory postsynaptic current ( mEPSC ) traces . ( B ) IL neurons had lower mEPSC frequency in the Reinstatement group ( n = 8 neurons from 6 mice ) than the Extinction group ( n = 8 neurons from 4 mice ) ( F ( 2 , 21 ) = 3 . 9 , p = 0 . 037; PExtinction vs Reinstatement = 0 . 030 ) . ( C ) mEPSC amplitude did not differ across groups ( F ( 2 , 21 ) = 1 . 9 , p = 0 . 38 ) . ( D ) mEPSC frequency negatively correlated with Δfreezing ( different degrees of freezing time between tests 1 and 2 ) in the Reinstatement group ( r = −0 . 83 , p = 0 . 040 ) . ( E ) Representative traces of EPSCs evoked by paired-pulse stimulation . ( F ) IL neurons had a higher paired-pulse ratio ( PPR ) in the Reinstatement group ( n = 8 neurons from 5 mice ) than the Extinction group ( n = 8 neurons from 4 mice ) ( t ( 14 ) = 2 . 2 , p = 0 . 049 ) . ( G ) PPR positively correlated with Δfreezing in the Reinstatement group ( r = 0 . 95 , p = 0 . 012 ) . *p < 0 . 05 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 00810 . 7554/eLife . 08274 . 009Figure 2—figure supplement 1 . Intrinsic excitability of infralimbic neurons did not change with fear reinstatement . ( A ) Representative traces showing spikes in response to a 400-pA current pulse . ( B ) The number of action potentials in response to depolarizing steps at different current intensities ( n = 21 , 20 , 20 neurons ) . ( C ) The maximum number of evoked spikes at any current step increased with fear extinction ( F ( 2 , 58 ) = 4 . 1 , p = 0 . 022; Tukey's test , p = 0 . 016 ) and did not change with fear reinstatement . *p < 0 . 05 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 009 To probe intrinsic neuronal excitability , the maximum number of action potentials generated during the current injections was also compared among the groups . The maximum number of action potentials in the Reinstatement group was not significantly different from either the Extinction group or the Fear group , while that of the Extinction group was higher than that of the Fear group , consistent with a previous study using auditory fear conditioning ( Santini et al . , 2008 ) ( Figure 2—figure supplement 1 ) . Other electrophysiological properties of IL neurons in the Reinstatement group were comparable to those in the Extinction and Fear groups ( Table 1 ) . Thus , intrinsic excitability of IL neurons did not change with fear reinstatement . 10 . 7554/eLife . 08274 . 010Table 1 . Electrophysiological properties of IL neuronsDOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 010FearExtinctionReinstatementResting potential ( mV ) −70 . 7 ± 1 . 1−72 . 0 ± 1 . 0−70 . 6 ± 0 . 7Input resistance ( MΩ ) 276 . 5 ± 24 . 6391 . 9 ± 32 . 4*362 . 6 ± 30 . 9Spike amplitude ( mV ) 75 . 5 ± 1 . 572 . 9 ± 2 . 076 . 3 ± 1 . 2First interspike interval ( ms ) 7 . 9 ± 0 . 58 . 8 ± 0 . 68 . 9 ± 0 . 4Rheobase ( pA ) 78 . 1 ± 5 . 765 . 5 ± 4 . 970 . 0 ± 6 . 1Spike threshold ( mV ) −37 . 3 ± 0 . 7−37 . 2 ± 1 . 0−35 . 1 ± 0 . 6Voltage sag ( mV ) −3 . 0 ± 0 . 3−3 . 1 ± 0 . 2−3 . 8 ± 0 . 4Half width of spike ( ms ) 1 . 01 ± 0 . 030 . 98 ± 0 . 021 . 04 ± 0 . 02fAHP ( mV ) −17 . 3 ± 0 . 7−16 . 7 ± 0 . 7−16 . 7 ± 0 . 7mAHP ( mV ) −1 . 6 ± 0 . 5−0 . 9 ± 0 . 4−1 . 0 ± 0 . 5*p < 0 . 05 vs Fear , Tukey's test . fAHP , fast afterhyperpolarization; mAHP , medium afterhyperpolarization . The mPFC , including the IL , receives dopaminergic innervation from the ventral tegmental area ( VTA ) . It is reported that aversive stimuli activate VTA dopaminergic neurons ( Matsumoto and Hikosaka , 2009; Brischoux et al . , 2009 ) and elevate dopamine concentration in the PFC ( Abercrombie et al . , 1989; Hamamura and Fibiger , 1993 ) . Additionally , dopamine application with electric stimulation suppresses transmitter release onto mPFC neurons via dopamine D1 receptors ( D1Rs ) ( Law-Tho et al . , 1994; Gao et al . , 2001 ) . Therefore , we hypothesised that a reminder shock activates the VTA-to-IL circuit and that dopamine D1 signalling in the IL contributes to reduction of synaptic input onto IL neurons and subsequent fear reinstatement . In order to assess this idea , we tested whether a reminder shock induces c-Fos expression in the VTA neurons projecting to the IL . We retrogradely labelled the neurons projecting to the IL by infusing Alexa 488-conjugated cholera toxin subunit B ( CTB ) into the IL ( Figure 3A ) . Of the retrogradely labelled cells ( CTBIL+ ) in the VTA , 59 . 1 ± 4 . 5% were immunopositive for a dopamine neuron marker , tyrosine hydroxylase ( TH+ ) , indicating that they were dopaminergic . This is within the range reported in previous studies ( Margolis et al . , 2006; Lammel et al . , 2011 ) . The mice underwent conditioning , extinction training , test 1 , and were exposed to chamber B with or without a reminder shock; their brains were removed 90 min later . c-Fos and TH were immunostained and observed in the VTA ( Figure 3B ) . We found that a reminder shock increased c-Fos expression in the CTBIL+ TH+ VTA neurons ( Figure 3C ) , but not in the CTBIL+ TH− VTA neurons ( Figure 3D ) . This result suggests that a reminder shock activates dopaminergic VTA neurons projecting to the IL . 10 . 7554/eLife . 08274 . 011Figure 3 . A reminder shock activates dopaminergic VTA neurons projecting to the IL . ( A ) Coronal brain section of a mouse with Alexa 488-conjugated cholera toxin subunit B ( CTB ) infusion into the IL . ( B ) A representative immunofluorescence image of the ventral tegmental area ( VTA ) neurons with c-Fos , tyrosine hydroxylase ( TH ) , and CTB . ( C ) A reminder shock increased the proportion of c-Fos+ neurons in IL-projecting TH+ VTA neurons ( no shock: n = 7 , reminder shock: n = 6 mice; t ( 11 ) = 4 . 3 , p = 0 . 0012 ) . ( D ) A reminder shock did not increase the proportion of c-Fos+ neurons in IL-projecting TH− VTA neurons ( no shock: n = 7 , reminder shock: n = 6 mice ) . **p < 0 . 01 , Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 011 To test whether D1R signalling is involved in reinstatement , we infused a D1R antagonist , SCH23390 , or vehicle 30 min before giving the mice a reminder shock and measured their freezing time in test 2 ( Figure 4—figure supplement 1 ) . SCH23390 infusions into the IL prevented reinstatement ( Figure 4A ) . On the other hand , SCH23390 infusions into the prelimbic cortex ( PL ) , a region adjacent to the IL , did not affect reinstatement ( Figure 4—figure supplement 2 ) . These results indicate a specific role of IL dopaminergic signalling in the induction of reinstatement . Next , we examined the effect of IL D1R blockage on reduction of synaptic input onto IL neurons associated with reinstatement . Brain slices were prepared after test 2 from the mice infused with SCH23390 or vehicle into the IL before the reminder shock . mEPSC frequency was higher in the SCH23390-infused mice than it was in the vehicle-infused mice ( Figure 4B , C ) , while mEPSC amplitude was comparable ( Figure 4D ) . Thus , IL D1R blockage attenuated reduction of synaptic input associated with reinstatement . Finally , we examined the effect of IL D1R blockage on c-Fos expression in the amygdala during test 2 . Brains were removed 90 min after test 2 from the mice infused with SCH23390 or vehicle into the IL before the reminder shock . SCH23390-infused mice showed higher and lower c-Fos expression than vehicle-infused mice in the ventral ITC and CeM , respectively ( Figure 4E ) . Thus , we concluded that IL D1R blockage prevents c-Fos expression changes in the amygdala associated with reinstatement . Taken together , these results indicate that D1R signalling in the IL is necessary for reduction of synaptic inputs , CeM disinhibition , and reinstatement . 10 . 7554/eLife . 08274 . 012Figure 4 . D1Rs in the IL mediate reinstatement . ( A ) SCH23390 infusions into the IL before a reminder shock suppressed reinstatement ( phosphate-buffered saline [PBS]: n = 15 , SCH23390: n = 14 mice; t ( 27 ) = 2 . 2 , p = 0 . 039 ) . ( B ) Representative mEPSC traces . ( C ) SCH23390-infused mice demonstrated higher mEPSC frequency ( n = 9 , 10 neurons; t ( 17 ) = 2 . 2 , p = 0 . 044 ) . ( D ) SCH23390 infusions had no effects on mEPSC amplitude . ( E ) SCH23390-infused mice demonstrated higher and lower c-Fos+ cell density in the ventral intercalated amygdala neurons ( ITCv ) and the central nucleus of the amygdala ( CeM ) , respectively ( n = 7–8 mice; ITCv: t ( 13 ) = 3 . 0 , p = 0 . 0093; CeM: t ( 14 ) = 2 . 9 , p = 0 . 011 ) . **p < 0 . 01 , *p < 0 . 05 . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 01210 . 7554/eLife . 08274 . 013Figure 4—figure supplement 1 . Histological verification of cannula placements in the experiment with SCH23390 infusions into the infralimbic ( A ) and the prelimbic ( B ) cortices . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 01310 . 7554/eLife . 08274 . 014Figure 4—figure supplement 2 . SCH23390 infusions into the prelimbic cortex had no effects on reinstatement ( n = 9 mice ) . Data represent mean ± standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 08274 . 014
Prevention of relapse is a challenge in treating anxiety disorders . Fear reinstatement can cause relapse after exposure therapy . Accordingly , we investigated the neural circuit mechanism of fear reinstatement . We found that a reminder shock decreased IL and ventral ITC activity and increased CeM activity as indexed by c-Fos expression . Reinstatement was accompanied by presynaptic depression of transmitter release onto IL neurons . Moreover , we found that a reminder shock activated IL-projecting dopaminergic neurons in the VTA , and the blocking of IL D1R signalling prevented reduction of synaptic inputs , activity changes of ventral ITC and CeM , as well as fear reinstatement . These findings suggest that a dopamine-dependent inactivation of extinction circuits underlies fear reinstatement . We compared c-Fos expression induced by an exposure to the experimental context before and after a reminder shock ( Extinction group and Reinstatement group ) in order to identify brain regions involved in processing fear reinstatement . Activity mapping with c-Fos is a well-established and often used method to identify brain regions that are involved in processing motivation behaviour , social behaviour , learning and memory , and so on ( Zhang and Kelley , 2000; Tronel and Sara , 2002; Frankland et al . , 2004; Veening et al . , 2005; Makino et al . , 2015 ) . Other studies reporting that neuronal activation patterns detected by c-Fos expression and by functional magnetic resonance imaging correlate well ( Lu et al . , 2004; Lazovic et al . , 2005 ) also support the utility of c-Fos immunohistochemistry to determine changes in neuronal activity . It is important to note , however , that c-Fos expression is not the same as neuronal firing activity . Because c-Fos expression is dependent on an increase in Ca2+ ( Lerea et al . , 1992 ) , firing activity does not always result in c-Fos elevation ( Clayton , 2000 ) . Further studies using in vivo electrophysiological recordings ( such as local field potential ) are needed to directly reveal the neuronal firing activity during reinstatement . We found a remarkable decrease in c-Fos expression followed by reinstatement in the IL but not in the PL ( Figure 1C , Figure 1—figure supplement 3 ) . Both the IL and PL of the mPFC have important roles in fear regulation . The PL promotes fear responses . As indicated by our data and previous reports , neuronal activity of the PL , which is elevated during fear retrieval , is not elevated ( or lowered ) during extinction retrieval ( Burgos-Robles et al . , 2009; Sotres-Bayon et al . , 2012 ) . On the other hand , the IL increases its activity during extinction retrieval ( Milad and Quirk , 2002 ) , suggesting the negative control of fear expression by the IL . In this study , we found that c-Fos expression in the Reinstatement group , compared with the Extinction group , was not changed in the PL , and it was lowered in the IL . Our findings suggest that fear reinstatement is regulated by downregulation of extinction circuits . Low IL activity may cause ITC downregulation and CeM upregulation to reinstate fear . We found that reinstatement was accompanied by low IL and ventral ITC c-Fos expression ( Figure 1C ) . Local injection of muscimol into the IL resulted in higher freezing ( Figure 1D ) , indicating that IL inactivation is sufficient for reappearance of extinguished fear . The IL has direct projections to the ITC ( Cho et al . , 2013 ) , which is critically involved in fear extinction ( Likhtik et al . , 2008; Busti et al . , 2011; Mańko et al . , 2011 ) . The ITC provides feedforward inhibition of output neurons in the CeM ( Royer et al . , 1999 ) . Therefore , low activity of the IL and ITC could result in disinhibition of the CeM . Accordingly , we found that reinstatement was accompanied by elevated c-Fos expression in the CeM ( Figure 1C ) . The CeM projects to brain structures controlling conditioned fear responses , including the periaqueductal grey and the ventromedial and lateral hypothalamus ( Hopkins and Holstege , 1978; Veening et al . , 1984; Cassell et al . , 1986 ) , and CeM activity is necessary and sufficient for expression of freezing ( Wilensky et al . , 2006; Ciocchi et al . , 2010 ) . Thus , disinhibition of the CeM by downregulation of IL–ITC pathway may underlie fear reinstatement . Another possible pathway is through the lateral subdivision of the central nucleus of the amygdala ( CeL ) , which also receives projections from the IL and contains two functionally distinct subpopulations of neurons ( CElon or somatostatin-expressing neurons , and CEloff or protein kinase C-δ-expressing neurons ) forming local inhibitory circuits to inhibit CeM activity ( Ciocchi et al . , 2010; Haubensak et al . , 2010; Li et al . , 2013 ) . Although we did not find a significant difference between the Extinction and Reinstatement groups in the CeL ( Figure 1—figure supplement 2 ) , further analysis with a distinction between those two populations might reveal the activation of CElon neurons , which suppress CEloff neurons and lead to disinhibition of the CeM . Our results provide a novel insight into the prefrontal dopaminergic modulation of an aversive memory . Many studies have shown that dopamine neurons are activated in response to appetitive stimuli , and that dopamine signalling affects reward-related behaviour and memory ( Schultz , 1997 ) . Although it has also been reported that aversive stimuli activate midbrain dopaminergic neurons ( Brischoux et al . , 2009; Matsumoto and Hikosaka , 2009 ) , the exact role of dopamine released in aversive situations is poorly understood . We found that reinstatement-inducing stimuli elevated c-Fos expression in the dopaminergic VTA neurons projecting to the IL ( Figure 3C ) . Moreover , blocking dopamine signalling in the IL prevented fear reinstatement ( Figure 4A ) , suggesting the critical role of dopamine in reviving the aversive memory . Further studies are required to identify how VTA dopaminergic neurons may be recruited by a reminder shock . Anatomically , VTA dopaminergic neurons that receive innervations from the lateral habenula preferentially project to the mPFC ( Lammel et al . , 2012 ) . The habenula receives input from limbic system and this circuit is implicated in aversive information processing ( Hikosaka , 2010; Okamoto et al . , 2012 ) . Thus , it is possible that a reminder shock activates the habenula–VTA circuit and subsequent dopaminergic signalling in the IL . We found that reinstatement was accompanied by dopamine D1R-dependent reduction of synaptic input in the IL ( Figure 4B , C ) . mEPSC frequency was low and paired-pulse ratio was high in the IL neurons of the Reinstatement group , and IL D1R blockage reversed the low mEPSC frequency . An additional experiment using minimal stimulation would be helpful to examine changes in release probability upon fear reinstatement . Previous in vitro studies also revealed the inhibitory effects of prefrontal D1Rs by pharmacological and genetic manipulations . Dopamine attenuates excitatory synaptic transmission in prefrontal neurons in a D1R-dependent manner ( Gao et al . , 2001; Mair and Kauer , 2007 ) . Application of dopamine , combined with electrical stimulation , can induce long-term synaptic depression in the mPFC ( Law-Tho et al . , 1995; Huang et al . , 2004 ) ; however , the phenomenon is not observed in heterozygous D1R knockout mice ( Huang et al . , 2004 ) . The signalling mechanisms underlying dopamine-mediated presynaptic depression remain to be determined . One possible mechanism is that dopamine triggers adenosine release and subsequently activates presynaptic adenosine A1 receptors . D1Rs and adenosine-mediated presynaptic depression have also been reported in the VTA of guinea pigs ( Bonci and Williams , 1996 ) and in the basal ganglia of zebra finches ( Ding et al . , 2003 ) . The present findings suggest a possible dopaminergic mechanism of fear reinstatement as follows: a reminder shock activates VTA dopaminergic neurons projecting to the IL , and dopamine D1R signalling lowers IL activity with presynaptic depression , which could result in low activity of the ventral ITC , thereby disinhibiting CeM and reinstating a once extinguished fear . Previous studies have shown that drugs of abuse increase dopamine release in both animals ( Di Chiara and Imperato , 1988; Chen et al . , 1990 ) and humans ( Laruelle et al . , 1995; Volkow et al . , 1999; Drevets et al . , 2001 ) . This dopaminergic mechanism of reinstatement may explain the high rate of comorbid substance use disorders with anxiety disorders ( Kendler et al . , 1996 ) .
All experiments were approved by the animal experiment ethics committee at The University of Tokyo ( approval number 24-10 ) and were in accordance with The University of Tokyo guidelines for the care and use of laboratory animals . Male C57BL/6J mice ( 8–15 weeks old; SLC , Shizuoka , Japan ) were housed in group cages of four under standard laboratory conditions ( 12-hr light/12-hr dark cycle , with light from 7 a . m . to 7 p . m . and free access to food and water ) . Mice were handled daily for 1 week and housed individually for 2 days before behavioural procedures . All behavioural procedures were performed between 9 a . m . and 2 p . m . Behavioural procedures for fear conditioning , extinction , and reinstatement were performed in accordance with our previous protocol ( Shen et al . , 2013 ) . For contextual fear conditioning , after a 150-s acclimation period in transparent rectangular conditioning chamber A ( 18 cm wide , 15 cm deep , 27 cm high ) with white light and a stainless steel grid floor , 3 shocks ( 1 mA , 2 s ) were delivered through a shock scrambler ( SGS-003DX; Muromachi Kikai , Tokyo , Japan ) with a 150-s interval between shocks . Mice were left in the chamber for an additional 60 s and then returned to their home cages . The entire duration of this session was 510 s . For extinction training and testing , mice were placed in chamber A without any shocks for 40 min and 5 min , respectively . A reminder shock ( 0 . 6 mA , 2 s ) was given immediately after the mice were placed in white triangular chamber B ( 22 cm wide , 19 cm deep , 27 cm high ) with red light and a stainless steel grid floor . The mice then returned to their home cages . Unless otherwise mentioned , testing sessions were 5 min . Mice in the Fear group underwent contextual fear conditioning on Day 1 and testing on Day 2 . Mice in the Extinction group underwent contextual fear conditioning on Day 1 , extinction training on Day 2 , and testing on Day 3 . Mice in the Extinction group of the PPR experiment underwent contextual fear conditioning on Day 1 , extinction training on Day 2 , testing on Day 3 , exposure to chamber B on Day 4 , and testing in chamber A on Day 5 . Mice in the Reinstatement group underwent contextual fear conditioning on Day 1 , extinction training on Day 2 , testing on Day 3 , a reminder shock on Day 4 , and testing on Day 5 . Each session was video recorded for automatic scoring of freezing according to a previously described method ( Nomura and Matsuki , 2008 ) . Freezing was defined as the absence of all movement except those related to breathing . Naive mice were kept in their home cages and were not exposed to the conditioning apparatus . The numbers of mice used in the c-Fos immunohistochemistry are as follows: ( naive , Fear , Extinction , Reinstatement ) = ( 9 , 11 , 11 , 9 ) in PL and IL; ( 8 , 10 , 11 , 8 ) in the lateral and basal nuclei of the amygdala , CeM and CeL; ( 8 , 11 , 11 , 8 ) in dorsal ITC; ( 8 , 11 , 10 , 8 ) in ventral ITC; and ( 8 , 8 , 9 , 8 ) in CA1 , CA2 , CA3 , and dentate gyrus . Mice were perfused intracardially with phosphate-buffered saline ( PBS ) followed by 4% paraformaldehyde 90 min after behavioural tests . Brains were removed and stored in the same fixative for 8 hr at 4°C and subsequently immersed in 20% and 30% sucrose for 24 hr and 48 hr at 4°C . The immunocytochemical staining was performed on 40-μm thick free-floating sections that were prepared using a cryostat ( HM520; Thermo Fisher Scientific , Waltham , MA , USA ) . For c-Fos staining with diaminobenzidine ( DAB ) , the sections were incubated in 0 . 2% Triton-X-100 for 15 min and 0 . 03% H2O2 for 30 min . The sections were incubated with a polyclonal anti-c-Fos antibody ( Anti-c-Fos ( Ab-5 ) ( 4–17 ) rabbit , 1:5000 , Calbiochem , San Diego , CA , USA ) for 48 hr at 4°C , goat anti-rabbit biotinylated secondary antibody ( BA-1000 , 1:500; Vector Laboratories , Burlingame , CA , USA ) for 2 hr , VECTASTAIN ABC Kit ( Vector Laboratories ) for 1 . 5 hr , and DAB solution ( 349-00903 , 0 . 03% , Wako , Osaka , Japan ) with 0 . 01% H2O2 for 7–10 min . The sections were mounted on slides , air-dried , dehydrated in ethanol solutions and xylene , and cover slipped with marinol . Images of the mPFC ( bregma 2 . 2 to 1 . 5 mm ) , amygdala ( bregma −1 . 2 to −1 . 8 mm ) , and hippocampus ( bregma −1 . 5 to −2 . 0 mm ) were acquired using a microscope ( Leica AF6000 , 10× objective lens [NA , 0 . 3] , Leica , Germany ) . All cell counting experiments were conducted blind to experimental group . The quantification of c-Fos-positive cells was performed with ImageJ software ( Scion , Frederick , MD , United States ) . c-Fos immunoreactive cells were counted bilaterally using at least three sections for each area . Sub-regions of the mPFC , amygdala , and hippocampus were outlined as a region of interest ( ROI ) according to the Paxinos and Franklin atlas . c-Fos-positive nuclei were counted relative to a counting threshold based on staining intensity and target size . The parameters of the counting threshold were set based on a standard control slide from each staining run . The mean density in each structure for each animal was divided by the mean density in that region for the naive control group in order to generate a normalized density for each animal . These normalized data were expressed as a percentage , and these percentages were averaged across mice in order to produce the mean of each group . For fluorescence immunohistochemistry , the sections were incubated with primary antibodies , including a polyclonal anti-c-Fos antibody ( 1:1000 ) and mouse anti-tyrosine hydroxylase antibody ( MAB318 , 1:500; Millipore , MA , United States ) , for 24 hr at 4°C , and secondary antibodies , including a goat anti-rabbit biotinylated antibody ( BA-1000 , 1:500; Vector Laboratories ) and Alexa Fluor 405 goat anti-mouse IgG secondary antibody ( A31553 , 1:400; Life Technologies , CA , United States ) for 2 hr , VECTASTAIN ABC Kit ( Vector Laboratories ) for 1 . 5 hr , and TSA-Cyanine 3 ( SAT704A001EA , 1:1000; Perkin–Elmer , Waltham , MA , USA ) for 1 hr . The sections were mounted in PermaFluor ( ThermoShandon , Pittsburgh , PA , United States ) . Images of the VTA ( bregma −2 . 9 to −3 . 4 mm ) were acquired using a confocal microscope ( CV1000 , 40× objective lens ( NA , 1 . 3 ) ; Yokogawa , Tokyo , Japan ) . All cell counting experiments were conducted blind to experimental group . The quantification of c-Fos-positive cells was performed with ImageJ software ( Scion ) . CTB positive and TH and c-Fos immunoreactive cells were counted bilaterally using at least five sections ( 374 cells from 13 mice ) . The VTA were outlined as an ROI according to the Paxinos and Franklin atlas . The number of c-Fos-positive cells in the CTB+ and TH+ cells was calculated by thresholding c-Fos immunoreactivity above background levels . The percentage for each animal was averaged across mice in order to produce the mean of each group . Under intraperitoneal xylazine ( 10 mg/kg ) and pentobarbital ( 2 . 5 mg/kg ) anaesthesia , 26-gauge stainless steel guide cannulas ( Plastics One , Roanoke , VA , United States ) were implanted aimed at the IL ( A/P 1 . 7 mm , L/M ±0 . 3 mm , D/V −3 . 0 mm ) or the PL ( A/P 2 . 0 mm , L/M ±0 . 3 mm , D/V −2 . 2 mm ) . These cannulas were secured to the skull using a mixture of acrylic and dental cement , and 33-gauge dummy cannulas were then inserted into each guide cannula to prevent clogging . Mice were given at least 7 days of postoperative recovery time . Mice underwent fear conditioning on Day 1 , extinction training on Day 2 , testing on Day 3 , and received bilateral infusions of PBS or muscimol ( 0 . 25 µg/side ) into the IL 30 min before testing on Day 4 . Mice underwent fear conditioning on Day 1 , extinction training on Day 2 , testing on Day 3 , received bilateral infusions of PBS or SCH23390 ( 1 µg/side ) into the IL or PL 30 min before the reminder shock on Day 4 , and testing on Day 5 . The numbers of mice used in the c-Fos immunohistochemistry are as follows: ( PBS , SCH23390 ) = ( 8 , 8 ) in CeM and ( 7 , 8 ) in ventral ITC . Alexa 488-conjugated CTB ( 0 . 5 μg/side , Life Technologies ) was infused into the IL 3 days before the beginning of behavioural procedures . Mice underwent fear conditioning on Day 1 , extinction training on Day 2 , testing on Day 3 , and exposed to the chamber B with or without reminder shock on Day 4 ( reminder shock group and no shock group , respectively ) . Brains were removed 90 min later . Infusions were made over 2 min , and the infusion cannulas ( 28 gauge , extending 0 . 5 mm below the guide cannula ) were left in place for at least 1 min afterwards . Mice were deeply anaesthetised with diethyl ether and decapitated 60–90 min after re-exposure to the conditioning context . Brains were removed quickly , and 300-µm thick coronal slices containing the IL were prepared with a vibratome ( VT 1200S , Leica ) in ice-cold , oxygenated ( 95% O2/5% CO2 ) modified artificial cerebrospinal fluid containing 222 . 1 mM sucrose , 27 mM NaHCO3 , 1 . 4 mM NaH2PO4 , 2 . 5 mM KCl , 0 . 5 mM ascorbic acid , 1 mM CaCl2 , and 7 mM MgSO4 . Picrotoxin ( 100 µM ) was added to ACSF ( artificial cerebrospinal fluid ) ( 127 mM NaCl , 1 . 6 mM KCl , 1 . 24 mM KH2PO4 , 1 . 3 mM MgSO4 , 2 . 4 mM CaCl2 , 26 mM NaHCO3 , 10 mM glucose ) to block inhibitory synaptic currents . Whole-cell patch-clamp recordings were performed with glass microelectrodes ( 3–6 MΩ ) filled with internal solution ( 120 K-gluconate , 5 KCl , 10 4- ( 2-hydroxyethyl ) -1-piperazineethanesulfonic acid , 1 MgCl2 , 10 phosphocreatine-Na2 , 2 MgATP , 0 . 1 Na2GTP , 0 . 2 ethylene glycol tetraacetic acid , pH 7 . 2–7 . 3 , 280–295 mOsm ) . For electrical stimulation , a pipette with a large tip ( ∼3 μm ) was filled with ACSF and placed in layer 2/3 . Brief current pulses ( 50 μs , 1–40 μA ) were delivered with a stimulation-isolation unit ( Nihon Kohden , Tokyo , Japan ) . Paired stimuli were given with an interstimulus interval of 50 ms , and the ratio between the amplitude of the second and first EPSCs was calculated . mEPSCs were recorded at a holding potential of −70 mV in the presence of tetrodotoxin ( TTX , 1 µM ) . mEPSCs were detected using an in-house MATLAB programme and were defined as inward currents with amplitudes greater than 7 pA ( Miura et al . , 2012 ) . To examine the intrinsic excitability , IL neurons were injected with 800-ms depolarizing current pulses ranging from 40 pA to 400 pA at 40-pA increments . The number of action potentials evoked by each current intensity was counted . The amplitude of fast afterhyperpolarization was calculated as the difference between the minimum potential after the second evoked spike within the 800-ms pulse and the spike threshold . To measure the medium afterhyperpolarization ( mAHP ) , cells were held at −70 mV , and the 800-ms pulse , which evoked two action potentials , was injected . The amplitude of mAHP was calculated as the difference between the negative peak of the potential after the end of the 800-ms pulse and the resting membrane potential ( −70 mV ) . To measure the voltage sag , a hyperpolarizing current pulse of 200 pA was injected in current-clamp mode . The voltage sag was calculated by subtracting the average steady-state voltage during a 100-ms period beginning 645 ms after the beginning of the hyperpolarizing step minus the peak of the hyperpolarization . Input resistance was calculated by current response to a 10-mV , 30-ms depolarizing pulse in voltage-clamp mode . Data were sampled at 20 kHz and filtered at 2 kHz using an Axopatch 700B amplifier ( Axon Instruments , Foster , CA , United States ) , Digidata 1440A ( Axon Instruments ) , and pClamp 10 . 2 ( Molecular Devices , Sunnyvale , CA , United States ) . All data were acquired , stored , and analysed using Clampex 10 , Clampfit , and MATLAB . All values are reported as mean ±SEM . Repeated measures analysis of variance ( ANOVA ) , Tukey's test after one-way ANOVA , Student's t-tests , and paired t-tests were performed to identify significant differences . | Anxiety disorders affect millions of people worldwide . While many people with anxiety disorders can recover with appropriate treatment , about 40% of these individuals will encounter a relapse of their condition . Researchers can investigate the causes of relapses by creating animal models of the processes involved . For example , if a mouse receives a small shock every time it enters a particular cage , it will learn to associate that cage with the shock . Once this association has been created , it can be ‘undone’ using a procedure called extinction . In the cage example , this may be performed by placing the mouse in the cage for a long time , but without giving it any shocks . Over time , the animal learns that the cage is no longer linked to an unpleasant outcome . However , if a mouse is given a reminder shock after extinction has occurred , the original association between the cage and the shock is re-established . This is known as fear reinstatement and is similar to a relapse . A number of brain regions are thought to be involved in fear reinstatement . One such region , the amygdala , is heavily involved in fear responses . It is thought that another part of the brain , the medial prefrontal cortex ( mPFC ) , can suppress the amygdala's responses , consequently reducing the animal's anxiety . While we have a good idea of which parts of the brain are involved in fear processing , we don't yet know how they work together to create a relapse . Hitora-Imamura et al . used the aforementioned method of selectively giving mice small shocks when they entered cages to induce fear , extinction , and fear reinstatement and examined how this affected the mice's brain activity . As expected , fear could be linked to activity in the amygdala . During extinction , high levels of activity in the medial prefrontal cortex suppressed the amygdala's response . When the mice experienced the reminder shock , a chemical called dopamine was released . When dopamine entered the medial prefrontal cortex , the region's activity was reduced , removing the ‘brakes’ from the amygdala and reinstating the mice's fear . The finding that dopamine is involved in fear reinstatement is particularly important , as many commonly abused drugs are known to increase levels of dopamine in the brain . Dopamine's role in fear reinstatement may explain why substance abuse is so closely linked to anxiety disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Prefrontal dopamine regulates fear reinstatement through the downregulation of extinction circuits |
Mitochondrial division is important for mitochondrial distribution and function . Recent data have demonstrated that ER–mitochondria contacts mark mitochondrial division sites , but the molecular basis and functions of these contacts are not understood . Here we show that in yeast , the ER–mitochondria tethering complex , ERMES , and the highly conserved Miro GTPase , Gem1 , are spatially and functionally linked to ER-associated mitochondrial division . Gem1 acts as a negative regulator of ER–mitochondria contacts , an activity required for the spatial resolution and distribution of newly generated mitochondrial tips following division . Previous data have demonstrated that ERMES localizes with a subset of actively replicating mitochondrial nucleoids . We show that mitochondrial division is spatially linked to nucleoids and that a majority of these nucleoids segregate prior to division , resulting in their distribution into newly generated tips in the mitochondrial network . Thus , we postulate that ER-associated division serves to link the distribution of mitochondria and mitochondrial nucleoids in cells .
The distribution of mitochondria and mitochondrial DNA ( mtDNA ) is accomplished through the engagement of multiple pathways , including mitochondrial division , fusion , motility , and tethering . Mitochondrial division is mediated by a dynamin-related protein , Dnm1 ( in yeast ) /Drp1 ( in mammals ) ( Lackner and Nunnari , 2009 ) , which self-assembles in a regulated manner on the surface of the mitochondrial outer membrane into helices that mediate mitochondrial scission ( Ingerman et al . , 2005; Mears et al . , 2011 ) . Dnm1 and Drp1 form helices in vitro , whose diameters are significantly smaller than the diameter of unconstricted mitochondria ( Ingerman et al . , 2005; Bossy et al . , 2010 ) . This observation suggests that a mechanism for Dnm1/Drp1-independent constriction may exist to facilitate helix assembly as a first step toward mitochondrial division . We recently discovered that ER tubules wrap around mitochondria and mark a majority of mitochondrial division sites in yeast and mammalian cells ( Friedman et al . , 2011 ) . These findings suggest that the process of ‘ER-associated Mitochondrial Division’ ( ERMD ) facilitates the creation of mitochondrial constriction sites , or geometric ‘hot spots’ , for Dnm1/Drp1 helix assembly . Consistent with this model , the association of ER tubules with mitochondrial constriction sites is independent of mitochondrial division components ( Friedman et al . , 2011 ) . ERMD is also independent of known ER tubule-shaping proteins and Mfn2-mediated ER–mitochondria contacts ( de Brito and Scorrano , 2008; Friedman et al . , 2011 ) . Indeed , the composition and biogenesis of ERMD sites are unknown . A candidate for tethering ER and mitochondria at ERMD sites is the ER-Mitochondria Encounter Structure ( ERMES ) —a multiprotein complex localized at an interface between the ER and mitochondria in budding yeast cells ( Kornmann et al . , 2009; Toulmay and Prinz , 2012 ) . The ERMES complex is composed of four core subunits , each of which is required for the formation of multiple ERMES foci per cell: Mdm10 and Mdm34 are integral to the mitochondrial outer membrane; Mdm12 is predicted to be cytosolic; and Mmm1 is an ER transmembrane protein ( Kornmann et al . , 2009; Stroud et al . , 2011 ) . ERMES is thought to function as a physical ER–mitochondria tether that serves to distribute mitochondria ( Boldogh et al . , 2003 ) and to facilitate the exchange of lipids between the two organelles ( Kornmann et al . , 2009; Voss et al . , 2012 ) . ERMES foci are localized adjacent to a subset of nucleoids , which are engaged in replicating DNA ( Hobbs et al . , 2001; Hanekamp et al . , 2002; Meeusen and Nunnari , 2003 ) , suggesting that ERMES plays an active role in nucleoid regulation . In agreement with this notion , deletion of core ERMES components disrupts nucleoid structure and transforms the tubular mitochondrial network into spherical mitochondria with relatively large diameters ( Burgess et al . , 1994; Youngman et al . , 2004 ) . The ERMES-deficient mitochondrial morphology phenotype is epistatic to the characteristic net-like mitochondrial structures that result from the deletion of genes encoding mitochondrial division components ( Youngman et al . , 2004 ) , consistent with ERMES functioning upstream of the mitochondrial division machinery in mitochondrial distribution . Cytological , biochemical , and genetic data indicate that the highly conserved Miro GTPase Gem1 is associated with ERMES at steady state ( Kornmann et al . , 2011; Stroud et al . , 2011 ) . In yeast cells lacking Gem1 , mitochondria form a distinct spectrum of abnormal structures , from tubules to clustered spheres , mitochondrial distribution into daughter cells is less efficient than that of wild-type mitochondria , and cells lose mitochondrial DNA at a significantly higher frequency ( Frederick et al . , 2004; Koshiba et al . , 2011 ) . Gem1 , however , is not an essential structural component of ERMES , as ERMES foci are observed in gem1Δ cells . However , in gem1Δ cells , it has been reported that ERMES foci are larger in size and fewer in number ( Kornmann et al . , 2011 ) , suggesting that Gem1 acts as a regulator of ERMES . Here , we show that ERMES and Gem1 are spatially and functionally linked to ERMD . Our data suggest a model where ERMES and Gem1 function in ERMD to facilitate the engagement and resolution of ER–mitochondria contacts , respectively . Our data indicate that Gem1 acts as a negative regulator of ER–mitochondria contacts at ERMD sites and is required for the spatial resolution of newly formed mitochondrial tips generated by a division event . We also provide evidence that ERMD positions division sites adjacent to mitochondrial nucleoids to bias their distribution into newly generated tips following division . Thus , we postulate that ERMD evolved to link the distribution of nucleoids and mitochondria .
To assess whether ERMES foci are spatially linked to mitochondrial division , we simultaneously imaged fluorescent protein ( FP ) –tagged ERMES components and mitochondria over time in living yeast cells . To this end , we expressed functional C-terminal GFP fusions of Mdm12 , Mdm34 , or Mmm1 from their respective endogenous loci and labeled mitochondria using mitochondrial matrix-targeted DsRed ( mito-DsRed ) . We observed a majority of mitochondrial division events associated with Mdm12 , Mdm34 , or Mmm1-labeled ERMES foci , which subsequent to the division segregate to only one of the two newly generated tips in the mitochondrial network ( Figure 1A , B ) . The fraction of division events associated with ERMES foci ( 54–60% ) is significantly higher than that predicted for a random association ( approximately 10% ) , based on the surface area of mitochondria associated with ERMES ( Figure 1B and ‘Materials and methods’ ) . The detection of ERMES foci by fluorescence imaging may underestimate ERMES foci associated with mitochondrial division as we observed a positive correlation between the percentage of mitochondrial divisions associated with a given ERMES subunit and its estimated whole-cell abundance ( Huh et al . , 2003 ) . However , we cannot rule out the possibility that non-ERMES–associated mitochondrial division sites exist in cells . 10 . 7554/eLife . 00422 . 003Figure 1 . ERMES marks sites of ER-associated mitochondrial division . ( A ) A timelapse series of Mdm34-yEGFP and mito-DsRed . For clarity , the images represent a maximum projection of a 2-μm section of a cell that accurately represents a mitochondrial division observed in a single plane . The first frame represents a whole-cell projection containing an inset that is shown in the remaining frames . Below each frame of the timelapse series is a plot of pixel intensity vs distance ( line-scan ) of mito-DsRed fluorescence signal as indicated by the dashed line at 0 s . The range of the y-axis is 260 arbitratry units and the range of the x-axis is 151 pixels . ( B ) Quantification of ERMES-associated mitochondrial divisions in wild-type cells using data collected as described above . ( C ) ERMES foci are associated with ER tubules at mitochondrial division sites . Cells expressing Mdm34-yEGFP , DsRed-HDEL , and mito-TagBFP were imaged in a single plane every 10 s . The first frame represents a whole-cell projection containing an inset that is shown in the remaining frames . ( D ) ERMES foci are adjacent to Dnm1 at the mitochondrial division sites . Cells expressing Mdm34-yEGFP , Dnm1-mCherry , and mito-TagBFP were imaged in a single focal plane every 10 s ( E ) gem1Δ cells expressing GFP-2xFLAG-Gem1 , Mdm34-mCherry , and mito-TagBFP were imaged in a single focal plane every 10 s . Expression of the mito-TagBFP construct in ( C–E ) was induced by growing the cells overnight to mid-log phase in synthetic media containing 2% galactose . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 00310 . 7554/eLife . 00422 . 004Figure 1—figure supplement 1 . ERMES subunits colocalize at mitochondrial division sites . ( A ) and ( B ) A wild-type strain expressing Mdm12-GFP ( A ) or Mmm1-GFP ( B ) and Mdm34-TdTomato from their endogenous loci and mito-TagBFP from an episomal plasmid was imaged over the course of 1 . 75 min on an OMX microscope as described in the ‘Materials and methods’ . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 004 In light of a previous report suggesting that the ERMES components Mdm34 and Mmm1 are not strictly colocalized ( Youngman et al . , 2004 ) , we also addressed whether different types of ERMES complexes/foci exist that could be preferentially associated with mitochondrial division events . Examination of differentially FP-tagged Mdm34 and Mmm1 or Mdm12 indicated that all components colocalize at mitochondrial division sites ( Figure 1—figure supplement 1 ) , consistent with the finding that each core ERMES subunit is necessary for assembly of the complex ( Kornmann et al . , 2009 ) . In addition , these two components associated with the same mitochondrial tip generated by division , indicating that the complex remains intact during the division process ( Figure 1—figure supplement 1 ) . Together , these observations demonstrate that mitochondrial division events are spatially linked to the ERMES complex . We assessed the relationship of ERMES to ERMD by imaging cells expressing Mdm34-yEGFP , an ER marker ( DsRed-HDEL ) , and a mitochondrial matrix-targeted blue fluorescent protein ( mito-TagBFP ) . As shown in a representative series of time lapse images in Figure 1C , we observed ERMES foci at interfaces between the ER and mitochondria and subsequently at ER-associated mitochondrial constriction and eventual division events . Consistent with this observation , the yeast mitochondrial division machine , marked by Dnm1-mCherry , was observed adjacent to Mdm34-labeled ERMES foci at mitochondrial division sites ( Figure 1D ) . Given that ERMES null mutations are epistatic to DNM1 null mutations ( Youngman et al . , 2004 ) , these observations suggest that ERMES functions early in ERMD at a step distinct from Dnm1 by bridging interactions between the ER and mitochondria . The conserved Miro GTPase Gem1 associates with the ERMES complex , potentially as a regulatory subunit ( Kornmann et al . , 2011; Stroud et al . , 2011 ) . Consistent with this , a functional GFP-2xFLAG-Gem1 fusion protein expressed in gem1Δ cells localized to mitochondrial-associated foci labeled by the ERMES subunit Mdm34-mCherry ( Figure 1E ) . Similar to ERMES , Gem1-labeled foci were spatially linked to mitochondrial division sites and segregated to only one tip following division ( Figure 1E ) . To gain insight into the function of Gem1 in ERMD , we characterized the behavior of ERMES and mitochondria in gem1Δ cells . As previously published , we observed that mitochondrial structure in gem1Δ cells is both aberrant and diverse , ranging from tubules to clustered spheres , with both having an increased mitochondrial diameter ( Figure 2A ) ( Frederick et al . , 2004 ) . In addition , regardless of the mitochondrial structure type , we observed that ERMES foci in gem1Δ cells were associated with apparent mitochondrial constriction sites , defined as a narrowing and/or resolved separation of the mitochondrial matrix labeled by mito-DsRed ( Figure 2A , arrow heads ) . A vast majority of ERMES-marked mitochondrial constriction sites in gem1Δ cells were stable , typically persisting for the duration of image capture ( 3–3 . 5 min ) ( Figure 2B and Figure 2—figure supplement 1B , C ) . This behavior was in contrast to wild-type cells , where ERMES-marked mitochondrial constriction sites resolved into two spatially separated tips within 15–60 s ( Figure 1A and Figure 2—figure supplement 1A , C ) . These observations indicate that Gem1 function is required for the resolution of mitochondrial constriction sites associated with division and/or in the subsequent segregation of mitochondrial tips generated by division events . 10 . 7554/eLife . 00422 . 005Figure 2 . Gem1 is required for mitochondrial distribution during ER-associated mitochondrial division . ( A ) ERMES foci are found at mitochondrial constriction sites in gem1Δ cells . Depicted is a whole-cell projection of gem1Δ cells expressing Mdm34-yEGFP and mito-DsRed . Arrowheads indicate mitochondrial constriction sites , which are also associated with Mdm34-yEGFP labeled foci . ( B ) The constriction sites associated with ERMES in gem1Δ cells are stable . gem1Δ cells expressing Mdm34-yEGFP and mito-DsRed were imaged as in Figure 1A . The first frame represents a whole-cell projection containing an inset that is shown in the remaining frames . Below each frame of the timelapse series is a plot of pixel intensity vs distance ( line-scan ) of mito-DsRed fluorescence signal as indicated by the dashed line at 0 s . The range of the y-axis is 260 arbitrary units and the range of the x-axis is 151 pixels . ( C ) Dnm1 is targeted to mitochondria in gem1Δ cells . A single focal plan in a timelapse series of a gem1Δ cell expressing GFP-HDEL , mito-tagBFP , and Dnm1-mCherry is shown . ER tubules also stably associate with mitochondrial constriction sites . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 00510 . 7554/eLife . 00422 . 006Figure 2—figure supplement 1 . ERMES-marked mitochondrial division is attenuated in gem1Δ cells . ( A ) Five additional examples of ERMES-linked mitochondrial divisions in wild-type cells . ( B ) Five additional examples of ERMES-linked mitochondrial constrictions in gem1Δ cells . Cells expressing Mdm34-yEGFP and mito-DsRed were imaged as described in Figures 1A and 2C . Scale bars , 2 μm . ( C ) Frequency profile of mitochondrial resolution times following ERMES-linked constrictions in wild-type and gem1Δ cells . The interval is defined as the time from initial mitochondrial constriction to the time when mitochondrial tips move independently . Twenty-nine randomly selected events from both wild-type and gem1Δ cells were analyzed . ERMES-linked mitochondrial constrictions that persisted throughout the length of the image capture ( 210 s ) are denoted as ‘never’ resolved . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 00610 . 7554/eLife . 00422 . 007Figure 2—figure supplement 2 . The first and second GTPase domains of Gem1 are required for its role in maintaining mitochondrial morphology . ( A ) Representative examples of mitochondrial morphology types that were used to categorize mitochondrial morphology phenotypes in gem1Δ mutants . ( B ) Schematic of Gem1 and the indicated mutations analyzed in this study . Blue boxes are GTPase domains; red boxes are EF-hand motifs; the green box is a transmembrane domain . ( C ) Percentages of each mitochondrial morphology class present in gem1Δ mutants and wild-type cells . ( D ) Representative examples of ERMES localization ( Mdm34-yEGFP ) and mitochondria ( mito-DsRed ) in wild-type cells and each of the gem1Δ mutants analyzed . Arrowheads mark ERMES-linked mitochondrial constrictions . The indicated mutations are specified in the top right corner of each panel . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 00710 . 7554/eLife . 00422 . 008Figure 2—figure supplement 3 . GTP hydrolysis by the second GTPase domain of Gem1 is required for mitochondrial distribution at ERMES-linked constrictions . ( A ) Timelapse series of Mdm34-yEGFP and mito-DsRed with Gem1 ( S462N ) expressed from pRS315 in a gem1Δ background . Cells were grown at 23°C in synthetic ethanol glycerol as in Figure 2—figure supplement 2B , C . Timelapse series are constructed as in Figure 2—figure supplement 2A , B . ( B ) Frequency profile of mitochondrial resolution times following ERMES-linked constrictions in gem1Δ + pRS315-Gem1 ( S462N ) cells as in Figure 2—figure supplement 2C . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 008 We addressed what structural features are important for Gem1’s role in the resolution of ERMES-associated mitochondrial constriction sites in cells . Under our conditions , expression of Gem1 mutants harboring abrogating mutations in either EF-hand I or II motifs or both rescued all mitochondrial phenotypes in gem1Δ cells to a similar extent as wild-type GEM1 ( Figure 2—figure supplement 2 ) . This observation is consistent with the published work demonstrating that the EF-hand regions play a relatively minor role in the maintenance of mitochondrial morphology and distribution , with only EF-hand I acting to stabilize Gem1 expression levels ( Frederick et al . , 2004; Koshiba et al . , 2011 ) . In contrast , expression of Gem1S19N or Gem1S462N , whose mutations abrogate the GTPase activity of GTPase domains 1 and 2 , respectively , failed to fully rescue mitochondrial morphological phenotypes in gem1Δ cells . The Gem1S19N mutation was the least functional , consistent with previously published observations ( Frederick et al . , 2004 ) . Abrogating mutations in the first GTPase domain , such as the Gem1S19N mutation , have been shown to reduce the steady-state localization of Gem1 to ERMES foci ( Kornmann et al . , 2011 ) . Thus , the severe phenotypes associated with Gem1S19N underscores the functional importance of Gem1’s localization to ERMES foci . Significantly , we also observed stable ERMES-associated mitochondrial constrictions in gem1Δ cells expressing either Gem1S19N or Gem1S462N ( Figure 2—figure supplements 2D and 3A , B ) . Thus , our structure function analysis of Gem1 indicates that similar features , specifically both Gem1 GTPase domains , are required for Gem1’s roles in both the maintenance of mitochondrial morphology and in the resolution of ERMES-marked mitochondrial constriction sites . Thus , the resolution of ERMES-associated mitochondrial constriction sites into segregated mitochondrial tips is a central function of Gem1 . We addressed whether the stability of ERMES-marked mitochondrial constriction sites is a consequence of defective mitochondrial division in gem1Δ cells by examining the behavior of the division machine , labeled with Dnm1-mCherry , over time . We observed that stable mitochondrial constriction sites in gem1Δ cells ( mito-TagBFP ) were persistently associated with ER tubules ( GFP-HDEL ) ( Figure 2C ) . In addition , Dnm1 puncta localized to these ER-associated mitochondrial constriction sites in a dynamic manner similar to that observed in wild-type cells , associating and dissociating on a 10-s time scale ( Figure 2C ) ( Frederick et al . , 2004 ) . These observations indicate that neither ER engagement with nor Dnm1 recruitment to mitochondrial constriction sites is significantly altered in gem1Δ cells , suggesting that mitochondrial division per se is not defective . This conclusion is supported by the observation that in gem1Δ cells lacking Fzo1 , an outer mitochondrial membrane dynamin-related protein required for mitochondrial fusion , mitochondria are fragmented , which is a morphology that requires mitochondrial division ( Frederick et al . , 2004 ) . Thus , taken together , these observations indicate that the stable ERMES-associated mitochondrial constriction sites in gem1Δ cells is a consequence of defective resolution of mitochondrial tips , perhaps as a consequence of altered ER–mitochondria contacts . To explore this idea , we analyzed wild-type and gem1Δ cells using three-dimensional electron tomography to determine whether mitochondrial constriction sites in gem1Δ cells represent division intermediates or fully resolved division events and to examine ER–mitochondria contact sites at higher resolution . Tomographic analysis revealed that , in contrast to tubular mitochondria in wild-type cells , mitochondria in gem1Δ cells are present in clusters containing fully resolved individual organelles ( Figure 3A–D and Videos 1–4 ) . Importantly , fully separate and adjacent mitochondria interacted with the same ER segment ( Figure 3B , C , B2 and C1 , ER–mitochondrial contact sites in red ) , suggesting that after division , resolved mitochondria remain tethered via this shared ER segment . 10 . 7554/eLife . 00422 . 009Figure 3 . Gem1 regulates ER–mitochondria contacts . ( A ) A tomograph and ( A1 and A2 ) corresponding three-dimensional tomogram of a mitochondrion ( purple ) and the ER ( green ) that closely apposes it ( contact sites in red , defined as <30 nm distance and ribosome excluded ) in a wild-type yeast cell . ( B ) Tomograph and ( B1 and B2 ) three-dimensional tomogram of mitochondria ( pink and purple ) and ER ( green , with contact shown in red ) for a gem1Δ cell . Mitochondria are shown in multiple hues of purple to indicate mitochondria that are discontinuous within the reconstructed volume . Magnified tomographs and three-dimensional tomograms of the purple mitochondria ( C and C1 ) and blue mitochondria ( D and D1 ) shown in the boxed regions of B1 are shown . ( E ) Tomograms in ( A , wild type ) and ( B , gem1Δ ) were used to calculate the area of mitochondrial surfaces closely apposed to the ER ( <30 nm , excludes ribosomes ) . The total mitochondrial surface area analyzed is similar for wild-type and gem1Δ cells ( 4 . 85 and 5 . 09 μm2 , respectively ) . Cells lacking Gem1 possess more clustered and smaller ER–mitochondria contacts . ( F ) The total surface area was calculated for the mitochondria that were modeled and shown in ( A , wild-type ) and ( B , gem1Δ ) , and the percent of this surface area that was in contact with the ER membrane was calculated for each . Mean diameter of the mitochondria shown in ( A , wild-type ) and ( B , gem1Δ ) . Scale bars , 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 00910 . 7554/eLife . 00422 . 010Figure 3—figure supplement 1 . Quantitative comparison of ER–mitochondria contacts in wild-type and gem1Δ cells . Three-dimensional tomograms of mitochondria in wild-type ( A ) and gem1Δ ( B ) cells . The total mitochondrial surfaces measured were 4 . 85 and 5 . 09 μm2 in wild-type and gem1Δ cells , respectively . Mitochondria are shades of purple , each shade representing a separate organelle , and ER–mitochondria contact surfaces are shaded red . Scale bars , 200 nm . ( C ) The surface areas of individual ER–mitochondria contacts shown in ( A ) and ( B ) . Although gem1Δ cells possess more ER–mitochondria contacts , the surface of each individual contact is smaller than the average ER-mitochondria contact in wild-type cells . ( D ) The shortest distance between pairs of ER–mitochondria contacts in wild-type and gem1Δ cells . Each category on the X-axis represents a different contact pair . ( E ) The diameters of mitochondria measured along their short axis . ( F ) Mitochondrial diameters were measured at and around ER-associated mitochondrial constriction sites in wild-type and gem1Δ cells . Although the diameter of mitochondria is larger in gem1Δ cells ( E ) and ( F ) , the diameter of constricted mitochondria was similar ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01010 . 7554/eLife . 00422 . 011Video 1 . Original tomographs and rotating three-dimensional models of the ER and mitochondria in a wild-type cell . ER ( green ) and a mitochondrion ( purple ) in a tomogram derived from three serial sections of a wild-type cell . Shown in red are regions of contact between the two organelles . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01110 . 7554/eLife . 00422 . 012Video 2 . Original tomographs and rotating three-dimensional models of ER and mitochondria in a gem1Δ cell . ER ( green ) and a mitochondrion ( purple ) in a tomogram derived from three serial sections of a gem1Δ cell . Shown in red are regions of contact between the two organelles . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01210 . 7554/eLife . 00422 . 013Video 3 . A rotating three-dimensional model of constricted , but continuous mitochondria in a gem1Δ cell . Rotating three-dimensional models of mitochondria ( purple ) that are globular and yet continuous with each other relative to the ER ( green ) and regions of contact between them ( red ) in the gem1Δ cell . Images correspond to the mitochondria shown in Figure 1B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01310 . 7554/eLife . 00422 . 014Video 4 . A rotating three-dimensional model of ‘tethered’ mitochondria in a gem1Δ cell . Globular and ‘tethered’ mitochondria in a gem1Δ cell are associated with the ER ( green ) and regions of contact between them ( red ) are noted . Images correspond to the mitochondria shown in Figure 1B , D . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 014 Quantitative analysis of an equivalent mitochondrial surface area in wild-type and gem1Δ cells ( approximately 5 μm2 ) revealed that gem1Δ cells harbor a significantly greater number of ER–mitochondria contacts than wild-type cells , which are also more clustered as compared to wild-type contacts ( Figure 3—figure supplement 1 ) . The surface area of individual ER–mitochondria contacts was , on average , smaller in gem1Δ cells as compared to wild-type cells ( Figure 3E and Figure 3—figure supplement 1C ) . However , the total ER–mitochondria contact surface was approximately three times greater than that observed for wild-type mitochondria ( Figure 3F ) . These findings are consistent with fluorescence data demonstrating that , in comparison to wild-type cells , ER tubules in gem1Δ cells were observed associated in a relatively stable manner with mitochondrial constriction sites ( Figure 2C ) . In addition , our EM analysis suggests that the reported larger ERMES foci observed by fluorescence imaging in gem1Δ cells may represent multiple unresolved ER–mitochondria contacts ( Kornmann et al . , 2011 ) ( Figure 3—figure supplement 1D ) . The average diameter of gem1Δ mitochondria , measured along their short axis , was also greater than that of wild-type cells ( Figure 3—figure supplement 1E ) , consistent with our fluorescence imaging analysis ( Figure 2 ) . However , the average diameter of ER-associated mitochondrial constriction sites in wild-type and gem1Δ cells were similar ( Figure 3—figure supplement 1F ) , consistent with our light imaging data demonstrating that Dnm1 is still efficiently targeted to mitochondrial constriction sites in gem1Δ cells ( Figure 2C ) . Together , these data suggest that Gem1 functions downstream of ER-associated mitochondrial constriction and division as a negative regulator of ER–mitochondria contacts , which is required for the efficient resolution and spatial distribution of mitochondria following division . To further investigate the functional significance of these observations , we explored the idea that ERMD serves to position division sites adjacent to actively segregating mitochondrial nucleoids , which is based on the observation that ERMES is spatially linked to actively replicating mitochondrial nucleoids ( Hobbs et al . , 2001; Meeusen and Nunnari , 2003 ) . To visualize nucleoids , we created a functional FP fusion to the nucleoid component , Yme2 ( Figure 4 and Figure 4—figure supplements 1 and 2 ) , an integral mitochondrial inner membrane protein with homology to exonucleases whose deletion alters nucleoid structure and copy number ( Hanekamp and Thorsness , 1996 , 1999; Park et al . , 2006 ) . Consistent with the previously published work , we observed that over time a subset of Yme2-GFP–labeled nucleoids persistently localize adjacent to ERMES foci , marked by Mdm34-tdTomato ( Figure 4A ) . In addition , ERMES-linked nucleoids were associated with mitochondrial division events ( Figure 4A ) . To characterize the behavior of yeast nucleoids at division sites , we simultaneously imaged mitochondria ( mito-dsRed ) and nucleoids ( Yme2-GFP ) . Quantitative analysis indicated that mitochondrial nucleoids are associated with over 80% of mitochondrial division events ( Figure 4B , C ) . Temporal analysis of Yme2-GFP–labeled nucleoids positioned at future sites of mitochondrial division indicated that they exhibit short-ranged oscillatory movements and often rapidly segregate and recoalesce prior to mitochondrial division ( Figure 4B and Figure 4—figure supplement 3 ) . Following mitochondrial division , a majority of Yme2-GFP foci were observed at the ends of both newly generated mitochondrial tips ( Figure 4B , C and Figure 4—figure supplement 3 ) . However , we also observed division events that resulted in a nucleoid in only one of the newly generated mitochondrial tips , indicating that division and nucleoid segregation are not obligatorily linked ( Figure 4C and Figure 4—figure supplement 3 ) . Consistent with a role of ERMES in nucleoid positioning , published data indicate that deletion of core ERMES components disrupts nucleoid structure ( Burgess et al . , 1994; Youngman et al . , 2004 ) . In contrast , we observed that nucleoids , as labeled by Yme2-GFP , maintained their discrete focal localization in gem1Δ cells , similar to wild-type cells ( Figure 4D ) . In addition , similar to ERMES distribution , Yme2-GFP foci were observed associated with stable mitochondrial constrictions in gem1Δ cells ( Figure 4D ) , consistent with the idea that the spatial resolution of ER–mitochondria contacts following division is fundamentally important for nucleoid distribution . Thus , our findings suggest that ERMD functions to position division sites in proximity to nucleoids to increase the probability of their distribution upon mitochondrial division . 10 . 7554/eLife . 00422 . 015Figure 4 . Nucleoid segregation is linked to mitochondrial division . ( A ) Yme2-GFP coaligns with ERMES foci , marked by Mdm34-TdTomato , at mitochondrial division events . ( B ) Nucleoid behavior , marked by Yme2-yEGFP , at mitochondrial ( mito-DsRed ) division sites . ( C ) Quantification of the relationship of nucleoids to division sites . ( D ) Representative images of gem1Δ cells expressing Yme2-GFP and mito-DsRed . Scale bars , 2 μm , except in the inset of ( B ) , which is 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01510 . 7554/eLife . 00422 . 016Figure 4—figure supplement 1 . Yme2-GFP is functional . ( A ) The functionality of Yme2-GFP was assessed by exploiting the synthetic lethal phenotype on rich dextrose media caused by deletion of both YME2 and GEP4 , which encodes a component of the cardiolipin biosynthetic pathway in mitochondria . Growth of gep4Δ cells on rich dextrose media was unaltered by expression of Yme2-GFP from its chromosomal locus , indicating that Yme2-GFP is functional . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01610 . 7554/eLife . 00422 . 017Figure 4—figure supplement 2 . Yme2-GFP localizes to nucleoids . ( A ) Chromosomally expressed Yme2-GFP in wild-type cells localized to punctate structures , which were also stained with the DNA selective vital dye DAPI and formed a beads-on-a-string pattern , characteristic of mitochondrial nucleoids . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 01710 . 7554/eLife . 00422 . 018Figure 4—figure supplement 3 . Additional examples of Yme2-GFP nucleoid behavior at a mitochondrial division sites in wild-type cells . Imaging was performed as described in Figure 4B . Scale bars , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00422 . 018
Here , we show that ERMES and the Miro GTPase Gem1 function in the process of ERMD , which serves to couple the segregation of mitochondria and mtDNA in cells . Specifically , our data support a model in which ERMES creates ER–mitochondria contacts along the mitochondrial network that serve to link actively replicating nucleoids to mitochondrial division and the subsequent Gem1-dependent spatial resolution of newly generated mitochondrial tips . Such a mechanism is likely required for efficient mtDNA distribution throughout the cell in addition to nucleoid segregation per se , as long-range movement of nucleoids within the organelle is limited ( Nunnari et al . , 1997; Okamoto et al . , 1998 ) . In mammalian cells , nucleoids are similarly localized at mitochondrial division sites and mitochondrial tips , suggesting that ERMD also plays this fundamental role in humans ( Garrido et al . , 2003; Iborra et al . , 2004 ) . In this context , although the Miro GTPase family is highly conserved in eukaryotes , Miro orthologs are not found in organisms that lack mtDNA , such as Giardia intestinalis and Trichomonas vaginalis , and possess mitochondrial-related mitosomes or hydrogenosomes , respectively ( Vlahou et al . , 2011 ) . This correlation is consistent with a fundamental role of Miro in mtDNA segregation . Also consistent with this view , gem1Δ cells lose mitochondrial DNA at a significantly higher frequency than wild-type cells ( Frederick et al . , 2004 ) . The mechanism of mitochondrial division site placement is apparently divergent from that utilized by ancestral bacteria , where division sites are determined in part by a nucleoid occlusion mechanism , which prevents cell division in the vicinity of the bacterial chromosome ( Wu and Errington , 2012 ) . However , we currently lack an understanding of what drives mitochondrial nucleoid segregation and of the mechanism underlying the spatial coupling of ER–mitochondria contact sites and nucleoids . Thus , alternative mitochondrial-specific mechanisms may exist to coordinate the timing of nucleoid replication and segregation with mitochondrial division . In ERMD , mitochondrial constrictions observed at the sites of ER–mitochondria contact , where the ER likely wraps around a mitochondrial tubule , are independent of the mitochondrial division dynamin ( Friedman et al . , 2011 ) . Our data are consistent with the model in which ERMES functions early in ERMD to mediate the biogenesis of this specialized region of ER–mitochondria contact . Thus , it is tempting to speculate that the ERMES complex functions to generate the ER tubules associated with constriction and/or to actively produce mitochondrial constriction at the sites of contact . ERMES has also been implicated as a bridge between mitochondria and the actomyosin network and thus may function at division sites to coordinate the recruitment of cytoskeletal and motor proteins , which could generate force required for mitochondrial constriction and/or distribution following division ( Boldogh et al . , 1998 , 2003 ) . Indeed , a recent study suggests that the actin cytoskeleton may be involved in mitochondrial division at ER–mitochondria contacts in mammalian cells ( Korobova et al . , 2013 ) . In contrast to ERMES , our data indicate that Gem1 functions relatively late in ERMD to promote the physical separation of mitochondrial tips generated by membrane scission . Specifically , our data point to a role for Gem1 post-scission in the negative regulation of ER–mitochondria contacts to facilitate the resolution of mitochondrial tips . Evidence from higher eukaryotes suggests that the Gem1 ortholog , Miro1/2 , functions in mitochondrial distribution by connecting mitochondria directly to a kinesin-1 adaptor protein Milton/TRAK to enable the microtubule based transport of mitochondria ( Guo et al . , 2005; Fransson et al . , 2006; Glater et al . , 2006; Wang and Schwarz , 2009; Misko et al . , 2010; Wang et al . , 2011 ) . Analogously , Gem1 in yeast may also function to negatively regulate ER–mitochondria contacts by facilitating the recruitment of motility factors to mitochondrial tips following division . However , while the Miro GTPase family is remarkably conserved ( Vlahou et al . , 2011 ) , the mechanisms of mitochondrial transport are divergent in eukaryotes . In many fungi , including budding yeast , mitochondrial distribution is actin dependent ( Hermann et al . , 1997; Simon et al . , 1997; Boldogh et al . , 1998 , 2003 ) , and in Dictyostelium , Miro is not required for microtubule-dependent mitochondrial transport ( Vlahou et al . , 2011 ) . Thus , alternatively , Gem1 may function in ERMD more directly to regulate the physical link between mitochondria and the ER via ERMES . Consistent with this idea , Gem1 associates with ERMES foci , and , in the absence of Gem1 , it has been reported that ERMES forms fewer and larger foci per cell ( Kornmann et al . , 2011 ) . This model of Gem1 function is also supported by the observation that mutations in the first GTPase domain of Gem1 abolish both Gem1’s function in ERMD and its association with ERMES ( Frederick et al . , 2004; Kornmann et al . , 2011; Koshiba et al . , 2011 ) . In addition , in mammalian cells , Miro1 is found at ER–mitochondria contacts ( Kornmann et al . , 2011 ) . In this context , the defect in mitochondrial segregation observed in gem1Δ cells at ER-associated mitochondrial constriction sites could result from the ER physically hindering mitochondrial motility . In this case , it is also possible that the mitochondrial motility defects in higher eukaryotes caused by Miro1/2 dysfunction are a secondary consequence of enhanced ER–mitochondria contacts .
All strains were constructed in either the W303 ( ade2–1; leu2–3; his3–11 , 15; trp1–1; ura3–1; can1–100 ) or BY4741 ( his3Δ1; leu2Δ0; met15Δ0; ura3Δ0 ) genetic background , both of which have been described previously ( Rothstein , 1983; Brachmann et al . , 1998 ) . ERMES components and Yme2 were tagged at their C-termini using PCR-based homologous recombination ( Janke et al . , 2004; Sheff and Thorn , 2004 ) . The red fluorescent protein TdTomato ( Shaner et al . , 2004 ) was cloned into pFA6 plasmids using standard cloning techniques . Haploid cells were transformed with PCR product using the lithium acetate method and plated on synthetic media or YPD + geneticin ( 300 μg/ml ) to select for the homologous recombination event . Correct integration of the cassette was confirmed by colony PCR , and western blotting of whole-cell protein extracts was used to verify correct protein expression . To generate gem1Δ MDM34-yEGFP or gem1Δ MDM34-mCherry strains , W303 MDM34-yEGFP::HIS5 or W303 MDM34-mCherry was mated to gem1Δ::HIS3 cells . The heterozygous diploids were sporulated , and the resulting tetrads were dissected . Two-to-two cosegregation of the histidine prototrophy was used to verify the gem1Δ MDM34-yEGFP genotype , and cosegregation of histidine prototrophy and kanamycin resistance was used to verify the gem1Δ MDM34-mCherry genotype . The yme2Δ gep4Δ and YME2-GFP gep4Δ strains were generated by mating haploid gep4Δ and yme2Δ or YME2-GFP strains , followed by tetrad dissection and identification of clones by cosegregating genetic markers . The plasmids pVT100U-DsRed ( mito-DsRed ) , pHS20-mCherry ( Dnm1-mCherry ) , pRS315-GFP-2xFLAG-GEM1 ( GFP-2XFLAG-Gem1 ) , and YIplac204/TKC-DsRed-HDEL ( DsRed-HDEL ) YIplac204/TKC-GFP-HDEL ( GFP-HDEL ) were described previously ( Westermann and Neupert , 2000; Rossanese et al . , 2001; Meeusen and Nunnari , 2003; Lackner et al . , 2009; Kornmann et al . , 2011 ) . The plasmid pYES-TagBFP ( mito-TagBFP ) was obtained by replacing BFP in pYES-BFP ( Westermann and Neupert , 2000 ) with TagBFP using standard cloning techniques . Plasmids harboring wild-type and mutant alleles of GEM1 were generated by amplifying the GEM1 locus from yeast genomic DNA with flanking regions 307-bp upstream and 644-bp downstream by PCR followed by insertion into pRS315 with standard restriction cloning techniques . Point mutations were generated using Quick-Change Mutagenesis , and all mutations were verified by DNA sequencing . Plasmids were transformed into gem1Δ MDM34-yEGFP pVT100U-DsRed grown to mid-log phase in synthetic ethanol glycerol media at 23°C . For all fluorescence microscopy described in the next two paragraphs , yeast cells were grown to mid-log phase in the appropriate synthetic media and were imaged live . Yeast cells grown to mid-log phase in the appropriate synthetic media were briefly sonicated , concentrated by centrifugation , and mounted on slides with a 4% agarose bed in synthetic dextrose growth medium . To evaluate the relationship of ERMES to mitochondrial division , wild-type or gem1Δ cells expressing Mdm12 , Mdm34 , or Mmm1-yEGFP and mito-DsRed were imaged using the spinning disc module of a Marianas SDC Real Time 3D Confocal-TIRF microscope ( Intelligent Imaging Innovations Denver , CO , 3i ) fit with a Yokogawa spinning disk head , a 100× 1 . 46 NA objective , and EMCCD camera . Z-stacks were taken at 0 . 4-μm increments over approximately 6 μm of the cell every 15 s for 3 . 5 min . Mitochondrial divisions were identified independently of ERMES foci by hiding the GFP channel and retrospectively analyzed for their association with ERMES . An ERMES focus was considered linked to mitochondrial division if its center was within 300 nm of the division site , which is 1 . 5 times the average radius of an ERMES focus ( 205 ± 45 nm , n = 68 ) . To calculate the ERMES to mitochondrial surface area ratio , Z-series of cells expressing Mdm34-yEGFP and mitoDsRed were collected at 0 . 4-μm increments . Separate channels for each marker were manually threshold in ImageJ , and the ‘Analyze Particles’ function was used to measure areas for each in all focal planes collected . Ratios were calculated for each plane and were averaged . This ratio ( 4 . 5% ) was used to calculate the percentage of ERMES foci associated with mitochondrial divisions due to random chance ( approximately 10% ) based on the requirement that mitochondrial division occurred less than 300 nm from the center of an ERMES focus . For three-color fluorescence imaging in Figures 1C-E and 2C , cells were imaged on DeltaVision-Real Time microscope ( IX70 DeltaVision; Olympus ) using a 60× 1 . 4 NA objective lens ( Olympus ) and a 100 W mercury lamp ( Applied Precision ) . Light microscopy images were collected using an integrated cooled charge-coupled device ( CCD ) –based camera ( CoolSNAP HQ; Photometrics ) equipped with a Sony Interline Chip . Datasets were processed using softWoRx’s ( Applied Precision ) iterative , constrained three-dimensional deconvolution method to remove out-of-focus light . To analyze GEM1 alleles by fluorescence microscopy , gem1Δ MDM34-yEGFP pVT100U-DsRed pRS315- ( empty/GEM1/gem1 alleles ) cells were grown to early log phase , concentrated by centrifugation and mounted on slides with 4% agarose beds and imaged on the Marianas SDC Real Time 3D Confocal-TIRF microscope as described above . For all imaging of Yme2-GFP , cells were grown to mid-log in synthetic dextrose medium supplemented with casamino acids . Cells were immobilized in glass bottomed culture dishes ( Bioptechs , Inc ) with Concanavalin A ( 1 mg/ml ) and overlaid with 2 ml of fresh medium . Cells were imaged on an OMX microscope equipped with a 100× 1 . 4 NA objective lens ( Carlton et al . , 2010 ) . Z-stacks were acquired over 7 µm in 0 . 2-µm increments . For higher temporal resolution ( 2 s per frame ) , only 3-µm-thick Z-stacks were collected . Images for each fluorophore were acquired simultaneously with independent EMCCD cameras ( iXON; Andor ) , and images were aligned post-capture using alignment parameters generated from images of 0 . 1-µm fluorescent microspheres ( TetraSpeck; Invitrogen ) . Images were processed with a denoising algorithm ( Boulanger et al . , 2010 ) and iterative , constrained three-dimensional deconvolution using the Priism software suite ( http://msg . ucsf . edu/IVE/ ) . Yme2-associated division events were scored as described above for Mdm34-yEGFP . For DAPI staining , cells were grown for 30 min in synthetic dextrose medium supplemented with casamino acids and 1 µg/ml DAPI and subjected to imaging . Haploid Saccharomyces cerevisiae cells were grown to log phase , harvested , and high-pressure frozen in a Balzers HPM 010 as previously described ( Nickerson et al . , 2010 ) . Automated Freeze Substitution was performed on a Leica AFS with 0 . 1% uranyl acetate and 0 . 25% glutaraldehyde in anhydrous acetone , embedded in Lowicryl HM20 ( Polysciences , Warrington , PA ) , and polymerized at −60°C ( Giddings , 2003 ) . A Leica Ultra-Microtome was used to cut 300-nm serial semi-thick sections; sections were stabilized using a formvar sandwich ( West et al . 2011 ) and labeled with fiduciary 15-nm colloidal gold ( British Biocell International ) ; dual-axis tilt series were collected of the samples from ±60° with 1° increments at 300 kV using SerialEM ( Mastronarde et al . , 1997 ) at 300 kV using a Tecnai 30 FEG ( FEI-Company , Eindhoven , the Netherlands ) . Tilt series were recorded at a magnification of 23 , 000 times using SerialEM with a 4 × 4K CCD camera ( Gatan , Inc . , Abingdon , United Kingdom ) as described ( West et al . 2011 ) . Individual tomograms were reconstructed using the IMOD package ( Kremer et al , 1996 ) , and its newest viewer 3DMOD 4 . 0 . 11 and serial tomograms were merged together in X- , Y- , and Z-direction to obtain a large continuous volume . The 3DMOD modeling software was used for the assignment of the outer leaflet of organelle membrane contours , and IMODINFO was used to obtain surface area and volume data of contour models . Images were further enhanced and manipulated in Adobe Photoshop 7 . We sorted , analyzed , and graphed the data using Microsoft Excel for Mac 2008 and Prism 5 for Mac OS X . Movies were made in 3DMOD , assembled in QuickTime Pro 7 . 5 , and movie size was reduced to less than 10 MB by saving movies as HD 720p in QuickTime . Mitochondrial surface area was scored as in contact with the ER membrane ( denoted as red objects in the three-dimensional models ) if it was within 30 nm , and ribosomes were excluded between the two membranes . | Mitochondria generate most of the energy used by cells , and they also play key roles in cellular growth , death , and differentiation . They are evolutionarily derived from bacteria and have retained their own DNA and protein translation system , but they are also dependent on the cell for their growth and replication . A significant portion of the outer membrane of a mitochondrion is in contact with the endoplasmic reticulum ( ER ) —an organelle that is the starting point for the synthesis of secreted proteins , and is also critical for the synthesis of lipids and other organelles . Recent work suggests that mitochondria–ER contact points mark sites of mitochondrial division , but it is unclear exactly how this process occurs . Here , Murley et al . use the budding yeast and model organism Saccharomyces cerevisiae to show that at mitochondrial division sites , a multiprotein complex called ERMES promotes the formation of ER–mitochondrial contact points , while an evolutionarily conserved enzyme , Gem1 , antagonizes these contacts to aid mitochondrial segregation . The contact points are found adjacent to nucleoids ( which are complexes of mitochondrial DNA and proteins ) —an observation suggesting that ER-associated mitochondrial division evolved to help distribute nucleoids between newly formed mitochondria . The present study also reveals a novel role for the conserved protein Gem1 and could lead researchers to reinvestigate the functions of Miro1/2—the equivalent of Gem1 in higher eukaryotes . Miro1/2 is thought to connect mitochondria to motor proteins , which transports them through the cell along microtubules . Dysfunction of Miro1/2 reduces the mobility of mitochondria , and the work of Murley et al . suggests that this could be a consequence of enhanced contacts between mitochondria and the ER . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2013 | ER-associated mitochondrial division links the distribution of mitochondria and mitochondrial DNA in yeast |
In contrast to transcriptional regulation , the function of alternative splicing ( AS ) in stem cells is poorly understood . In mammals , MBNL proteins negatively regulate an exon program specific of embryonic stem cells; however , little is known about the in vivo significance of this regulation . We studied AS in a powerful in vivo model for stem cell biology , the planarian Schmidtea mediterranea . We discover a conserved AS program comprising hundreds of alternative exons , microexons and introns that is differentially regulated in planarian stem cells , and comprehensively identify its regulators . We show that functional antagonism between CELF and MBNL factors directly controls stem cell-specific AS in planarians , placing the origin of this regulatory mechanism at the base of Bilaterians . Knockdown of CELF or MBNL factors lead to abnormal regenerative capacities by affecting self-renewal and differentiation sets of genes , respectively . These results highlight the importance of AS interactions in stem cell regulation across metazoans .
Stem cells are found in all animals and are defined by their capacity to self-renew and to differentiate into different cell types ( Sánchez Alvarado and Yamanaka , 2014 ) . In mammals , embryonic stem cells ( ESCs ) derived from pre-implantation embryos can be cultured in vitro and differentiated into virtually any cell type ( Martello and Smith , 2014 ) ; however , a similarly potent cell type has not been found in adults . In contrast , in other animals , pluripotent stem cells are maintained during the entire life , and are often associated with extraordinary regenerative capabilities ( Solana , 2013; Tanaka and Reddien , 2011 ) . One of the most extreme examples are freshwater planarians , from which almost any body part can regenerate a complete organism in a few days . This ability relies on a large number of stem cells present in the adult , called neoblasts . Illustrating their pluripotency , single neoblasts transplanted into lethally irradiated hosts can rescue this lethality , restore tissue turnover , generate all cell types of the adult planarian and completely transform the genotype and phenotype of the host into that of the donor ( Wagner et al . , 2011 ) . However , recent analyses at single-cell resolution showed that the neoblast pool is highly heterogeneous , also including multiple lineage-committed precursors ( van Wolfswinkel et al . , 2014 ) . Despite significant progress , how neoblasts are regulated and enable planarian cell turnover as well as regeneration upon wounding is still largely unknown . Initial transcriptomic analyses of planarian neoblasts have revealed hundreds of genes that are differentially enriched in both planarian and mammalian stem cells compared to all differentiated cell types despite 500 million years of independent evolution ( Labbé et al . , 2012; Onal et al . , 2012; Reddien et al . , 2005; Resch et al . , 2012; Rouhana et al . , 2010; Solana et al . , 2012 ) , suggesting the existence of universal regulatory features across animal pluripotent cells . However , this conservation does not include the major transcriptional regulators of mammalian stem cells . In ESCs , pluripotency is maintained by a core set of transcription factors that include OCT4 , NANOG and SOX2 , but these factors and their interactions are largely not conserved beyond the vertebrate lineage ( Fernandez-Tresguerres et al . , 2010; Gold et al . , 2014; Onal et al . , 2012 ) . For instance , no homolog of NANOG has been described to date in any invertebrate species , despite extensive search ( Scerbo et al . , 2014 ) . Therefore , elucidating how the regulation of pluripotency in invertebrates occurs in the absence of this core set of factors is crucial to understand the biology of animal stem cells . Post-transcriptional regulation is more recently emerging as another key mechanism for controlling ESC biology ( Ye and Blelloch , 2014 ) . In particular , various reports have established the importance of alternative splicing ( AS ) for ESCs and somatic cell reprogramming ( Han et al . , 2013; Ohta et al . , 2013; Venables et al . , 2013; Ye and Blelloch , 2014 ) . AS is the process by which introns and exons are selectively included or excluded from the pre-mRNA to produce multiple mRNA and protein isoforms . AS can therefore expand transcriptomic complexity in a cell type- or developmental stage-specific manner , adding an extra layer of regulation to the control of gene expression . Moreover , highly regulated alternative exons often encode disordered regions of proteins that embed binding motifs , and thus have the potential to rewire protein-protein interactions in a context-specific manner ( Buljan et al . , 2012; Ellis et al . , 2012 ) . AS is chiefly regulated by RNA binding proteins ( RBPs ) , which are themselves often differentially expressed in a cell type-regulated manner . These factors typically bind to pre-mRNAs in a sequence- and position-specific fashion , thereby modulating inclusion or exclusion of the target alternative sequence . For example , members of the MBNL family of RBPs are lowly expressed in mammalian ESCs , but show higher levels of expression in all profiled differentiated samples – ranging from transformed cell lines ( e . g . HeLa , 293T , etc . ) to highly histologically and functionally diverse adult tissue types ( including multiple brain regions , muscle , liver , kidney , testis , etc . ) – where they repress a program of alternative exons that are characteristic of ESCs ( Han et al . , 2013 ) . Knockdown of MBNL in differentiated cultured cells thus induces ESC-like AS patterns and this is sufficient to enhance somatic cell reprogramming ( Han et al . , 2013 ) . MBNL targets are involved in diverse cellular processes , from cytoskeletal dynamics to gene regulation , and include disparate actors , from protein kinases to transcriptional regulators . For instance , MBNL-regulated AS of the transcription factor FOXP1 in mouse and human embryonic stem cells changes its DNA-binding properties , so that its stem cell specific isoform ( FOXP1-ES ) promotes transcription of pluripotency genes and represses differentiation genes , while its canonical isoform activates genes involved in differentiation ( Gabut et al . , 2011 ) . Consistently , FOXP1-ES is repressed by MBNL upon differentiation by direct binding to intronic regions in the pre-mRNA ( Han et al . , 2013 ) . Intriguingly , members of another RBP family , CELF , have been shown to regulate AS in an antagonistic manner to that of MBNL factors in mammals . For example , during heart development , CELF factors promote embryonic AS patterns , which are later replaced by adult heart AS patterns promoted by MBNL proteins ( Kalsotra et al . , 2008 ) . Multiple exons have been shown to antagonistically respond to CELF and MBNL factors in a variety of mammalian differentiation , disease and cell culture systems ( Dasgupta and Ladd , 2012; Kalsotra et al . , 2008; Lee and Cooper , 2009; Wang et al . , 2015 ) . CELF factors , however , have not been linked so far to regulation of AS in stem cells . Despite these studies , the degree to which regulation by AS is conserved or whether AS plays any role in non-mammalian stem cells is still entirely unknown . To address these questions , we here investigated the regulatory role and functional importance of AS in planarian stem cells and regeneration in vivo . Our results show that AS and the factors involved in its regulation , as well as their specific interactions ( i . e . functional antagonism between MBNL and CELF proteins ) , are crucial for planarian stem cell biology and regeneration and likely a deeply conserved feature of animal stem cells .
We used a recent de novo transcriptome assembly ( Liu et al . , 2013 ) to annotate intron and exons in the genome of Schmidtea mediterranea ( assembly version 3 . 1; Figure 1—figure supplement 1A–B , Materials and methods ) . Most planarian protein-coding genes ( 75% ) are multiexonic ( Figure 1—figure supplement 1C ) . Intron length displays a sharp bimodal distribution: whereas most introns are relatively small , with lengths centred on 57 bp , another subset ranges between 1 and 10 Kbps ( Figure 1—figure supplement 1D ) . Next , we produced and compiled over 30 samples of novel and available deep coverage RNA sequencing ( RNA-Seq ) data from multiple sources ( Figure 1—source data 1 ) . These consist of FACS-isolated cell populations – including neoblast-enriched ( X1 ) , neoblast progeny-enriched ( X2 ) and differentiated cell-enriched ( Xins ) fractions ( Figure 1A ) – as well as wild type and neoblast-depleted whole animals . We employed these data and previously described methodologies to comprehensively identify all types of AS in planarians and quantify their alternative sequence inclusion levels using the ‘Percentage Spliced In’ ( PSI ) metric ( [Braunschweig et al . , 2014; Irimia et al . , 2014] , methods ) . These approaches yielded 12 , 276 AS events , the majority of which ( 56 . 2% ) corresponded to alternatively retained introns ( Figure 1—figure supplement 2A ) . We also identified 2529 alternative exons that can be either fully included or skipped from the mRNAs , including single and multi-cassette events , as well as 262 microexons ( exons of length ≤27 nucleotides ( nt ) ( Irimia et al . , 2014 ) , 72 of which had length ≤15 nt ) ( Figure 1—figure supplement 2A ) . 10 . 7554/eLife . 16797 . 003Figure 1 . Alternative splicing is differentially regulated between planarian stem cells and differentiated cells . ( A ) Planarian Stem cells ( 'Neoblasts' ) , their differentiating progeny , and differentiated cells are purified with FACS ( 'X1' , 'X2' , 'Xins' , respectively ) . RNA-seq and computational analyses were subsequently used to identify X1-differential AS at a genome-wide scale . ( B ) Distribution of AS events with increased/decreased inclusion of the alternative sequence in X1 . Alt3/5 , alternative splice site acceptor/donor selection; IR , intron retention; AltEx , cassette exons . ( C ) Heatmap of inclusion level values for 293 representative X1-differential AS events . Bars in the dendrogram correspond to AS types in B . ( D ) Representative RT-PCR assays monitoring AS patterns in FACS isolated cell fractions . Red and blue exons indicate those exons with higher and lower inclusion in X1 compared to Xins fractions , respectively . Scatter plot shows correspondence between PSI estimates by RNA-Seq and RT-PCR in whole worms and X1 , X2 and Xins fractions for 22 events ( R2 = 0 . 92 , n = 88 ) . ( E ) Proportion of alternatively spliced exons by length class with increased inclusion in X1 ( 'X1-inc' ) or Xins ( 'X1-exc' ) fractions , or not differentially regulated between X1 and Xins fractions ( 'Non-X1 reg' ) DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00310 . 7554/eLife . 16797 . 004Figure 1—source data 1 . Schmidtea mediterranea RNA-seq samples used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00410 . 7554/eLife . 16797 . 005Figure 1—source data 2 . List of neoblast-differential AS events . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00510 . 7554/eLife . 16797 . 006Figure 1—figure supplement 1 . Genome annotation pipeline and summary statistics . ( A ) Genome annotation pipeline used in this study . ( B ) Number and percentage of canonical introns ( with GY-AG splice sites ) in all genes , coding genes , and introns within coding regions . ( C ) Whole genome intron length distribution shows a marked bimodal distribution . ( D ) Number of protein-coding genes according to their intron density . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00610 . 7554/eLife . 16797 . 007Figure 1—figure supplement 2 . Identification and analysis of neoblast differential AS in planarians . ( A ) Number of alternative splicing events identified by type . Left: number of AS events based on RNA-seq data by type . Alt3/5 , Alternative acceptor /donor splice site choice; IR , intron retention; AltEx , alternative exon skipping events ( cassette exons ) . Middle: number of exon skipping events according to the number of neighboring alternative exons that form the event . Right: number of alternative exons according to their length . Microexons are defined as those exons with lengths between 3 and 27 nucleotides . ( B ) Distribution by type of planarian AS events with increased/decreased neoblast inclusion of the alternative sequence . 'Alt3/5' , alternative splice site acceptor/donor selection; 'IR' , intron retention; 'MIC' , microexons ( length between 3 and 27 nucleotides ) ; 'Single' , single cassette long ( >27 nucleotides ) exon; 'Multi' , long exon that is part of an array of multiple neighbouring cassette exons . ( C ) Percent of neoblast-differential , alternative or constitutive exons with high ( average disorder rate >0 . 67 ) , mid ( between 0 . 33 and 0 . 67 ) and low ( <0 . 33 ) disorder calculated using Disopred2 . p-values correspond to 3-way Fisher tests . ( D ) Percent of residues that overlap a PFAM protein domain . p-values correspond to proportion tests . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00710 . 7554/eLife . 16797 . 008Figure 1—figure supplement 3 . RT-PCR validation of neoblast-differential AS events . RT-PCR validation assays for 22 and 14 representative AltEx and IR events that are predicted to be differentially regulated between neoblasts and differentiated cells by RNA-Seq analyses . Red and blue exons/introns indicate those exons/introns with higher and lower inclusion in X1 compared to Xins fractions , respectively . Scatter plots show correspondence between inclusion level estimates by RNA-Seq and RT-PCR in whole worms and X1 , X2 and Xins fractions ( AltEx , R2 = 0 . 92 , n = 88; IR , R2 = 0 . 68 , n = 56 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00810 . 7554/eLife . 16797 . 009Figure 1—figure supplement 4 . Gene Ontology analysis of X1-differential AS events . ( A-C ) Gene Ontology analysis of neoblast-differential AS events: Gene ontology enrichment analysis for 'Biological Process' , 'Molecular Function' and 'Cellular Component' terms was performed for neoblast-differential events that are predicted to ( I ) generate distinct protein isoforms in X1 and Xins ( black ) , ( J ) disrupt ORF in X1 ( dark grey ) , and ( K ) disrupt ORF in Xins ( light grey ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 00910 . 7554/eLife . 16797 . 010Figure 2 . Abundant IR in planarian neoblasts . ( A ) Representative RT-PCR assays monitoring IR patterns in FACS isolated cell fractions . Red and blue introns indicate those introns with higher and lower inclusion in X1 compared to Xins fractions , respectively . Scatter plot shows correspondence between PSI estimates by RNA-Seq and RT-PCR in whole worms and X1 , X2 and Xins fractions for 14 events ( R2 = 0 . 68 , n = 56 ) . ( B ) Intron length distributions for all introns ( black ) , all retained introns ( grey ) , X1-included introns ( red ) and X1-excluded introns ( blue ) . Median length ( m ) is indicated for each set . ( C–D ) Gene expression measured using the cRPKM metric for genes containing X1-included ( C ) or excluded ( D ) introns in X1 , X2 and Xins cell fractions . P-value corresponds to a Wilcoxon Sum Rank test between X1 and Xins expression values . ( E ) Percent of X1-differential AS events by type that are predicted to generate alternative ORF-preserving isoforms ( black ) , disrupt the ORF in neoblasts or differentiated cells ( dark/light grey ) , or overlap non-coding sequences ( white ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 010 By comparing neoblasts versus differentiated cells and whole worms depleted of neoblasts ( methods ) , we identified 246 and 256 AS events with increased and decreased alternative sequence inclusion levels in neoblasts , respectively ( X1-included and X1-excluded , together referred to as 'neoblast-differential'; Figure 1B and Figure 1—figure supplement 2B , Figure 1—source data 2 ) . Most of these events had intermediate inclusion levels in X2 samples ( which comprise a mixture of stem cells and their early differentiation progeny , Figure 1C ) , while inclusion levels in whole worm ( WW ) samples largely matched those of Xins samples . This latter correlation is expected since ~80% of the cells in adult planarians are differentiated ( Baguñà and Romero , 1981 ) . In contrast to the global AS pattern ( Figure 1—figure supplement 2A ) , the majority of neoblast-differential AS events involved cassette exons ( 61 . 4% , Figure 1B ) . RT-PCRs using primers ( Supplementary file 1 ) flanking alternative exon sequences confirmed all ( 22/22 ) tested neoblast-differential exons and showed a high correlation between RT-PCR and RNA-Seq inclusion level estimates ( Figure 1D and Figure 1—figure supplement 3; R2 = 0 . 92 , n = 88 ) . Among these cassette exons , we detected 64 microexons that were differentially regulated between X1 and Xins fractions , the majority of which ( 82 . 8% ) were more included in differentiated cells ( Figure 1E ) . Consistent with this bias towards differentiated cells , vertebrate microexons have been recently shown to display enrichment in neural and muscle tissues compared to ESCs ( Irimia et al . , 2014 ) . The majority of neoblast-differential exons ( 68 . 5% ) overlapped protein-coding regions and conserved the reading frame; therefore , they were predicted to generate distinct protein isoforms in neoblasts and differentiated cells . This set of AS events were significantly enriched in genes involved in cytoskeleton and cell signalling functions ( Figure 1—figure supplement 4A ) , gene ontology categories that were also enriched among ESC-differential alternatively spliced genes in mammals ( Han et al . , 2013 ) . Planarian genes with neoblast-differential exons were associated with a wide range of functions , and included cytoskeleton regulators ( e . g . add3 ) , membrane trafficking proteins ( e . g . tpd52 ) , metabolic enzymes ( e . g . pcca ) , translation factors ( e . g . eif3h ) and protein kinases ( e . g . map4k3 ) , among others . Interestingly , neoblast-differential exons significantly more often overlapped disordered regions of proteins and avoided structured domains compared to general alternatively spliced and constitutive exons ( Figure 1—figure supplement 2C , D; p ≤ 4 . 7 × 10−5 for all comparisons , 3-way Fisher test and proportion tests , respectively ) . A similar pattern was described for tissue-specific exons in mammals ( Buljan et al . , 2012; Ellis et al . , 2012 ) , suggesting that neoblast-differential exons may also contribute to modulate protein-protein interactions in planarian cells . 10 . 7554/eLife . 16797 . 011Figure 3 . The neoblast-specific AS program is extensively conserved in D . japonica . ( A ) Left: representative RT-PCR assays monitoring AS patterns for five representative neoblast-differential AS events in FACS isolated cell fractions from D . japonica . Scatter plot shows correspondence between ΔPSI ( X1-Xins ) estimates by RT-PCR in S . mediterranea and D . japonica for 21 cassette exons ( circles ) and 6 retained introns ( diamonds ) . Conservation of regulation is observed for both alternative sequences with higher ( 'X1-inc' , red ) and lower ( 'X1-exc' , blue ) inclusion in neoblasts . ( B ) Schematic examples of the occurrence of neoblast-differential AS ( blue bars ) with respect to protein domain organization in two pairs of gene homologues in human and planarian . The examples show that AS events fall in similar protein regions in both human and planarian orthologs . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01110 . 7554/eLife . 16797 . 012Figure 3—source data 1 . Orthologous gene groups with neoblast/ESC-differential exons in both planarian and humans . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01210 . 7554/eLife . 16797 . 013Figure 3—source data 2 . Conservation of stem cell-differential AS events . ( A ) Comparison between S . mediterranea and D . japonica AS . ( B ) Comparison between planarian and human stem cell regulated AS ( Refers to Figure 3—source data 1 ) DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01310 . 7554/eLife . 16797 . 014Figure 4 . Identification of bruli and mbnl as major regulators of neoblast-specific AS . ( A ) Scattered plot highlighting differential gene expression for selected tissue-specific AS factors in planarians . X-axis: differential gene expression in X1 vs Xins cell fractions for selected AS factors ( color dots ) and core spliceosomal components ( grey dots ) . The red line is the median enrichment value for the core spliceosomal components . An AS factor is thus considered to be enriched in X1 or Xins fractions if it is located to the right or to the left of this line , respectively . Y-axis: gene expression levels in whole worms ( WW ) , using the cRPKM metric . ( B ) Test of regeneration speed after head ablation upon AS factor knockdown . Identification of normally looking eyespots was used as proxy for complete regeneration for 10 individuals per experimental condition . ( C ) ΔPSI estimates in X1 cell fractions by RT-PCR for two X1-included exons ( red ) , one X1-included retained intron ( c17orf39 , orange ) and five X1-excluded exons ( blue ) 10 days after RNAi treatment with dsRNA coding for AS factor combinations . ΔPSI values for each event and experiment are calculated respect to the average PSIs in three wild type samples . ( D ) Percent of neoblast-differential exons with sufficient read coverage that change their inclusion levels ( ΔPSI ≥ 15 ) in whole worms towards the X1 pattern ( as expected for a negative regulator of neoblast-differential AS ) upon knockdown of each AS factor . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01410 . 7554/eLife . 16797 . 015Figure 4—source data 1 . Annotation of RBPs in planarians . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01510 . 7554/eLife . 16797 . 016Figure 4—source data 2 . Annotation of spliceosomal components in planarians . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01610 . 7554/eLife . 16797 . 017Figure 4—source data 3 . Single and multiple splicing factor knockdown RNAi groups . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01710 . 7554/eLife . 16797 . 018Figure 4—figure supplement 1 . Identification of bruli and mbnl as major regulators of neoblast-specific AS . ( A ) FACS scatter plots for bruli and control RNAi treated animals 10 days after RNAi inyection . ( B ) Proportion of cells in X1 , X2 and Xins fractions 10 days after bruli and control RNAi treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 01810 . 7554/eLife . 16797 . 019Figure 4—figure supplement 2 . Enrichment of RBP binding motifs associated with neoblast-differential exons . Motif enrichment analyses for all RNAcompete-derived motifs ( Ray et al . , 2013 ) for neoblast-differential alternatively spliced exons with increased or decreased inclusion in X1 fraction . Each intronic box corresponds to a 20-nucleotide bin at the indicated location with respect to the AS exon . The whole exonic sequence was evaluated as a single bin . In each box , counts of the top 10% most strongly bound 7-mers were compared between the AS set of interest and a non-regulated background set . Colors encode significance of enrichment: log10 P-value Fisher's exact test . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 019 Intron retention ( IR ) consists in the selective retention of a given intron in the mature mRNA . Notably , 107 IR events were differentially enriched in X1 fractions compared to only 41 in Xins ( Figure 1B ) . RT-PCRs confirmed all ( 11/11 ) tested X1-included introns compared to half ( 2/4 ) of X1-excluded ( Figures 2A and Figure 1—figure supplement 3; R2 = 0 . 68 , n = 56 , for all neoblast-differential introns tested ) . X1-included introns were usually longer than other intron types ( Figure 2B , P = 7 . 1 × 10−10 , Wilcoxon Sum Rank test ) , and belonged to genes that were more highly expressed in differentiated cells than in neoblasts and their progeny ( Figure 2C–D ) . Importantly , in contrast to cassette exon events , which were predicted to generate different protein isoforms in neoblast and differentiated cells , nearly all X1-included IR events were predicted to disrupt the reading frame specifically in neoblasts ( Figure 2E; p<2 . 2 × 10−16 , chi-squared test ) . Moreover , these introns preferentially affected genes that were significantly enriched in functions related to cell differentiation , negative regulation of proliferation and cell type-specific metabolic pathways ( Figure 1—figure supplement 4B–C ) . Therefore , altogether these data suggest that neoblast-specific IR may be operating to ensure that a subset of early-transcribed differentiation genes is not active in neoblasts . These genes might then be selectively activated upon differentiation by splicing out the 'detained' intron , similar to recent reports in mammalian ESCs ( Boutz et al . , 2015; Braunschweig et al . , 2014 ) . To investigate whether the neoblast-differential AS program of S . mediterranea has been conserved during evolution , we mapped our validated set of AS exons ( Figure 1—figure supplement 3 ) to a transcriptome assembly of the planarian species Dugesia japonica ( Nishimura et al . , 2015 ) , which has diverged from S . mediterranea approximately 85 million years ago ( Lazaro et al . , 2011 ) , similar to the estimated time for the human-mouse divergence . We used this information to design D . japonica-specific primers ( Supplementary file 1 ) , and then performed RT-PCRs in FACS isolated populations from D . japonica and compared them to S . mediterranea . Strikingly , 20/21 ( 95 . 2% ) of the probed cassette exons were also present and alternatively spliced in D . japonica , and 16/20 ( 80% ) of these showed conserved differential regulation between X1 and Xins fractions in the two species ( Figure 3A and Figure 3—source data 2 ) . Moreover , half of the IR events that could be probed despite the lack of a reference genome sequence in D . japonica also displayed conserved neoblast-differential regulation ( 3/6; Figure 3A ) . This level of neoblast-differential conservation is even higher than that observed between ESC-differential exons in mammals ( Han et al . , 2013 ) , and strongly argues for the functional relevance of this AS program in planarians . We next asked whether neoblast-differential exons were conserved between neoblasts and human ESCs . We identified 15 orthologous gene groups with neoblast/ESC-differential exons in both planarian and humans ( Figure 3—source data 1 , Materials and methods ) . Remarkably , in ten of these orthologous groups , planarian and human stem cell regulated exons fall in similar regions of the protein , and in at least four of them ( add3 , eml4 , fmnl3 and ppfibp1 ) the intervening sequences are partly orthologous , despite extensive rearrangement of intron-exon structures ( Figure 3B and Figure 3—source data 2 ) . Therefore , although neoblast and ESC AS programs are largely lineage-specific , AS may impact some proteins in a similar manner in both species . We next sought to identify trans-acting RBP factors that regulate neoblast-differential AS . Based on homology searches and presence of RNA binding domains , we identified over 300 putative RNA binding proteins in planarians ( Figure 4—source data 1 ) . Homologs of well-known mammalian tissue-specific AS factors were then selected as potential regulators of the planarian neoblast AS program ( Figure 4A ) . To first assess their differential enrichment in neoblast or differentiated cell fractions , we compared their gene expression levels across our RNA-Seq panel with those of single-copy core spliceosomal components ( genes associated with KEGG terms for spliceosomal complex A , B and C; Figure 4—source data 2 ) . Strikingly , a member of the CELF family ( Smed-bruno-like , bruli ) of AS regulators showed the strongest differential expression enrichment in X1 fractions among the multiple probed AS factors ( Figure 4A and Figure 4—source data 1 ) , and had been previously reported to be needed for planarian stem cell self-renewal and regeneration ( Guo et al . , 2006 ) . Most other investigated AS factors showed no differential enrichment in X1 fractions compared to the core spliceosome ( e . g . Smed-ptbp-1 , -2 , Smed-rbfox-1 ) , or were differentially enriched in Xins fractions ( e . g . Smed-mbnl-1 , Smed-esrp-1 , 2 , etc . ) ( Figure 4A and Figure 4—source data 1 ) . To query the role of these factors in the regulation of the neoblast AS program and in regeneration , we performed RNAi of these genes separately or in combinations ( for paralogs with similar enrichment in X1 or Xins fractions ) . A total of eight RNAi groups comprising two single RNAi and six multiple RNAi combinations ( Figure 4—source data 3 ) were knocked down in parallel with a control RNAi . Tests of regeneration capability upon head ablation in these knockdowns showed diverse defects for several AS factor groups ( Figure 4B ) . As previously reported , bruli ( RNAi ) showed a dramatic phenotype with no regeneration . A similar phenotype was also observed for the ptbp ( RNAi ) group . Combined knockdown of the three other CELF factors , enriched in Xins fractions , induced elongation and movement defects , but did not seem to affect regeneration time dynamics . Finally , a significant delay in regeneration was observed for the mbnl group . To evaluate the transcriptomic impact of these factors in regulating neoblast-differential AS in stem cells , we FACS-sorted planarian X1 cell populations from these eight RNAi groups 10 days after RNAi ( when the neoblast population was still not decreased upon bruli RNAi treatment , Figure 4—figure supplement 1 ) , and performed semi-quantitative RT-PCR for eight representative neoblast-differential AS events ( Figure 4C ) . Only bruli ( RNAi ) treatment induced dramatic changes in inclusion levels in X1 fractions compared to the control RNAi in the expected direction ( i . e . towards the differentiated pattern; Figure 4C ) , suggesting that bruli may act as a positive regulator of the neoblast AS program . Next , to assess the importance of the AS factors that were enriched in differentiated cells , we performed RNA-Seq of whole worms from these eight RNAi groups . Knockdown of the mbnl group [a four-gene knockdown: Smed-mbnl-1;Smed-mbnl-like-1 , -2 , -3 ( RNAi ) , hereafter mbnl ( RNAi ) ] ( Figure 4—source data 3 ) showed the most widespread changes in the X1-differential AS program towards the neoblast pattern ( Figure 4D ) , suggesting that MBNL proteins may act as negative regulators of the neoblast AS program . Remarkably , RNA sequence motif enrichment analyses using a library of RNAcompete-derived binding profiles for over 100 animal RBPs ( Ray et al . , 2013 ) showed that MBNL consensus motifs are also the most significantly enriched in the downstream introns of X1-excluded cassette exons ( Figure 4—figure supplement 2 ) . This result is consistent with planarian MBNL proteins enhancing X1-excluded exons via binding to their downstream introns , as described for mammalian MBNL proteins ( Han et al . , 2013; Wang et al . , 2012 ) . Based on these data , we decided to investigate the role of BRULI and MBNL factors as putative positive and negative regulators of the neoblast-differential AS program , respectively . We found four potential mbnl orthologs in the planarian transcriptome , comprising a canonical MBNL ortholog with two pairs of zinc finger domains ( Smed-mbnl-1 , hereafter mbnl-1 ) , as well as three orthologs with only one pair ( Smed-mbnl-like-1 , 2 , 3 , hereafter mbnl-like-1 , 2 , 3 ) ( Figure 5—figure supplement 1 ) . All mbnl orthologs displayed expression enrichment in Xins fractions ( Figure 4A ) . Whole mount in situ hybridizations showed that mbnl-1 has a widespread expression pattern , while mbnl-like-1 and mbnl-like-2 are expressed mainly in gut tissue ( Figure 5A; no expression was detected for mbnl-like-3 ) . Analysis of recently published single-cell sequencing data further revealed expression of mbnl-1 , mbnl-like-1 and mbnl-like-2 in epidermis , which is commonly lost in whole-mount in situ protocols , and confirmed the widespread expression of mbnl-1 ( Wurtzel et al . , 2015 ) . In contrast , as previously reported ( Guo et al . , 2006 ) , bruli was specifically expressed in neoblasts , as shown both by in situ hybridization ( Figure 5A ) and single-cell sequencing data ( Wurtzel et al . , 2015 ) . 10 . 7554/eLife . 16797 . 020Figure 5 . Combined mbnl knockdown has stronger effects than individual knockdown . ( A ) Whole worm in situ hybridization for bruli , mbnl-1 and mbnl-like1 , 2 and 3 . Scale bar is 0 . 5 mm . ( B ) ΔPSI in whole worm estimates by RT-PCR for two X1-inc exons ( red ) , one X1-inc retained intron ( c17orf39 , orange ) and five X1-exc exons ( blue ) 10 days after RNAi treatment with dsRNA coding for bruli , mix of dsRNAs against the four mbnl genes [mbnl ( RNAi ) ] , mbnl-1 , mbnl-like-1 , mbnl-like-2 , and mbnl-like-3 . Two independent controls samples were included; ΔPSI values are relative to the first control sample . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 02010 . 7554/eLife . 16797 . 021Figure 5—figure supplement 1 . BRULI and MBNL factors domain architecture and effects on regeneration and alternative splicing . ( A-B ) Amino acid sequence ( A ) and schematization of protein domains ( B ) of BRULI , MBNL-1 , MBNL-LIKE-1 , MBNL-LIKE-2 and MBNL-LIKE-3 . RRM1 domains are depicted in green and Zinc-finger domains in magenta , and their CCCH aminoacids indicated in red . The X1-inc alternative exon of mbnl-1 is depicted in cyan . ( C ) Sequence identity matrix between planarian and human CELF RRM1 ( top ) and MBNL ZnF ( bottom ) domains . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 02110 . 7554/eLife . 16797 . 022Figure 5—figure supplement 2 . bruli and mbnl factors effects on regeneration and alternative splicing . ( A ) The effects of bruli and mbnl knockdown and mbnl-1 single knock down are reproducible and dependent on the level of cut . Top: Workflow of the experiment: 10 animals per experimental group and time point were injected with dsRNA coding for bruli , gfp as a control and a mix of dsRNAs against the four mbnl genes and cut 22 days after RNAi; trunk and tail pieces were monitored independently . Two independent dsRNAs coding for different regions of mbnl-1 were used ( mbnl-1 and mbnl-1* ) . The experiment was replicated twice independently . Bottom: Results of the scoring for trunk and tail regeneration experiments; color key is indicated below . ( B ) The effects of bruli and mbnl knockdown on AS events are detectable 5 days after RNAi and are permanent for at least 30 days: quantifications for two X1-inc exons ( red ) , one X1-inc retained intron ( c17orf39 , orange ) and five X1-exc exons ( blue ) 5 , 10 , 15 , 20 , 25 and 30 days after RNAi treatment with dsRNA coding for bruli , gfp as a control and a mix of dsRNAs against the four mbnl genes . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 022 To further investigate the role of CELF and MBNL factors in the regulation of the neoblast AS program and regeneration , we performed RNAi of these genes , separately or in combinations ( in the case of the mbnl paralogs ) . In the case of mbnl , a significant delay in regeneration was observed for the combined knockdown of all four mbnl homologues . These regeneration defects were even stronger when regenerating tails were examined ( Figure 5—figure supplement 2A ) . When knocked down individually , only Smed-mbnl-1 had minor effects on regeneration ( Figure 5—figure supplement 2A ) . These observations were mirrored at the transcriptomic level , as shown by RT-PCRs for several representative AS events in whole worms ( Figure 5B ) . Importantly , while the regeneration phenotype was observed 22 days after the RNAi injection , effects on AS inclusion levels where observable as soon as five days after the initiation of the treatment , plateauing after day 10 post-injection ( Figure 5—figure supplement 2B ) . In order to investigate the concerted transcriptomic-wide impact of CELF and MBNL factors in the regulation of neoblast-differential AS , we next FACS-sorted X1 and Xins populations from bruli ( RNAi ) , mbnl ( RNAi ) and control ( RNAi ) animals 10 days after RNAi in duplicates and subjected them to RNA-Seq . Consistent with our RT-PCR results ( Figure 4C ) , bruli knockdown induced strong inclusion level changes in X1 fractions towards the differentiated pattern ( Figure 6A; P = 1 . 4 × 10−19 , binomial test ) . In particular , X1-included alternative sequences became less included after bruli knockdown ( Figure 6A , red dots ) while X1-excluded ones were more included ( Figure 6A , blue dots ) . On the other hand , mbnl knockdown induced changes in the opposite direction in Xins fractions ( Figure 6B; P = 2 . 1 × 10−11 , binomial test ) . These changes in inclusion levels obtained by RNA-seq analyses were independently validated by RT-PCR assays ( 19/19 changes , 100%; R2 = 0 . 87 , n = 225; Figure 6C and Figure 6—figure supplement 1 ) . Remarkably , 33/87 ( 37 . 9% ) mbnl-regulated AS events ( 14 cassette exons , 11 microexons , and 8 retained introns ) were also affected by bruli knockdown in the opposite direction ( Figure 6D; P = 2 . 2 × 10−7 , binomial test; corresponding to 33/141 ( 23 . 4% ) bruli-regulated AS events ) . As expected , the number of AS changes of bruli and mbnl knockdown in Xins and X1 fractions respectively was lower and was not significantly associated with neoblast-differential regulation ( Figure 6—figure supplement 2A , B ) . Of note , 37/110 ( 33 . 6% ) of X1-retained introns showed decreased retention upon bruli knockdown ( ΔPSI ≥ 15 ) , suggesting that BRULI regulates coordinated IR events in planarian stem cells . On the other hand , mbnl knockdown affected a large fraction of Xins-enriched microexons ( 20/54 , 37% ) . 10 . 7554/eLife . 16797 . 023Figure 6 . bruli and mbnl antagonistically regulate neoblast-specific AS . ( A ) High negative association ( p<1 . 4 × 10−19 , one-sided binomial test ) between differences in inclusion levels ( ΔPSI ) of X1-differential AS events in X1 versus Xins fractions , and differences in bruli and control RNAi treated X1 fractions . Red/Blue dots correspond to X1-included/excluded AS events . ( B ) High positive association ( p<2 . 1 × 10−11 , one-sided binomial test ) between differences in inclusion levels ( ΔPSI ) of X1-differential AS events in X1 versus Xins fractions , and differences in mbnl and control RNAi treated Xins fractions . Only AS events with sufficient read coverage in the control and KD samples and an absolute ΔPSI>15 are plotted . ( C ) RT-PCR assays monitoring AS patterns for 5 representative X1-differential AS events in FACS isolated cell fractions treated with bruli , control or mbnl RNAi . ( D ) Most X1-differential AS events that are affected by both bruli and mbnl knockdown are regulated in an antagonistic manner ( p<2 . 2 × 10−7 , one-sided binomial test ) . ( E ) 2-dimensional histogram of RNA-compete 7-mer Z-scores , comparing the sequence-specific binding of planarian and human proteins . Unspecific 7-mers with Z-score <0 for both RBPs were excluded . ρ , Spearman rank correlation; R , Pearson correlation . Top 10 planarian motifs are highlighted . ( F ) Motif-enrichment analysis . Each intronic box corresponds to a 20-nucleotide bin at the indicated location relative to the AS exon ( middle box ) . Color encodes the significance of enrichment ( Fisher's exact test , Bonferroni corrected for the number of tested bins ) of high affinity 7-mers for BRULI or MBNL , comparing each differentially spliced exon set with an unaffected background set . Exon sets ( I-IV ) correspond to those in quadrants indicated in panels B and C . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 02310 . 7554/eLife . 16797 . 024Figure 6—figure supplement 1 . Examples of neoblast-differential AS events regulated by bruli and/or mbnl . RT-PCR assays for 25 representative neoblast-differential exons and retained introns in X1 , X2 and Xins fractions of worms treated with bruli , control or mbnl RNAi . Red and blue exons/introns indicate those exons/introns with higher and lower inclusion in X1 compared to Xins fractions , respectively . Scatter plot shows correspondence between inclusion level estimates by RNA-Seq and RT-PCR ( R2 = 0 . 87 , n = 225 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 02410 . 7554/eLife . 16797 . 025Figure 6—figure supplement 2 . bruli and mbnl antagonistically regulate neoblast-specific AS . ( A-B ) bruli and mbnl have a minor impact on neoblast-differential AS events in Xins and X1 , respectively: ( A ) Lack of association ( p = 0 . 206 , one-sided binomial test ) between differences in inclusion levels ( ΔPSI ) of neoblast-differential AS events in X1 versus Xins fractions , and differences in bruli and control RNAi treated Xins fractions . ( B ) Lack of association ( p = 0 . 714 , one-sided binomial test ) between differences in inclusion levels ( ΔPSI ) of neoblast-differential AS events in X1 versus Xins fractions , and differences in mbnl and control RNAi treated X1 fractions . ( C ) RNAcompete-derived motifs for BRULI and MBNL proteins . Top-scoring RNA-binding 7mers and corresponding logos identified by RNAcompete analysis are shown for MBNL-1 ( excluding a neoblast-differential exon; Xins isoform ) , MBNL-1 ( including a neoblast-differential exon; X1 isoform ) , MBNL-LIKE-1 , MBNL-LIKE-2 , and BRULI . Scatterplots show the correlation between 7-mers Z-scores derived from microarray data splits , SetA and SetB . Logos are created after aligning the top 10 highest scoring 7-mers . Both isoforms of MBNL-1 display similar binding preferences . RNAcompete retrieved simpler 4-mer or 3-mer containing GCU motifs for MBNL-LIKE-1 and MBNL-LIKE-2 , which contain only one pair of zinc fingers . ( D ) Conserved binding specificities for planarian and human CELF and MBNL splicing factors: Venn diagrams showing the overlap of RNA-compete 7-mers , comparing the top 100 ( top ) and the top 1000 ( bottom ) 7-mers of planarian and human proteins . P-values correspond to hypergeometric tests . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 025 To investigate the mechanisms by which BRULI and MBNL regulate neoblast-differential AS , we performed RNAcompete assays ( Ray et al . , 2009; Ray et al . , 2013 ) to identify their consensus RNA binding motifs . Purified GST-tagged RBPs were incubated with a 75-fold excess of an RNA pool and the binding preferences of the RBP elucidated by analysing bound RNAs by microarray analyses . The top scoring motifs ( Figure 6E and Figure 6—figure supplement 2C ) for BRULI and MBNL-1 contained the 3-mer UGU and 6-mer GCUUGC , respectively , consistent with previous reports for their mammalian homologs . Moreover , the binding specificities of the planarian and mammalian homologs strongly correlated ( Figure 6E ) and highly significantly overlapped ( Figure 6—figure supplement 2D; p<5 . 5 × 10−89 for all pairwise homolog comparisons , hypergeometric test ) across all measured 7-mers in RNAcompete assays . Thus , the planarian MBNL and CELF factors and its mammalian homologs have conserved binding specificities . Next , to investigate if planarian MBNL and CELF binding motifs were significantly associated with neoblast-differential alternative exons that show changes upon splicing factor knockdown ( Figure 6A , B ) , we analysed the presence of these motifs in the regulated exons and their flanking introns compared to sets of control exons ( see Materials and methods ) . Both BRULI and MBNL-1 motifs were highly significantly enriched in the downstream introns of exons downregulated upon the respective KD , whereas MBNL-1 motifs were also significantly enriched in exonic sequences of exons upregulated upon mbnl KD ( Figure 6F ) . These locations are consistent with the enhancing or repressive functions , respectively , from RNA regulatory maps described for these factors in mammals ( Han et al . , 2013; Wang et al . , 2012 ) , and strongly suggest a direct regulatory role in AS regulation in planarians . Next , to investigate how knockdown of bruli and mbnl impairs planarian regeneration at the cellular level , we sequenced mRNA extracted from whole worms at different time points after RNAi injections and evaluated the levels of cell type-specific markers ( Figure 7A ) . It was previously described that bruli knockdown induces a conspicuous neoblast loss phenotype ( Guo et al . , 2006 ) . Consistently , we observed a strong downregulation of neoblast markers after bruli knockdown ( Figure 7—figure supplement 1 ) , which were not affected by mbnl knockdown . Strikingly , the progeny markers prog-1 and prog-2-1 , which are expressed in postmitotic epidermal progenitors ( Eisenhoffer et al . , 2008; van Wolfswinkel et al . , 2014 ) , changed their expression levels in opposing directions upon bruli/mbnl knockdown . More precisely , bruli knockdown induced loss of these markers , whereas mbnl knockdown led to an increase in their expression ( Figure 7B and Figure 7—figure supplement 1 ) . Similar results were found with recently described markers for lineage-committed subclasses of neoblasts ( van Wolfswinkel et al . , 2014 ) ( Figure 7—figure supplement 1 ) and epidermal progenitors ( Zhu et al . , 2015 ) ( Figure 7B , Figure 7—figure supplement 1 ) , which were confirmed by qPCR for prog-1 , prog-2-1 and prog-1-1 ( Figure 7C ) . This concomitant decrease and increase in the expression of multiple progeny markers for bruli and mbnl knockdown , respectively , is likely due to the loss and accumulation of progenitors , respectively , rather than to specific changes in the expression levels of these markers in an invariant pool of progenitor cells . 10 . 7554/eLife . 16797 . 026Figure 7 . bruli and mbnl antagonistically regulate neoblast biology . ( A ) Schematic representation of phenotypic analysis experiments . ( B ) Gene expression changes estimated by RNA-Seq of several progeny markers 25 days after bruli ( green ) or mbnl ( purple ) RNAi treatment compared to controls . ( C ) Quantification of gene expression levels by qPCR of prog-1 , prog-2-1 , and prog-1-1 in whole worms 25 days after treatment with bruli , control or mbnl RNAi . ( D ) Proportion of genes belonging to each of the 6 clusters defined in ( Onal et al . , 2012 ) based on their X1vs Xins enrichment that are decreased 20 , 25 and 30 days after bruli and mbnl RNAi treatment . P-values correspond to hypergeometric tests for Clusters 1 ( P = 5 . 6 × 10−56 , P = 4 . 0 × 10−113 , P = 2 . 7 × 10−178 , for bruli ( RNAi ) days 20 , 25 and 30 , respectively ) and 6 ( P = 1 . 2 × 10−3 , P = 1 . 0 × 10−14 , P = 1 . 2 × 10−16 , for mbnl ( RNAi ) days 20 , 25 and 30 , respectively ) . All other tests were not significant . Black histograms on the right side indicate schematically , for each cluster , the relative gene expression levels in each cell fraction . ( E ) In situ hybridization of three representative Cluster 1 ( red bars ) and Cluster 6 ( blue bars ) genes in whole worms . ( F ) qPCR-based gene expression estimates of three representative Cluster 1 ( red bars ) and Cluster 6 ( blue bars ) genes in whole worms 25 days after bruli , control or mbnl RNAi . ( G ) Principal Component 1 separates transcriptomes from bruli RNAi treated samples together with Smed-H2B ( RNAi ) ( which affects neoblast self-maintenance ) from mbnl RNAi treated samples together with Smedwi-2 ( RNAi ) ( which impairs neoblast differentiation ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 02610 . 7554/eLife . 16797 . 027Figure 7—figure supplement 1 . bruli and mbnl knockdown have contrasting effect of specific gene markers . Log2 fold change in gene expression of key markers 20 , 25 and 30 days after bruli ( left ) or mbnl ( right ) treatment , measured from RNA-Seq . Red/Green intensities indicate different levels of down/up regulation . bruli and mbnl knockdown have distinct effects on different types of markers . In particular , they show antagonistic effects of 'Early progeny' markers . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 02710 . 7554/eLife . 16797 . 028Figure 7—figure supplement 2 . Gene Ontology analysis of genes downregulated upon bruli and mbnl knockdown . Gene ontology enrichment analysis for 'Biological Process' , 'Molecular Function' and 'Cellular Component' terms of genes downregulated upon bruli and mbnl knockdown was performed for genes that were significantly downregulated after treatment with bruli ( green ) or mbnl ( red ) RNAi . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 028 Next , we compared the effect of bruli and mbnl knockdown on the levels of gene sets previously defined by clustering common patterns of expression in X1 , X2 and Xins fractions ( Onal et al . , 2012 ) . Cluster 1 comprises genes with strong differential expression enrichment in neoblasts , while Cluster six genes are conversely expressed mainly in differentiated tissues and excluded from the neoblast compartment . Consistent with a prominent neoblast loss ( Guo et al . , 2006 ) , Cluster 1 genes were highly significantly decreased upon bruli knockdown ( Figure 7D ; p<5 . 6 × 10−56 for the three time points , hypergeometric test ) . In contrast , mbnl knockdown impacted predominantly Cluster six genes ( Figure 7D ; p<0 . 01 for the three time points , hypergeometric test ) . The expression patterns of three decreased genes from Clusters 1 and the from Cluster 6 were confirmed by in situ hybridization ( Figure 7E ) . The three Cluster 1 transcripts were observed in neoblasts as previously described; on the other hand , the three Cluster six genes impacted by mbnl knockdown were expressed only in differentiated cells ( comprising gut and secretory-like cells ) . We then confirmed the depletion of these markers after bruli and mbnl RNAi by qPCR . ( Figure 7F ) . The three Cluster 1 transcripts were strongly decreased after bruli knockdown , while the three Cluster six genes were decreased only after mbnl knockdown . These results are thus consistent with prominent neoblast loss after bruli RNAi , on the one hand , and progressive scarcity of differentiated cell types due to impaired neoblast differentiation after mbnl knockdown , on the other hand . These opposed effects of bruli and mbnl were also reflected in the gene functions altered upon knockdown , with stem cell function-related terms enriched in bruli ( RNAi ) decreased genes , compared to differentiated tissue function associated terms for mbnl ( RNAi ) ( Figure 7—figure supplement 2 ) . In addition , a Principal Component Analysis ( PCA ) showed that the first PC sharply separated bruli samples together with Smed-H2B ( RNAi ) ( which affects neoblast self-maintenance [Solana et al . , 2012] ) from mbnl samples together with Smedwi-2 ( RNAi ) ( which impairs neoblast differentiation [Reddien et al . , 2005] ) ( Figure 7G ) , further showing their opposing effects in stem cell properties . Taken together our results show that both CELF and MBNL factors are required for regeneration in planarians in vivo , and that they act by antagonizing each other in the control of neoblast self-renewal and differentiation .
In this study we systematically characterized AS in planarian stem cells . We find widespread and strong differences in the pre-mRNA processing of genes expressed in both stem and differentiated cells . Most of these AS differences were evolutionarily conserved between two planarian species , strongly arguing for functional importance ( Irimia et al . , 2009 ) . Furthermore , even several human genes undergo stem cell-specific AS in similar regions to those of their planarian orthologs . Moreover , we also found a large number of introns that are specifically retained in neoblast transcripts and that reduce the production of functional proteins in this cell type . Our results thus provide evidence that IR may play an important and distinct role in the regulation of stem cell gene expression . Our results further indicate that BRULI and MBNL proteins functionally interplay to regulate planarian regeneration . bruli is highly expressed in stem cells and contributes to shape neoblast-specific transcriptomes at least in part through regulation of exon skipping and intron retention . Depletion of bruli results in neoblast loss and thus in a complete lack of regeneration . On the other hand , mbnl factors are more expressed in differentiated cells , where they contribute to establish differentiated gene expression . Therefore , loss of mbnl function is likely to affect neoblast differentiation , thereby reducing and/or slowing down regeneration . Our computational , biochemical and functional experiments indicate that CELF and MBNL proteins functionally antagonize by directly binding to their targets to control stem cell biology and regeneration in planarians ( Figure 8 ) . Thus , these two families of RBPs not only play antagonistic roles in the regulation of multiple mammalian differentiation systems ( Kalsotra et al . , 2008; Lin et al . , 2006; Ward et al . , 2010 ) , but directly control stem cell-specific AS in an evolutionarily distant organism . Moreover , MBNL proteins have been shown to be major direct negative regulators of mammalian ESC-differential AS ( Han et al . , 2013; Venables et al . , 2013 ) , similarly to what we describe here for planarian neoblast AS . While MBNL targets are largely not shared between planarians and humans ( only seven orthologous gene groups have MBNL-regulated , non-homologous , exons in both species ) , the upstream role of MBNL proteins as repressors of stem cell-differential AS appears equivalent in both species . This suggests that direct negative regulation of stem cell transcriptomes by MBNL proteins , and perhaps positively by CELF factors , is a deeply conserved ancestral feature of animal stem cells . Indeed , a recent study has shown that sponges – the most basal metazoan lineage –also differentially express a CELF homolog in their stem cells and an MBNL homolog in differentiated cells ( Alié et al . , 2015 ) . However , since the true homology status of planarian , poriferan and human stem cells is unclear , it also possible that CELF and MBNL proteins form an ancestral developmental regulatory module that has been deployed to regulate pluripotent vs . differentiated transcriptomes multiple times across evolution . For instance , in Drosophila melanogaster a Bruno homolog ( aret ) is only highly expressed in the early pluripotent embryo stages ( from 0 to 4 hr post fertilization ) during development , while the embryonic expression of the single Mbnl ortholog ( mbl ) gradually increases from 6 hr post fertilization ( Graveley et al . , 2011 ) . Furthermore , we have shown that not only the antagonistic expression but also the binding specificities of BRULI and MBNL proteins are very similar to those of their human counterparts , an observation that is in line with the high level of conservation of binding specificities and positional regulatory codes described for several other RBPs ( Brooks et al . , 2011; Irimia et al . , 2011; Lareau and Brenner , 2015; Ray et al . , 2013; Wang et al . , 2012; Loria et al . , 2003 ) . The dual regulation by CELF and MBNL suggests that other interactions between factors in the control of AS might exist . 10 . 7554/eLife . 16797 . 029Figure 8 . Model for BRULI and MBNL regulation of neoblast-specific AS . Schematic representation of BRULI and MBNL regulation of AS in different planarian cell fractions and respective RNAi mediated knockdown phenotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 16797 . 029 In summary , we have unveiled an ancient post-transcriptional antagonistic regulatory switch formed by CELF and MBNL factors that is associated with pluripotency in multiple animal systems . This evolutionary conservation contrasts with the apparent high turnaround of stem cell regulation at the transcription factor level , in which the core set of TFs that regulate mammalian ESCs is likely a lineage-specific innovation ( Fernandez-Tresguerres et al . , 2010 ) . Future investigation of post-transcriptional regulation in other in vivo systems of stem biology and regeneration should further contribute to establishing ancestral mechanisms for the control of pluripotency in metazoans .
We used the publically available assembly version 3 . 1 of the genome of the freshwater planarian Schmidtea mediterranea . Given the relatively fragmentary nature of this genome assembly , we used a recent de novo genome-independent transcriptome assembly based on multiple RNA-Seq data sources ( Dresden transcriptome - http://planmine . mpi-cbg . de/ ) as transcript reference for the annotation . The 26 , 562 transcripts from this transcriptome were mapped to the genome using blat with the following parameters:–q = rna –fine ( to search for initial and terminal exons in high quality transcripts ) . The output was then sorted using pslSort , and then filtered using pslCDnaFilter with the following parameters: -minCover = 0 . 75 -minId = 0 . 96 -globalNearBest = 0 . 005 –filterWeirdOverlapped . If multiple redundant mappings ( i . e . of overlapping sequences in the query ) of a single transcript fulfilled these conditions , only the hit with the best blat score was kept . If different regions of a transcript mapped to different scaffolds ( 936 transcripts ) , the transcript was split into two or more gene models , one in each scaffold ( gene sub-models were designated with suffixes following '_2' , '_3' , etc . ) . Next , a series of filters were applied to remove usual annotation errors . First , the strand of a gene model was flipped if there were more CT-AC than GT-AG introns ( often occurring in gene chimeras in opposite orientations; 301 gene models ) . Second , extra multiple-mapping transcript regions within the same scaffold were filtered out ( in 281 gene models ) . Third , exons separated by annotated introns shorter than 30 nucleotides ( usually 1–5 nucleotides ) were collapsed into single exons ( these often reflected UTR polymorphisms or discrepancies between the transcriptome and genome assemblies; 7665 merged exons ) . This resulted in a GTF containing 23 , 161 single-transcript gene models . We next used TransDecoder ( version 1 ) ( Haas et al . , 2013 ) with default parameters and compared against the Pfam A database to identify coding sequences ( CDS ) . A total of 15 , 008 gene models had detected open reading frames ( ORFs ) . For those cases in which TransDecoder detected more than one ORF in a given transcript ( usually due to assembly errors that introduce PTCs ) , we generated multiple transcript identifiers ( designated with suffixes following '_b' , '_c' , etc . ) . Custom perl scripts were then used to add start_codon and stop_codon lines to the GTF . Finally , to provide meaningful gene names , we performed tblastx ( e-value 0 . 001 , without low complexity filter ) of the original planarian transcriptome against human mRNAs from Ensembl v71 and assign the gene symbols of best human hits ( for 12 , 171 cases ) . This genome annotation was used as reference for our pipeline for alternative splicing ( AS ) identification , which includes multiple steps to identify non-annotated exons and splice sites ( see below ) . Libraries for RNA sequencing generated in the present study were produced with poly-A selected RNA and according to the manufacturer’s directions . All sequences were generated using Illumina HiSeq2000 machines in high yield mode . Read lengths , number of reads and percentage of reads mapped to the genome , as well accession number and sources for public RNA-Seq data are provided in Figure 1—source data 1 . Identification and quantification of major types of AS events , including exon skipping ( comprising events with single or multiple cassette exons and microexons between 3–27 nucleotides ) , intron retention ( IR ) , and alternative 3’ and 5’ splice sites ( Alt3 and Alt5 ) were performed using vast-tools , a recently described multi-module pipeline that has been applied to human , mouse and chicken ( Gueroussov et al . , 2015; Irimia et al . , 2014 ) . Here , we provide a summary of its main features and the specific modifications related to the planarian annotation . For exon skipping ( AltEx ) , we used three complementary approaches , as described in ( Irimia et al . , 2014 ) . First , a 'transcript-based' approach uses full transcript information from multiple sources to identify single or groups of multiple cassette exons ( consecutive alternative internal exons flanked by constitutive exons ) . Given the limited availability of Expressed Sequence Tags ( ESTs ) and cDNAs for planarians , we used other published de novo transcriptome assemblies as ESTs ( Adamidi et al . , 2011; Blythe et al . , 2010; Labbé et al . , 2012; Onal et al . , 2012; Resch et al . , 2012; Rouhana et al . , 2012 ) . In addition , we also used RNA-Seq-based transcript annotations , derived from our multiple samples ( Figure 1—source data 1 ) individually using STAR ( Dobin et al . , 2013 ) and cufflinks ( Trapnell et al . , 2010 ) . Novel junctions obtained in the 'splice site-based' module ( see below ) were also incorporated . Finally , 215 additional cassette exons identified from previous transcriptomic assemblies ( Labbé et al . , 2012; Onal et al . , 2012 ) were incorporated manually ( indicated as 'CASSETTEd' in the Sme . EXSK . Template . * . txt files from vast-tools , see below ) . With this information , we identified individual or groups of neighbouring internal exons that are skipped in certain transcripts . Next , to quantify their inclusion levels , we mapped RNA-Seq reads to combinations of exon-exon junctions ( EEJs ) that define each splicing event . For the case of single exon skipping events , we generated EEJs for C1A , AC2 and C1C2 ( A represents the alternative exon and C1 and C2 represent the neighboring constitutive exons ) , requiring a minimum of eight positions from each exon . For multi-exon events we generated all possible forward combinations from C1 to C2 exons . If multiple alternative 5’ and/or 3’ splice sites were associated with any alternative , C1 or C2 exons , they were also included in the combinations . Second , a 'splice site-based' module utilizes the joining of all hypothetically-possible EEJ forward combinations from annotated and de novo splice sites ( as described in ( Han et al . , 2013 ) , where it was used to identify Embryonic Stem Cell [ESC]-differential exons ) . To identify splice sites de novo , for each annotated splice site donor/acceptor , we scanned two downstream/upstream introns for potential splice site acceptors/donors that would constitute a novel EEJ . Next , after subtracting the reads that map to the genome , we mapped our RNA-Seq data ( Figure 1—source data 1 ) to this library of all potential EEJs , and considered 'novel splice sites' those supported by at least five reads mapped to multiple positions of the EEJ . Identification and quantification of cassette exons was done as described in ( Han et al . , 2013 ) . Third , a 'microexon module' includes de novo searching of pairs of donor and acceptor splice sites in intronic sequence to detect novel , very short ( i . e . 3–15 nt ) microexons and subsequent quantification of inclusion levels using exon-microexon-exon junctions ( EEEJ ) ( Irimia et al . , 2014 ) . The outputs from the three AltEx modules were combined to produce a non-redundant list of cassette exons and associated quantifications . For exons that are identified by more than one module , the representative with the highest overall read coverage is kept . In case of equal coverage , priority is given to events derived from the 'transcript-based' module , followed by those from the 'microexon' module . For intron retention ( IR ) , we used the approach described in ( Braunschweig et al . , 2014 ) . Finally , to detect and quantify Alt3 and Alt5 events , we used the output from mapping RNA-Seq reads to the EEJ library generated by the 'splice site-based module' , which provides information on the usage of every hypothetical splice site donor and acceptor , as described in ( Irimia et al . , 2014 ) . In all modules , quantification of alternative sequence inclusion in the transcripts is derived only from junction reads ( either EEJs or EIJs ) . Raw RNA-seq reads were processed and EEJ/EIJ read counts corrected for mappability obtained as previously described ( Irimia et al . , 2014 ) . These counts were used to derive alternative sequence inclusion levels using the ‘Percent Splice In’ metric ( PSI; percent of transcripts from a given gene that include the alternative sequence ) . The different modules to detect and quantify AS have been integrated into vast-tools ( https://github . com/vastgroup/vast-tools; species key 'Sme' ) . Associated VASTDB files can be downloaded at http://vastdb . crg . eu/libs/vastdb . sme . 31 . 1 . 15 . tar . gz . Finally , steady state mRNA levels for each gene were quantified using the metric 'corrected ( for mappability ) Reads Per Kilobase pair per Million mapped reads' ( cRPKMs , [Labbé et al . , 2012] ) using the original de novo transcriptome employed for genome annotation ( Dresden transcriptome -http://planmine . mpi-cbg . de/ ) , and it is also implemented in vast-tools ( option --expr in vast-tools align ) . For all types of events , we used the same definition for a given sequence to be considered alternatively spliced: 10 ≤ PSI/PSU/PIR ≤ 90 in at least 10% of the samples with sufficient read coverage and/or a range of PSI/PSU/PIRs ≥ 25 across all samples with sufficient read coverage . A given event was considered to have sufficient read coverage in a particular RNA-Seq sample according to the following criteria: - For AltEx ( except for those quantified using the microexon pipeline ) : ( i ) ≥10 actual reads ( i . e . before mappability correction ) mapping to the sum of exclusion EEJs , OR ( ii ) ≥10 actual reads mapping to one of the two inclusion EEJs , and ≥5 to the other inclusion EEJ . - For microexons: ( i ) ≥10 actual reads mapping to the sum of exclusion EEJs , OR ( ii ) ≥10 actual reads mapping to the sum of inclusion EEEJs . - For IR: ( i ) ≥10 actual reads mapping to the sum of skipping EEJs , OR ( ii ) ≥10 actual reads mapping to one of the two inclusion EIJs , and ≥5 to the other inclusion EIJ . - For Alt3 and Alt5: ≥10 actual reads mapping to the sum of all EEJs involved in the specific event . The total number of AS events by type identified using all the RNA-Seq data described above are depicted in Figure 1—figure supplement 2 . To identify AS events that were differentially regulated between X1 and Xins cell fractions , we first required that any given AS event has sufficient read coverage ( see above ) in a minimum of 3 out of 6 X1 and 3 out of 8 differentiated samples ( either Xins or neoblast-depleted whole worms; referred below as 'diff' ) . We implemented a set of non-mutually exclusive definitions to maximize the detection of different patterns of neoblast AS regulation , including both quantitative ( i . e . with large PSI differences between the samples from the X1 and diff groups ) and qualitative ( i . e . in which one of the isoforms is present only in the set of X1 or differentiated cell samples , even if with relatively small PSI differences between both groups ) . The following definitions were employed: where Total_samplesX1/diff is the number of X1/differentiated samples with enough read coverage; MeanX1/diff is the mean PSI/PSU/PIR for all X1/differentiated samples; Mean80diff are the mean PSI/PSU/PIR for the differentiated samples excluding the 20% of samples with the most distant PSI/PSU/PIRs from the X1 mean value ( i . e . ninth to 10th deciles ) ; MinX1/diff is the minimum PSI/PSU/PIR value for X1/differentiated samples; MaxX1/diff is the maximum PSI/PSU/PIR value for X1/differentiated samples; RangeX1/diff is the difference between MaxX1/diff and MinX1/diff; and SDX1/diff is the standard deviation of PSI/PSU/PIR values for X1/differentiated samples . In addition , we required bona fide neoblast-differential AS events to have a p<0 . 05 when comparing X1 and differentiated samples using the B-statistic , i . e . the empirical Bayes log-odds of differential PSI/PSU/PIRs ( Smyth , 2004 ) ( as implemented in 'ebayes' , from the limma package in R ) . AS events not differentially regulated between X1 and differentiated cells ( 'AS_nonX1' events ) were defined as those AS events ( as defined in the section 'Alternative splicing definition and minimum read coverage' ) that had ( i ) enough read coverage in at least 3 X1 samples and 3 differentiated samples , ( ii ) | ( MeanX1 - Meandiff ) | < 10 , and ( iii ) p ≥ 0 . 05 on the Bayesian test . GO terms were mapped and extracted from an InterProScan ( Zdobnov and Apweiler , 2001 ) annotation . Enrichment was performed using the library GOstats ( Falcon and Gentleman , 2007 ) ( testDirection = 'over' , conditional = TRUE ) . The background was defined as all expressed transcripts ( log2 ( TPM+1 ) >1 . 5 ) in all samples any given experiment . For time course experiments the test set was defined as the union of downregulated transcripts in days 20 , 25 and 30 ( threshold log2 FC <−0 . 7 ) . KEGG annotation was performed using the KEGG Automatic Annotation Server ( Moriya et al . , 2007 ) ( KAAS ) with single best hit and a custom set of species ( Homo sapiens ( hsa ) , Mus musculus ( mmu ) , Gallus gallus ( gga ) , Xenopus laevis ( xla ) , Danio rerio ( dre ) , Strongylocentrotus purpuratus ( spu ) , Drosophila melanogaster ( dme ) , Caenorhabditis elegans ( cel ) , Helobdella robusta ( hro ) , Lottia gigantea ( lgi ) , Schistosoma mansoni ( smm ) , Nematostella vectensis ( nve ) , Amphimedon queenslandica ( aqu ) , Saccharomyces cerevisiae ( sce ) ) . KEGG enrichment was assessed using the same subsets as in the cluster enrichment analyses and using the GOstats library ( testDirection = 'over' , conditional = TRUE ) . We annotated RNA binding proteins ( RBPs ) in our reference transcriptome . RBPs were identified using the InterProScan annotation by filtering PFAM ( Finn et al . , 2014 ) ids associated with known RNA binding domains [RRM ( PF05172 , PF08675 , PF00076 , PF04059 , PF08777 , PF13893 , PF14259 , PF03467 , PF03468 , PF10598 , SM00360 , PS50102 ) ; KH ( PF00013 , PF07650 , SM00322 ) ; helicases ( PF00270 ) and other domains ( PF12171 , PF05741 , PF12251 , PF14709 , PF00035 , PF02037 , PF00806 , PF01423 , PF02171 , PF12701 , PF14438 ) ] and top BLAST ( Camacho et al . , 2009 ) hits against complete proteomes ( Uniprot [UniProt 2015] ) of several species ( Human , mouse , C . elegans , fruitfly , S . mansoni , Dugesia japonica , zebrafish , chicken and xenopus ) filtered for relevant key terms ( RNA-binding , RNA , binding ) . From this list , a shortlist of candidate AS factors known to regulate AS in other organisms were selected ( Figure 4—source data 1 ) and those with the highest evidence were selected for RNAi studies ( Figure 4—source data 1 , colored ) . We grouped those in clusters of orthology using KEGG annotations . When the X1/Xins ratio of different orthologs of the same group differed greatly ( indicating differential expression of the orthologs ) , the group was split into two groups to separate the orthologues highly specifically expressed in X1 from the highly specifically expressed in Xins ( Figure 4—source data 3 ) . Since in a previous version of the transcriptome ( BIMSB [Adamidi et al . , 2011] ) the mbnl-1 locus had two different transcripts IDs ( isotig19687 and isotig20952 ) , one dsRNA was designed against both of them ( noted as mbnl-1 and mbnl-1* , Supplementary file 1 ) . Comparison to spliceosomal components was done to gain insight of housekeeping gene expression in stem cells vs . differentiated cells . We selected core spliceosomal components from Complexes A , B and C by searching the relevant KEGG terms in our KEGG annotation . When two or more paralogs with the same KEGG term were found those were removed since their expression patterns were frequently more divergent , indicating possible subfunctionalization of paralog genes . The final list of transcripts used and their expression values is in Figure 4—source data 3 . RNAi was done as described below . Regeneration tests were performed 22 days after RNAi treatment . RNAs from FACS samples and whole worm samples for sequencing were extracted 10 days after RNAi treatment as described below . Since only one replicate was performed or this experiment , we used a more stringent minimum read coverage: - For AltEx ( except for those quantified using the microexon pipeline ) : ( i ) ≥15 actual reads ( i . e . before mappability correction ) mapping to the sum of exclusion EEJs , OR ( ii ) ≥15 actual reads mapping to one of the two inclusion EEJs , and ≥10 to the other inclusion EEJ . - For microexons: ( i ) ≥15 actual reads mapping to the sum of exclusion EEJs , OR ( ii ) ≥15 actual reads mapping to the sum of inclusion EEEJs . For each AS factor group , we quantified the percent of X1-differential alternative exons with sufficient read coverage across the 8 RNAi treated samples ( n = 173 ) that showed a ∆PSI ≥15 between the treated and the control samples towards the PSI in X1 . Alternative sequences were first mapped to coding ( CDS ) or non-coding ( UTRs ) sequences based on our TransDecoder-based genome annotations . For those alternative sequences that are not present in our reference annotation , mapping was projected based on the information of the upstream exon . Then , sequences that contained in-frame stop codons ( based on annotations and in-frame sequence translation from the upstream exon ) or start codons ( based on annotations ) were flagged . Following a recent study ( Irimia et al . , 2014 ) , AS events were predicted to generate alternative protein ( ORF-preserving ) isoforms upon both inclusion and exclusion when they fall in the CDS , do not disrupt the reading frame ( i . e . they have lengths multiple of three nucleotides ) , and do not contain in-frame stop codons predicted to trigger non-sense mediated decay ( NMD; those stop codons that fall further than 50 nucleotides away from a downstream EEJ ) . Furthermore , if an alternative sequence is predicted to introduce an in-frame stop codon that would not trigger NMD but would generate a protein >100 aminoacids shorter than the annotated isoform , it was also considered as an ORF-disrupting event . For multiexonic cassette events ( arrays of more than one alternative exon ) , NMD/ORF-disruption was assessed for all exons as a group based on their inclusion patterns in neoblast and differentiated samples ( e . g . if one frame-shifting exon is downregulated in neoblast samples and another upregulated , both the neoblast and differentiated isoforms can have intact reading frames ) . For ORF-disrupting AS events , two main categories were defined: ORF-disruption upon sequence inclusion or upon sequence exclusion . Therefore , depending on the neoblast inclusion pattern , we defined events that cause ORF disruption in neoblast samples , but not in other tissues ( 'ORF disruption in X1' ) , or the opposite , AS events that disrupt ORFs only in differentiated samples ( 'ORF disruption in Xins' ) . To investigate the overlap of alternative exons with disordered regions and protein domains we first mapped the alternative exons as well as the upstream ( C1 ) and downstream ( C2 ) exons to the proteome as annotated in our genome annotation ( see above ) . Next , to include unannotated exons , we recreated novel protein isoforms by introducing the exonic sequence downstream of the upstream C1 exon . The final set of proteins was used to run de novo predictions using Disopred2 ( for structural disorder , [Ward et al . , 2004] ) , and Pfam ( for protein domains , ( Finn et al . , 2014 ) ; only domains from the A module were used ) , both with default parameters . For both disorder and Pfam domains , the fraction of residues from alternative exons overlapping those features are reported . For consistency , for all exon classes , only exons that would generate protein isoforms both when included and skipped ( i . e . internal exons with lengths multiple of three nucleotides without in-frame stop codons ) were analyzed . With the aim of designing D . japonica-specific primers for all validated neoblast-differential AS events ( Figure 1—figure supplement 4 ) , we first blasted the S . mediterranea alternative sequences , as well as the neighbouring upstream and downstream exons against a recently published transcriptome assembly of D . japonica . If at least two of these sequences mapped against the same contig as the best hit , this was designated as the best ortholog and primers were designed in equivalent D . japonica sequences . Next , for those cases that we could not match using this strategy , we blasted the translated S . mediterranea full protein sequence again the D . japonica transcriptome using tblastn . The best D . japonica hit was considered as the potential ortholog , and further manual annotation through exon-specific alignments was performed to identify the region homologous to the S . mediterranea AS event . All primer D . japonica sequences are provided in Supplementary file 1 . We used two approaches to assign orthology between planarian and human genes with neoblast-differential and ESC-differential , respectively . First , Inparanoid v8 . 0 ( Ostlund et al . , 2010 ) was run for one reference protein per gene in planarian and human using default parameters . If the planarian and human proteins were grouped in the same cluster , they were considered putative orthologs . Second , we blasted the planarian proteins against the human reference proteome and the first three hits were matched against the subset of genes with ESC-differential exons . From these two complementary approaches , orthology calls were manually verified and dubious mis-assignments were discarded . In total 15 orthologous groups comprising 17 planarian genes ( with neoblast-differential 30 exons ) and 20 human genes ( harbouring 22 ESC-differential exons ) . Next , to assess orthology at the exonic level , we mapped intron positions into protein sequence alignments , as previously described ( D'Aniello et al . , 2008 ) ; in addition , conserved protein domains were used as milestones to define homologous protein regions ( Figure 3—source data 2 ) . With this information , we evaluated whether ( i ) the exons fall in equivalent protein regions ( e . g . within two protein domains ) , ( ii ) the sequence was orthologous , and ( iii ) the exons were orthologous ( based on the intron position alignments ) . The exons were defined to be in 'similar protein regions' ( Figure 3—source data 1 ) if they fall in the same relative region of the protein ( as per ( i ) ) and the distance in the alignment is lower than 100 residues . The Full-length version of bruli , mbnl-1 ( both the X1- and Xins-enriched isoforms ) , mbnl-like-1 and mbnl-like-2 were cloned into GST-tagged vector ( pTH6838 ) using primers listed in Supplementary file 1 . GST-tagged proteins were expressed and purified from Escherichia coli as previously described ( Ray et al . , 2009 , 2013 ) . The RNA pool generation , RNAcompete pull-down assays , and microarray hybridizations were performed as previously described ( Ray et al . , 2009 , 2013 ) . Briefly , GST-tagged RBPs ( 20 pmoles ) and RNA pool ( 1 . 5 nmoles ) were incubated in 1 mL of Binding Buffer ( 20 mM HEPES pH 7 . 8 , 80 mM KCl , 20 mM NaCl , 10% glycerol , 2 mM DTT , 0 . 1 mg/mL BSA ) containing 20 mL glutathione sepharose 4B ( GE Healthcare ) beads ( pre-washed 3 times in Binding Buffer ) for 30 min at 4°C , and subsequently washed four times for two minutes with Binding Buffer at 4°C . One-sided Z-scores were calculated for the motifs as described previously ( Ray et al . , 2013 ) . For the alternatively spliced exons from each of the test sets ( I-IV in Figure 5 ) and corresponding background sets ( 'AS_nonX1' events [see above] that showed a ΔPSI < 5 between bruli or mbnl KD and control ) , the sequences of the flanking introns , as well as the alternative exons themselves , were fetched from the planarian genome . The sub-sequences corresponding to the first and last 80nt of each intron were further subdivided into 20-nt bins . Introns shorter than 160-nt were excluded from the analysis ( see below ) . Next , all 7-mer occurrences within each bin were counted and summed for corresponding bins over all events in the signal and background sets . We then selected the top 10% 7-mers with the highest affinity for an RNA-binding protein , as measured by RNAcompete ( z-score ) . For each bin , we computed the sum-total occurrences of the high affinity 7-mers in the signal and control events . If enrichment was observed , a P-value was calculated using Fisher's exact test . If significance was reached ( p<0 . 01 ) the bin is represented as a filled rectangle , with the saturation of the color proportional to −log10 ( P-value ) . Non-significant bins are represented as empty rectangles . Similar results were obtained when 80 nucleotides upstream and downstream the exon were assayed , without discarding shorter introns . To calculate enrichment of general RBPs in sequences associated with X1-included and X1-excluded exons ( Figure 4—figure supplement 1 ) , we used the entire CISBP-RNA database ( Ray et al . , 2013 ) and performed a similar analysis for each of the available RBP records , using all 'AS_nonX1' events as background . We sorted the RBPs by the strongest significance in any of the bins and employed a cutoff of p<10−4 to restrict the output to the most significant hits . All animals belonged to the Berlin-1 strain of asexual type Schmidtea mediterranea , recently generated from one single individual . For RNAi experiments , dsRNAs were synthesized as previously described ( Solana , 2013 ) . Animals were injected with dsRNA against the coding region of the gene of interest ( control ( RNAi ) planarians were injected with dsRNA coding for GFP ) for three consecutive days ( days one , two and three after RNAi ) and kept at 20°C , as previously described . dsRNAs were delivered at a concentration of 1 µg/µl . When multiple dsRNAs were used simultaneously , each dsRNA was injected at a concentration of 1 µg/µl in the same solution . Since in a previous version of the transcriptome ( BIMSB ) the mbnl-1 locus had two different transcripts IDs ( isotig19687 and isotig20952 ) , one dsRNA was designed against both of them ( Supplementary file 1 ) . When appropriate , the concentration of negative control dsRNA coding for gfp was adjusted in the control samples to the maximum concentration of dsRNA injected in the experimental groups . FACS experiments were performed as previously described ( Hayashi et al . , 2006; Onal et al . , 2012 ) . Essentially , planarians were cut into little pieces on ice and in the presence of trypsin to help cell dissociation . Cells were then sequentially filtered through 40 µm and 20 µm filters and stained with the cytoplasmic dye Calcein-AM ( BD Biosciences , at a final concentration of 0 . 5 µg/ml ) and the nuclear dye Hoechst 33 , 342 ( Fluka Biochemika , at a final concentration of 20 µg/ml ) . Propidium Iodide was used to discard dead cells . Cells were then sorted with a BD FACSAria III directly into Trizol LS containing tubes . X1 gating is achieved by sorting the population of cells with double content of DNA . From the single content , those cells with low calcein staining were gated as X2 and those with high calcein staining were gated as Xins . Control extractions with irradiated planarians were used to assist correct gating . RNA was extracted using Trizol LS ( Ambion ) following manufacturer instructions . For Dugesia japonica experiments , minor adjustments in the nuclear staining and the FACS gating were applied since this species has a larger DNA content ( Nishimura et al . , 2015 ) . For RT-PCR and qPCR analyses , RNA was extracted with self-made Trizol reagent ( modified from [Chomczynski and Sacchi , 1987] ) reverse transcribed with an oligodT primer using Maxima H Minus Reverse Transcriptase ( Thermo Scientific , Waltham , MA ) . RT-PCRs were visualized in 2 . 5–3 . 3% agarose gels . Relative ratios between the two isoforms were then calculated based on the relative intensity of the PCR bands , measured using Image J . qPCR experiments were technically replicated twice , and performed with 2 biological replicates of each condition . Each sample was always loaded in triplicates . In situ hybridization was performed as previously described ( King and Newmark , 2013 ) using an Intavis Vsi Pro robot . Probes were synthesized from PCR amplicons as previously described ( Solana , 2013 ) . All primers used in each experiment and provided in Supplementary file 1 . For regeneration tests , 10 planarians per time point and condition were selected , cut at various time points , and observed and scored daily under the scope . At each scoring time point , identification of normally looking eyespots was used as proxy for complete regeneration . Trunk ( amputated head and tail ) and trunk/tail ( only head was amputated ) pieces were observed for 10 days , tail pieces were observed for 12 days as their regeneration time is slower . All control trunk/tail , trunk and tail pieces used in this study had visible eyespots by day 6 , 7 and 10 respectively . RNA-Seq reads were processed and filtered for low quality and 3' and 5' adapter removal using Flexbar v2 . 5 ( Dodt et al . , 2012 ) . in-silico ribosomal depletion was performed with bowtie2 ( Langmead and Salzberg , 2012 ) in local mode ( bowtie2 –local –very-sensitive-local -x rRNA_INDEX -p 8 -U FASTQ --un FILTERED . fastq ) against a pool of platyhelminthes rRNA index . Filtered reads were then mapped to the reference planarian transcriptome using bowtie2 default parameters . Transcript quantification was performed using htseq-count ( Anders et al . , 2015 ) . To assess differential expression we resorted to previously described clustering of genes by gene expression profiles in FACS sorted populations ( Onal et al . , 2012 ) . Transcripts from the original source were first mapped to our reference transcriptome using BLAT ( Kent , 2002 ) and pslReps with the options -minCover = 0 . 5 and -minAli = 0 . 1 -nearTop = 0 . 005 filtering parameters . The clusters were assigned by matching 1-to-many . In case of multiplicity , the following rules were applied: if a transcript was matched to cluster 1 and 2 or 5 and 6 , cluster 2 and cluster 6 were assigned , respectively . Other combinations were considered ambiguous and discarded and the cluster left blank . Next , the control samples were normalized using quantile normalization with the function normalize . quantiles ( R-base library - http://www . R-project . org ) to reduce biases by rRNA contamination . The dataset was then filtered for reliably detected transcripts using a threshold of log2 ( TPM+1 ) >1 . 5 laying a total of 12 , 599 transcripts . Two conditional sets ( bruli KD downregulated , mbnl KD downregulated ) were defined as transcripts below a threshold of −0 . 7 log2 FC over the corresponding control sample . Differential enrichment of downregulated genes from specific clusters compared to the global transcriptome was assessed using hypergeometric test , using all expressed transcripts ( 12 , 599 ) as background and performing multiple testing Bonferroni correction on the p-value . | Stem cells are specialized cells found in all animals that can develop into several different types of mature cells . Stem cells are therefore well suited for maintaining organs that are in heavy use , such as the intestine , and for regenerating tissues that are prone to injury , like the skin . One reason why stem cells differ from mature cell types is because they activate , or “express” , different sets of genes . In addition , many genes can be expressed as one of several versions . These variants , also known as isoforms , are generated by a process called alternative splicing . In mature cells in mammals , a group of proteins called the MBNL proteins help to prevent the expression of gene isoforms that are characteristic to stem cells . The adult flatworm Schmidtea mediterranea contains stem cells that can regenerate any part of the body . Solana , Irimia et al . have now investigated whether alternative splicing is important for controlling how the worm’s stem cells behave . After establishing which gene isoforms are expressed in the stem cells and the mature cells , the levels of different sets of proteins that control alternative splicing were experimentally reduced . The results indicate that just as seen in mammals , the MBNL proteins reduce the expression of stem cell-related gene isoforms in the flatworms . Furthermore , Solana , Irimia et al . found that another protein called CELF counteracts MBNL proteins by helping to express gene isoforms that are active in stem cells . The interplay between the MBNL and CELF proteins has also been observed in human cells . Thus , it appears that this way of controlling alternative splicing is common to flatworms and mammals and is therefore evolutionarily ancient . This suggests that other similar ways of controlling stem cells by interactions between regulatory proteins might be working in all animal stem cells . Further studies are now needed to investigate these control proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2016 | Conserved functional antagonism of CELF and MBNL proteins controls stem cell-specific alternative splicing in planarians |
Asia is considered an important source of influenza A virus ( IAV ) pandemics , owing to large , diverse viral reservoirs in poultry and swine . However , the zoonotic origins of the 2009 A/H1N1 influenza pandemic virus ( pdmH1N1 ) remain unclear , due to conflicting evidence from swine and humans . There is strong evidence that the first human outbreak of pdmH1N1 occurred in Mexico in early 2009 . However , no related swine viruses have been detected in Mexico or any part of the Americas , and to date the most closely related ancestor viruses were identified in Asian swine . Here , we use 58 new whole-genome sequences from IAVs collected in Mexican swine to establish that the swine virus responsible for the 2009 pandemic evolved in central Mexico . This finding highlights how the 2009 pandemic arose from a region not considered a pandemic risk , owing to an expansion of IAV diversity in swine resulting from long-distance live swine trade .
Our ability to predict outbreaks of zoonotic pathogens requires an understanding of their ecology and evolution in reservoir hosts . At the onset of the 2009 influenza pandemic , whole-genome sequencing revealed that pdmH1N1 was a novel reassortant virus comprised of segments from three major swIAV lineages ( Garten et al . , 2009 ) : segments PB2 , PB1 , PA , NP and NS were derived from a triple reassortant H3N2 swine virus ( TRsw ) that originated in North American swine during the mid-1990s; the HA ( H1 ) segment derived from the classical swine H1N1 lineage ( Csw ) that has circulated in North American swine since the 1918 pandemic; and the NA ( N1 ) and MP segments were related to the avian-like Eurasian swine lineage ( EAsw ) that emerged in European pigs in the late 1970s . Genetic and epidemiological evidence indicate that the first outbreak of pdmH1N1 occurred in humans in Mexico ( Brockmann and Helbing , 2013; Chowell et al . , 2011; Lemey et al . , 2009 ) . However , it remained unclear whether Mexico also was the site of the zoonotic transmission event that gave rise to the 2009 outbreak , given that EAsw viruses had never been detected in swine in any part of the Americas , including Argentina , Brazil , Chile , Mexico , Peru , Canada , and the intensively sampled swine herds in the United States ( Anderson et al . , 2013; Nelson et al . , 2015b; Pereda et al . , 2011; Tinoco et al . , 2015 ) . At the time , Asia was the only region where TRsw , Csw , and EAsw viruses were known to co-circulate in swine ( Zhu et al . , 2013 ) . Asia is a global source of novel human seasonal influenza viruses ( Lemey et al . , 2014; Russell et al . , 2008 ) , avian viruses associated with the pandemics of 1957 and 1968 ( Kawaoka et al . , 1989 ) , and emerging avian viruses with pandemic potential ( Lam et al . , 2015; Taubenberger Morens , 2009 ) . Since 2009 , several studies have implicated Asian swine as a possible source of the TRsw/Csw/EAsw reassortant swine virus that gave rise to pdmH1N1 . Asia is the only region where TRsw , Csw , and EAsw are known to exchange segments through reassortment ( Lam et al . , 2011; Poonsuk et al . , 2013; Takemae et al . , 2008; Vijaykrishna et al . , 2011 ) . At the onset of the pandemic , IAVs identified in swine ( swIAVs ) in Hong Kong , SAR , were more closely related to pdmH1N1 than any other swIAVs available globally ( Smith et al . , 2009 ) . One Hong Kong swIAV was identified with a genotype similar to the pdmH1N1 virus except for the NA segment , and transmitted via respiratory droplet in ferrets ( Yen et al . , 2011 ) . However , the time of the most recent common ancestor ( tMRCA ) of the Hong Kong swIAVs and pdmH1N1 was still approximately 10 years ( Smith et al . , 2009 ) . Combined with the lack of any complete swIAV genomes from Mexico’s large swine populations at the start of the pandemic , and the difficulty of understanding how a virus that evolved in Asian swine caused its first outbreak in humans in Mexico , the geographical location of the swine-to-human transmission event that gave rise to pdmH1N1 has remained uncertain . In the years following the pandemic , new surveillance in Mexican swine has identified Csw , TRsw , and pdmH1N1 and seasonal H3N2 viruses of human origin ( Nelson et al . , 2015a ) , but no EAsw viruses . However , the majority of swIAVs in previous analyses were collected in northern and eastern Mexico , and no studies to date have included viruses from major swine-producing states in central Mexico , including Jalisco , Guanajuato , and Puebla ( Figure 1 ) . 10 . 7554/eLife . 16777 . 003Figure 1 . Live swine production in Mexico . Each Mexican state is shaded according to the density of pigs ( the number of pigs per square kilometer , light green = lower and dark blue = higher ) . The six Mexican states where influenza viruses were collected for this study are indicated , and the number of pigs in the state is provided below the state’s name . Source data from Mexico’s Secretariat of Agriculture , Livestock , Rural Development , Fisheries and Food ( SAGARPA ) is available in Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 003 10 . 7554/eLife . 16777 . 004Figure 1—source data 1 . Swine population sizes in Mexican states in 2014 . Data available from Mexico's Secretariat of Agriculture , Livestock , Rural Development , Fisheries and Food ( SAGARPA ) ( www . siap . gob . mx ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 004 Resolving whether the viral precursors of pdmH1N1 evolved in swine in Mexico or Asia has important implications for understanding pandemic emergence and informing risk assessment . We therefore undertook expansive surveillance efforts in Mexico , isolating the virus from pigs with respiratory symptoms in farms from six Mexican states with high swine production , including Sonora in northern Mexico , Yucatan in eastern Mexico , and previously unsampled states in central-east Mexico ( Puebla ) , and central-west Mexico ( Jalisco , Guanajuato , and Aguascalientes , Figure 1 ) . These efforts yielded 58 complete viral genomes ( sequence accession numbers are available in the Dryad data repository under Doi:10 . 5061/dryad . m550m ) that were used to identify swIAVs that are closely related to pdmH1N1 , resolving the origins of pdmH1N1 in its reservoir host .
Among the 58 swIAVs newly sequenced from Mexico , genome segments were identified from all three major swIAV lineages ( Csw , TRsw , and EAsw ) and all three major human IAV lineages of the 21st century ( H3N2 , pdmH1N1 , and pre-pandemic H1N1 ) , evidence of extensive gene flow between Mexican swine and ( a ) humans , ( b ) US swine ( Csw and TRsw viruses ) , and ( c ) European swine ( EAsw viruses ) ( Figure 2 ) . Except for TRsw ( genotype 5 ) and pdmH1N1 ( genotype 6 ) , no other IAV lineages were identified intact in Mexican swine , evidence of frequent reassortment between viral lineages . Of the 12 genotypes identified , 10 were reassortants . Four genotypes were triple reassortants: genotype 1 contains all three swine lineages ( Csw/TRsw/EAsw ) ; genotype 9 contains all three human-origin lineages ( pre-pandemic H1N1/human H3N2/pdmH1N1 ) ; and genotypes 2 and 11 contain one human-origin and two swine lineages ( pre-pandemic H1N1/TRsw/EAsw and pdmH1N1/Csw/TRsw , respectively ) . Although human-origin viruses were observed in all regions of Mexico , the importance of human-to-swine transmission in the evolution of swIAV diversity in Mexico has been characterized recently in detail ( Nelson et al . , 2015a ) . Instead , this study will focus on evolutionary processes that are directly involved in the emergence of pdmH1N1 precursors in Mexican swine , including reassortment and viral migration driven by long-distance movements of live swine . 10 . 7554/eLife . 16777 . 005Figure 2 . Genetic diversity of IAVs in Mexican swine , 2010–2014 . Twelve genotypes were identified by surveillance in Mexican swien herds during 2010–2014 . Each oval represents one of the eight segments of the viral genome . The surface antigens HA and NA are listed first , followed by the six internal gene segments . The shading of each oval corresonds to the genetic lineage of IAVs found in humans and swine globally . The number of swIAVs with a given genotype is indicated for each region in Mexico : Sonora ( northern Mexico ) , Jalisco/Aguascalientes/Guanajuato ( central-west Mexico ) , Puebla ( central-east Mexico ) , and Yucatan ( eastern Mexico ) . The genotype and additional characteristics of each of the 58 swIAVs collected and sequenced from Mexico for this study are provided in Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 005 10 . 7554/eLife . 16777 . 006Figure 2—source data 1 . Characteristics of the 58 svIAVs collected in Mexico for this study . Viruses are listed in order of date of collection . Mexican state is provided ( see Figure 1 for a map of Mexico ) . The genotype ( Figure 1B ) and lineage and clade Figure 2 ) of each segment for each virus is provided . Internal genes are categorized as triple reassortant internal genes ( trig ) , classical ( NP segment of AVX-30 ) , pdmH1N1 ( pdm ) or Eurasian ( eur ) . HA segments are categorized as classical swine ( H1c ) , human seasonal H1N1 ( H1hu ) , H3 , pdmH1N1 ( pdm ) , or Eurasian ( eur ) . NA segments are categorized as classical swine ( N1c ) , N2 ( 1998 or 2002 lineage ) , pdmH1N1 ( pdm ) , or Eurasian ( eur ) . Within these lineages , clades are differentiated by location: Jalisco ( J ) , Sonora ( S ) , Yucatan ( Y ) , Puebla ( P ) , Guanajuato ( G ) . Singleton viruses are indicated ( sing ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 006 A high degree of population structure was observed among swIAV populations in Mexico’s northern , central , and eastern regions . Viruses collected in Sonora , Yucatan , Puebla , and Jalisco/Guanajuato/Aguascalientes consistently were positioned in different sections of the phylogenetic trees inferred for individual genome segments ( Figure 3 ) , evidence of independent viral introductions into these regions from humans , US swine , and Eurasian swine ( summarized in Figure 4A ) . Viral gene flow was not frequently observed between Mexico’s northern , eastern , and central regions ( Figures 3 , 4A ) , a contrast to the frequency of inter-regional IAV migration observed in US swine herds , driven by ongoing long-distance movement of animals ( Nelson et al . , 2011 ) . The only evidence for a migration event between two Mexican regions that is observed for multiple segments involves a single Yucatan virus ( AVX-57 ) that is closely related to Jalisco viruses , evidence of Jalisco-to-Yucatan migration ( Figure 4A , Figure 3—source data 1 ) . Viral migration occurred frequently between states in Mexico’s central-west region , as evidenced by the clustering of viruses from Jalisco , Guanajuato , and Aguascalientes within the same phylogenetic clades ( Figure 3—source data 1 ) . Jalisco’s large swine herds were found to be an important source of viruses in the neighboring states of Guanajuato and Aguascalientes ( Figure 4A , Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 16777 . 007Figure 3 . Evolutionary relationships between swIAVs collected in Mexico and pdmH1N1 . Time-scaled Bayesian MCC trees inferred for the eight segments of the IAV genome , for the lineages found in pdmH1N1: TRIG ( PB2 , PB1 , PA , NP , and NS ) , classical ( H1 ) , and avian-like Eurasian ( N1 and MP ) . Trees include the 58 swIAVs collected in Mexico for this study , representative pdmH1N1 viruses , and other related swIAVs collected globally . The color of each branch indicates the most probable location state . For clarity , the pdmH1N1 clade is depicted as a triangle , the color of which represents the inferred location state of the node representing the inferred common ancestor of pdmH1N1 and the most closely related swIAVs ( indicated by an open circle ) . Posterior probabilities are provided for key nodes . More detailed phylogenies including tip labels are provided in Figure 3—source data 1 . along with trees inferred for lineages not shown here . Similar phylogenies inferred using maximum likelihood methods are provided in Figure 3—source data 2 . Similar phylogenies that use genotype instead of geographic location as a trait are provided in Figure 3—source data 3 . Nine Mexican swIAVs were excluded from the phylogenetic analysis because they were outliers in root-to-tip divergence ( Figure 3—figure supplement 1 ) . More detailed phylogenies of the pdmH1N1 viruses reveal multiple independent introductions from humans into swine in Mexico ( Figure 3—source data 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 007 10 . 7554/eLife . 16777 . 008Figure 3—source data 1 . MCC trees presenting the evolutionary relationships between swIAVs collected in Mexico and swIAVs and human IAVs collected globally , for each IAV segment as well as for each IAV lineage found in Mexican swine: PB2 ( TRIG/pdmH1N1 ) , PB1 ( TRIG/pdmH1N1 ) , PA ( TRIG/pdmH1N1 ) , H1 ( classical/pdmH1N1 ) , H1 ( avian-like Eurasian ) , H1 ( human seasonal/human- like swine ) , H3 ( human seasonal/human-like swine ) , NP ( TRIG/classical/pdmH1N1 ) , N1 ( classical ) , N1 ( avian-like Eurasian/pdmH1N1 ) , N2 ( human seasonal/human-like swine ) , MP ( TRIG/classical ) , MP ( avian-like Eurasian/pdmH1N1 ) , NS ( TRIG/classical/pdmH1N1 ) . The color of each branch indicates the most probable location state , similar to Figure 3 . Clades of Mexican viruses are labeled . Posterior probabilities >80 are provided for key nodes . The 95% HPD values for the estimated tMRCA also are provided for key nodes with light blue bars . Viruses with genotype 1 similar to pdmH1N1 are indicated with green stars on the PB2 tree; a more detailed presentation of the evolution of genotypes in Mexico in all trees is provided in Figure 3—source data 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 008 10 . 7554/eLife . 16777 . 009Figure 3—source data 2 . Maximum likelihood trees with tip labels . Evolutionary relationships of pdmH1N1 and swIAVs sampled in Mexico and globally for the eight segments of the IAV genome . Phylogenies and color scheme are similar to Figure 3 , except the trees are inferred using maximum likelihood methods and all horizontal branch lengths are drawn to scale ( nucleotide substitutions per site ) . Trees are midpoint rooted for clarity and bootstrap values >70 are provided for key nodes . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 009 10 . 7554/eLife . 16777 . 010Figure 3—source data 3 . Time spent in a genotype using Markov rewards . Similar phylogenies as presented in Figure 3—source data 1 , but in this case Mexican swine viruses are specified by genotype instead of geography . Non-Mexican viruses remain specified by geographical location . Markov rewards representing the time spent in a genotype between state transitions are provided in the upper right . Genotype 1 , which is similar to pdmH1N1 ( Figure 2 ) , is shaded gold . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 010 10 . 7554/eLife . 16777 . 011Figure 3—source data 4 . Detailed phylogenetic analysis of pdmH1N1 in Mexican swine . Evolutionary relationships of pdmH1N1 viruses collected from humans and swine , 2009–2014 for the representative PB2 and HA segments . Separate viral introductions from humans into Mexican swine are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 011 10 . 7554/eLife . 16777 . 012Figure 3—figure supplement 1 . Mexican swIAVs excluded from the phylogenetic analysis . A residual analysis was used to identify outliers from the linear regression line describing the relationship between the date of collection and genetic distances of sequences in our analysis . Nine swIAVs from Mexico were identified as outliers and removed from our analysis . These viruses were collected in 2010 but are closely related to reference triple reassortant viruses from 1998–1999 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 01210 . 7554/eLife . 16777 . 013Figure 4 . Heat-maps of IAV gene flow between locations . 'Markov jump' counts measure the number of inferred transitions , modeled by a continuous-time Markov chain process , that occur along the branches of the phylogeny , providing a measure of gene flow . The intensity of the color ( red = high; white = low ) reflects the number of Markov jump counts from a swine population in a location of origin ( y-axis ) to swine in one of six Mexican states ( destination , x-axis; asymmetrical , summarized across all lineages and segments ) , and between human populations in Mexico , the United States , and globally during the early spatial dissemination of pdmH1N1 in humans during March–May 2009 . Asterisks indicate the geographical source ( y-axis ) of the highest number of Markov jump counts for a particular destination ( x-axis ) . Similar spatial linkages were observed using a Bayes factor ( BF ) test ( Figure 4—figure supplement 1 ) . A phylogenetic tree depicting the evolutionary relationships between human pdmH1N1 viruses is provided in Figure 4—figure supplement 2 . Source data for both heat-maps is provided in Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 013 10 . 7554/eLife . 16777 . 014Figure 4—source data 1 . Expected number of location state transitions ( 'Markov jump' counts ) along the branches of inferred phylogenies , summarized for ( a ) all segments and lineages identified in Mexican swine and ( b ) human pdmH1N1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 014 10 . 7554/eLife . 16777 . 015Figure 4—figure supplement 1 . Supported rates of viral migration . Bayes factor ( BF ) test for significant non-zero rates of swIAV migration in Mexico . Rates supported by a BF greater than 5 are indicated . The color of the line represents the relative strength of support for a rate; red lines indicate strong support and pale yellow lines indicate weaker support . For the 5 TRIG segments ( PB2 , PB1 , PA , NP , and NS ) , the average BF is provided . The direction of viral flow is indicated by the line’s gradient , with the white end indicating the source location . The dotted lines indicate that rates between Asia and Mexico likely arise from gaps in sampling in Eurasian swine prior to 1995 , and are unlikely to represent actual viral movements from Asia to Mexico . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 01510 . 7554/eLife . 16777 . 016Figure 4—figure supplement 2 . Phylogeography of pdmH1N1 in humans . Evolutionary relationships between pdmH1N1 viruses collected in humans globally during March 1 , 2009–May 31 , 2009 , inferred for 422 concatenated genome sequences . The color of each branch indicates the most probable location state: red = Asia; orange = Canada; yellow = Europe; light green = Mexico , for which the state of location was not known; green = Mexico City , Mexico; greenish blue = San Luis Potosi , Mexico; light blue = Veracruz , Mexico; blue = South America; bluish purple = Northeastern United States , including New York; purple = Midwestern United States , including Wisconsin; pink = Southern United States , including Texas; and coral = Western United States , including California . Select spatial clusters are labeled . Posterior probabilities >80 are provided for key nodes . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 016 Time-scaled phylogenies suggest that viral introductions from US and European swine into Mexico have occurred periodically for at least two decades , coinciding with a sharp increase in the number of reported imports of live swine into Mexico in the late 1980s ( Figure 5 ) . Notably , EAsw viruses have been introduced from Eurasian swine into central Mexico on at least two occasions . Although the US and Canada are Mexico’s primary trade partners for the import of live swine , live swine imports into Mexico also were reported from several European countries during the late 1990s , including the United Kingdom and Denmark ( Figure 6 ) . We estimate that at least three TRsw viruses and two Csw viruses were introduced from US swine into Mexican swine herds in the northern and central regions . Estimates of the number and timing of independent swIAV introductions into Mexico are complicated by gaps in global background swIAV sequence data in the 1990s/early 2000s and topological differences in the phylogenies inferred for different segments arising from reassortment . Our conservative approach of defining introductions based on both high posterior probabilities ( >90 ) on the phylogeny ( Figure 3—source data 1 ) and Bayes factor ( BF ) support for significant rates of migration ( BF > 6 ) ( Figure 4—figure supplement 1 ) has likely underestimated the number of introductions . We also recognize the need to draw phylogeographic inferences with an awareness of missing data . Major gaps in surveillance in swine in Europe and Asia during the 1990s have resulted in long branch lengths between swIAVs collected in Mexico and Eurasia ( Figure 3—source data 1 ) . Although it is difficult to infer from the tree whether the EAsw MP identified in Mexico was introduced from Asian or European swine ( Figure 4—figure supplement 1 ) , the lack of live swine imports reported between Mexico and any Asian country ( Figure 6 ) , and the low export of Asian swine globally ( Nelson et al . , 2015c ) suggest that Europe may be a more likely source of the Mexican viruses , similar to the H1 and N1 segments . 10 . 7554/eLife . 16777 . 017Figure 5 . Import of live swine and IAVs into Mexico . The value of live swine imported into Mexico ( USD , y-axis , left ) from all countries during 1969–2010 is presented in the background in pink , based on trade data reported to the Food and Agricultural Organization ( FAO ) of the United Nationals , available in Figure 5—source data 1 . Each horizontal line represents an introduction of an IAV segment into Mexican swine , the timing of which is inferred from the MCC trees . The shading of each line indicates the inferred location of origin of an introduction , consistent with Figure 3: dark pink = USA/Canadian swine , black = Eurasian swine , grey = humans . The length of the line indicates uncertainty in the timing of an introduction . Triangles represent clades resulting from onward transmission in Mexico and extend forward as far as the most recently sampled virus . The shading of each triangle indicates the destination location , consistent with Figure 3: dark purple = Yucatan , dark blue = Sonora , light blue = Puebla , and green = Jalisco , Aguascalientes , and Guanajuato ( central-west Mexico ) . Lines without triangles represent singletons . A similar figure annotated with the segment associated with each introduction is provided in Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 017 10 . 7554/eLife . 16777 . 018Figure 5—source data 1 . Reported value of live swine imports into Mexico from all countries during 1969–2010 . Data available from the Food and Agriculture Organization ( FAO ) of the United Nations Datasets repository , http://data . fao . org/datasets . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 018 10 . 7554/eLife . 16777 . 019Figure 5—figure supplement 1 . Similar to Figure 5 , but annotated with the segment associated with each introduction . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 01910 . 7554/eLife . 16777 . 020Figure 6 . Sources of live swine imports into Mexico . Total imports of live swine into Mexico ( USD ) during 1996–2012 from nine reported trade partners: United States , Canada , Ireland , United Kingdom , Denmark , Spain , Guatemala , Guyana , and Peru . The shade of the line indicates the volume of imports ( red = high , purple = low ) . The shading countries is for purposes of clarity only . Trade data is available from the UN Commodity Statistics Database ( Figure 6—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 020 10 . 7554/eLife . 16777 . 021Figure 6—source data 1 . Pairwise information on imports of live swine from specific countries is available from 1996–2012 . Data available from the United Nations' Commodity Trade Statistics Database , http://comtrade . un . org . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 021 An important consequence of the long-distance dissemination of EAsw , Csw , and TRsw viruses into central Mexico was the identification of 18 swIAVs in central-west Mexico with the same combination of EAsw , Csw , and TRsw segments as pdmH1N1 ( referred to as genotype 1 , Figure 2 ) . Genotype 1 viruses are distinct from the reverse zoonosis introductions of pdmH1N1 from humans to swine ( genotype 6 ) that have been observed globally ( Nelson and Vincent , 2015 ) . Phylogenetically , genotype 1 viruses from Mexico form a sister lineage to pdmH1N1 on the PB2 , PB1 , PA , HA , NP , and NA trees ( Figure 3 , Figure 3—source data 2 ) . Genotype 2 viruses from Jalisco ( genotype 2 viruses are similar to genotype 1 , but contain a pre-pandemic human-origin H1 segment [Figure 2] ) are closely related to pdmH1N1 on the MP tree . No Mexican swIAVs were closely related to pdmH1N1 on the NS tree . However , NS is the segment with the highest tMRCA between pdmH1N1 and the most closely related swIAVs ( Figure 7 ) , an indication of larger gaps in sampling and missing data . It is therefore likely that the NS precursor in Mexico was not detected by surveillance or has been replaced in intervening years by reassortment . The tMRCA analysis therefore provides further indication that Mexico is the likely origin of the swine virus that gave rise to pdmH1N1 . However , it is difficult to determine the precise location within central-west Mexico of zoonotic transmission , owing to high rates of gene flow and reassortment between swIAVs in Jalisco and Guanajuato and resulting differences in tree topologies ( Figures 3 , 4A ) . Swine viruses from Guanajuato are most closely related to pdmH1N1 on the phylogenies inferred for the PB2 , PB1 , H1 , and N1 segments . Jalisco viruses are most closely related to pdmH1N1 on the PA and MP trees . Both Guanajuato and Jalisco viruses are related to pdmH1N1 on the NP tree . Therefore , the swine-to-human transmission event could have occurred in Guanajuato , Jalisco , or possibly another state in central Mexico that has not been sampled . 10 . 7554/eLife . 16777 . 022Figure 7 . Timing and location of swine ancestors of pdmH1N1 . The color of each dot represents the inferred location of the node representing the common ancestor of pdmH1N1 viruses and the most closely related swine ivurses ( indicated by open circles on the MCC trees in Figure 3 ) , for a posterior distribution of ~2000 trees inferred for each segment . A high proportion of blue dots indicates a higher proportion of trees with Mexico as the inferred location state . The x-axis indicates the tMRCA of the same node , again for each tree . The 95% HPD is provided in brackets . Older tMRCAs are associated with longer phylogenetic branch lengths and gaps in sampling . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 022 10 . 7554/eLife . 16777 . 023Figure 7—source data 1 . Times to the most recent common ancestor ( tMRCA ) and posterior probabilities ( >0 . 01 ) for the location state of the node representing the ancestor of the pdmH1N1 clade and the most closely related swine viruses ( indicated with open circles in Figure 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 023 Of the 10 reassortant genotypes identified in Mexican swine , four genotypes were identified in central Mexican swine that contain at least one EAsw segment , indicative of multiple reassortment events involving EAsw viruses . Genotype 1 viruses emerged multiple times in central Mexico and sustained transmission longer than any other genotype in central Mexican swine , as estimated using ‘Markov rewards’ ( Figure 3—source data 3 ) . No EAsw PB2 , PB1 , PA , NP , and NS segments were detected in any swine in Mexico , suggesting replacement by reassortment with Csw and TRsw viruses in central Mexico . One Puebla virus ( AVX-47 , genotype 4 ) was identified with a HA ( H1 ) from EAsw . The tMRCA of this singleton virus and EAsw viruses from Europe is estimated to be in the early 1990s ( Figure 3—source data 1 ) . Although gaps in surveillance in European swine in the 1990s result in uncertainty about the precise timing of viral introduction from Europe-to-Mexico , it is possible that the EAsw H1 has circulated in Mexican swine undetected for many years , either in Puebla or another under-sampled state ( Figure 3—source data 1 ) . Notably , EAsw NA and MP segments were identified frequently in central Mexican swine . A conserved internal gene constellation of TRIG PB2 , PB1 , PA , NP , and NS segments and an EAsw MP segment was observed in over 97% of the swIAVs in Jalisco , Guanajuato , and Aguascalientes ( Figure 2 ) . A separate analysis of pdmH1N1 viruses collected globally in humans during the first wave of the 2009 pandemic ( March–May 2009 ) provides additional support for an origin of the pdmH1N1 outbreak in central Mexico , consistent with a previous demonstration of the centrality of Mexico in the 2009 pandemic using epidemiological data ( Brockmann and Helbing , 2013 ) . The use of genetic data provides more refined within-country spatial details . Location state posterior probabilities for the root of the tree were highest in Mexico City ( 0 . 81 ) , consistent with an origin of the pandemic outbreak in Mexico’s capital city ( Figure 4—figure supplement 2 ) . Texas shares a long and heavily-trafficked border with Mexico ( Weinberg et al . , 2003 ) , and the southern United States was the largest recipient of pdmH1N1 introductions from Mexico during the early stages of the 2009 pandemic ( Figure 4B ) . However , the southern US was not a major source of pdmH1N1 viruses to the United States’ Midwest , West , Northeast regions or globally . Rather , viruses identified in New York , Wisconsin , California , and other US states represent independent introductions of pdmH1N1 from Mexico . Notably , the Northeast region of the United States , including New York , had the largest role in the onward global dissemination of the pandemic virus to Europe , Asia , and South America ( Figure 4B ) , perhaps driven by high volumes of air traffic from New York City and the size of the city’s epidemic during the spring wave ( Viboud et al . , 2014 ) .
Overall , our findings resolve a long-standing question of where the pandemic H1N1 virus originated in swine . Moreover , they underscore the importance of understanding the global evolution of IAVs in pigs , and the pandemic threat presented by the large number of independently evolving swIAV populations worldwide that periodically experience invasions of new viruses . Our findings demonstrate the central importance of inter-hemispheric swine movements in the long-range dissemination of swIAV diversity , which allowed divergent Eurasian and North American viruses to co-circulate in central Mexico and reassort into genotype 1 viruses ( Figure 8 ) . It remains unclear whether conditions in Mexico were uniquely suitable for the emergence of genotype 1 viruses , or whether it is possible that similar viruses also circulate undetected in Asia , the only other location where Csw , EAsw , and TRsw viruses and various reassortants are known to be present ( Lam et al . , 2011 ) . The inability of researchers to produce genotype 1 viruses experimentally in cell cultures or pigs ( Ma et al . , 2014 ) implies strain-specific differences in reassortment capabilities . The likelihood of pandemic emergence is therefore enhanced by the seeding of diverse swIAV lineages in countries with varying swine production practices and fitness pressures . The independent evolution of multiple variants increases the probability that one will evolve the capacity to transmit to humans . 10 . 7554/eLife . 16777 . 024Figure 8 . Origins of pdmH1N1 . Summary of the migration and reassortment events leading to the emergence of pdmH1N1 precursor viruses in central-west Mexican swine . Segments from classical and Eurasian viruses for which there is no evidence of onward transmission in central-west Mexican swine are indicated with short horizontal lines . DOI: http://dx . doi . org/10 . 7554/eLife . 16777 . 024 Although understanding the positive and negative interactions between fitness in human and swine hosts is central to understanding pandemic emergence , this area of research remains poorly understood . Strong patterns of reassortment between EAsw and North American viruses in Mexican swine are an indication of competitive interactions . The apparent fitness of the TRIG + EAsw MP internal gene constellation in central Mexican swine was central to the emergence of the genotype 1 viruses that gave rise to pdmH1N1 , but is not well understood . The success of genotype 1 viruses for many years in central-west Mexico is notable , given the low levels of persistence of pdmH1N1 viruses as an intact genome in swine following reverse zoonosis in many countries ( Liang et al . , 2014; Nelson and Vincent , 2015; Nelson et al . , 2015d; Pereda et al . , 2011; Watson et al . , 2015 ) , a pattern also observed in Sonora . The lower levels of onward transmission of pdmH1N1 surface genes in swine could relate to competition with existing H1s and N1s . The identification of EAsw NA and MP segments in central Mexican swine is notable , given that these segments were associated with enhanced respiratory droplet transmission in animal models ( Lakdawala et al . , 2011; Yen et al . , 2011 ) . The EAsw MP segment also is associated with zoonotic transmission of swIAVs at agricultural fairs in the US ( Bowman et al . , 2012 ) . A number of studies have examined the transmission potential of naturally occurring or lab-generated avian and swine viruses in ferret or other animal models , as a measure of pandemic potential ( Herfst et al . , 2012; Sorrell et al . , 2009; Yang et al . , 2016 ) . However , a fundamental understanding of fitness landscapes and trade-offs in multi-host systems has been constrained by the a lack of final resolution on the ethics of various experiments that involve novel viruses that present a potential threat to humans ( Frank et al . , 2016 ) . The identification of viruses related to the 2009 pandemic precursors invites new experimental research into central questions of pandemic emergence at the human-swine interface . The detection of EAsw in Mexico suggests that even relatively low levels of livestock movement can disseminate a highly transmissible virus such as IAV long distances , bringing attention to the importance of surveillance and mitigation strategies , including quarantine . Presently , imported live swine are not routinely tested for influenza . The emergence of the 2009 pandemic virus was closely linked to the increase in Mexico's imports of live swine during the 1990s and the influx of new influenza virus lineages from the United States and Europe . At this time we cannot rule out Asia-to-Mexico viral migration , owing to possible omissions in reported imports and exports of swine and gaps in sequencing of viruses in swine globally . However , all that is known about the nature of international swine production and trade suggests that this direction of viral movement is unlikely . Long-distance trade of pigs from Europe , the United States , and Canada to countries in Asia and South America occurs primarily to import high-efficiency European and North American breeding lines . Although illegal trade of poultry is important in the spread of AIVs , particularly within the East Asian region , there are no known economic incentives ( and strong economic disincentives ) for the illegal trade of pigs from Asia and Mexico . Globally , Asia is overwhelmingly a net importer of swine from Europe and North America , with extremely little export of Asian pigs or their viruses to any other continents ( Nelson et al . , 2015c ) . Another outstanding question is whether genotype 1 viruses have disseminated from Mexico to other countries in Latin America , where surveillance is low or non-existent . Mexico has reported export of live swine to several countries in Latin America ( Belize , Colombia , Costa Rica , Cuba , El Salvador ) . However , the total volume of Mexico’s live swine exports during 1996–2012 amounts to only one-tenth of Costa Rica’s over the same time period , and one-thirtieth of Brazil’s . Mexico is therefore considered an ecological sink for swine influenza viruses , with limited opportunity for viral export to other countries ( Nelson et al . , 2015c ) . The unexpected detection of EAsw in Mexico highlights the need to revisit outdated assumptions about the spatial distribution of IAVs in swine populations globally , given the capacity of long-distance live swine movements to disseminate viral diversity between continents . Although surveillance of swIAVs has been initiated in many new regions since 2009 , it remains highly imbalanced on a global scale and low compared to avian surveillance ( Nelson and Vincent , 2015 ) . Recent detections of novel swIAVs in regions that traditionally have not conducted surveillance in swine , including Latin America and South-east Asia , underscore the importance of expanding surveillance in the large number of under-sampled regions with high densities of swine ( Cappuccio et al . , 2011; Nelson et al . , 2015a; Ngo et al . , 2012; Perera et al . , 2013 ) . China remains an important source of zoonotic influenza viruses with pandemic potential , and the identification of strong regional variations in swIAV diversity in Mexico highlights the importance of capturing swIAV diversity across China’s heterogeneous regions . Although the importance of regional live swine movements in the evolution of swIAVs in the United States has been documented ( Nelson et al . , 2011 ) , many countries lack data on the intra-country movements of swine , limiting study of interactions between animal movements and regional patterns of swIAV diversity . Data on poultry movements within and between countries are similarly lacking , limiting our understanding of the role of trade in the spatial dynamics of IAVs in domestic birds . Our ability to predict future pandemics will require intensified viral surveillance and an understanding of how economic forces and international trade policies affect changes in animal movements and production practices that drive viral emergence .
Swine influenza isolates were obtained from samples submitted to Avimex laboratories ( Mexico ) for confirmation diagnostics of pigs with respiratory syndrome . The samples covered years 2010 ( 1 sample ) , 2011 ( 6 samples ) , 2012 ( 34 samples ) , 2013 ( 14 samples ) and 2014 ( 3 samples ) . All samples came from large commercial facilities from 6 states: Sonora ( 8 different farms ) , Jalisco ( 6 farms ) , Guanajuato ( 3 farms ) , Puebla ( 2 farms ) , Yucatan ( 2 farms ) and Aguas Calientes ( 1 farm ) . Virus was isolated in specific pathogen free embryonated eggs or in cell culture; HA , NA subtype was determined and samples from first or second passage were sent to Mount Sinai ( New York ) for RNA extraction and whole genome sequencing . When multiple isolates of the same subtype were obtained from the same farm and outbreak the sample with highest viral titer was submitted . GenBank accession numbers of the 58 genomes are available in the Dryad data repository under Doi:10 . 5061/dryad . m550m . Figure 2—source data 1 describes the date , location , subtype and lineage assignment of each viral segment . Viral RNA was obtained from 140 µl of the original sample , or after amplification in MDCK cells , using the QIAamp Viral RNA minikit QIAGEN ) . Next , 5 µl of viral RNA were used as template in a 50 µl multi-segment RT-PCR reaction ( Superscript III high-fidelity RT-PCR kit ) with influenza-specific universal primers complementary to the conserved 12–13 nucleotides at the end of all 8 genomic segments . Primer sequences and final concentrations in the reaction were as follows ( influenza-complementary sequences are underlined ) : Opti1-F1 – 5’ GTTACGCGCCAGCAAAAGCAGG ( 0 . 1 μM ) ; Opti1-F2 - 5’ GTTACGCGCCAGCGAAAGCAGG ( 0 . 1 μM ) ; Opti1-R1 - 5’ GTTACGCGCCAGTAGAAACAAGG ( 0 . 2 μM ) . Amplicons were purified with 0 . 45x volume AMPure XP beads ( Beckman Coulter ) and 0 . 5–1 µg was sheared to an average fragment size of 150 bp on a Bioruptor Pico sonicator ( Diagenode ) . Next , amplicon sequence libraries were prepared using the end repair , A-tailing , and adaptor ligation NEBNext DNA library prep modules for Illumina from New England Biolabs according to the manufacturer’s protocol . Following final size-selection with 1x volume Ampure XP beads , and secondary PCR ( 8 cycles ) to introduce barcoded primers , multiplexed libraries were sequenced on the Illumina HiSeq 2500 platform in a single-end 100 nt run format . Single-end 100 nt reads were first filtered with cutadapt ( Martin , 2011 ) to remove low-quality sequences and adapters . Reads were then mapped against a non-redundant copy of IAV sequences in the Influenza Research Database ( IRD ) using STAR ( Dobin et al . , 2013 ) and chimeric reads with non-contiguous alignments to reference segments ( typically originating from defective interfering particles containing segments with internal deletions ) were removed . Assembly of IAV genomic segments was performed using a custom pipeline ( available from the authors ) in multiple stages . First , an initial assembly was done using the inchworm component of Trinity ( Grabherr et al . , 2011 ) , and viral contigs bearing internal deletions were identified by BLAT ( Kent , 2002 ) mapping against non-redundant IRD reference sequences . In the second stage the inchworm assembly was repeated but now removing breakpoint-spanning kmers from the assembly graph . Resulting IAV contigs were then oriented , and trimmed to remove low-coverage ends and any extraneous sequences beyond the conserved IAV termini . In the final stage the CAP3 ( Huang and Madan , 1999 ) assembler was used to improve contiguity by merging contigs originating from the same segment type if their ends overlapped by at least 25 nt . Assembly quality and contiguity was assessed for all segments by mapping sequence reads back to the final assemblies using the Burrows-Wheeler Alignment ( BWA ) tool ( Li and Durbin , 2009 ) and complete segments were annotated using the NCBI Influenza Virus Sequence Annotation Tool ( Bao et al . , 2007 ) . In addition to the 58 whole-genome sequences generated for this study from Mexico , background data sets were included that were recently used in previous studies of IAV evolution in swine ( Nelson et al . , 2014 , 2015a , 2015b , 2015c ) . The vast majority of swIAV sequences from the Americas are deposited in NCBI’s GenBank , but to ensure that no additional sequences of importance were available from the Global Initiative on Sharing All Influenza Data’s ( GISAID ) Epiflu database from Europe and Asia , sequences for the EAsw lineage were downloaded from GISAID ( acknowledgement of the authors and research laboratories that contributed these data are provided in Supplementary files 1–2 ) . A large number of whole-genome sequences from swIAVs in European swine were published by Watson et al . ( 2015 ) after we completed our analysis . To ensure that none were closely related to pdmH1N1 or any of the Mexican swIAVs , particularly on the trees inferred for the Eurasian N1 and MP segments , we inferred neighbor-joining trees including the new European swIAV sequences deposited into GenBank by these authors ( data available from the authors ) . Sequence alignments were constructed for each of the six internal gene segments ( PB2 , PB1 , PA , NP , MP , and NS ) and for the H1 , H3 , N1 , and N2 antigenic segments separately using MUSCLE v3 . 8 . 3 ( Edgar , 2004 ) , with manual correction in Se-Al v2 . 0 ( available at http://tree . bio . ed . ac . uk/software/seal/ ) . The program Path-O-Gen v1 . 4 . 0 ( available at http://tree . bio . ed . ac . uk/software/pathogen/ ) was used to identify potential sequencing errors that substantially deviated from the linear regression of root-to-tip genetic distance against time , which were subsequently removed from the study . We excluded nine triple reassortant Mexican swIAVs ( e . g . , A/swine/Mexico/Mex19/2010 ( H1N1 ) , Figure 3—figure supplement 1 ) that had shorter branch lengths than expected given their year of sampling , an indication of possible errors in sequencing or assembly . For the purposes of lineage assignment , initial phylogenetic trees were inferred using the neighbor-joining method available in PAUP v4 . 0b10 for each of the ten alignments ( available at http://paup . csit . fsu . edu ) . For the PB2 , PB1 , PA , NP , and NS segments , each virus was categorized as related to the triple reassortant internal genes ( ‘TRIG’ ) lineage , classical lineage ( for NP and MP ) , avian-like Eurasian ( MP ) , or pdmH1N1 . Reference sequences for pdmH1N1 from humans and swine , including A/California/04/2009 ( H1N1 ) , were included to assist identification of the pdmH1N1 lineage . H1 and N1 segments were categorized as ( a ) classical , ( b ) avian-origin Eurasian , ( c ) pdmH1N1 , or ( d ) human seasonal H1N1 virus origin . All H3 and N2 segments belonged to the same lineage: human seasonal H3N2 virus-related . The sequence alignment for each segment was further separated into each of these lineages , and trees were inferred independently for each alignment that contained at least one Mexican virus: PB2-TRIG ( n = 273 sequences ) , PB1-TRIG ( n = 329 sequences ) , PA-TRIG ( n = 309 sequences ) , H1-classical ( n = 352 sequences ) , H1-human seasonal ( n = 213 sequences ) , H1-eurasian ( n = 305 sequences ) , H3 ( n = 331 sequences ) , NP- TRIG/classical ( n = 291 sequences ) , N1-classical ( n = 287 sequences ) , N1-Eurasian ( n = 314 sequences ) , N2 ( n = 726 sequences ) , MP-TRIG/classical ( n = 256 sequences ) , MP-Eurasian ( n = 300 sequences ) , and NS-TRIG/classical ( n = 479 sequences ) . For the NP , MP , and NS segments , the TRIG lineage is a continuation of the classical lineage , as classical NP , MP , and NS segments were incorporated into triple reassortant viruses during the reassortment ( Zhou et al . , 1999 ) . Phylogenetic relationships were inferred for each of the data sets separately using the time-scaled Bayesian approach using MCMC available via the BEAST v1 . 8 . 2 package ( Drummond et al . , 2012 ) and the high-performance computational capabilities of the NIH HPC Biowulf Linux cluster at the National Institutes of Health , Bethesda , MD ( http://hpc . nih . gov/ ) . A relaxed uncorrelated lognormal ( UCLN ) molecular clock was used , with a constant population size , and a general-time reversible ( GTR ) model of nucleotide substitution with gamma-distributed rate variation among sites . For viruses for which only the year of viral collection was available , the lack of tip date precision was accommodated by sampling uniformly across a one-year window from January 1st to December 31st . The MCMC chain was run separately at least three times for each of the data sets and for at least 100 million iterations with sub-sampling every 10 , 000 iterations , using the BEAGLE library to improve computational performance ( Suchard and Rambaut , 2009 ) . All parameters reached convergence , as assessed visually using Tracer v . 1 . 6 , with statistical uncertainty reflected in values of the 95% highest posterior density ( HPD ) . At least 10% of the chain was removed as burn-in , and runs for the same lineage and segment were combined using LogCombiner v1 . 8 . 0 and downsampled to generate a final posterior distribution of 1000 trees that was used in subsequent analyses . In order to visualize branch lengths between swIAVs and the pdmH1N1 clade in terms of genetic distance , rather than time , the phylogenetic relationships of each of the eight data sets containing the pdmH1N1 clade were inferred using the maximum likelihood ( ML ) method available in the program RAxML v7 . 2 . 6 ( Stamatakis , 2006 ) , incorporating a general time-reversible ( GTR ) model of nucleotide substitution with a gamma-distributed ( Γ ) rate variation among sites . To assess the robustness of each node , a bootstrap resampling process was performed ( 500 replicates ) , again using the ML method available in RAxML v7 . 2 . 6 . Although not the focus of this study , phylogenetic trees also were inferred for the pdmH1N1 lineage to identify recent human-to-swine transmission events in Mexico ( inferred for PB2 and H1 , as representative segments , Figure 3—source data 4 ) , using methods described above and background data sets used in previous studies ( Nelson et al . , 2015b; 2015d ) . The phylogeographic analysis considered 11 locations: Asia ( including China , Japan , South Korea , Thailand , Vietnam ) , USA/Canada , Europe ( including Belgium , Czech Republic , Denmark , France , Germany , Italy , Netherlands , Poland , Spain , and the United Kingdom ) , South America ( including Argentina and Brazil ) , Mexico ( Aguascalientes ) , Mexico ( Guanajuato ) , Mexico ( Jalisco ) , Mexico ( Puebla ) , Mexico ( Sonora ) , and Mexico ( Yucatan ) , as well as humans ( globally ) . The location state was specified for each viral sequence , allowing the expected number of location state transitions in the ancestral history conditional on the data observed at the tree tips to be estimated using ‘Markov jump’ counts ( Minin and Suchard , 2008a ) , which provided a quantitative measure of asymmetry in gene flow between regions . The location of viruses in the pdmH1N1 clade was left uninformed , allowing the reconstruction of the location state of the common ancestor of pdmH1N1 to be unbiased by human pdmH1N1 data . For computational efficiency the phylogeographic analysis was run using an empirical distribution of 1000 trees ( Lemey et al . , 2014 ) , allowing the MCMC chain to be run for 25 million iterations , sampling every 1000 . A Bayesian stochastic search variable selection ( BSSVS ) was employed to improve statistical efficiency for all data sets containing more than four location states . Maximum clade credibility ( MCC ) trees were summarized using TreeAnnotator v1 . 8 . 0 and the trees were visualized in FigTree v1 . 4 . 2 . Heat-maps were constructed using the R package to summarize Markov jump counts inferred over the totality of phylogenies ( all segments , all swIAV lineages ) . Waiting times between location state transitions ( 'Markov rewards' ) ( Minin and Suchard , 2008b ) also were estimated , allowing us to estimate ‘Markov jump’ counts normalized for the percent time of locations in the tree . All available whole-genome sequence data for pdmH1N1 viruses collected globally in humans during March 1–May 31 , 2009 were downloaded from the National Center for Biotechnology Information’s Influenza Virus Resource ( Bao et al . , 2008 ) ( http://www . ncbi . nlm . nih . gov/genomes/FLU/FLU . html ) on December 12 , 2015 . The over-sampled locations of New York , USA and Wisconsin , USA were randomly subsampled to 50 genomes each , resulting in a final data set of 422 complete concatenated whole-genome sequences ( Supplementary file 3 ) . The sequence data were aligned using MUSCLE v3 . 8 . 3 ( Edgar , 2004 ) . Phylogenetic relationships were inferred using the time-scaled Bayesian approach using MCMC available via the BEAST v1 . 8 . 2 package , in this case assuming a model of exponential growth in the number of infections , as this model is more realistic during the early growth stage of new pandemic virus ( Smith et al . , 2009 ) . The alignment and phylogeographic analysis , using the methods described above , considered 12 locations: Asia ( China , Japan , South Korea , Malaysia , Singapore , Thailand ) , Canada , Europe ( United Kingdom , Finland , Italy , Russia , Netherlands , France , Germany ) , South America ( Argentina , Chile , as well as the Dominican Republic ) , Mexico ( Mexico City ) , Mexico ( San Luis Potosi ) , Mexico ( Veracruz ) , Mexico ( state unknown ) , United States - Northeast ( Maryland , Massachusetts , New Jersey , New York , Rhode Island ) , United States – Midwest ( Indiana , Iowa , Kansas , Michigan , Nebraska , Ohio , Wisconsin ) , United States – South ( Alabama , Texas , Florida , Mississippi , South Carolina , Tennessee ) , and United States – West ( Arizona , California , Colorado , New Mexico ) . Heat-maps were constructed using the R package to summarize Markov jump counts inferred over the phylogeny . Waiting times between location state transitions ( 'Markov rewards' ) ( Minin and Suchard , 2008b ) also were estimated , allowing us to estimate ‘Markov jump’ counts normalized for the percent time of locations in the tree . To estimate the extent of reassortment and onward transmission of different genotypes in swine in Mexico , we used a similar phylogeographic approach , but in this case specified a location state for each Mexican virus based on genotype . For viruses collected for this study , we used the 12 genotypes defined in Figure 2 . For the six additional whole-genome sequences available from Mexico that were published previously ( Nelson et al . , 2015c ) we defined two new genotypes: genotypes 13 and 14 . In addition to estimating transitions between locations ( Markov jumps ) , we also estimated the waiting times between transitions ( 'Markov rewards' ) ( Minin and Suchard , 2008b ) as an indication of the period of time Mexican lineages circulate as different genotypes . The trade value ( USD ) for live swine trade between other countries and Mexico for the years 1996–2012 was obtained from the United Nations’ Commodity Trade Statistics Database ( available at http://comtrade . un . org , accessed March 20 , 2014 ) ( Figure 6—source data 1 ) . Estimates of the total number of live swine imports into Mexico ( country of origin unknown ) were obtained for a longer time period ( 1969–2010 ) from the Food and Agriculture Organization ( FAO ) of the United Nations Datasets repository ( available at http://data . fao . org/datasets , accessed March 21 , 2014 ) ( Figure 5—source data 1 ) . Data on the size of live swine population in each of Mexico’s states was obtained from Mexico’s Secretariat of Agriculture , Livestock , Rural Development , Fisheries and Food ( SAGARPA ) ( www . siap . gob . mx ) , accessed December 30 , 2015 ( Figure 1—source data 1 ) . | In 2009 a new influenza virus jumped from pigs to humans and spread very rapidly , causing an initial outbreak in Mexico and becoming a global pandemic in just a few months . Although the most straightforward explanation is that the virus originated in swine in Mexico , several studies suggested that this was unlikely because key genetic components of the virus had never been detected in the Americas . Determining the source of the disease is critical for predicting and preparing for future influenza pandemics . Mena , Nelson et al . sought to better characterize the genetic diversity of influenza viruses in Mexican swine by obtaining the entire genetic sequences of 58 viruses collected from swine in Mexico , including some from previously unsampled regions in central Mexico . The sequences revealed extensive diversity among the influenza viruses circulating in Mexican swine . Several viruses included genetic segments that originated from viruses from Eurasia ( the landmass containing Europe and Asia ) and had not previously been detected in the Americas . The new sequences contained key genetic components of the 2009 pandemic virus . Furthermore , the sequences suggest that viruses with a similar genetic composition to the 2009 pandemic virus have been circulating in pigs in central-west Mexico for more than a decade . Thus , this region is the most likely source of the virus that started the 2009 pandemic . Mena , Nelson et al . also found that the movement of viruses from Eurasia and the United States into Mexico closely follows the direction of the global trade of live swine . This highlights the critical role that animal trading plays in bringing together diverse viruses from different continents , which can then combine and generate new pandemic viruses . A potential next step is to perform experiments that investigate how well the swine viruses can replicate and pass between different animal models . Comparing the results of such experiments with the findings presented by Mena , Nelson et al . could identify factors that make the viruses more likely to spread to humans and produce a pandemic . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Origins of the 2009 H1N1 influenza pandemic in swine in Mexico |
Nonribosomal peptides represent a large class of metabolites with pharmaceutical relevance . Pteridines , such as pterins , folates , and flavins , are heterocyclic metabolites that often serve as redox-active cofactors . The biosynthetic machineries for construction of these distinct classes of small molecules operate independently in the cell . Here , we discovered an unprecedented nonribosomal peptide synthetase-like-pteridine synthase hybrid biosynthetic gene cluster in Photorhabdus luminescens using genome synteny analysis . P . luminescens is a Gammaproteobacterium that undergoes phenotypic variation and can have both pathogenic and mutualistic roles . Through extensive gene deletion , pathway-targeted molecular networking , quantitative proteomic analysis , and NMR , we show that the genetic locus affects the regulation of quorum sensing and secondary metabolic enzymes and encodes new pteridine metabolites functionalized with cis-amide acyl-side chains , termed pepteridine A ( 1 ) and B ( 2 ) . The pepteridines are produced in the pathogenic phenotypic variant and represent the first reported metabolites to be synthesized by a hybrid NRPS-pteridine pathway . These studies expand our view of the combinatorial biosynthetic potential available in bacteria .
Nonribosomal peptides are a structurally and functionally privileged class of natural products constructed from a highly diverse pool of potential proteinogenic and nonproteinogenic amino acid building blocks ( Walsh et al . , 2013; Walsh , 2016 ) . Members of the family include well-known pharmacologically relevant agents , such as vancomycin , daptomycin , penicillin , cyclosprin , and many others . The core catalytic domains of a minimal nonribosomal peptide synthetase ( NRPS ) extender module include condensation ( C ) , adenylation ( A ) , and peptidyl-carrier protein ( PCP , a . k . a . , thiolation , T ) domains . The NRPS first selects its cognate amino acid from the available substrate pool , a step controlled by the selectivity of the A domain , activates it as an aminoacyl adenylate , and subsequently loads it onto the PCP domain . The C domain then typically catalyzes the formation of a trans-amide bond establishing individual peptide backbone linkages through nucleophilic attack of the free amino group present on a downstream aminoacyl-PCP on the upstream peptidyl-PCP . NRPSs can also engage in ‘hybrid’ pathways to dramatically expand their biocatalytic capabilities . Of note are the NRPS-polyketide synthase ( PKS ) hybrid pathways , which produce important molecules , including rapamycin , bleomycin , epothilone , colibactin , and others . In contrast to the ‘assembly line’ logic of the majority of NRPSs , pteridines , such as the cofactors biopterin and folate , are derived from guanosine triphosphate ( GTP ) ( Brown , 2006 ) . Pteridines are composed of fused pyrimidine and pyrazine rings , and natural pteridines are typically functionalized at the C-6 position of the pteridine core ( vide infra ) . The electronic properties of this scaffold underlie its role in redox active cofactors critical to a host of metabolic transformations , such as hydroxylation of aromatic compounds and generation of the neurotransmitter nitric oxide ( Groehn et al . , 2000 ) . Additionally , functionalized pteridines serve as reactants in a number of one-carbon group transfer reactions in metabolism and , more recently , were implicated in catalysis of 1 , 3-dipolar cycloaddition-mediated decarboxylation reactions ( Payne et al . , 2015; White et al . , 2015 ) . The increase in genomic sequence information from diverse microbial sources has highlighted enormous untapped metabolic potential for discovery of novel nonribosomal peptides and pteridines among other metabolite groups , including groups synthesized by new enzyme classes ( Cimermancic et al . , 2014 ) . As a result of horizontal gene transfer events during evolution , it is known that the bacterial secondary metabolic pathways encoding many of these metabolites often reside on genomic islands ( Shankar et al . , 2006; Penn et al . , 2009; Ziemert et al . , 2014 ) . Genome synteny analyses platforms enable in silico visualization of co-localized genetic loci among phylogenetically related organisms and represent general tools to aid in identifying genomic islands . We have previously used genome synteny analysis to aid in identification of ‘atypical pathways’ and to determine the biocatalytic functions of hypothetical proteins located on genomic islands ( Guo and Crawford , 2014; Vizcaino et al . , 2014b ) . Here , we employ genome synteny analysis to identify a genomic island in the entomopathogen Photorhabdus luminescens TT01 harboring an unprecedented combination of nonribosomal peptide synthetase ( NRPS ) -like and pteridine synthase biosynthetic machineries , suggesting a novel type of hybrid pathway . We demonstrate that this pathway encodes new metabolites dependent on the hybrid enzymatic machinery , the pepteridines . P . luminescens is a Gram-negative Gammaproteobacterium that undergoes stochastic phenotypic variation , and through the use of genetically ‘locked’ variants ( Somvanshi et al . , 2012 ) , we show that the pepteridines are produced in a specific variant associated with pathogenesis . Allelic-exchange mutagenesis in P . luminescens and comparative quantitative proteomic analysis further reveal that the genetic locus affects production of several groups of proteins related to the biosynthesis of known quorum sensing and secondary metabolite systems .
P . luminescens is a Gammaproteobacterium mutualistically associated with nematodes , and the pair prey on insect larvae ( Waterfield et al . , 2009 ) . P . luminescens produces an assortment of bioactive molecules and antibiotics to regulate its mutualistic and pathogenic interactions ( Vizcaino et al . , 2014b; Challinor and Bode , 2015 ) . Consequently , this genus , which includes one human pathogen , rivals the Streptomyces genus in terms of the number of secondary metabolic pathways in a given genome ( Duchaud et al . , 2003; Tobias et al . , 2016 ) . Using the MicroScope bioinformatics platform , we identified a genomic island ( plu2792-plu2799 , Figure 1—figure supplement 1 ) , harboring mixed NRPS-pteridine synthase machinery ( Figure 1 ) . Protein sequence homology analysis demonstrated predicted pteridine biosynthetic enzymes , such as GTP cyclohydrolase ( GTPCH ) Ι , 6-pyruvoyltetrahydropterin synthase , pteridine reductase , and pteridine pyrophosphokinase . Interestingly , a NRPS carrier protein ( thiolation domain , T ) and a condensation domain ( C ) were genetically fused to a pyruvate dehydrogenase E2-like subunit ( plu2796 ) . The pyruvate dehydrogenase complex ( E1 , E2 , and E3 subunits ) is a well-studied set of enzymes that converts pyruvate to acetyl-CoA ( Patel and Roche , 1990 ) , serving as a key metabolic bridge between glycolysis and the citric acid cycle . A complementary E1-like subunit is also encoded in the pathway . This pathway was not detected by early versions of the antiSMASH algorithm ( Blin et al . , 2013 ) used for identifying the biosynthetic pathways of secondary metabolites; however , recent integration of the ClusterFinder algorithm which aids in identifying divergent biosynthetic gene clusters of unknown classes ( Cimermancic et al . , 2014; Weber et al . , 2015 ) allowed detection of part of the pathway ( plu2796-plu2798 ) as a likely pathway of secondary metabolic enzymes . 10 . 7554/eLife . 25229 . 003Figure 1 . The pepteridine biosynthetic locus . Green , pteridine synthesis genes; Blue , pyruvate dehydrogenase-like genes; Red , NRPS-like genes . T , thiolation domain; C , condensation domain . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 00310 . 7554/eLife . 25229 . 004Figure 1—figure supplement 1 . Genome synteny analysis using MicroScope . Genome synteny analysis revealed a genomic island in P . luminescens TT01 ( note absence of biosynthetic genes in phylogenetically related species , lower panel ) harboring a number of enzymes possessing homology to pteridine and NRPS biosynthetic machineries ( +1 , +2 , and +3 reading frames , upper panel ) . The regulator in the genomic island was not included in our design , as the pathway was placed under the control of a T7 promoter . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 004 We first cloned the hybrid NRPS-pteridine genomic island without its clustered regulatory protein ( Plu2792 ) and placed it under the control of a phage T7 promoter for heterologous expression in Escherichia coli BAP1 ( Pfeifer et al . , 2001 ) . Comparison of the culture broths from E . coli harboring the pathway relative to that of an empty vector control ( pET28a ) revealed a pathway-dependent yellow phenotype ( Figure 2B ) . Initial comparative single quadrupole liquid chromatography-mass spectrometry ( LC/MS ) analysis of their butanol extracts revealed two dramatically increased peaks in the broth harboring the pathway ( Figure 2A ) . Through extensive 1D- and 2D-NMR , high-resolution-electrospray ionization-quadrupole-time-of-flight-mass spectrometry ( HR-ESI-QTOF-MS ) , and LC/MS co-injections with authentic standards ( Figure 2C–E and Figure 2—figure supplements 1 and 2 ) , the chemical structures of the small molecules attributed to these peaks were assigned as known 2-amino-4-oxopteridine small molecules , 7 , 8-dihydroxanthopterin ( 3 ) and pterin ( 4 ) . 10 . 7554/eLife . 25229 . 005Figure 2 . Production of 7 , 8-dihydroxanthopterin ( 3 ) and pterin ( 4 ) by heterologous expression . ( A ) HPLC traces ( 310 nm ) of butanol extracts showing compounds 3 and 4 being over-produced in the heterologous expression strain . ( B ) The edge of a culture flask growing E . coli BAP1 carrying an empty vector pET28a ( top ) , and a culture flask growing E . coli BAP1 carrying the wild-type pepteridine pathway ( bottom ) . ( C ) HPLC traces ( 310 nm ) from HPLC co-injection with authentic pterin ( 4 ) , and UV absorption spectral comparison between natural and authentic pterin ( inset ) . ( D ) UV-vis absorption spectra of compounds 3 and 4 . ( E ) 1H NMR comparison of natural ( blue ) and standard ( black ) 3 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 00510 . 7554/eLife . 25229 . 006Figure 2—figure supplement 1 . 1H and 13C NMR spectra of compound 3 . ( A ) 1H NMR spectrum of compound 3 in DMSO-d6 . ( B ) 13C NMR spectrum of compound 3 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 00610 . 7554/eLife . 25229 . 007Figure 2—figure supplement 2 . 2D- ( gCOSY and gHMBCAD ) NMR spectra of compound 3 . ( A ) gCOSY NMR spectrum of compound 3 in DMSO-d6 . ( B ) gHMBCAD NMR spectrum of compound 3 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 007 To identify novel lower abundance advanced metabolites dependent on both the NRPS and pteridine biosynthetic enzymes , we individually constructed nonpolar genetic deletions of every biosynthetic gene in the pathway , maintaining transcriptional and translational control elements for comparative metabolomics . Relative to the control vector , we identified 224 molecular features in butanol extracts that were dependent on the presence of the wild-type pathway ( Supplementary file 1A ) . These include small molecules that are encoded by the pathway , shunt metabolites emerging from the pathway , and host metabolites that are enhanced by the pathway from an undetectable to a statistically significant level . Of these , 114 were dependent on the GTPCH Ι homolog , Plu2793 , which is predicted to convert GTP to 7 , 8-dihydroneopterin triphosphate , initiating pteridine synthesis ( Burg and Brown , 1968 ) . The predicted functional redundancy with primary metabolic enzymes is expected to account for the larger number of metabolic perturbations observed for plu2793 . Thirty-seven molecular features were dependent on the atypical dehydrogenase E2-NRPS fusion enzyme , Plu2796 , and only 12 were dependent on both . The number of wild-type molecular features that were dependent on the remaining biosynthetic enzymes , Plu2794 , Plu2795 , Plu2797 , Plu2798 , and Plu2799 , are listed in Table 1 . 10 . 7554/eLife . 25229 . 008Table 1 . Genetic dependency of molecular features . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 008GeneDependent molecular featuresplu2793114plu279431plu279519plu279637plu279710plu279821plu279926plu2793/plu279612Of the wild-type molecular features , this table denotes how many of them were deleted ( i . e . , not detected ) in each mutant strain . We then conducted tandem MS analysis on the pathway-dependent molecular features for pathway-targeted molecular networking ( Vizcaino et al . , 2014a ) . Molecular networking is a powerful approach to network molecules based on tandem MS fragmentation similarities ( Watrous et al . , 2012 ) . Pathway-targeted molecular networking focuses on networking of metabolites dependent on the presence of a functional pathway ( e . g . , wild-type versus a secondary metabolic pathway mutant ) . By assessing and mapping the relative production levels of the metabolites onto the larger wild-type pepteridine network ( Figure 3—figure supplements 1 and 2 ) , we could visualize how individual genetic mutations affect overall metabolite distributions at a systems level ( Figure 3 ) . White nodes represent metabolites abolished in a given mutant strain , thereby dramatically focusing metabolite discovery efforts . Some deletions led to a decrease in , rather than complete abolishment of , select molecular feature production , indicating that substantial metabolic crosstalk occurs between this specific hybrid secondary metabolic pathway and primary metabolism . With the observed crosstalk , we focused our structural characterization efforts on the smaller number of molecular features that were dependent on the atypical E2-NRPS machinery . We also included pterin ( 4 , green node in Figure 3 ) as a standard in our network generation to aid in defining metabolites possessing pteridine structural scaffolds . Through this analysis , two prominent metabolites were selected for NMR-based structural characterization ( Figure 3 ) . 10 . 7554/eLife . 25229 . 009Figure 3 . Relative abundances of wild-type pathway-dependent metabolites in wild-type and mutant pepteridine pathways . The average ionization intensity is depicted for each molecular feature under a given genetic condition ( wild-type and Δplu2793 through Δplu2799 ) . Alterations in abundance are correlated with changes in nodal color intensity among genetic constructs , allowing visual assessment of product distributions for a given mutation . See Figure 3—figure supplement 2 for information on mass of pathway-dependent metabolites . Structural characterization of compounds 1 and 2 is shown in Figure 3—figure supplements 4–10 , Table 2 , and the Materials and methods section . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 00910 . 7554/eLife . 25229 . 010Figure 3—figure supplement 1 . Untargeted molecular networking of culture extracts from cells harboring the wild-type pepteridine biosynthetic pathway . The node with the green hue represents pterin ( 4 ) networked into the untargeted wild-type map . This network includes abundant molecules from the pathway in addition to primary metabolites and media components . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01010 . 7554/eLife . 25229 . 011Figure 3—figure supplement 2 . Pathway-targeted molecular networking of the wild-type pepteridine biosynthetic pathway . Comparative metabolomic analysis between control and experimental culture samples enabled targeting ( MS2 ) of molecular features dependent on the presence of the biosynthetic pathway . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01110 . 7554/eLife . 25229 . 012Figure 3—figure supplement 3 . LC/MS Extracted Ion Count ( EIC ) chromatograms of compounds 1 ( A ) and 2 ( B ) from butanol extracts of the pepteridine heterologous expression strain . LC/MS traces were extracted with m/z 224 corresponding to compound 1 ( A ) and m/z 210 corresponding to compound 2 ( B ) from butanol extracts of the E . coli BAP1 culture broths carrying the wild-type pepteridine pathway ( green ) or the empty vector ( pET28a , red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01210 . 7554/eLife . 25229 . 013Figure 3—figure supplement 4 . 1H and 13C NMR spectra of compound 1 . ( A ) 1H NMR spectrum of compound 1 in DMSO-d6 . ( B ) 13C NMR spectrum of compound 1 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01310 . 7554/eLife . 25229 . 014Figure 3—figure supplement 5 . 2D- ( gCOSY and gHSQCAD ) NMR spectra of compound 1 . ( A ) gCOSY NMR spectrum of compound 1 in DMSO-d6 . ( B ) gHSQCAD NMR spectrum of compound 1 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01410 . 7554/eLife . 25229 . 015Figure 3—figure supplement 6 . 2D- ( gHMBCAD and NOESY ) NMR spectra of compound 1 . ( A ) gHMBCAD NMR spectrum of compound 1 in DMSO-d6 . ( B ) NOESY NMR spectrum of compound 1 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01510 . 7554/eLife . 25229 . 016Figure 3—figure supplement 7 . 1H and 13C NMR spectra of compound 2 . ( A ) 1H NMR spectrum of compound 2 in DMSO-d6 . ( B ) 13C NMR spectrum of compound 2 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01610 . 7554/eLife . 25229 . 017Figure 3—figure supplement 8 . 2D- ( gCOSY and gHSQCAD ) NMR spectra of compound 2 . ( A ) gCOSY NMR spectrum of compound 2 in DMSO-d6 . ( B ) gHSQCAD NMR spectrum of compound 2 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01710 . 7554/eLife . 25229 . 018Figure 3—figure supplement 9 . 2D- ( gHMBCAD and NOESY ) NMR spectra of compound 2 . ( A ) gHMBCAD NMR spectrum of compound 2 in DMSO-d6 . ( B ) NOESY NMR spectrum of compound 2 in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 01810 . 7554/eLife . 25229 . 019Figure 3—figure supplement 10 . Amide conformation of compounds 1 and 2 . ( A ) Plausible cis- ( a ) and trans- ( b ) amide conformations of compound 1 ( top ) and interpretation of 2D NOESY spectral data of compound 1 ( bottom ) , ( B ) Plausible cis- ( a ) and trans- ( b ) amide conformations of compound 2 ( top ) and interpretation of 2D NOESY spectral data of compound 2 ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 019 Compounds 1 and 2 with molecular ions at m/z 224 . 1141 ( [M+H]+ , calcd m/z 224 . 1147 , C9H14N5O2 ) and m/z 210 . 0987 ( [M+H]+ , calcd m/z 210 . 0991 , C8H12N5O2 ) , respectively , appeared to be structurally related in the molecular network with a 14 Da mass difference likely attributable to CH2 . A bacterial culture in M9-minimal medium ( 6 l ) supplemented with casamino acids ( 5 g/l ) was initiated for structural characterization . LC/MS analysis of the accumulated butanol extracts led to detection of compounds 1 ( tR = 7 . 3 min ) and 2 ( tR = 5 . 8 min ) ( Figure 3—figure supplement 3 ) . However , initial isolation attempts proved challenging because of the low solubility and high polarity of these molecules , which co-eluted with the amino acid supplements . Consequently , isolation of 1 and 2 proceeded through larger scale cultivation in M9-minimal medium ( 32 l ) . Although a much lower yield was observed , the isolation was streamlined in the absence of amino acid supplements . The clarified medium was lyophilized , and the dried residue was extracted with 50% aqueous methanol ( 2 l ) . Flash column chromatography ( C18 ) followed by several rounds of reverse-phase liquid chromatography separation led to isolation of pure compounds 1 ( 0 . 8 mg ) and 2 ( 1 . 2 mg ) , which we named pepteridines A and B , respectively . Structural elucidation of pepteridine A ( 1 ) and B ( 2 ) was achieved through interpretation of 1D- ( 1H and 13C ) and 2D-NMR ( gCOSY , gHSQCAD , gHMBCAD , and NOESY ) spectral data ( Figure 3—figure supplements 4–9 and Table 2 ) . Briefly , 1H NMR spectral data coupled with gHSQCAD of 1 were used to show that three NH or NH2 exchangeable protons , three methylene protons , and one methyl signal were present , suggesting that compound 1 is composed of one hydrogenated pteridine scaffold and one distinguished acyl group . Interpretation of gCOSY and gHMBCAD NMR spectra established the two partial structures to be 2-amino-5 , 6 , 7 , 8-tetrahydropteridin-4 ( 3H ) -one and a propionyl group . Key HMBC correlations from the propionyl protons and the methylene protons at C-6 in the pteridine scaffold demonstrated the sharing of an amide carbonyl . These correlations established the connectivity of the propionyl group at N-5 , thereby characterizing 1 as a 2-amino-5-propionyl-5 , 6 , 7 , 8-tetrahydropteridin-4 ( 3H ) -one ( Figure 3 ) . Interpretation of NOESY NMR data and comparison of the proton chemical shifts at the pteridine C-6 to cis- and trans-amide analogs supported the cis-amide conformation of the acyl group at N-5 in solution ( Figure 3—figure supplement 10 ) ( Lanyon-Hogg et al . , 2015 ) . The critical difference between 1 and 2 was the absence of a CH2 signal in the NMR spectral data , supporting the assumption that 2 contains an acetyl group in place of the propionyl group in 1 , and further HMBC NMR analysis and HR-ESI-QTOF-MS data confirmed this . 10 . 7554/eLife . 25229 . 020Table 2 . NMR spectral data of pepteridine A ( 1 ) and B ( 2 ) in DMSO-d6 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 020Pepteridine A ( 1 ) No . δCatypeδHbMult ( J in Hz ) HMBC1N2154 . 6C3NH10 . 07br s4157 . 3C4a93 . 1C5N638 . 5CH24 . 54dd ( 12 . 2 , 3 . 6 ) C-1′ , C-4a2 . 33m741 . 9CH23 . 30d ( 12 . 2 ) 2 . 97dt ( 12 . 0 , 4 . 2 ) C-68NH6 . 96br sC-4a , C-6 , C-78a153 . 1C9N1′174 . 0C2′26 . 5CH22 . 57dt ( 15 . 0 , 7 . 4 ) C-1′ , C-3′2 . 15dt ( 14 . 8 , 7 . 3 ) C-1′ , C-3′3′9 . 7CH30 . 88t ( 7 . 4 ) C-1′ , C-2′NH26 . 25br sPepteridine B ( 2 ) No . δCatypeδHbMult ( J in Hz ) HMBC1N2155 . 2C3NH10 . 04br s4157 . 3C4a93 . 5C5N638 . 3CH24 . 52dd ( 12 . 1 , 3 . 5 ) C-1′ , C-4a2 . 32dt ( 11 . 7 , 2 . 4 ) 741 . 8CH23 . 30d ( 12 . 0 ) 2 . 98dt ( 11 . 9 , 4 . 1 ) C-68NH6 . 96d ( 4 . 0 ) 8a153 . 3C9N1′170 . 6C2′22 . 5CH31 . 97sC-1′NH26 . 22br sNMR spectra were recorded at b600 MHz for 1H NMR and a100 MHz for 13C NMR , respectively . Combining the bioinformatic , genetic , and comparative metabolomic analyses with the new structures suggests a hybrid biosynthetic route to the pepteridines , consisting of pteridine synthesis from GTP , acyl-synthesis via oxidative decarboxylation of α-keto acids , and NRPS-dependent condensation of these distinct substrates ( Figure 4 ) . The presence of predicted GTPCH Ι , 6-pyruvoyltetrahydropterin synthase , and pteridine reductase homologs in the pathway supports the formation and redox control of a tetrahydropterin substrate . The presence of the pyruvate-dehydrogenase-like E1 and E2-NRPS fusion enzymes supports derivation of the pepteridine acyl-appendages from α-keto acids through an analogous dehydrogenase mechanism . It is likely , however , that the pathway interacts with other pteridine and dehydrogenase biosynthetic enzymes encoded in the genome of the heterologous host , E . coli , such as the E3 subunit of the pyruvate dehydrogenase complex , which is required for lipoamide regeneration but absent from the pathway . Rather than producing an acyl-CoA , as in the pyruvate dehydrogenase reaction , the CoA-derived carrier protein arm ( T-domain ) could be directly primed in place of CoA ( Figure 4 ) . Analogous carrier protein priming mechanisms have been proposed for branched chain fatty acid substrate utilization in the formation of N-acyl-amides and the pristinamycin IIa streptogramin antibiotic ( Craig and Brady , 2011; Brachmann et al . , 2012 ) ; and for a glycolicacyl-NRPS extender unit in formation of the naphthyridinomycin antitumor antibiotic ( Peng et al . , 2012 ) . In contrast to these pathways , we propose that the resulting loaded acyl-carrier protein in pepteridine biosynthesis would be unusually condensed with a free tetrahydropterin substrate by the atypical NRPS C domain to install the cis-amide acyl-linkage . It is currently unclear whether the C domain catalyzes direct cis-amide bond formation or whether a trans-amide is formed and then isomerized to the observed cis-conformation in the pepteridines . 10 . 7554/eLife . 25229 . 021Figure 4 . Proposed pepteridine biosynthesis . α-Ketobutyrate is processed by the atypical dehydrogenase-NRPS-like complex to generate a propionyl-loaded carrier protein ( blue ) . Pteridine enzymes interact with the metabolism of the host to generate tetrahydropterin ( green ) . The NRPS C domain couples these substrates to form the cis-amide bond ( red ) as illustrated for 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 021 To gain additional support for our proposed biosynthesis , we analyzed pepteridine production in our full series of pathway mutant strains . Differential LC/HR-ESI-QTOF-MS analysis of butanol extracts demonstrated that pepteridine A ( 1 ) detection was completely abolished in the GTPCH Ι homolog mutant ( Δplu2793 ) , the 6-pyruvoyl-tetrahydropterin synthase homolog mutant ( Δplu2794 ) , and the atypical E2-NRPS mutant ( Δplu2796 ) , indicating a genetic requirement for both pteridine and NRPS biosynthetic machineries ( Figure 5A ) . Production of pepteridine B ( 2 ) was substantially reduced in the Δplu2793 and Δplu2794 strains , and completely abolished in the Δplu2796 mutant strain ( Figure 5C ) . The minor residual production of 2 in these specific pteridine knockout strains is consistent with our observation that pterin ( 4 ) substrates can be detected at low abundance in the control strain ( pET28a ) , further highlighting the metabolic crosstalk between this pathway and primary metabolism . To further demonstrate the biosynthetic dependence on the NRPS machinery , a Plu2796 S434A point mutant was generated in the wild-type pathway construct to site-specifically inactivate the NRPS carrier protein domain . Assessment of pepteridine production from this construct in comparison with wild-type , Δplu2796 , and control constructs confirmed dependency on the NRPS machinery ( Figure 5B , D ) . 10 . 7554/eLife . 25229 . 022Figure 5 . Gene deletion and NRPS inactivation analyses on 1 ( A–B ) and 2 ( C–D ) . ( A–B ) Extracted ion chromatograms of 1 for wild-type and mutants are shown . ( C–D ) Extracted ion chromatograms of 2 for wild-type and mutants are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 022 We then supplemented bacterial cultures harboring the wild-type pathway with varying concentrations of free α-ketobutyrate and pyruvate and , as expected , observed enhanced production of the respective pepteridines in a dose-dependent manner ( Figure 6A , B and Figure 6—figure supplement 1 ) . Additionally , universally 13C-labeled α-ketobutyrate ( 13C4 ) supplementation led to 13C3-labeling of pepteridine A ( 1 ) , as determined by HR-ESI-QTOF-MS ( m/z 227 . 1244 ( [M+H]+ , calcd m/z 227 . 1248 , 13C312C6H14N5O2 ) , further supporting the proposed biosynthesis ( Figure 6C and Figure 6—figure supplement 2 ) and providing a forward path for protein biochemistry studies on this unprecedented enzymatic system . Importantly , these studies link pepteridine production to α-ketoacid substrate availability ( i . e . , pyruvate and α-ketobutyrate ) rather than free acyl-CoA substrates ( i . e . , acetyl- and propionyl-CoA ) . 10 . 7554/eLife . 25229 . 023Figure 6 . α-Ketobutyrate feeding studies on 1 . ( A–B ) Production enhancement of 1 in α-ketobutyrate supplementation studies ( Arel , relative integration value ) . ( C ) 13C4-α-ketobutyrate supplementation leads to 13C3 incorporation in 1 consistent with the proposed biosynthesis . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 02310 . 7554/eLife . 25229 . 024Figure 6—figure supplement 1 . Production of compounds 1 and 2 in dose-dependent α-ketobutyrate ( A ) and pyruvate ( B ) feeding studies . HPLC/MS traces were extracted with m/z 224 corresponding to compound 1 ( A ) and m/z 210 corresponding to compound 2 ( B ) from butanol extracts of the culture broths fed with varying concentrations of α-ketobutyrate ( A ) and pyruvate ( B ) . Dose-response plots for 1 and 2 were determined using extracted peak integration values . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 02410 . 7554/eLife . 25229 . 025Figure 6—figure supplement 2 . Characterization of 13C4-α-ketobutyrate incorporation in compound 1 . Incorporation of 13C4-α-ketobutyrate into compound 1 ( B ) led to a clear 3 Da mass shift of native 1 ( A ) via oxidative decarboxylation of α-ketobutyrate . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 025 Photorhabdus bacteria are both pathogens to insects and mutualists to a specific nematode host ( Clarke , 2008; Waterfield et al . , 2009; Clarke , 2014 ) . Photorhabdus asymbiotica can also cause infections in humans ( Gerrard et al . , 2004 ) . To enhance their fitness for these variable host-bacteria objectives , Photorhabdus bacteria undergo phenotypic variation , which is controlled by a stochastic invertible promoter switch ( Somvanshi et al . , 2012 ) . The orientation of the promoter regulates the formation of P- and M-form phenotypic variants . The P-form , typically the dominant variant in wild-type cultures , is pathogenic to insects . The P-form switches to the M-form , a small colony variant , which adheres to specific cells in the nematode intestine . It is thought that the M-form phenotype participates in colonization of its mutualistic nematode host . To determine whether the pepteridines are produced in P . luminescens and in which variant , we analyzed butanol extracts of P . luminescens genetically ‘locked’ in the M- and P-forms using LC/HR-ESI-QTOF-MS . In the M9-base medium , both pepteridines A and B could be detected in the pathogenic P-form phenotypic variant ( Figure 7 ) . However , under identical conditions , no production was observed in the M-form . These studies link pepteridine structure to phenotypic variant status and suggest that the pepteridines may participate in P-form biological activities . 10 . 7554/eLife . 25229 . 026Figure 7 . Extracted ion counts LC/HR-ESI-QTOF-MS analysis of pepteridines A ( left panel ) and B ( right panel ) from two genetically engineered P . luminescens strains locked in the phenotypic variants M-form ( top ) and P-form ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 02610 . 7554/eLife . 25229 . 027Figure 7—figure supplement 1 . Extracted ion counts chromatograms from LC/HR-ESI-QTOF-MS analysis of pepteridines A ( left panel ) and B ( right panel ) from the butanol extracts of P-form culture broth ( top ) , standard compounds ( middle ) , and co-injection ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 027 To initiate functional cellular studies , we deleted the pepteridine genetic locus in P . luminescens TT01 on a wild-type background ( Δlocus ) and a ΔhexA background ( ΔhexA/locus ) using allelic-exchange mutagenesis . HexA is a LysR-type transcriptional repressor that regulates stilbene production and participates in the insect pathogen , nematode mutualist transition ( Joyce and Clarke , 2003; Kontnik et al . , 2010 ) , and global transcriptomic analysis indicates that pepteridine genes are upregulated in ΔhexA/hfq strains ( Tobias et al . , 2016 ) . We cultivated the wild-type and mutant strains in a culture medium based on the high concentrations of free proteinogenic amino acids found in insect hemolymph ( Crawford et al . , 2010 ) . Twenty-four hour cultures were centrifuged , the cells were rapidly lysed , and the protein fractions were trypsinized for quantitative proteomic analysis . Relative fold changes ( wild-type vs Δlocus and ΔhexA vs ΔhexA/locus ) were calculated based on quantitative LC/MS-MS analyses ( Figure 8 ) . On a wild-type background , deletion of the locus had little effect on the proteome ( Figure 8A ) . However , on a ΔhexA background and consistent with the transcriptomic data , more dramatic proteomic effects were observed , supporting regulation of the pepteridine genomic locus by HexA ( Figure 8B ) . Proteins that participate in pyrone quorum sensing and secondary metabolism were downregulated in the ΔhexA/locus mutant relative to the ΔhexA control . For example , deletion of the pepteridine genetic locus led to a 5-fold decrease of the enzyme Plu4844 in the ΔhexA background ( Figure 8B ) . Plu4844 participates in synthesis of pyrone autoinducers ( Brachmann et al . , 2013 ) . We similarly observed a 3–4-fold decrease of the enzymes Plu2817 , Plu2204 , and Plu4187 . These enzymes are involved in biosynthesis of the potent phenoloxidase inhibitor rhabduscin ( Plu2817 ) ( Crawford et al . , 2010 , 2012 ) ; cinnamic acid ( Plu2204 , 2207 , 2208 ) , a key substrate of the multipotent stilbenes ( Joyce et al . , 2008 ) ; and the polyketide anthraquinone pigments ( Plu4186 , 4187 , 4188 , 4192 ) ( Brachmann et al . , 2007 ) . Collectively , our proteomic data support that the pepteridine genomic locus is regulated by HexA and positively affects pyrone quorum sensing and select secondary metabolic pathways . 10 . 7554/eLife . 25229 . 028Figure 8 . Quantitative proteomic analysis of a Δlocus strain in a wild-type background ( A ) and ΔhexA background ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 02810 . 7554/eLife . 25229 . 029Figure 8—source data 1 . Proteins increased in WT vs . WTΔlocus strains by LC-MS/MS . Intensities from label-free quantification calculated using MaxQuant , averaged for biological triplicate samples , then log2 transformed . Proteins presented exhibited ≥2 fold increased average intensity in WT compared with WTΔlocus . Two-tailed t-test performed using a cutoff of FDR = 0 . 01 and 2-fold signal difference , or p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 02910 . 7554/eLife . 25229 . 030Figure 8—source data 2 . Proteins increased in WTΔlocus vs . WT strains by LC-MS/MS . Intensities from label-free quantification calculated using MaxQuant , averaged for biological triplicate samples , then log2 transformed . Proteins presented exhibited ≥2 fold increased average intensity in WTΔlocus compared with WT . Two-tailed t-test performed using a cutoff of FDR = 0 . 01 and 2-fold signal difference , or p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 03010 . 7554/eLife . 25229 . 031Figure 8—source data 3 . Proteins increased in ΔhexA vs . ΔhexAΔlocus strains by LC-MS/MS . Intensities from label-free quantification calculated using MaxQuant , averaged for biological triplicate samples , then log2 transformed . Proteins presented exhibited ≥2 fold increased average intensity in ΔhexA compared with ΔhexAΔlocus . Two-tailed t-test performed using a cutoff of FDR = 0 . 01 and 2-fold signal difference , or p<0 . 05 ( + indicates statistical significance ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 03110 . 7554/eLife . 25229 . 032Figure 8—source data 4 . Proteins increased in ΔhexAΔlocus vs . ΔhexA strains by LC-MS/MS . Intensities from label-free quantification calculated using MaxQuant , averaged for biological triplicate samples , then log2 transformed . Proteins presented exhibited ≥2 fold increased average intensity in ΔhexAΔlocus compared with ΔhexA . Two-tailed t-test performed using a cutoff of FDR = 0 . 01 and 2-fold intensity difference , or p<0 . 05 ( + indicates statistical significance ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 03210 . 7554/eLife . 25229 . 033Figure 8—source data 5 . All proteins observed in WT and WTΔlocus strains by LC-MS/MS . Intensities from label-free quantification calculated using MaxQuant and log2 transformed . Intensities for biological triplicate samples are presented . Average fold change shown for WT compared with WTΔlocus . Two-tailed t-test performed using a cutoff of FDR = 0 . 01 and 2-fold signal difference , or p<0 . 05 ( + indicates statistical significance ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 03310 . 7554/eLife . 25229 . 034Figure 8—source data 6 . All proteins observed in ΔhexA and ΔhexAΔlocus strains by LC-MS/MS . Intensities from label-free quantification calculated using MaxQuant and log2 transformed . Intensities for biological triplicate samples are presented . Average fold change shown for ΔhexA compared with ΔhexAΔlocus . Two-tailed t-test performed using a cutoff of FDR = 0 . 01 and 2-fold signal difference , or p<0 . 05 ( + indicates statistical significance ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25229 . 034 Bacterial natural products represent a rich source of lead structures in small molecule drug discovery efforts and serve as excellent molecular probes in biology . Novel classes of putative biosynthetic enzymes , including enzymes belonging to new types of “hybrid" metabolic pathways , can be identified in expanding genome sequence databases through genome mining approaches . Integration of two or more distinct types of metabolic pathways provides combinatorial biosynthetic routes to expand the structural and functional properties of natural products . In this study , using genome synteny analysis , we identified an unprecedented genomic island in P . luminescens that harbored both NRPS-like and pteridine biosynthetic machineries . This hybrid biosynthetic pathway is responsible for synthesis of new pteridine metabolites , the pepteridines . Pepteridine A ( 1 ) and B ( 2 ) , which respectively harbor atypical cis-amide functionalized C3 and C2 acyl-chains off a free pterin substrate , a particularly unusual feature for a NRPS , were extensively characterized by NMR . The genetic determinants of pepteridine production were assessed in a heterologous host and networked at the systems level , providing global pathway-dependent maps to visualize relative metabolite distributions among wild-type and mutant strains . The pepteridine metabolites were also identified in the wild-type host , P . luminescens , specifically associated with its phenotypic variant linked to pathogenesis ( the P-form ) . Quantitative proteomic analysis further demonstrated that when the LysR-type transcriptional repressor is derepressed ( ΔhexA ) , the pepteridine genetic locus positively affects pyrone qurourm sensing and select secondary metabolic pathways ( i . e . , these pathways were downregulated in the ΔhexA/locus strain relative to the ΔhexA strain ) . The pepteridines likely participate in chemical signaling . Indeed , in the Gram-negative plant pathogen Agrobacterium tumefaciens , pterin signaling was recently implicated in biofilm regulation ( Feirer et al . , 2015 ) . The pepteridines discovered here by targeting atypical genomic sequence space , to the best of our knowledge , represent the first metabolites to be characterized from a hybrid NRPS-pteridine biosynthetic gene cluster . Characterization of these types of atypical hybrid pathways expands our view on the combinatorial biosynthetic potential available in nature’s metabolic toolbox .
The genomic island spanning genes plu2793 to plu2799 from Photorhabdus luminescens TT01 were amplified via PCR using Phusion High-Fidelity DNA Polymerase ( New England Biolabs ( NEB ) , USA ) and primers 2796-cluster-5 and 2796-cluster-3 ( Supplementary file 1B ) . Reactions ( 50 µl ) were assembled according to the manufacturer’s protocol with inclusion of 5% ( v/v ) dimethyl sulfoxide and the use of 1 µl of confluent P . luminescens TT01 as template . Thermal cycling was carried out on a C1000 Touch thermal cycler equipped with a Dual 48/48 Fast Reaction Module ( Bio-Rad , USA ) . Amplification was assessed using a 0 . 75% agarose gel in 1 × TAE ( 40 mM Tris pH 7 . 6 , 20 mM Acetic Acid , and 1 mM EDTA ) stained with GelGreen ( Biotium , USA ) . Amplification products were purified using the QIAquick PCR Purification Kit ( Qiagen , USA ) according to the manufacturer’s protocol . The purified product and pET24b ( EMDMillipore - Novagen , USA ) were individually digested with NdeI and XhoI ( NEB ) , and purified using the QIAquick PCR Purification Kit . Ligation of the products was carried out using T4 DNA ligase ( NEB ) . Ligation mixtures were directly transformed into 50 µl of MAX Efficiency DH5α chemically competent cells ( Invitrogen , USA ) and recovered in 200 µl super optimal broth with catabolite repression ( SOC , 2% ( w/v ) tryptone , 0 . 5% ( w/v ) yeast extract , 10 mM NaCl , 2 . 5 mM KCl , 10 mM MgCl2 , 10 mM MgSO4 , and 20 mM glucose ) following the manufacturer’s protocol . Successful transformants were selected by plating 150 µl of the transformation outgrowth onto lysogeny broth ( LB ) agar ( BD , USA; 1% ( w/v ) tryptone , 0 . 5% ( w/v ) yeast extract , 1% ( w/v ) NaCl , and 1 . 5% ( w/v ) agar ) plates supplemented with 25 µg/ml kanamycin ( American Bioanalytical , USA ) and overnight growth at 37°C . Single , well-defined colonies were picked and grown overnight as suspension cultures at 37°C and 250 rpm in 5 ml LB ( BD; 1% ( w/v ) tryptone , 0 . 5% ( w/v ) yeast extract , and 1% ( w/v ) NaCl ) supplemented with 25 µg/ml kanamycin . Plasmids were harvested using the QIAprep Spin Miniprep Kit ( Qiagen ) according to the manufacturer’s protocol . Incorporation of the biosynthetic pathway encompassing genes plu2793-plu2799 was assessed by end sequencing using the T7 promoter and T7 terminator primers ( Genewiz , USA; or Keck Foundation Biotechnology Resource Laboratory at Yale University , USA ) . Primer walking fully validated the inserted genetic sequence along with the plasmid’s transcriptional and translational control elements proximal to the region of insertion . The construct was named pEplu2796 . Sequencing primers are listed in Supplementary file 1C . Escherichia coli BAP1 ( Pfeifer et al . , 2001 ) chemically competent cells were prepared according to standard molecular biology protocols and transformed using 1 µl of the desired construct . Transformants were selected via overnight growth at 37°C on LB agar plates supplemented with 25 µg/ml kanamycin . Resulting single colonies were picked for overnight liquid culture at 37°C and 250 rpm in LB media supplemented with 25 µg/ml kanamycin . Glycerol stocks were subsequently prepared for long-term storage . The E . coli BAP1 strain carrying the selected pathway was spread onto an LB agar plate supplemented with 25 µg/ml kanamycin and incubated at 37°C overnight . A single colony was inoculated into 5 ml LB liquid media containing 25 µg/ml kanamycin and grown overnight at 37°C and 250 rpm . 5 ml of M9 minimal medium ( Amresco , USA; 0 . 6% ( w/v ) Na2HPO4 , 0 . 3% ( w/v ) KH2PO4 , 0 . 05% ( w/v ) NaCl , 0 . 1% ( w/v ) NH4Cl , 0 . 2% ( w/v ) glucose , 2 mM MgSO4 , and 0 . 1 mM CaCl2 ) supplemented with 5 g/l casamino acids ( Amresco ) and 25 µg/ml kanamycin was inoculated at 1:200 with the overnight growth , and grown at 37°C and 250 rpm until the optical density ( OD600 ) reached 0 . 5–0 . 6 . Growth was slowed by placing the culture on ice . Production was induced via the addition of isopropyl-1-thio-β-D-galactopyranoside ( IPTG , American Bioanalytical ) at a final concentration of 0 . 1 mM and growth proceeded at 25°C and 250 rpm . After 72 hr , the culture was centrifuged ( 2000 × g , 20 min , 4°C ) and the supernatant was then extracted 1:1 with water-saturated butanol ( 1 × 5 ml ) . The organic fraction was dried under reduced pressure on a Genevac ( USA ) HT-4X evaporation system for 2 hr . The crude material , which was re-suspended in 100 µl of 50% aqueous methanol , was analyzed via single quadrapole LC/MS ( Column: Phenomenex Kinetex C18 ( 100 Å ) 5 µm ( 4 . 6 × 250 mm ) column; Flow rate: 0 . 7 ml/min; Mobile phase composition: water:acetonitrile ( ACN ) gradient solvent system containing 0 . 1% formic acid: 0–30 min , 5–100% ACN; hold for 5 min , 100% ACN; 0 . 1 min , 100–5% ACN; hold for 1 . 9 min , 5% ACN; 6 . 1 min re-equilibration post-time , 5% ACN ) . All of the genes in the pepteridine gene cluster were individually deleted using PCR to yield a complete series of clean deletion constructs designed to minimize polar effects on the operon and preserve necessary internal transcriptional and translational elements . PCR reactions ( 50 µl ) were prepared using the pEplu2796 construct DNA as a template ( ~120 ng/µl ) . The primer pairs used for these reactions are listed in Supplementary file 1B ( e . g . , Delplu2793F and Delplu2793R - primer pair for the creation of a Δplu2793 construct ) . Primers were designed according to established protocols ( Liu and Naismith , 2008 ) . Briefly , each primer comprised two regions: one 3’ region , which binds immediately after the genetic region to be deleted ( bold-faced ) , and one 5’ region , which binds immediately before the genetic region to be deleted . Standard molecular biology protocols as described above were employed . Deletion of the desired gene was validated by sequencing using the sequencing primer ( Supplementary file 1C ) immediately upstream of the deleted region ( Δplu2793 , pET28aUpstream; Δplu2794 , Seq T7P*1; Δplu2795 , Seq 2; Δplu2796 , Seq 6; Δplu2797 , Seq 11; Δplu2798 , Seq 14; Δplu2799 , Seq 15 ) . Constructs possessing the desired deletion sequences were then fully sequence validated . Comparative metabolomic analyses and pathway-targeted molecular networking were performed following established workflows with minimal modifications , detailed subsequently ( Figure 3 , Figure 3—figure supplements 1 and 2 , Table 1 and Supplementary file 1A ) ( Vizcaino et al . , 2014a; Vizcaino and Crawford , 2015 ) . Biosynthetic pathway expression was carried out in M9 minimal media supplemented with 5 g/l casamino acids , 0 . 5 g/l L-phenylalanine ( Sigma-Aldrich , USA ) , and 25 µg/ml kanamycin . Expression cultures were inoculated at 1:1000 from stationary phase cultures grown overnight . Seven expression cultures were prepared . Six of these were biological replicates and one was a technical replicate that was solely used to monitor growth . As each sample set reached the desired OD600 of 0 . 5–0 . 6 , it was briefly placed on ice to cool . Once all sample sets had reached density and had sufficiently cooled , expression was then induced with 0 . 1 mM IPTG , and the culture sets were allowed to grow for 72 hr at 25°C and 250 rpm . The cell mass was then pelleted via centrifugation at 2000 × g for 20 min at 4°C . The supernatant from each sample was collected ( ~5 ml ) and subsequently extracted with 6 ml of water-saturated butanol . The organic layer was collected and dried under reduced pressure on a Genevac HT-4X evaporation system . Dried extracts were stored under nitrogen at −80°C until use . Samples were prepared for HR-ESI-QTOF-MS analysis by resuspension in 200 µl methanol ( LC-MS grade , Fisher , USA ) . Insoluble debris was removed by centrifugation at 20 , 000 × g for 5 min and 50 µl of the supernatant was placed in an HPLC vial for analysis . The remaining sample was dried and stored as previously described for future analysis . The Agilent iFunnel 6550 QTOF system was used for sample analysis . 2 µl of sample was injected and analyzed at 25°C and 0 . 7 ml/min on a Phenomenex Kinetex C18 ( 100 Å ) 5 µm ( 4 . 6 × 250 mm ) column with a water:ACN gradient solvent system containing 0 . 1% formic acid: 0–30 min , 5–100% ACN; hold for 5 min , 100% ACN; 0 . 1 min , 100–5% ACN; hold for 1 . 9 min , 5% ACN; 6 . 1 min re-equilibration post-time , 5% ACN . Mass spectra were acquired in the range of 25–1700 m/z at a scan rate of 1 spectra/s using Dual Agilent Jet Stream ( AJS ) ESI in positive mode . Source parameters were set as follows: drying gas temp , 225°C; drying gas flow , 12 l/min; nebulizer pressure , 35 psig; sheath gas temperature , 275°C; sheath gas flow , 12 l/min; fragmentor voltage , 125 V; skimmer voltage , 65 V; OCT 1 RF Vpp , 750 V; capillary voltage , 3500 V; nozzle voltage , 1000 V . Data were acquired and analyzed using MassHunter Workstation Data Acquisition ( Version B . 05 . 01 , Build 5 . 01 . 5125 . 1 , Agilent Technologies ) and MassHunter Qualitative Analysis ( Version B . 06 . 00 , Agilent ) , respectively . Comparison of the six biological replicates for each sample set allowed for selection of the five with the highest overall chromatographic similarity for further processing . Molecular feature extraction was performed with the following parameter alterations: peak spacing tolerance , 0 . 0025 m/z plus 7 ppm; limit assigned charge states to a maximum of , 1; restrict charge states to , 1 . In importing data into Mass Profiler Professional ( MPP , Agilent ) , ‘minimum number of ions’ was set to 1 . In performing the initial abundance analysis on the samples , a minimal normalized abundance of 18 was set . For controls , a molecular feature found in any one of the five samples was tabulated . Conservatively for wild-type , all five biological replicates had to possess the molecular feature for it to be tabulated . To ensure that all molecular features present in mutant strains throughout the five biological replicates were tabulated , a feature was tabulated if found in any one of the five biological replicates . Molecular features found in control samples were removed from both the wild-type and individual mutant feature lists . Wild-type pathway molecular features were subsequently compared withmutant pathway molecular features . Features present in the wild-type and absent in a given mutant were deemed to be dependent on that given mutant . An inclusion list was generated around these features for subsequent pathway-targeted tandem MS ( MS2 ) analysis . MS2 analysis was run following the previously described acquisition method with some changes: nebulizer pressure , 50 psig; fragmentor voltage , 200 V . Auto MS2 data collection was used with spectra being acquired at 1 spectra/s in the mass range of 25–1700 m/z . Fixed collision energies were set at 0 , 25 , 40 , 45 , 50 , and 100 V . Auto MS2 precursors were limited to those determined to be fully pathway-dependent or dependent on a given enzyme in the pathway . A maximum of 20 precursor ions were analyzed per cycle and a minimum precursor threshold was set at 10 , 000 counts ( absolute ) or 0 . 01% ( relative ) . Isotope models were inactivated and scan speed was varied based on precursor abundance with a target of 25 , 000 counts/spectrum . The MS2 accumulation time limit was employed . All target masses were allowed within a 10 ppm range and a 0 . 5 min retention time margin of error . Additionally , Auto MS2 was used without a preferred ion list on the wild-type sample to establish an untargeted molecular network . The collision energy dependence of the fragmentation patterns of select molecular features was assessed , and we determined that 40 V was optimal for fragmentation of the pepteridines . After data conversion to the mzXML file format , each file was edited to remove non-optimal collision energies while maintaining the preferred 40 V collision energy . This was done as the presence of multiple fragmentation patterns for a single molecular feature can result in the presence of said feature multiple times in the molecular network . In establishing molecular networks through the Global Natural Products Social Molecular Networking Platform ( GnPS , http://gnps . ucsd . edu ) , the following parameters were varied from their default values: parent mass tolerance , 0 . 001 Da; ion tolerance , 0 . 5 Da; min cos , 0 . 5; all filtering was disabled and filter below STD DEV and min peak intensity were both set to 0 . Cytoscape ( Cytoscape Consortium , USA ) was used to visualize and edit the final networks . Parent ion masses networked in GnPS that did not match those present on our initial inclusion list based on precise retention time and high-resolution mass data , were removed from the network . Additionally , all features present in the wild-type inclusion list were validated against the control . Any features that were discovered to be false positives were subsequently removed from the network . When single nodes lacking connectivity were generated during this process , these nodes were removed from the network . The charge state for each feature was evaluated manually against the computationally generated charge state and any inaccuracies were corrected . For all networks , the thickness of the edges connecting nodes is representative of the strength of interaction between the nodes , with thicker lines denoting stronger interactions . The minimum thickness was set at 2 pts representing a cosine score of 0 . 5 while the thickest possible edge possessing a cosine score of 1 was set to 15 pts . To decouple the NRPS machinery from that of the fused dehydrogenase E2 subunit in Plu2796 , a point mutation in the wild-type pathway was generated ( pEplu2796-S434A ) . In NRPS logic , the biosynthetic pathway intermediates are ferried through the various rounds of enzymatic modification via attachment to a phosphopantetheine arm . This phosphopantetheine tether is post-translationally coupled to a conserved Ser residue . Removal of this Ser destroys carrier protein activity . Bioinformatic assessment of Plu2796 demonstrated a phosphopantetheinyl binding site in the T domain . This conserved Ser was identified as S434 by pattern matching with known amino acid motifs indicative of a phosphopantetheinyl attachment site . Using the mutagenesis strategy outlined in ‘Construction of scar-less deletion constructs’ with the modifications detailed subsequently , S434 was mutated to A434 . This mutation prevents the attachment of phosphopantetheine thereby inactivating the NRPS functionality of the Plu2796 enzyme . The primers PluS434AF and PluS434AR were used in the PCR reactions . The wild-type pEplu2796 vector was used as a template . Successful point mutagenesis was assessed via sequencing using the Seq 8 primer ( Supplementary file 1C ) . The pEplu2796-S434A construct was then fully sequence validated over the region harboring the biosynthetic pathway . This construct was transformed into E . coli BAP1 ( see ‘Preparation of constructs for heterologous expression studies’ ) . Subsequently , LC-HR-ESI-QTOF-MS analysis was performed on this construct in comparison with the wild-type ( pEplu2796 ) , negative control ( pET28a ) , and Δ2796 constructs according to the instrument parameters and chromatography method outlined in ‘Comparative metabolomic profiling and molecular networking . ’ Comparison of extracted ion chromatograms demonstrated the dependence of pepteridine A and B on the NRPS thiolation domain present in Plu2796 . Extracted ion chromatograms of the S434A mutant match with those for the full deletion of Plu2796 . All experiments were conducted in triplicate . A 5 ml LB liquid culture supplemented with 25 µg/ml kanamycin was initiated by inoculation of a single colony of the E . coli BAP1 strain carrying the wild-type pepteridine construct , pEplu2796 . After overnight growth at 37°C and 250 rpm , the culture was used to seed additional 32 × 5 ml LB cultures , which were further incubated at 37°C and 250 rpm for 18 hr . Each of the 32 , 5 ml cultures was used to inoculate one of 32 , 1 l cultures containing M9 minimal medium supplemented with 25 µg/ml kanamycin . These cultures were incubated at 37°C and 250 rpm until the OD600 reached 0 . 5–0 . 6 . Pathway expression was initiated via the addition of 0 . 1 mM IPTG , and the cultures were further incubated at 25°C and 250 rpm for 72 hr . The whole 32 l culture volume was centrifuged at 14 , 000 g and 4°C for 30 min , and the supernatant was lyophilized ( ~7 days ) . The dried sample was extracted with a total of 2 l of 50% aqueous methanol , filtered , and evaporated under reduced pressure to yield the crude material ( approximately 4 . 0 g ) . The crude organic extract ( 4 . 0 g ) was subjected to flash C18 column chromatography ( 300 g ) with a step gradient elution ( 0% , 10% , 20% , 40% , 60% , and 100% methanol in water ) . The 20% methanol fraction was further fractionated using an Agilent Prepstar HPLC system ( Agilent Polaris C18-A 5 µm ( 21 . 2 × 250 mm ) column ) with a linear gradient elution ( 5–50% methanol in water over 60 min , 8 ml/min , 1 min fraction collection window ) . A combined fraction ( 27 + 28 ) was subsequently isolated by reversed-phase HPLC ( Phenomenex Luna C18 C8 ( 2 ) ( 100 Å ) 10 µm ( 10 . 0 × 250 mm ) column ) with a linear gradient elution ( 5–100% ACN in water over 60 min , 4 ml/min ) to yield impure compound 1 . Compound 1 ( tR = 19 . 85 min , 0 . 8 mg ) was then purified by reversed phase C8-HPLC ( Phenomenex Luna C8 ( 2 ) ( 100 Å ) 10 µm ( 10 . 0 × 250 mm ) column ) with isocratic separation ( 10% methanol in water , 2 ml/min ) , followed by final purification over an Agilent Phenyl-Hexyl 5 µm ( 9 . 4 × 250 mm ) column using isocratic separation ( 5% ACN in water , 2 ml/min ) to yield 0 . 8 mg of pure 1 . For compound 2 , the 10% methanol fraction from the flash C18 column chromatography step was separated on a semi-preparative reversed-phase HPLC system ( Agilent Polaris C18-A 5 µm ( 21 . 2 × 250 mm ) column ) using isocratic separation ( 5% ACN in water with 0 . 01% TFA , 8 ml/min , 1 min fraction collection window ) . A pooled fraction ( 13 + 14 ) was then isolated by reversed-phase HPLC ( Phenomenex Luna C18 ( 2 ) ( 100 Å ) 5 µm ( 4 . 6 × 150 mm ) column ) using an isocratic solvent system ( 5% ACN in water over 30 min , 2 ml/min ) to afford impure compound 2 ( tR = 15 . 86 min ) . Final purification of compound 2 ( 1 . 2 mg , tR = 6 . 51 min ) was carried out over an Agilent Phenyl-Hexyl 5 µm ( 9 . 4 × 250 mm ) column using isocratic separation ( 5% ACN in water , 2 ml/min ) . Additionally , a combined fraction ( 10 + 11 ) containing compound 3 was further isolated by reversed-phase HPLC ( Phenomenex Luna C8 ( 2 ) ( 100 Å ) 10 µm ( 10 . 0 × 250 mm ) column ) using an isocratic solvent system ( 10% ACN in water over 30 min , 2 ml/min ) . Compound 3 ( tR = 9 . 5 min ) was then purified via an Agilent Phenyl-Hexyl 5 µm ( 9 . 4 × 250 mm ) column eluting with an isocratic solvent system ( 3% methanol in water , 2 ml/min ) to yield 3 mg of pure material . Pepteridine A ( 1 ) : white powder; UV ( CH3OH ) λmax ( log ε ) 282 ( 3 . 68 ) , 222 ( 3 . 85 ) nm; 1H and 13C NMR spectra , see Table 2; HR-ESI-QTOF-MS [M+H]+m/z 224 . 1141 ( calcd for C9H14N5O2 , 224 . 1147 ) . Pepteridine B ( 2 ) : white powder; UV ( CH3OH ) λmax ( log ε ) 282 ( 3 . 48 ) , 222 ( 3 . 75 ) nm; 1H and 13C NMR spectra , see Table 2; HR-ESI-QTOF-MS [M+H]+m/z 210 . 0985 ( calcd for C8H12N5O2 , 210 . 0991 ) . 7 , 8-Dihydroxanthopterin ( 3 ) : white powder; UV ( CH3OH ) λmax ( log ε ) 310 ( 2 . 99 ) , 272 , ( 3 . 25 ) , 222 ( 3 . 54 ) nm; 1H NMR ( DMSO-d6 , 600 MHz ) δ 10 . 32 ( 1H , br s , 3-NH ) , 8 . 80 ( 1H , br s , 5-NH ) , 6 . 60 ( 1H , br s , 8-NH ) , 6 . 14 ( 2H , br s , 2-NH2 ) , 3 . 73 ( 2H , s , H-7 ) ; 13C NMR ( DMSO-d6 , 100 MHz ) δ 161 . 9 ( C-6 ) , 154 . 4 ( C-4 ) , 152 . 0 ( C-2 ) , 150 . 7 ( C-8a ) , 93 . 4 ( C-4a ) , 46 . 0 ( C-7 ) ; HR-ESI-QTOF-MS [M+H]+m/z 182 . 0631 ( calcd for C6H8N5O2 , 182 . 0678 ) . Pterin ( 4 ) : pale yellow powder; UV ( CH3OH ) λmax ( log ε ) 330 ( 2 . 48 ) , 272 ( 3 . 26 ) , 226 ( 3 . 46 ) nm; HR-ESI-QTOF-MS [M+H]+m/z 164 . 0574 ( calcd for C6H6N5O , 164 . 0572 ) . The chemical structures of metabolites 1–4 were identified by analyses of NMR and HR-ESI-QTOF-MS data; and spectral comparison with validated standards . The structure of compound 3 , which was elucidated by 1D- and 2D- ( gCOSY and gHMBCAD ) NMR experiments , was identified as the 7 , 8-dihydroxanthopterin ( 3 ) and supported by comparing its NMR spectral data with a commercial standard of 3 . The structure of pterin ( 4 ) was unambiguously characterized by HR-ESI-QTOF-MS data , HPLC co-injection with standard pterin , and by comparison of the UV absorption spectral data with that of a pterin commercial standard . Compound 1 was isolated as a white powder . The molecular formula was assigned as C9H13N5O2 ( [M+H]+ at m/z 224 . 1141 ) based on HR-ESI-QTOF-MS spectroscopic data . The 1H NMR spectral data recorded in DMSO-d6 displayed two NH protons [δH 10 . 07 ( 1H , br s ) , 6 . 96 ( 1H , br s ) ] , an NH2 proton [δH6 . 25 ( 2H , br s ) ] , three methylene groups [ ( δH 4 . 54 , 2 . 33 ) , ( δH 3 . 30 , 2 . 79 ) , ( δH 2 . 57 , 2 . 15 ) ] , and one methyl triplet ( δH 0 . 88 ) . Interpretation of the HSQC data coupled with 13C NMR spectral data showed a total of nine signals , which allowed us to assign all protons to the four directly bonded carbons ( δC 41 . 9 , 38 . 5 , 26 . 5 , and 9 . 7 ) , together with the resonances of five quaternary carbons ( δC 174 . 0 , 157 . 3 , 154 . 6 , 153 . 1 and 93 . 1 ) , suggesting the presence of a 2-amino-4-oxo-tetrahydropteridine moiety and a propionyl group . The sequential COSY correlations from a NH proton ( δH 6 . 96 ) to a methylene group ( δH 4 . 54 , 2 . 33 ) established a dimethylene diamine-type partial structure and the COSY cross-peaks between a triplet methyl signal ( δH 0 . 88 ) and a methylene group ( δH 2 . 57 , 2 . 15 ) also supported the presence of the propionyl group . The HMBC correlations from a NH proton ( δH 6 . 96 ) and a methylene group ( δH 2 . 57 , 2 . 15 ) to a quaternary carbon ( δC 93 . 1 ) allowed us to construct a 1 , 2 , 3 , 4-tetrahydropyrazine ring system and the three bond HMBC correlation sharing a carbonyl amide ( δC 174 . 0 ) from a methyl signal ( δH 0 . 88 ) and a methylene group ( δH 4 . 54 , 2 . 33 ) led to attachment of the propionyl group to the tetrahydropyrazine ring via N-acylation at the 5-position of the pteridine ring . NOESY interpretation and the presence of the shifted methylene protons ( δH 4 . 54 , 2 . 33 ) , which are adjacent to the amide group , supported formation of the N-acyl group corresponding to the cis-amide conformation in the rotamer system . The full structure of compound 1 was unambiguously characterized by the HR-ESI-QTOF-MS data analysis . Compound 2 was also isolated as a white powder . HR-ESI-QTOF-MS analysis showed that the molecular formula of 2 was C8H11N5O2 ( [M+H]+m/z 210 . 0985 ) possessing a 14 Da mass difference , which could arise from loss of CH2 from 1 . The 1D- ( 1H and 13C ) NMR spectral data of 2 were almost identical to that of 1 except for the absence of the CH2 signal , indicating the presence of an acetyl group instead of a propionyl group characterized from the NMR interpretation of compound 1 , which is supported by HMBC correlation from a singlet methyl proton ( δH 1 . 97 ) and a methylene group ( δH 4 . 52 , 2 . 32 ) to a carbonyl amide ( δC 170 . 6 ) . α-Ketobutyrate and pyruvate were purchased from Sigma-Aldrich . E . coli BAP1 transformed with the pepteridine pathway ( pEplu2796 ) was plated onto LB agar containing 25 µg/ml kanamycin and grown overnight at 37°C . A single , well-defined colony was selected and cultured in 5 ml LB medium supplemented with 25 µg/ml kanamycin overnight at 37°C and 250 rpm . Five milliliters of M9 minimal media supplemented with 25 µg/ml kanamycin and either filter sterilized α-ketobutyrate or pyruvate at multiple concentrations were inoculated with 25 µl of overnight culture ( 1:200 ) and incubated at 37°C until the OD600 was between 0 . 5 and 0 . 6 . The M9 cultures were then induced by addition of 0 . 1 mM IPTG and incubated at 25°C and 250 rpm for 72 hr . The cultures were centrifuged ( 2000 g , 20 min , 4°C ) , and the supernatants were extracted with water-saturated butanol ( 1 × 5 ml ) . The organic layers were dried under reduced pressure using a Genevac HT-4X evaporation system , and the crude materials were resuspended in 100 µl of 50% aqueous methanol . The resuspensions were then analyzed by single quadrapole LC/MS ( Column: Phenomenex Kinetex C18 ( 100 Å ) 5 µm ( 4 . 6 × 250 mm ) column; flow rate: 0 . 7 ml/min; mobile phase composition: water:ACN gradient solvent system containing 0 . 1% formic acid: 0–30 min , 5–100% ACN; hold for 5 min , 100% ACN; 0 . 1 min , 100–5% ACN; hold for 1 . 9 min , 5% ACN; 6 . 1 min re-equilibration post-time , 5% ACN ) . The feeding experiments were conducted in triplicate . 13C4-α-ketobutyrate was purchased from Cambridge Isotope Laboratories ( USA ) . A single colony of E . coli BAP1 carrying the pepteridine pathway ( pEplu2796 ) was inoculated into 5 ml LB medium supplemented with 25 µg/ml kanamycin . Three biological replicates were prepared , and the cultures were incubated at 37°C and 250 rpm overnight . Five milliliters of M9 minimal media supplemented with 25 µg/ml kanamycin and filter sterilized 13C4-α-ketobutyrate ( 1 . 2 g/l ) was inoculated with 25 µl of overnight culture and incubated at 37°C until the OD600 was between 0 . 5 and 0 . 6 . The M9 cultures were then induced by addition of 0 . 1 mM IPTG and incubated at 25°C and 250 rpm for 72 hr . The cultures were centrifuged ( 2000 × g , 20 min , 4°C ) , and the supernatants were extracted with water-saturated butanol ( 1 × 5 ml ) . The organic layers were dried under reduced pressure using a Genevac HT-4X evaporation system , and the crude materials were resuspended in 100 µl of 50% aqueous methanol . The resuspensions were then analyzed on the Agilent iFunnel 6650 QTOF system ( Column: Phenomenex Kinetex C18 ( 100 Å ) 5 µm ( 4 . 6 × 250 mm ) column; flow rate: 0 . 7 ml/min; mobile phase composition: water:ACN gradient solvent system containing 0 . 1% formic acid: 0–30 min , 5–100% ACN; hold for 5 min , 100% ACN; 0 . 1 min , 100–5% ACN; hold for 1 . 9 min , 5% ACN; 6 . 1 min re-equilibration post-time , 5% ACN ) . Genetically locked P . luminescens in the M- and P- forms ( Somvanshi et al . , 2012 ) were grown on Luria-Bertani agar plates at 30°C . Single colonies of M- and P-forms were inoculated into individual 5 ml LB liquid medium and cultivated in a shaking incubator for 48 hr ( 30°C , 250 rpm ) . Five milliliters of each culture broth were then transferred to 1 l scale of M9 minimal medium ( 0 . 2% ( w/v ) glucose , 2 mM MgSO4 , and 0 . 1 mM CaCl2 ) supplemented with 5 g/l casamino acids and cultivated at 30°C and 250 rpm . After 72 hr , 1 l each of M- and P-form culture broths were centrifuged at 14 , 000 × g for 20 min , and the supernatant was then extracted with butanol ( 2 × 1 l ) . The butanol-soluble fractions were separately dried under reduced pressure . The samples were subsequently resuspended in 5 ml methanol and 1–2 µl of each sample was injected for HR-ESI-QTOF-MS analysis ( Column: Phenomenex Kinetex C18 ( 100 Å ) 5 µm ( 4 . 6 × 250 mm ) column; flow rate: 0 . 7 ml/min; mobile phase composition: water:acetonitrile ( ACN ) gradient solvent system containing 0 . 1% formic acid: 0–30 min , 5–100% ACN; hold for 5 min , 100% ACN; 0 . 1 min , 100–5% ACN; hold for 1 . 9 min , 5% ACN; 6 . 1 min re-equilibration post-time , 5% ACN ) . Extracted ion count chromatograms were obtained by extracting with m/z 224 . 1147 and 210 . 0991 corresponding to pepteridines A and B , respectively , with a 10 ppm mass window . The butanol-soluble fraction sample from genetically locked P-form was used in co-injection experiments with standards for confirmation by HR-ESI-QTOF-MS using the same analytical methods . Allelic-exchange mutagenesis was used to excise the pepteridine genetic locus in P . luminescens TT01 ( pluCDS3313775R-plu2799 ) ( Crawford et al . , 2010; Kontnik et al . , 2010 ) . Approximately 1 . 5 kB of both upstream and downstream genomic sequences were amplified using the primer pairs Up-F and Up-R; and Dwn-F and Dwn-R , respectively . PCR reactions ( 50 µl ) were prepared using Q5 High-Fidelity DNA Polymerase according to the manufacturer’s protocols with the inclusion of 5% ( v/v ) DMSO and 1 μl of confluent , wild-type P . luminescens TT01 as template . Products were purified using the QIAquick PCR Purification Kit following the manufacturer’s protocol . Overlap extension PCR was then used to fuse the fragments ( Ho et al . , 1989 ) . The homologous recombination cassette and pDS132 were both digested with SacI-HF ( NEB ) ( Philippe et al . , 2004 ) . rSAP ( NEB ) was included in the pDS132 digest to dephosphorylate the vector . The linearized vector and cassette were ligated using T4 DNA ligase . Ligation products were transformed into chemically competent E . coli DH5α λpir using a standard heat-shock transformation procedure . Positive transformats were selected by plating 150 μl of the outgrowth onto LB agar plates supplemented with 25 μg/ml chloramphenicol ( American Bioanalytical ) and grown overnight at 37°C . Colony PCR ( cPCR ) was used to screen for positive constructs . cPCR reactions were carried out as described above using the primer pair pDS-F and pDS-R . A positive construct was subsequently sequence validated using primers pDS-F , pDS-R , CDSSeq-1 , CDSSeq-2 , 99Seq-1 , and 99Seq-2 . This construct was transformed by heat-shock into the diaminopimelic acid ( DAP ) auxotrophic donor strain E . coli WM6026 λpir ( Blodgett et al . , 2007 ) as described above with additional inclusion of 0 . 3 mM DAP in the LB agar plates . This strain along with wild-type P . luminescens TT01 and P . luminescens RifrΔhexA ( generously provided by Professor David Clarke ) were grown to confluency in LB media at 37°C and 30°C , respectively , and 250 rpm . All overnight growths were subcultured 1:1000 into LB and grown to OD600 ≈ 0 . 6; mixed at 1:1 and 1:4 ( donor:recipient ) ratios; and filtered through a 0 . 4 µm sterile filter . Filter mating was carried out overnight at 30°C on LB agar supplemented with 0 . 3 mM DAP . The outgrowth was then resuspended in LB and streaked onto LB agar containing chloramphenicol ( 25 μg/ml ) . Colored colonies indicative of Photorhabdus were selected and streaked onto LB agar supplemented with 5% sucrose for SacB counterselection . Positive colonies were re-streaked three times and single colonies were validated by cPCR . Positive cPCR products were purified using the QIAquick PCR Purification Kit and sequence validated using the primers LocUp , LocDown , CDSSeq-1 , CDSSeq-2 , 99Seq-1 , and 99Seq-2 . P . luminescens wild-type and ΔhexA strains encoding or lacking the pepteridine synthesis pathway were grown in 5 ml hemolymph-mimetic medium ( 5 g yeast extract , 10 g NaCl , and proteinogenic amino acids based on hemolymph concentrations ) for twenty-four hr at 30°C and 250 rpm . 200 μl pellets of strains were prepared in biological triplicate and frozen at −80°C . Upon thawing , sample processing ( protein extraction , alkylation , and trypsin digest ) was performed as in Lajoie et al . ( 2013 ) . Peptides were desalted using a C18 MacroSpin column ( The Nest Group ) . Samples were then dried in a centrifugal vacuum concentrator and dissolved in 22 μl 70% formic acid:0 . 1% trifluoroacetic acid mixed 3:8 volumetrically . In the same buffer , samples were further diluted to 0 . 5 μg/μl based on A280 measurements , and 2 . 5 μg of peptides ( 5 μl ) were analyzed using an ACQUITY UPLC M-Class ( Waters ) paired with a Q Exactive Plus ( Thermo ) mass spectrometer . Column parameters , gradient profiles , and settings for mass spectrometry are described in Ferdaus et al . ( 2016 ) . The P . luminescens annotated proteome was downloaded from Uniprot ( proteome ID UP000002514 ) and used for mass spectra searches using MaxQuant v . 1 . 5 . 1 . 2 ( Cox & Mann ) with Cys carbamidomethyl fixed modification and Asn/Gln deamidation , Met oxidation , Ser/Thr/Tyr phosphorylation , and N-terminal acetylation variable modifications . Search parameters considered peptides resulting from two or fewer missed tryptic cleavages and peptides at least five residues long with a 1% false discovery rate . Using Perseus software v . 1 . 5 . 0 . 15 ( Tyanova et al . , 2016 ) , averaged label-free quantification intensities for sample replicates were compared between two samples by two-tailed T test , with significance cutoffs for protein intensity differences established by either permutation-based false discovery rate ( FDR = 0 . 01 with a 2-fold minimum intensity difference ) or p<0 . 05 , both reported in Figure 8—source data 1–6 . UV/Vis spectra were obtained on an Agilent ( Agilent Technologies , USA ) Cary 300 UV-visible spectrophotometer with a path length of 10 mm . 1H and 2D- ( gCOSY , gHSQCAD , gHMBCAD and NOESY ) NMR spectral data were recorded on an Agilent 600 MHz NMR spectrometer equipped with a cold probe . 13C NMR spectral data were recorded at 100 MHz on an Agilent NMR spectrometer . Flash column chromatography was carried out on Lichroprep RP-18 ( 40–63 µm , Merck , USA ) . Routine HPLC analysis was performed on an Agilent 1260 Infinity system with a Phenomenex ( USA ) Luna C18 ( 2 ) ( 100 Å ) 5 µm ( 4 . 6 × 150 mm ) column and a Photo Diode Array ( PDA ) detector . The separation and purification of metabolites were performed using an Agilent Prepstar HPLC system with an Agilent Polaris C18-A 5 µm ( 21 . 2 × 250 mm ) column , a Phenomenex Luna C18 ( 2 ) or C8 ( 2 ) ( 100 Å ) 10 µm ( 10 . 0 × 250 mm ) column , and an Agilent Phenyl-Hexyl 5 µm ( 9 . 4 × 250 mm ) column . Low-resolution electrospray ionization mass spectrometry ( ESI-MS ) data were measured on an Agilent 6120 Quadrupole LC/MS system with a Phenomenex Kinetex C18 ( 100 Å ) 5 µm ( 4 . 6 × 250 mm ) column . High-resolution ESI-MS data were obtained using an Agilent iFunnel 6550 QTOF ( quadrupole time-of-flight ) MS instrument fitted with an electrospray ionization ( ESI ) source coupled to an Agilent 1290 Infinity HPLC system . | Many bacteria produce small molecules that can be used as a basis for developing new drugs . The instructions for the pathways that make these molecules are stored in the genome – the complete set of genetic material – of the bacteria . With the advancements in genome sequencing over the last decade , these instructions are becoming much more readily available in sequence databases . “Genome mining” is a strategy that involves searching these databases to identify unknown biochemical pathways and help to characterize them . This strategy could help us to discover many more ecologically and medically relevant molecules . A bacterium called Photorhabdus luminescens produces many antibiotics and other molecules that play a variety of roles in the bacterium’s lifecycle . The bacteria live in the gut of roundworms , and the two species have a mutually beneficial relationship where they help each other to acquire food . However , the bacteria are less friendly to the insects that the roundworms infect . When P . luminescens is released into the body of an insect , it takes on a disease-causing form and releases toxic molecules that kill the insect . Park , Perez et al . have now used genome mining to identify a biochemical pathway in P . luminescens that combines the pathways used to create two different types of small molecules produced by the bacteria . This “hybrid” pathway produces a new set of molecules – called pepteridines – that are released by the disease-causing form of the bacteria . Park , Perez et al . also identified a regulator protein that controls the hybrid pathway . This regulator is known to help the bacteria to change into the form that kills insects . The pathway also affects the production of proteins known to be involved in “quorum sensing” . In this process , bacteria use a diverse set of chemical signals to report how many other bacteria are nearby , which enables the bacteria to launch coordinated biological responses – for example , releasing toxic molecules – when their numbers are great enough . In the future , further experiments will be pursued to rigorously characterize how the components of the new hybrid pathway work together . While the hybrid pathway responsible for the production of the pepteridines serves as one example of the utility of genome mining , there is still much room for further discovery . Applying a similar strategy to different organisms has the potential to uncover other pathways of biomedical relevance . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"genetics",
"and",
"genomics"
] | 2017 | Genome mining unearths a hybrid nonribosomal peptide synthetase-like-pteridine synthase biosynthetic gene cluster |
Over two-thirds of integral membrane proteins of known structure assemble into oligomers . Yet , the forces that drive the association of these proteins remain to be delineated , as the lipid bilayer is a solvent environment that is both structurally and chemically complex . In this study , we reveal how the lipid solvent defines the dimerization equilibrium of the CLC-ec1 Cl-/H+ antiporter . Integrating experimental and computational approaches , we show that monomers associate to avoid a thinned-membrane defect formed by hydrophobic mismatch at their exposed dimerization interfaces . In this defect , lipids are strongly tilted and less densely packed than in the bulk , with a larger degree of entanglement between opposing leaflets and greater water penetration into the bilayer interior . Dimerization restores the membrane to a near-native state and therefore , appears to be driven by the larger free-energy cost of lipid solvation of the dissociated protomers . Supporting this theory , we demonstrate that addition of short-chain lipids strongly shifts the dimerization equilibrium toward the monomeric state , and show that the cause of this effect is that these lipids preferentially solvate the defect . Importantly , we show that this shift requires only minimal quantities of short-chain lipids , with no measurable impact on either the macroscopic physical state of the membrane or the protein's biological function . Based on these observations , we posit that free-energy differentials for local lipid solvation define membrane-protein association equilibria . With this , we argue that preferential lipid solvation is a plausible cellular mechanism for lipid regulation of oligomerization processes , as it can occur at low concentrations and does not require global changes in membrane properties .
Lipid bilayers are the most common means of chemical compartmentalization in biology . The bilayer interior , formed by the acyl chains , is a ≈30 Å layer of low-dielectric fluid oil ( Fricke , 1925 ) that provides a natural electrostatic barrier for the passage of charged and polar species . This insulating core enables the cell to generate trans-bilayer chemical and electrical potential-energy gradients that fuel essential metabolic functions . While the macroscopic structure of the lipid bilayer is shared across nearly all species and organelles , their chemical compositions are remarkably diverse . For example , phospholipids can vary in their headgroup moieties , in the length and degree of saturation of the acyl chains , and in the chain-headgroup linkage , i . e . ester vs . ether ( van Meer et al . , 2008 ) . Acyl chains can feature modifications such as branching , or even form covalent bonds across monolayers , as in tetraether lipids ( Valentine , 2007 ) . Lipidomics studies indeed show that cellular membranes include hundreds of lipid types ( Brügger , 2014 ) . It has been proposed that this diversity is in part explained by the ‘homeoviscous adaptation’ of cells , that is , the need to maintain an appropriate membrane fluidity under a wide variety of environmental conditions ( Sinensky , 1974 ) . For example , a recent study indicates that under varying dietary fatty acid input , mammalian cells alter the lipid composition of their membranes to regulate this key property ( Levental et al . , 2020 ) . Yet , some of these compensatory chemical changes appear to be excessively redundant . For example , under cold growth temperatures E . coli generates unsaturated lipids to increase membrane fluidity , but it also increases production of short-chain lipids ( Marr and Ingraham , 1962; Sanders and Mittendorf , 2011 ) . Do these different chemical strategies target others cellular processes that change coincidentally with variations in fluidity ? Is there more to the vast diversity in lipid compositions observed across different types of membranes and conditions , beyond the basic requirement of a fluid lipid bilayer ? One possibility is that this lipid diversity reflects a coupled relationship with the other major constituent of all cellular membranes , namely integral membrane proteins ( Phillips , 2018 ) . The mechanisms of these proteins are , fundamentally , not unlike those of water-soluble proteins , and entail processes such as molecular recognition , conformational exchange and catalyzed chemistry . For membrane proteins , however , the lipid bilayer provides a distinct reaction environment where lipid molecules are the primary solvent . In any biological equilibrium reaction , the solvent plays a major role in defining the energetic landscape; it seems therefore logical to hypothesize that the variability in the chemical composition of physiological membranes might reflect adaptive mechanisms of regulation of protein structure and function . A key question is , however , how this kind of regulation can be sufficiently targeted and specific , rather than globally disruptive . Here , we examine the molecular role of the lipid solvent in a highly prevalent reaction in membrane biology , namely protein oligomerization . Indeed , among membrane-protein classes of known structure , approximately 70% are found as homo- or hetero-oligomers ( Aleksandrova et al . , 2020 ) , compared with about 55% of water-soluble proteins . This comparison is striking because the principal driving force for the formation of protein oligomers in water , that is , the hydrophobic effect ( Tanford , 1978 ) , cannot be a dominant factor in the membrane , as its interior is largely dehydrated . Membrane protein complexes do bury large non-polar surfaces , bringing many hydrophobic side-chains into close proximity , in the range of van der Waals interactions . Yet , it is unclear whether these kinds of protein-protein contacts are the main drivers for the association of integral membrane proteins ( Cristian et al . , 2003 ) as these side chains also form numerous , similarly favorable contacts with lipids in the dissociated states . Likewise , it is not evident that interfacial tensions at the protein-lipid boundary are a dominant factor; while the acyl-chain core would favor association to reduce the total area of the protein-lipid interface , the head-group layer has an opposite effect ( Dixit and Lazaridis , 2020; Marsh , 2008 ) . Nonetheless , it has long been recognized that the complementarity between membrane proteins and their lipid environment is imperfect , resulting in different kinds of perturbations in the structure and dynamics of the bilayer ( Mouritsen and Bloom , 1993; Marsh , 2008 ) . In the context of protein-protein association , local perturbations in membrane thickness are particularly noteworthy; this effect , referred to as ‘hydrophobic mismatch’ , is a key factor in the dimerization equilibrium of helical peptides such as Gramicidin A ( Goforth et al . , 2003; Andersen and Koeppe , 2007 ) and WALP ( Sparr et al . , 2005 ) , and has also been proposed to explain the organization of various rhodopsins and other GPCRs ( Mondal et al . , 2013; Pearson et al . , 1983; Soubias et al . , 2015 ) . This type of perturbation results from a suboptimal match between the exposed non-polar surface of a transmembrane protein and the intrinsic thickness of the acyl-chain core of the bilayer , for a given composition . This mismatch typically forces the bilayer to deform , which translates into an energetic penalty; thus , oligomeric states that minimize this penalty are favored , at least in regard to the membrane energetics . Furthermore , because this energetic penalty will depend on the unperturbed bilayer thickness , variations in lipid composition might provide a means for the cell to regulate oligomerization processes ( Andersen and Koeppe , 2007 ) . However , the potential for this seems limited , as there is an inherent biological drive for cells to maintain the basic biophysical properties of their membrane through homeostatic adaptation ( Levental et al . , 2020 ) . Yet , in a mixed lipid bilayer that better reflects a biological membrane , it is important to consider that lipids will exhibit differential distributions depending on their energetics . Early lattice models of model membrane systems demonstrated that differential energy terms for lipids around protein surface vs . the bulk would lead to enrichment around proteins , that would in turn provide a lipid-mediated driving force for protein association ( Marcelja , 1976; Sperotto and Mouritsen , 1991; Sperotto and Mouritsen , 1993 ) . Thus , we hypothesize that the physiological mechanism of lipid regulated oligomerization equilibrium will involve a molecular mechanism such as this , occurring at low concentrations of regulatory lipids within the membrane , targeting local membrane defects around the protein surface that are introduced by hydrophobic mismatch , and occur in the absence of macroscopic changes in membrane structure . A key to evaluating the dominant driving forces for membrane protein oligomerization is to develop assays that quantify this kind of equilibria in lipid bilayers with sufficient accuracy and sensitivity to variations in lipid composition . Previously , we established such an assay based on single-molecule fluorescence microscopy , and carried out measurements of the free-energy of dimerization of the E . coli Cl-/H+ antiporter CLC-ec1 ( Figure 1A ) in 2:1 palmityl , oleoyl phosphatidyl-ethanolamine/phosphatidyl-glycerol ( 2:1 POPE/POPG ) lipid bilayers ( Chadda et al . , 2016 ) . These membranes are a synthetic mimic of the E . coli polar-lipid content , and consist of C16:0/18:1 acyl-chains , the most commonly found in biological membranes ( Phillips , 2018 ) . While CLC-ec1 had been known to exist as a homodimer in detergent and membranes ( Maduke et al . , 1999; Dutzler et al . , 2002 ) , our measurements revealed this complex results from association of two functionally competent monomers ( Robertson et al . , 2010 ) , via an interface of about 1200 Å2 , most of which is inside the membrane ( Figure 1B ) . Specifically , the measured equilibrium dimerization free energy for this complex is −10 . 9 kcal/mole , relative to a standard state of 1 monomer/lipid ( Chadda et al . , 2018 ) . This is a remarkable finding , in that it implies that the population of dissociated monomers at biological protein-expression levels is virtually zero . This interaction is thus reminiscent of obligate water-soluble homomeric complexes , whose association is dominated by the hydrophobic effect ( Bahadur et al . , 2003; Yan et al . , 2008 ) ; by analogy , it is reasonable to infer that the energetics of the lipid solvent might also be key for CLC-ec1 dimerization . Indeed , examination of the dimerization interface shows that the two central helices are much shorter than what is typical in transmembrane segments ( Figure 1B ) . In the monomeric state , a significant hydrophobic mismatch might therefore exist between this protein surface and the surrounding membrane ( Figure 1C ) , which would be completely eliminated upon dimerization , possibly explaining the remarkable stability of this complex . CLC-ec1 thus appears to be an excellent system to examine the fundamental questions outlined above . That is , can protein-induced membrane deformations contribute to explain the structure and stability of obligate membrane protein complexes ? What is the extent of the changes in membrane lipid composition that are necessary to influence these oligomerization reactions , what is the underlying mechanism , and importantly , are those changes physiologically viable , that is , do they preserve or impair protein function ? To address these questions , we first use molecular dynamics simulations of monomeric and dimeric CLC-ec1 in 2:1 POPE/POPG , to evaluate the lipid bilayer structure in each state . This analysis informs a series of new experimental assays , namely small-angle neutron scattering measurements , Cl- transport assays and single-molecule photobleaching analyses , with which we determine how varying quantities of short-chain C12:0 , di-lauryl ( DL ) lipids alter the dimerization equilibrium as well as the activity of this transporter . To obtain a molecular level interpretation for these new experimental results , we return to molecular dynamics simulations in membrane mixtures that mimic the experimental conditions . These studies lead to a perspective of the dimerization reaction as primarily controlled by the energetics of local lipid solvation of the associated and dissociated states ( Marsh , 1995; Marsh , 2008 ) , and underscore the essential role of molecular-scale heterogeneity of the lipid bilayer in defining membrane protein association equilibria .
As mentioned , the features of the CLC dimerization interface ( Figure 1B ) suggest that , when exposed in the monomeric state , there might be a hydrophobic mismatch with the surrounding membrane ( Figure 1C ) . If so , the energetic cost associated with solvating the monomer could translate into an effective driving force toward dimerization; by burying these ‘problematic’ interfaces away from the lipids , the system would gain free energy upon association ( Figure 1D ) . To begin to validate or refute this hypothesis , we first studied the structure of the lipid bilayer around the CLC-ec1 monomer and dimer , using coarse-grained molecular dynamics ( CGMD ) simulations , and evaluated whether the exposed dimerization interface indeed appears to cause the membrane to adopt a higher energy state , relative to the other regions of the protein surface . For both monomer and dimer , we used 2:1 POPE/POPG membranes ( Figure 2—figure supplement 1A ) , corresponding to the C16:0/18:1 acyl chains used in our previous experimental measurements of the reversible dimerization reaction ( Chadda et al . , 2016 ) . It is worth noting that any simulation of a non-homogenous membrane necessarily presupposes an initial spatial distribution of the lipid components , which is not only arbitrary but also may not be representative of the equilibrium condition . Prior to examining the structure of these membranes , it is therefore key to ascertain that the simulations are long enough for the two lipid components to mix fully and spatially re-distribute according to the free-energy landscape of the molecular system . One way to examine this process of mixing is to quantify , for each lipid in the simulation box , what fraction of all other lipids in the same leaflet are at some point part of their first solvation shell . In our case , this analysis shows that each lipid , on average , is in direct contact with 80% of all other lipids in the course of each of our simulations ( i . e . over 1100 molecules ) ( Figure 2—figure supplement 1B ) . Given that at any given timepoint , a solvation shell consists of fewer than 10 lipids , this result implies extensive mixing and hence no concern in regard to the starting condition . We also examined the orientation of the protein in the bilayer , which is also an arbitrary initial condition . This analysis indicates that the simulations broadly explore orientation space , resulting in clearly defined probability distributions for both monomer and dimer ( Figure 2—figure supplement 1C ) . With sufficient lipid exchange over the time-scale of the simulations , we proceeded to analyze the shape of the lipid bilayer near the monomer and the dimer as well as other structural descriptors ( Figure 2—figure supplement 2 ) . From the simulated trajectories , we calculated 3D density maps reflecting the spatial distribution of both acyl chains and ester linkages in the protein vicinity ( Figure 2A , B ) . The results for the monomer show that the membrane shape is deformed at the dimerization interface , thinning near the two shorter helices at the center . Elsewhere along the monomer perimeter the membrane is largely unperturbed , and its shape is nearly identical to what we observe for the dimer , confirming the simulations are probing the membrane structure reliably . Quantitative analysis of bilayer thickness , measured by the separation between the outer and inner ester layers and represented on a 2D heat map , shows that the magnitude of this thinning defect is about 8 Å relative to the bulk ( Figure 2C ) , that is , nearly a quarter of the unperturbed hydrophobic thickness of this 2:1 POPE/POPG membrane . This defect is also clearly specific to the dimerization interface in the monomeric state , consistent with the 3D density maps . Smaller defects are discernable elsewhere but , as noted , they are indistinguishable if monomer and dimer are compared . Membrane thickness deformations are sometimes conceptualized as resulting from spring-like compressions or extensions of the lipid chains ( Andersen and Koeppe , 2007; Brown , 2017 ) . In this case , however , the mean acyl chain end-to-end distance near the protein is only a fraction of 1 Å smaller than the bulk value ( Figure 2—figure supplement 3A ) ; this minor perturbation is also not specific to the dimerization interface , but is present at other regions . Therefore , the thinning defect that we observe does not arise due to a significant compression of the acyl chains . Instead , our simulations show that increased lipid-chain tilt is in large part what leads to the membrane thinning . This effect is clear from analysis of the orientation of the acyl chains in terms of the coarse-grained equivalent of a second-rank order parameter , which reveals a clear change at the dimerization interface ( Figure 2—figure supplement 3B ) . To quantify this effect more directly , we evaluated the mean lipid-chain tilt angle across the system , relative to the membrane normal ( Figure 2D ) . In the bulk , this angle averages to 0° for one leaflet and 180° for the other , as one would expect , as the lipid dynamics are isotropic . Approaching the dimerization interface , however , this angle increases gradually and is maximally deflected by 60° , in both leaflets . This drastic change in orientation can be clearly visualized in 3D by analyzing the ‘average structure’ of the lipid molecules residing at different positions along the membrane ( Figure 2E ) . In the bulk , this average yields a linear structure , perfectly perpendicular to the membrane mid-plane , again due to the isotropy of the lipid configurational dynamics . However , inthe lipids that are closest to the dimerization interface ( yellow helices ) , the acyl chains adopt tilted , non-bilayer configurations in order to optimally solvate the protein . Alongside this drastic change in tilt angle , we also observe that near the dimerization interface the acyl chains in one leaflet show a greater degree of inter-digitation with those in the other leaflet ( Figure 2—figure supplement 3C ) , compared to the bulk or elsewhere along the protein perimeter . In summary , our simulation data clearly shows that when the dimerization interface of CLC-ec1 is exposed to the lipid solvent , it deforms the surrounding membrane by thinning and twisting the bilayer structure ( additional effects in lipid density and hydration will be discussed later below ) . To solvate this ‘problematic’ interface , C16:0/18:1 lipids must adopt non-bilayer configurations that are significantly tilted and more entangled with lipids in the opposite leaflet . Interestingly , the perturbations we observe are in all cases symmetric with respect to the bilayer midplane , consistent with the fact that the CLC-ec1 monomer consists of two topologically inverted structural repeats; this observation further underscores that it is the protein structure that dictates the morphology of the adjacent bilayer . Altogether , these results clearly indicate that optimal lipid solvation of the monomeric state in C16:0/18:1 lipids requires the membrane to adopt a high-energy conformation . Because dimerization completely eliminates this membrane defect , the cost of lipid solvation of monomeric CLC-ec1 must therefore translate into an attractive force . Although the precise magnitude of this stabilizing effect is not directly revealed by the results presented thus far , our single-molecule TIRF assays enable us to evaluate its significance experimentally . That is , if solvation of the membrane defect caused by monomeric CLC-ec1 indeed implies a dominant energetic penalty , then the dimerization equilibrium should be shifted toward the monomeric state by introducing lipids that are a ‘better’ solvent for this defect; this shift should be reflected in the measured free-energy of dimerization . Since the membrane defect induced by monomeric CLC-ec1 is constructed by hydrophobic thinning , we decided to test this hypothesis by introducing short-chain di-lauryl ( DL ) C12:0 lipids into the C16:0/18:1 PO lipid membranes while keeping the overall 2:1 PE/PG headgroup composition constant ( Figure 3A ) . DL lipids are shorter than PO lipids by 4–6 carbons per chain , and are also fully saturated , losing the ω−9 double bond in one chain . Before drawing any conclusions in regard to the CLC-ec1 dimerization , we sought to characterize the intrinsic properties of these quaternary lipid bilayers . To do so , we first measured phase-transition thermograms for different DL/PO ratios by differential scanning calorimetry ( Figure 3B ) . The mixtures show broad profiles; however , the membranes are fluid at room temperature , with the exception of the 100% DL condition . Plotting the peak Tm as a function of DL shows eutectic behavior with a minimum Tm at about 30% DL ( Figure 3C ) . Next , we examined the structure of the DL/PO bilayers at 25°C with small-angle neutron scattering ( SANS ) . Using a spherical , multi-lamellar liposome model to fit the scattering spectra ( Figure 3D , E ) , we observe a gradual decrease in the bilayer thickness as the DL content is increased ( Figure 3F ) ; at 70% DL , the membrane is about 6 Å thinner than that with no DL . This change is consistent with published SANS measurements for POPC vs . DLPC ( Kučerka et al . , 2011 ) and POPG vs . DLPG ( Pan et al . , 2014 ) . It also approximately matches the magnitude of the defect created by the CLC-ec1 monomer ( Figure 2C ) , indicating that the DL lipids in these mixtures might be suitable for solvating the dimerization interface . To understand how addition of DL impacts the bilayer thickness at the molecular level , we also carried out coarse-grained molecular dynamics ( CGMD ) simulations for the pure PO/DL membranes . As the DL content is increased , the observed change in thickness in the simulations reproduce the experimental trend , despite the approximations inherent to the CG forcefield ( Figure 3G ) . Further analysis indicates a high degree of cooperativity between the two lipid-chain types: for example , if the bilayer thickness is quantified by the distance between the two ester layers , there is virtually no difference when this distance is evaluated only for DL vs . PO lipids , at any % DL ( Figure 3G ) . This observation indicates that as DL is added , their lipid headgroups remain aligned with those of the PO lipids so as to minimally perturb the degree of hydration of the headgroup layer . It is worth noting that the average number of contacts formed between the acyl chains in one leaflet and those in the other is also a conserved quantity , regardless of the PO/DL content ( Figure 3H ) . For example , comparing the 0% and 10% DL membranes , we observe that a given DL chain can form only half of the interleaflet contacts seen for the PO chains in the absence of DL . However , to counter this destabilizing effect , PO lipids slightly increase the number of interactions they form across the membrane midplane and become more interdigitated; as a result , the number of chain contacts is , on average , unchanged , but this translates into a thinning of the bilayer . It appears , therefore , that an optimal degree of headgroup-layer hydration and interleaflet contacts dictates the thickness of the pure PO membrane; as DL lipids are added these PO/DL membranes adapt to preserve these two quantities , which requires them to become thinner . The significance of these conserved quantities will be discussed again further below . Following these results , we investigated whether the monomer-dimer equilibrium can be influenced by addition of DLPE/PG lipids to the POPE/PG membranes , using our previously established single-molecule subunit-capture approach ( Chadda et al . , 2016; Chadda and Robertson , 2016 ) . In this method , the protein is site-specifically labeled with a Cy5-maleimide fluorophore and reconstituted into lipid bilayers that are fused into large multilamellar vesicles by freeze/thaw cycles . In this state , the membrane area is sufficiently large to permit monomer and dimer populations to equilibrate according to the association constant of the reaction and the protein to lipid mole fraction , χprotein . The oligomeric-state distribution resulting from this equilibrium condition is quantified by fragmenting the membranes into fixed liposome compartments via extrusion , and by counting the probability distribution of subunit capture by single-molecule photobleaching analysis using TIRF microscopy . The photobleaching probability distribution follows a Poisson distribution provided one considers heterogenous compartments and multiple protein species ( Cliff et al . , 2020 ) , and thus the population of oligomeric species can be quantified using this approach . However , it is also important to control for other factors that affect the probability distribution such as the protein labeling yield and liposome size distribution . Our site-specific labeling procedure ( Chadda et al . , 2016; Chadda and Robertson , 2016 ) provides a consistent labeling yield , PCy5 , WT = 0 . 663 ± 0 . 005 ( mean ± sem , n = 27 independent purifications with Cy5 labeling , Figure 4—source data 7 ) ; thus , as long as we know the liposome size distribution , we can determine any changes of CLC-ec1 dimerization equilibrium in different lipid environments . Using this approach , we set out to study the degree of CLC dimerization in a single mixed lipid condition in which we observe thinner membranes , namely 20% DL . Our first step was to examine the mixed DL/PO liposomes using cryo-electron microscopy imaging . 2:1 PE/PG membranes containing 20% DL and 80% PO were prepared , freeze-thawed into multi-lamellar vesicles , extruded through 400 nm filters and then imaged and analyzed to measure the size distribution directly . The liposomes and membranes appear similar to those in the 0% DL condition ( i . e . 2:1 POPE/POPG ) , with comparable radius and fractional surface area distributions ( Figure 4A–C ) . There is a significant proportion of multilamellar vesicles in both compositions , 44% for 20% DL and 25% for 100% PO samples . Next , WT-Cy5 20% DL liposomes were imaged by single-molecule TIRF microscopy . Example images and raw data for the photobleaching traces for PO and 20% DL liposomes ( Figure 4D , E ) demonstrate no changes in the quality of images obtained in the different lipid conditions . While the cryo-EM imaging indicated no significant differences in liposome size distributions , we also examined the photobleaching probability distributions of two experimental controls: I201W/I422W , referred to as ‘WW’ , a version of the protein with two tryptophan substitutions at the dimerization interface ( Robertson et al . , 2010 ) that reports the fixed monomer probability distribution; and R230C/L249C , or ‘RCLC’ , a disulfide cross-linked constitutive dimer ( Nguitragool and Miller , 2007 ) that reports the fixed dimer probability distribution ( Chadda et al . , 2018 ) . Photobleaching analysis of these controls in 20% DL liposomes show dependencies on the protein mole fraction comparable to those observed for the 2:1 POPE/POPG composition ( Figure 4—figure supplement 1A , B ) . Thus , our analyses indicate that the 20% DL , 80% PO 2:1 PE/PG liposome population is comparable to the 100% PO condition , allowing us to attribute changes in the single-molecule photobleaching distributions to specific changes in CLC dimerization . With our quantification method benchmarked , we analyzed the photobleaching probability distribution of WT CLC-ec1 in 20% DL 2:1 PE/PG lipid bilayers and compared it to the WW and RCLC control data in the same lipid condition ( Figure 4F ) . Calculation of the fraction of dimer in these protein populations , from least-squares fitting to the WW and RCLC reference distributions , shows that dimerization is significantly destabilized , that is the equilibrium is shifted toward the monomeric state ( Figure 4G ) . By fitting to an equilibrium dimerization isotherm , we estimate a lower-limit of the KD , Dimer > 4 . 2 ± 1 . 3 x 10−6 subunits/lipid , as the reaction falls out of the dynamic range for these measurements leading to an insufficient fit of the reaction . Still , the limited reaction indicates that the 20% DL condition destabilizes dimerization by at least +3 kcal/mole . To verify that this shift reflects a new equilibrium , we also examined whether the mostly dimeric population of CLC-ec1 in 2:1 POPE/POPG lipid bilayers is driven toward the monomer state when fusing the proteo-liposomes with DL-containing membranes . Figure 4H shows the resultant distribution of diluting χprotein= 2 x 10−6 subunits/lipid proteoliposomes 1:1 via freeze-thawed fusion with either 0% DL or 40% DL ( i . e . final DL proportion is 20% ) . Indeed , after incubating the fused samples for 5 days at room temperature , the probability distribution showed a significant shift toward monomers , indicated by an increase in single steps , P1 ( Figure 4—figure supplement 1C , D ) . Therefore , alternative approaches consistently demonstrate that the short-chain DL lipid shifts the oligomeric distribution of CLC-ec1 toward the monomeric form . Finally , we examined whether CLC-ec1 remained functional in this new membrane environment . To do so , we carried out chloride efflux measurements from CLC-ec1 proteo-liposomes ( Walden et al . , 2007 ) and quantified the chloride transport activity as a function of DL in the membrane . The protein remained effective at transporting chloride in 20% DL ( Figure 4I ) , with no difference in the fraction of inactive vesicles ( Figure 4J ) , and a modest twofold decrease in chloride efflux rate ( Figure 4K ) . Therefore , CLC-ec1 is significantly destabilized toward the dissociated monomeric form in 20% DL , 80% PO 2:1 PE/PG membranes , yet remains a functionally competent chloride transporter in the new lipid composition . Next , we examined how dimerization depends on the DL/PO ratio by carrying out a titration experiment , studying the monomer-dimer population as a function of DL , from 10−8 to 80% ( Figure 5A ) . Note these experiments were conducted at dilute protein densities within the membrane , with 1 subunit per million lipids ( χprotein= 1 x 10−6 subunits/lipid ) , where WT CLC-ec1 is ≈ 80% dimeric in 2:1 POPE/POPG . Based on our experiments of WW and RCLC controls ( Figure 4F , Chadda et al . , 2018 ) , we know a dimeric population is expected to yield comparable probabilities of single and double steps ( P1 ≈ P2 ) , while a monomeric population will exhibit mainly single steps ( P1 > P2 ) . The reason why a dimeric population includes a significant observation of single steps is because our experimental labeling yield is PCy5 = 0 . 66 , and binomial statistics predicts a nearly equal proportion of singly and doubly labeled Cy5 dimers , as demonstrated by our previous theoretical simulations ( Chadda et al . , 2016; Chadda and Robertson , 2016; Cliff et al . , 2020 ) . The raw photobleaching probability distributions show a population shift from nearly all dimers in 0% DL to all monomers at 80% DL , which resembles the distribution of WW in 2:1 POPE/POPG . Calculation of Fdimer from these data shows that the impact of DL on dimerization follows two phases with an inflection point around 1% DL ( Figure 5B ) . We also examined the dependency of CLC activity on the presence of DL in greater detail by measuring CLC dependent chloride efflux while titrating DL in the membrane ( Figure 5—figure supplement 1 ) . For samples with 10% DL or less , we observed no change in chloride transport activity; by contrast , at 40% DL there is approximately a 70% reduction in transport rate . To compare how dimerization and function relate to bilayer structure , we plotted the normalized change in the bilayer thickness from the SANS data , ΔdB , with the normalized change in the dimeric population , ΔFdimer and the normalized change in transport rate ΔkP ( Figure 5C ) . From this plot , we can see that > 60% of the dimerization changes occur below 10% DL , that is , before there are any major changes in the macroscopic structure of the membrane . However , the change in chloride efflux rate correlates directly with the change in bilayer thickness . Therefore , while function appears to be impacted by global changes in membrane thickness following a simple trend , our photobleaching results demonstrate that dimerization equilibrium is coupled to the membrane in a more complex manner . Converting FDimer to the change in free energy of dimerization relative to the zero DL condition , ΔΔG , highlights the two types of molecular linkage observed . At DL >1% , the complex is destabilized by 0 . 8 ± 0 . 3 kcal/mole for every addition of 10% DL in the membrane ( Figure 5D ) . As this corresponds to the range where we observe membrane thinning , it is reasonable to assume this change is linked to the bulk properties of the membrane . However , at DL <1% , ΔΔG shows a linear dependency with the logarithm of DL , with a destabilization of 0 . 14 ± 0 . 07 kcal/mole for every Log10 change in % DL ( Figure 5E , Figure 5—source data 1 ) . This coupling describes a microscopic process effected by DL , detectable even when DL is present in minimal amounts . It also indicates the molecular mechanism , as a linear dependency of ΔΔG on the logarithm of the co-solvent activity corresponds to a thermodynamic linkage model of preferential solvation , as described by Tanford , 1969; Record and Anderson , 1995; Marsh , 1995; Timasheff , 2002a . To investigate whether preferential solvation is involved in the mechanism by which DL shifts CLC-ec1 dimerization equilibrium toward monomers , we again turned to CGMD simulations . Specifically , we carried out simulations of the CLC monomer in bilayers of DLPE/DLPG/POPE/POPG lipids with a 2:1 PE/PG ratio and a DL content of either 1% , 10% , 30% , or 50% ( Figure 6—figure supplement 1A ) . Before drawing any conclusions from these simulations , we again ascertained that these complex bilayers do appropriately mix in the timescale of the trajectories , using a metric identical to that considered for the POPE/POPG simulations . Taking the 50% DL membrane as an example , we observe that by the end of the simulations , any one PO or DL lipid has been in direct contact with about 90% of all other lipid molecules in the same leaflet ( Figure 6—figure supplement 1B ) , indicating near-ideal equilibration . We then proceeded to examine the structure of these membranes using the same descriptors as those employed above . Interestingly , the thinned-membrane defect around the dimerization interface is still observed in the mixed PO/DL simulations , even for 50% DL ( Figure 6A ) , and is comparable to what we find in the 2:1 POPE/POPG simulations ( Figure 2C ) . That is , the defect is observed even when the ‘macroscopic’ thickness of the lipid bilayer is reduced; for example , for 50% DL , the thinning is about 5 . 5 Å . Examination of the average lipid structure across the membrane show that both PO and DL lipids solvate the totality of the protein surface and that , as in the 100% PO condition , the thinned-membrane defect results from both types of lipids becoming increasingly tilted as they approach the dimerization interface ( Figure 6B , Figure 6—figure supplement 2 ) . However , 3D density maps for the first lipid solvation shell indicate these two lipid types are not distributed identically ( Figure 6C ) . At the two interfaces not involved in dimerization , the DL density signal weakens at the center of the membrane , revealing the DL chains are too short to solvate the hydrophobic span of the protein , which is better matched to PO lipids . Conversely , the density signal for PO is weaker than DL at the dimerization interface , and weaker than that seen in the 100% PO membrane , indicating PO is depleted here . This depletion , and the corresponding enrichment in DL lipids , becomes apparent in 2D projections of the percent difference between the observed lipid density ratio ( DL/PO ) and the expected bulk ratio ( Figure 6D ) . These data show that the normalized probability of observing DL rather than PO at the dimerization interface is higher than elsewhere in the membrane , irrespective of PO/DL composition . Conversely , DL is depleted at the other two interfaces , consistent with the 3D density analysis discussed above . Quantification of the DL enrichment as a function of the distance from the protein surface reveals this effect extends for up to 30 Å from the dimerization interface and confirms that it is largely independent of the PO/DL ratio ( Figure 6E ) . As noted above , increasing DL content ultimately results in a change in the overall thickness of the PO/DL bilayers . Thus , it could be reasonably argued that this ‘macroscopic’ effect would reduce the energetic cost of the thinned-membrane defect caused by monomeric CLC-ec1 , irrespective of whether one lipid type or another is preferentially enriched , and thereby cause a shift in the dimerization equilibrium . It is important to note , however , that the enrichment effect we report is discernable at 1% DL , that is , in conditions where there is virtually no change in the global thickness of the bilayer , relative to the PO condition , both in experiment and simulation ( Figure 3F , G ) . Yet , 1% DL has a profound impact on the dimerization equilibrium ( Figure 5 ) . Limitations in computing speed currently preclude us from verifying this effect for even smaller quantities of DL lipids with adequate statistically significance . Nevertheless , the existing results underscore that a process distinct from a change of the global properties of the membrane is dominant in this regime , which we posit is that of preferential lipid solvation . The observation of near complete lipid mixing in our simulations implies that the enrichment of DL at the dimerization interface is neither artifactual nor transient but rather a minimum free-energy state of the lipid-solvent structure . What are the molecular factors that explain this observation , that is , what drives the preferential residence of DL over PO in this specific region of the membrane ? As mentioned above , our simulation data for the pure PO/DL bilayers indicates that the collective degree of interdigitation between opposing leaflets is a conserved quantity that dictates membrane structure . This interdigitation can be quantified by metrics such as the average number of interleaflet contacts formed by the acyl-chains ( Figure 3G , H ) . With this observation in mind , 2D maps of the number of interleaflet lipid contacts for the CLC-ec1 systems reveal key differences that seem to explain why DL is enriched at the dimerization interface ( Figure 7A ) . In the bulk , there is no difference between the 100% and 50/50 PO/DL conditions , for example , when all lipids are averaged . The number of interleaflet contacts is again a conserved quantity , as observed for the pure bilayers . At the CLC-ec1 dimerization interface , however , this conserved quantity cannot be matched by the PO lipids; whether for 100% PO or 50/50 PO/DL , PO lipids create an excessive , clearly non-native overlap between leaflets . By contrast , when at the dimerization interface , DL chains very closely reproduce the conserved bulk values . Thus , by segregating PO lipids away from this interface , and accumulating DL lipids instead , the system minimizes the negative impact of the thinned-membrane defect created by the CLC monomer , shifting the equilibrium toward the dissociated state . The mitigating effect of the DL lipids is also apparent from other more conventional descriptors of bilayer structure . In Figure 7B , for example , we quantify how the number of lipid near-neighbors varies across the membrane . The data shows that in 100% PO the CLC monomer introduces a clear lipid-density defect at the dimerization interface , particularly at the level of the headgroup and ester-linkage layers . This perturbation , in turn , leads to a marked increase in the degree of water penetration of the hydrocarbon interior of the bilayer , by almost threefold relative to the bulk-membrane values ( Figure 7C ) . As noted earlier , these perturbations impact both leaflets , in a manner that reflects the internal symmetry of the CLC monomer , and very likely signify a major energetic cost . The preferential solvation of the dimerization interface by DL does not entirely eradicate these defects , but it is clear from our data that they are greatly minimized ( Figure 7B , C ) . In view of these results , we can plausibly infer that in the PO/DL conditions the energetic cost of lipid solvation of the monomer is reduced , relative to the pure PO condition . This preferential solvation effect would in turn explain the shift in dimerization equilibrium observed experimentally , particularly at low DL concentrations .
Our computational studies demonstrate that the CLC-ec1 monomer in 2:1 POPE/POPG introduces a non-native thinned defect in the surrounding membrane , due to the exposure of the shorter central helices that form the core of the dimerization interface . Experimentally , we have measured that the free energy of CLC-ec1 dimerization in 2:1 POPE/POPG lipid bilayers is −10 . 9 ± 0 . 1 kcal/mole ( one subunit/lipid standard state ) . Given the nature of the perturbations caused by the monomer , and the fact that this free-energy value can be drastically shifted by addition of minimal amounts of DLPE/DLPG lipids , we believe it is very likely that this dimerization reaction is driven largely by the energetics of the membrane , with protein-protein interactions contributing on a smaller scale . Conclusive evaluation of this hypothesis will however require further experimental and computational investigations of the dimerization equilibrium for a range of protein constructs in different lipid bilayers and conditions , and a direct quantification of the anticipated differences in lipid solvation energetics of associated and dissociated states , in each case . The concept that protein-induced membrane defects can translate into an effective driving force toward oligomerization presents a generalizable solution to the problem of membrane protein self-organization , while allowing for evolutionary adaptations in amino acid sequence that might be advantageous . In this perspective , association primarily depends on the overall protein architecture and the general chemical features of the protein surface . Strict conservation of specific amino acids at specific sites on the protein surface is thus not critical though not entirely inconsequential . Membrane protein complexes may thus evolve high shape-complementarity , for example to maximize the exclusion of lipids in the complexed form of the protein ( Li et al . , 2013 ) , and thereby achieve greater stability . Indeed , previous analysis showed that CLC-ec1 exhibits high shape-complementarity , comparable to other high-affinity antigen antibody complexes ( Robertson et al . , 2010 ) . As noted , membrane perturbations appear to influence the association of other systems as well . Assembly of Gramicidin A peptides into functional ion-channel dimers results in a hydrophobic mismatch with the surrounding bilayer , and thus dimerization can be inhibited by increasing the global membrane thickness ( Goodall , 1971; Mobashery et al . , 1997 ) . FRET measurements for reconstituted Rhodopsin have indicated the formation of higher order assemblies when the membrane is thicker or thinner than a certain range ( Botelho et al . , 2006 ) . Computational studies have rationalized this kind of spatial organization as resulting from anisotropic defects in membrane thickness or curvature , which become mitigated upon association in specific geometries ( Mondal et al . , 2014; Kahraman and Haselwandter , 2019 ) . A striking illustration of this concept is found in the inner mitochondrial membranes , where ATP synthases , each a dimeric complex , spontaneously assemble into micrometer-scale linear arrays , priming the membrane to invaginate and form cristae ( Anselmi et al . , 2018 , Blum et al . , 2019 ) . Thus , there is a growing body of evidence that suggests that membrane-dependent forces are a key factor in the self-assembly and organization of membrane protein complexes . To our knowledge , however , this study is the first to probe how such forces can dictate the oligomerization equilibrium of a strongly-bound integral membrane protein complex , in the absence of global physical membrane changes , and without compromising its biological functionality . The notion that dissociated and associated states of a membrane protein oligomer can perturb the bilayer in distinct ways , as a result of hydrophobic mismatch , implies it is conceivable that cellular mechanisms exist through which variations in lipid composition can regulate this type of equilibria . However , different mechanisms can be envisaged . The relative energetics of solvation of the dissociated and associated states would logically be altered if there is a global change in membrane thickness ( Andersen and Koeppe , 2007 ) . Alternatively , a different mechanism could involve that certain lipids bind to the 'problematic’ protein-membrane surface , in a manner similar to conventional agonists or antagonists ( Song et al . , 2013 ) . Neither of these mechanisms , however , explain our experimental data . The first effect that we observe is when short-chain saturated DL lipids are added to PO bilayers at extremely low DL activities , ranging from 1 DL per 1010 PO up to 1% DL . In this regime , we determined that the bilayer thickness is identical to that of PO membranes , and we measure no change in protein function . Yet , upon increasing the amount of DL , we detect a gradual correlated increase in the monomeric proportion of CLC-ec1 , and this effect appears to be linear with respect to the logarithm of DL over six orders of magnitude ( Figure 5F–H ) . At first , it seems intuitive to interpret this data as a process of competitive inhibition , that is , one or more DL-specific binding sites might exist at the dimerization interface , with a lipid affinity of KD , DL , the occupancy of which precludes dimerization . However , we can immediately see that our data do not agree with this type of linkage . This type of model would lead to complete saturation of the population of the monomeric state over a much narrower increase in % DL , at most thousand-fold , and centered at the hypothetical KD , DL . The gradual , linear relationship of the decay of the dimeric population with the logarithm of % DL that we observe in our data , over six orders of magnitude , is simply not in agreement with a model of site-specific competitive binding . Lipid-composition effects can be however conceptualized beyond the paradigms of bimolecular recognition or global morphological changes . If we consider lipids as solvent molecules ( Marsh , 1995 ) , a different type of linkage model , used to examine mixed aqueous solvent systems , explains our observations . The stability of soluble proteins , both as oligomeric assemblies or folds , is known to be dependent on the relative activities of the co-solvents present , due to preferential solvation effects ( Tanford , 1969; Record and Anderson , 1995; Schellman , 1987; Schellman , 2003; Timasheff , 2002b; Lee and Timasheff , 1981 ) . That is , one state of the protein might be ‘preferentially solvated’ by a given co-solvent , and so an increase in that co-solvent activity shifts the reaction equilibrium to that state . This is a form of linkage that describes how proteins can be stabilized or destabilized by salts , glycerol , sugar , or chaotropic denaturants . It does not involve specific binding , but rather non-specific affinities that lead to a linear dependency of the change in free energy with the log-activity of the co-solvent , that is , log ( Keq ) ∝ log ( aDL ) , as we observe in our experiments ( Figure 5E ) . While preferential solvation alone , that is , without bulk-membrane changes , had not been previously experimentally demonstrated to impact the formation of obligate complexes of integral membrane proteins , the notion that the features of the protein-lipid interface can dictate the spatial distribution of different lipid types in its vicinity has been previously documented . Early Monte Carlo simulations of membrane lattice models showed that consideration of differential energetics for annular lipids , and different lipid types , accounts for lateral enrichment of lipids around proteins , introducing a potential lipid-mediated driving force for aggregation in membranes ( Marcelja , 1976; Sperotto and Mouritsen , 1991; Sperotto and Mouritsen , 1993 ) . More recently , CGMD simulations of a wide set of membrane proteins in highly complex bilayers , it was observed that each protein induces a unique lipid solvation structure , akin to a ‘lipid fingerprint’ ( Corradi et al . , 2018 ) . Similarly , a simulation study of the Gramicidin A dimer in a two-component bilayer with C16:1 and C24:1 acyl-chains , reported that the latter become underrepresented in the first solvation shell , as the C16:1 chains better match the hydrophobic thickness of the dimer ( Beaven et al . , 2017 ) . The experiments and simulations described in our study indeed confirm that preferential solvation effects alone can dictate the energetics of oligomerization reactions for integral membrane proteins , even those that are assumed to be obligate oligomers , where the associated form appears to be derived from evolutionarily pressures , for example CLC . That the monomer causes a thinned-membrane defect that is eliminated upon dimerization is key . We observe how at this defect the distribution of PO and DL lipids diverges from what would be expected based on their bulk ratio , and the shorter DL lipids become enriched while the PO lipids are depleted . This enrichment is specific to the dimerization interface , and therefore also specific for the monomeric state that exposes this interface . And importantly , it is observed irrespective of the PO/DL content of the bilayer , as could be expected for an effect that is dictated by the protein itself . Thus , although any deviation from the bulk-membrane homogeneity does entail a free-energy penalty , the larger gains resulting from a more optimal solvation of the exposed dimerization interface ultimately translate into a strong shift in favor of the monomeric state . As is logical , the preferential solvation effect is ultimately superseded by more global changes in the state of the membrane; according to our SANS experiments , these changes begin to take place when the DL proportion exceeds 10% , which is the kind of change in lipid composition that has been typically evaluated in previous studies of membrane-driven organization processes . In this regime , we do observe an additional depletion of the dimer population , corresponding to the second phase of our dimerization data ( Figure 5D ) , because the energetic penalty of solvating the monomer is further reduced as the membrane becomes thinner . It is possible that this change in energetic relationship reflects a change in the relative lipid energetics , however , in this high-DL range we also observe a correlated decrease in CLC-ec1 transport activity . That is , while the thinner DL/PO membranes match the exposed dimerization interface better , they also compromise the functional integrity at high DL . While we do not have direct structural information about the protein under these conditions , one plausible interpretation for this functional degradation is that the structural mechanism of the protein is somehow impaired in globally thinner membranes . Preferential solvation would thus appear to be a more viable mechanism of lipid regulation of oligomerization reactions of specific species under physiological conditions where biology strives to maintain global membrane properties . Preferential solvation is a generalizable effect that could modulate any equilibrium whereby one or more protein states introduce a local morphological defect into the membrane . Besides oligomerization reactions , this effect likely defines the energetics of the intermediate conformational states that are encountered during membrane protein folding , as well as gating , transport , and signaling . Given the highly complex composition of real biological membranes , one can therefore envisage that each of these conformational states will have a different local lipid composition , optimized to stabilize the structure of the membrane in that state . The relative free-energies of the combined protein-membrane system in each state will therefore be dependent on the lipid types that are available , which the cell can alter through for example regulation of lipid synthesis and degradation pathways . While membrane protein reactions can be severely influenced in laboratory conditions through drastic changes in the chemical and physical state of membrane , a critical point to note is that a plausible regulatory mechanism must be effective in the actual range that is physiologically viable . Cell membranes are known to undergo changes in lipid composition due to many external factors ( Marr and Ingraham , 1962; Sanders and Mittendorf , 2011 ) ; yet , it is rare for a membrane to change its composition so much that its macroscopic structure is significantly altered . For instance , the membranes of E . coli cells grown at colder temperatures will primarily decrease chain saturation , with only minor changes to the amount of short-chain lipids . The resulting changes are presumed to maintain fluidity while maintaining an appropriate thickness of the membrane so that the majority of membrane proteins remain properly solvated and can still function optimally . Homeostatic adaptation of macroscopic membrane properties have been shown for mammalian cells also ( Levental et al . , 2020 ) . As far as we know , there is no situation where a cell will change the overall macroscopic thickness of its membranes due to a physiological stimulus . Therefore , when contemplating possible mechanisms of physiological regulation within the membrane , and particularly with chain-length in mind , we must consider that they should be consistent with low-level changes of these types of lipids within the membrane . Our experiments show that CLC-ec1 dimerization is sensitive to the amount of the short-chain DL lipid in the membrane; even at low levels , from 10−8 to 1% , we observe a gradual and non-saturating impact , indicating that dimerization is tunable without a global change in the state of the membrane . The resulting change in the dimer population , from 80% to 50% , could certainly impart a physiological effect if it was linked to a cell signaling function . It is equally important to note that the phenomenon of preferential solvation naturally allows for this gradual tuning , as opposed to what would be expected for a process of site-specific lipid-ligand binding , which would inhibit dimerization in a switch-like manner over a much narrower range of DL-lipid concentrations . While site-specific binding mechanisms may be at play for some types of processes and specific lipid types , we anticipate that preferential solvation effects will be found to control diverse kinds of membrane protein equilibria in physiological settings . Our examination of the impact of short-chain lipids on CLC-ec1 dimerization sheds light on a potentially ubiquitous mechanism of action by regulatory molecules within the membrane . In simulations , we observe the CLC-ec1 monomers force PO lipids to adopt non-native conformations , and many of the features of the bilayer near the dimerization interface , ranging from lipid tilt-angle to interleaflet contacts or water exclusion , are radically different from those in the bulk . When DL lipids are present , they disproportionately accumulate at this defect , spontaneously , while maintaining non-specific interactions . In doing so , DL lipids restore some of the bulk-like features to the bilayer near the dimerization interface . That is , although the defect remains , DL is a better solvent for it , and thereby stabilizes the dissociated monomeric state . A small lipid like DL could thus be considered a chemical that drives disaggregation , analogous to chaotropic denaturants stabilizing the un-folded states of proteins in aqueous solution . Many regulatory molecules in membranes are also small fatty acids; pharmacological agents like general anesthetics are small non-polar molecules as well . It is possible that these small lipoidal factors act similarly to DL in the problem examined here , and that they preferentially solvate and stabilize the membrane in states where local deformations and defects created by a protein become exposed . This may promote protein disaggregation , especially at high enough densities , and shift membrane protein equilibria to optimize activity . Altogether , our findings lead to the hypothesis that the complexity of lipid compositions found in biological cell membranes , leveraged through mechanisms such as preferential solvation , permits the cell to regulate and fine-tune the reactions of membrane proteins within – folding , oligomerization , and conformational changes – amidst the extremely variable conditions that life faces . It will be fascinating to continue to unravel the nature of these processes through further experimental and simulation studies . This study provides fundamental insights into an ubiquitous process in membrane physiology , namely protein oligomerization . It also yields a novel perspective of the mechanism by which cells could regulate the stability of membrane protein complexes through subtle variations in the lipid composition of the bilayer . Specifically , we have posited that a principal driving force for the oligomerization of membrane proteins stems from differences in the lipid solvation energetics of the associated and dissociated states . Such differences arise when one of the states in equilibrium introduces a perturbation in the bilayer , for example due to hydrophobic mismatch , which would not be naturally observed otherwise , therefore implying a significant free-energy cost from the membrane standpoint . A driving force that originates in the energetics of lipid solvation is by definition highly sensitive to the composition of the membrane . In this regard , the perspective that emerges from this study differs from models that postulate site-specific binding or global changes in the state of the membrane . In our perspective , the lipid bilayer is a system of co-solvents that can alter their spatial distribution so as to preferentially solvate one or more of the states of any given reaction . A particular state might be therefore favored or disfavored , statistically speaking , depending on the energetics of the solvation structure that is achievable by a given co-solvent mixture . It follows that minimal changes in the lipid composition of the membrane can have a profound effect on specific oligomerization reactions , without any global morphological changes that might broadly compromise protein functionality , that is , what is expected for a physiologically realistic regulatory process .
All simulations were calculated with GROMACS 5 . 2 . 1 ( Abraham et al . , 2015 ) using the MARTINI 2 . 2/ElNeDyn22 forcefield ( Wassenaar et al . , 2015 ) . Temperature and pressure were maintained constant at 303 . 15 K and 1 bar , using the velocity-rescale thermostat and the Parrinello-Rahman semi-isotropic barostat , respectively . Equations of motion were integrated using the leapfrog algorithm with a time step of 20 fs . Electrostatics were treated with the reaction field method using a cutoff of 1 . 2 nm . To ensure statistical significance , several independent runs were performed for each system ( see Figure 2—source data 1 for further details ) . The simulations are based on the crystal structure of wild-type CLC-ec1 dimer deposited in the Protein Data Bank , entry 1OTS ( resolution 2 . 51 Å ) ( Dutzler et al . , 2003 ) . Chloride ions were included at sites Scen and Sint , and E113 ( chains A and B ) and D417 ( chain A ) were protonated as indicated from electrostatics analysis ( Faraldo-Gómez and Roux , 2004 ) In the monomer state the N-terminus was truncated up to residue 30 , as this cytoplasmic helix , which domain-swaps in the dimer , is highly flexible and able to adopt alternate conformations ( Robertson et al . , 2010 ) The atomic structure was coarse-grained using the martinize tool; different mixtures of POPE , POPG , DLPE and DLPG lipids were then added around the protein and the systems solvated . The total system charge was neutralized by addition of Na+ ions , and the system buffered with NaCl to a concentration of 150 mM . The preparatory stages included an 15 , 000-steps energy minimization using the steepest-descent method , and a 5-ns equilibration to bring the system to desired temperature and pressure . To simplify the visualization and analysis of trajectories , the protein was not permitted to rotate around the Z-axis ( i . e . the membrane perpendicular ) or to diffuse away from the membrane center . Note this is strictly equivalent to re-defining the laboratory frame as the molecular frame for each snapshot , and thus these restrictions have no impact on the sampling of the internal configurational space . These orientational/translational restraints were implemented with PLUMED ( Bonomi et al . , 2009 ) ; specifically , two centers-of-mass , A and B , were defined using elements of helices H/P ( residues 406–409 , 411 , 412 , 194–196 , 197 , 198 ) and the linker regions between M/N and E/F ( residues 138 , 143–145 , 347 , 348 , 351–353 ) . In the monomer simulations , harmonic potentials were used to keep both A and B on the YZ plane and equidistant from the membrane center . In the dimer , centers A and B were combined into a single center per monomer , C , and the same restraints were applied to keep the dimer on the YZ plane . The vertical drift of the membrane was also removed prior to trajectory analysis , by re-centering each snapshot so that the midpoint of centers A and B is fixed in place . For the pure bilayer simulations , the same was accomplished by holding fixed the z-component of the membrane center . To map any given descriptor of the lipid structure onto the x-y plane , a grid consisting of square cells each with an area of 0 . 005 Å2 was constructed . Data derived from analysis of individual lipid molecules in each simulation snapshots were mapped onto specific grid points based on the XY position of the corresponding ester beads ( GL1 and GL2 ) ; specifically , data was added to all grid points contained within the van der Waals radius of the beads . The grid-point data was then averaged over the all trajectory snapshots . To ascertain which grid-points reflect statistically significant data , the frequency with which each grid point was assigned to any lipid , referred to hereafter as the occupancy number , was annotated . Grid points with less than 40% of the average occupancy were considered to be not statistically significant and excluded from graphical representations and/or global averages . Occupancy numbers were also used to quantify the enrichment of DL lipids in the mixed PO/DL systems , relative to the bulk ratio . Specifically , the percent enrichment at grid point i was computed as ( 1 ) %Ei= 100DLPOi-DLPOBDLPOBwhere ρDL and ρPO refer to the lipid occupancy number , for DLPX and POPX lipids respectively , and the subscripts i and B indicates the ratio at grid point i or the expected ratio in the bulk given the condition simulated , that is , if both lipid types were distributed evenly across the box . To compute the enrichment as a function of distance d from the protein ( or a specific interface ) , grid points within a mask centered at that distance and 10 Å in width were selected . The percent enrichment was then computed as ( 2 ) %Ed= 100DLPOM-DLPOBDLPOBwhere the subscript M refers to the sum of the occupancy numbers over the grid points found within each mask . All grid-based analysis tools are in-house software , available for download in https://github . com/TMB-CSB/Membrane-Analysis-Tools-Gromacs ( Bernhardt and Faraldo-Gómez , 2021 ) , with the exception of the 3D density maps , which were calculated using the volmap plugin of VMD ( Humphrey et al . , 1996 ) . For more details on the grid-based lipid metrics analysis see Figure 2—figure supplement 2 . Detergent solubilized lipids were prepared as described before ( Chadda et al . , 2016 ) with the modification that dry lipids were solubilized in 2:1 chloroform:methanol followed by two washes in 3:1 pentane:dichloromethane . This was done due to the fact that DL lipids ( DLPE or 12:0 PE; 1 , 2-dilauroyl-sn-glycero-3-phosphoethanolamine and DLPG or 12:0 PG; 1 , 2-dilauroyl-sn-glycero-3-phospho- ( 1'-rac-glycerol ) ( sodium salt ) ) , unlike PO lipids , were found to be insoluble in chloroform or pentane alone . For a typical preparation , 4 mL of POPE , and 2 mL of POPG ( 25 mg/mL stocks in chloroform , Avanti Polar Lipids Inc ) were combined in a glass vial ( 22 mm; RPI , Malvern , PA ) . The chloroform was evaporated under a continuous stream of 0 . 22 μm filtered N2 gas ( Ultra High Purity Nitrogen 5 . 0 Grade; Airgas ) . The dried lipid mass , was dissolved at least once in 2:1 chloroform/methanol followed by 1–2 washes in 3:1 pentane/dichloromethane and drying while rotating , leaving a thin film of lipids along the walls and the bottom of the glass vial . The lipid film , containing 150 mg total lipids ( 100 mg POPE +50 mg POPG ) was dried under continuous stream of N2 , approximately 10–12 min . Next , after addition of 161 . 3 mg ( 21 . 5 mg/ml ) CHAPS and 7 . 5 ml Dialysis Buffer ( DB: 300 mM KCl , 20 mM citrate pH 4 . 5 ( adjusted with NaOH ) ) sonication was performed leading to a translucent suspension of the CHAPS/POPE/POPG mixture . The final concentration of components was 20 mg/mL 2:1 POPE/POPG ( mass ratio ) and 35 mM CHAPS . DL lipids were prepared as follows: 40 mg DLPE , and 20 mg DLPG ( powder , Avanti Polar Lipids Inc ) were added to a glass vial . The solids were solubilized in 2:1 chloroform/methanol and then were taken through the identical washing procedure as PO lipids until a thin , uniform lipid film resulted after drying . Next 64 . 5 mg CHAPS and 3 mL DB was added followed by sonication . The final concentration of components was 20 mg/mL 2:1 DLPE/DLPG ( mass ratio ) and CHAPS at 35 mM . Finally , the PO and DL master stocks were mixed in different ratios ( volume/volume ) resulting in the quaternary lipid mixtures . For instance , to prepare 1 mL of the 20% DL ( w/w ) lipid mixture , 0 . 8 mL of the PO lipid stock was mixed with 0 . 2 mL of the DL lipid stock and used immediately for CLC reconstitution . For reference , the conversion of % DL ( w/w ) to mole fraction and molality are presented in Figure 5—source data 3 . The 2:1 PE/PG - PO and DL 25 mg/mL lipid stocks were solubilized in 2:1 chloroform:methanol as described above , and then mixed together to yield the following titration - 0 , 10 , 30 , 50 , 70 , 90 , 100 %DL . After mixing , the lipids were dried under N2 gas , and solubilized in DB ( 10–15 mg/mL ) by sonication yielding the formation of small unilamellar vesicles . These samples were freeze-thawed 7x to form multi-lamellar vesicles , which were stored at room temperature and examined by DSC days-weeks after preparation . Samples were degassed prior to measurement , and data was collected using a MicroCal VP-DSC differential scanning calorimeter . Data were collected at multiple scan rates to ensure that there was minimal influence of the scan rate on the measured melting transition . Presented data were collected on heating from 2°C to 50°C with a scan rate of 30°C/hr and were baseline corrected . Source data is provided in Figure 3—source data 3 . –source data and 1 . Cryo-electron microscopy ( EM ) imaging and analysis of images was performed as described earlier ( Chadda et al . , 2018; Cliff et al . , 2020 ) . Briefly , liposomes were freeze-thawed seven times , and then extruded through a 400 nm nucleopore filter ( GE Life Sciences ) 21 times . Three µL of the undiluted sample was loaded onto a glow-discharged Lacey carbon support film ( Electron Microscope Sciences ) , blotted , and plunged into liquid ethane using a Vitrobot System ( FEI ) . Images were collected on a FEI Titan Krios G3 300kV Cryo-TEM microscope with a Gatan K2 Summit Direct electron detector ( GATAN ) . Magnifications of 6500x , 33 , 000x , and 53 , 000x were used . For size determination , liposomes were manually outlined in Fiji and ImageJ ( Schindelin et al . , 2012; Schindelin et al . , 2015 ) to measure the outer radii of all liposomes , including those located on the carbon . Multilamellarity was manually counted as the fraction of vesicles containing more than one bilayer . Liposome size distribution source data is provided in Figure 4—source data 1 . Liposomes were prepared by drying as described previously , then sonicating the dried lipid films in reconstitution buffer prepared with 99 . 9% pure D2O ( Cambridge Isotopes ) . Note , the pD of the buffer was measured by soaking the pH electrode in pure D2O for several minutes and then adjusted with NaOD for a final pD of 4 . 5 ( Krezel and Bal , 2004 ) . Prior to measurement , liposomes were freeze-thawed following the procedure described previously , then extruded in two steps , first through 400 nm filters and then through 100 nm nucleopore membranes . SANS data were collected on the NGB30SANS instrument at the NIST Center for Neutron Research at the National Institute of Standards and Technology ( NIST ) . Data were collected using a neutron wavelength ( λ ) 6 Å and a wavelength spread ( Δλ/λ ) of 0 . 12 with sample to detector distances of 1 m , 4 m , and 13 m . Additional data were collected using λ = 8 . 4 Å with a sample to detector distance of 13 m . These instrument configurations provided access to a q-range of 0 . 001 Å−1 < q < 0 . 04 Å−1 where q is the scattering vector and is defined as q = 4πλ−1sin ( θ/2 ) and θ is the scattering angle . Samples were sealed in titanium cells with quartz windows and sample temperature was controlled at 25°C ( ±0 . 1°C ) during data acquisition . Data were reduced to absolute intensity using the macros provided by NIST ( Kline , 2006 ) . SANS data were analyzed with the multilamellar form factor in the SasView application . The data showed a broad shoulder at q ≈ 0 . 06 Å−1 due to the presence of a mixture of unilamellar and multilamellar vesicles ( Scott et al . , 2019 ) . SANS data were fit with an array distribution of N , where N is the number of lamellar shells and the reported results are for the distribution that gave the best fit to the data , defined as the minimum χ2 value . Approximately 85–90% of the vesicle population contained a single lamella which was in good agreement with Cryo-EM experiments that confirmed the presence of ≈85% unilamellar , ≈ 10% bi-lamellar ( vesicles containing two bilayers ) and ≈5% multilamellar vesicles ( vesicles with three or more bilayers ) . Cryo-EM imaging of the liposomes also showed a bimodal distribution of vesicle sizes . The SANS analysis fixed the distribution of outer vesicle radii based on the cryo-EM results and only fit the data for q > 0 . 015 Å−1 where the form factor contribution from the vesicle radii were constant ( Pencer et al . , 2006 ) . The parameters fit during the analysis were the bilayer thickness ( db ) , the water layer thickness ( dw ) and the scattering length density of the bilayer ( results not shown ) . Source data is provided in Figure 3—source data 2 . DNA constructs for CLC-ec1 C85A/H234C ( WT ) , C85A/H234C/I201W/I422W ( WW ) ( Chadda et al . , 2016 ) and C85A/H234C/R230C/L249C ( RCLC ) were described previously ( Chadda et al . , 2018 ) . Expression and purification of these CLC-ec1 variants was carried out as described earlier ( Chadda et al . , 2016 ) . Briefly , proteins were overexpressed in BL21-AI E . coli competent cells and extracted into 2% n-Decyl-β-D-Maltopyranoside ( DM; Anatrace , Maumee OH ) containing 5 mM TCEP ( Tris ( 2-carboxyethyl ) phosphine; Soltec Bioscience , Beverly , MA ) . After removing cellular debris by centrifugation , the protein was affinity purified using TALON cobalt affinity resin ( Clontech Laboratories , Mountain View , CA ) followed by size exclusion chromatography on Superdex 200 10/30 GL size exclusion column ( GE Healthcare , Little Chalfont , UK ) into size exclusion buffer ( SEB ) : 150 mM NaCl , 20 mM MOPS pH 7 . 0 , 5 mM analytical-grade DM . Addition of TCEP during purification ensures that the engineered cysteine at the residue H234C remains reduced and available for maleimide labeling . This can be quantitatively estimated after reacting the purified protein with Ellman’s reagent ( DNTB , 5 , 5’-Dithio-bis ( 2-nitrobenzoic acid ) ; Sigma-Aldrich ) as described before ( Chadda et al . , 2016 ) . Fluorescent labeling of the protein is conducted in presence of 5X Cy5-maleimide followed by separation of unreacted dye using affinity and size-exclusion chromatography . Quantification of the Cy5 labeling yield per subunit , PCy5 , was carried out as described previously ( Chadda et al . , 2016; Chadda and Robertson , 2016 ) . For reconstitution , Cy5-labeled protein is mixed 20 mg/mL lipids in DB + 35 mM CHAPS and then dialyzed in independent buckets to prevent the possibility of cross-contamination between different lipid compositions . Note , the effect of contamination during dialysis appears negligible in our experiments , as we quantified it by photobleaching analysis and observed a small , non-significant difference ( Figure 4—figure supplement 1E ) . Chloride transport assays from 400 nm extruded liposomes were performed as described earlier ( Walden et al . , 2007; Chadda et al . , 2016 ) . Functional measurements were carried out 6 . 4 ± 6 . 1 days ( mean ± std , n = 2–5 ) after freeze/thaw and sample incubation in the dark , at room temperature . Chloride transport was quantified in two ways , by fitting the initial slope by linear regression , kinit . , or fitting the full transport trace to the following exponential association function:norm . Cl-=F0 , vol . 1-e-kleakt+ ( 1-F0 , vol . ) 1-e-kPt All data are listed in Figure 4—source data 2 . Proteoliposomes samples were extruded , imaged on TIRF microscope , and the videos analyzed for counting single-molecule photobleaching steps as described earlier ( Chadda et al . , 2016; Chadda and Robertson , 2016; Chadda et al . , 2018 ) . Briefly , dialyzed proteoliposomes were freeze-thawed seven times leading to formation of large multilamellar vesicles ( MLVs ) . The samples were stored at room temperature , in the dark , with 0 . 02% NaN3 until extrusion and single-molecule imaging . Overall , the Cy5 labeling yield was PCy5 = 0 . 663 ± 0 . 005 ( mean ± sem , n = 27 ) for wild-type CLC-ec1 samples . Imaging was carried out 3–15 days after freeze-thaw and sample incubation in the dark , at room temperature . Images were analyzed as described previously using the imscroll software in MATLAB ( Friedman and Gelles , 2015 ) . To quantify the underlying dimerization reaction from the photobleaching data , the same methods described in were followed ( Chadda et al . , 2016; Chadda and Robertson , 2016; Chadda et al . , 2018 ) . Briefly , photobleaching probability distributions ( P1 , P2 , P3+ ) were determined for each construct as a function of protein density and lipid composition . The fraction of dimer in the protein population , FDimer , was estimated by least-squares fitting of the linear combination of the probability distributions for the monomer and dimer controls under similar conditions . Equilibrium constants were obtained by fitting the data to an equilibrium dimerization isotherm , ( 3 ) FDimer=1+4Keqχ*-1+8Keqχ*4Keqχ*and then converted to ΔG∘=−RTln ( Keq ) , standard state = 1 subunit/lipid . All data is listed in Figure 4—source data 3–6 and Figure 5—source data 1 . | A cell’s outer membrane is made of molecules called lipids , which band together to form a flexible thin film , just two molecules thick . This membrane is dotted with proteins that transport materials in to and out of cells . Most of these membrane proteins join with other proteins to form structures known as oligomers . Except , how membrane-bound proteins assemble into oligomers – the physical forces driving these molecules to take shape – remains unclear . This is partly because the structural , physical and chemical properties of fat-like lipid membranes are radically different to the cell’s watery interior . Consequently , the conditions under which membrane oligomers form are distinct from those surrounding proteins inside cells . Membrane proteins are also more difficult to study and characterize than water-soluble proteins inside the cell , and yet many therapeutic drugs such as antibiotics specifically target membrane proteins . Overall , our understanding of how the unique properties of lipid membranes affect the formation of protein structures embedded within , is lacking and warrants further investigation . Now , Chadda , Bernhardt et al . focused on one membrane protein , known as CLC , which tends to exist in pairs – or dimers . To understand why these proteins form dimers ( a process called dimerization ) Chadda , Bernhardt et al . first used computer simulations , and then validated the findings in experimental tests . These complementary approaches demonstrated that the main reason CLC proteins ‘dimerize’ lies in their interaction with the lipid membrane , and not the attraction of one protein to its partner . When CLC proteins are on their own , they deform the surrounding membrane and create structural defects that put the membrane under strain . But when two CLC proteins join as a dimer , this membrane strain disappears – making dimerization the more stable and energetically favorable option . Chadda , Bernhardt et al . also showed that with the addition of a few certain lipids , specifically smaller lipids , cell membranes become more tolerant of protein-induced structural changes . This might explain how cells could use various lipids to fine-tune the activity of membrane proteins by controlling how oligomers form . However , the theory needs to be examined further . Altogether , this work has provided fundamental insights into the physical forces shaping membrane-bound proteins , relevant to researchers studying cell biology and pharmacology alike . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2021 | Membrane transporter dimerization driven by differential lipid solvation energetics of dissociated and associated states |
The majority of human breast cancer is estrogen receptor alpha ( ER ) positive . While anti-estrogens/aromatase inhibitors are initially effective , resistance to these drugs commonly develops . Therapy-resistant tumors often retain ER signaling , via interaction with critical oncogenic coregulator proteins . To address these mechanisms of resistance , we have developed a novel ER coregulator binding modulator , ERX-11 . ERX-11 interacts directly with ER and blocks the interaction between a subset of coregulators with both native and mutant forms of ER . ERX-11 effectively blocks ER-mediated oncogenic signaling and has potent anti-proliferative activity against therapy-sensitive and therapy-resistant human breast cancer cells . ERX-11 is orally bioavailable , with no overt signs of toxicity and potent activity in both murine xenograft and patient-derived breast tumor explant models . This first-in-class agent , with its novel mechanism of action of disrupting critical protein-protein interactions , overcomes the limitations of current therapies and may be clinically translatable for patients with therapy-sensitive and therapy-resistant breast cancers .
Endocrine therapies for estrogen receptor alpha ( ER ) -positive breast cancer involve modulation of ER signaling using either antiestrogens ( AE ) or aromatase inhibitors ( AI ) . However , most patients develop resistance to these drugs , and disease progression is common , with progression to metastases ( Musgrove and Sutherland , 2009; Ma et al . , 2015 ) . ER signaling is complex and involves coregulators ( McDonnell and Norris , 2002; O'Malley and Kumar , 2009 ) . Therapy-resistant tumors often retain ER-expression and ER-signaling . While multiple mechanisms maintain ER signaling in therapy-resistant tumors , ER signaling is mediated by the interactions between activated ER and critical coregulator proteins ( Dasgupta and O'Malley , 2014; Lonard et al . , 2007 ) . Alterations in the concentration or activity of selective coregulators enable ER-signaling from AE-ER complexes , effectively converting the antagonist to an agonist ( O'Hara et al . , 2012; Kurebayashi , 2003 ) . Over a third ( 38% ) of ER coregulators identified in breast cancer are over-expressed ( Lonard et al . , 2007; Lonard and O'Malley , 2012; Cortez et al . , 2014 ) , such as SRC3 ( AIB1 ) ( List et al . , 2001; Azorsa et al . , 2001 ) , SRC2 ( Kurebayashi et al . , 2000 ) , and PELP1 ( Habashy et al . , 2010 ) . These deregulated coregulators contribute to mammary tumorigenesis ( Cortez et al . , 2014 ) , therapy resistance and metastases ( Kumar et al . , 2009; Shou et al . , 2004; Burandt et al . , 2013; Girard et al . , 2014 ) . Recent studies revealed that breast tumors acquire mutations in the ER ligand binding domains ( L536N , Y537S , Y537N and D538G ) that facilitate constitutive activity of these mutant ER ( MT-ER ) in the absence of ligand ( Toy et al . , 2013; Robinson et al . , 2013; Jeselsohn et al . , 2014; Merenbakh-Lamin et al . , 2013 ) . Tumors with MT-ER interact with oncogenic coregulators to drive ER-dependent transcriptional programs and proliferation and are poorly responsive to AEs and AIs ( Toy et al . , 2013; Robinson et al . , 2013; Jeselsohn et al . , 2014; Merenbakh-Lamin et al . , 2013; Toy et al . , 2017 ) . Thus , there is a strategic need for drugs that disrupt interactions between ER and critical coregulators to block ER signaling . In this study , we have synthesized a series of small organic molecules to emulate the nuclear receptor ( NR ) box motif , important for ER coregulator interactions . We have identified a small molecule named as ER coregulator binding modulator-11 ( ERX-11 ) , with potent anti-proliferative activity against ER-driven breast tumors . ERX-11 interacts with ER and blocks the interaction between ER and coregulators . In ER-expressing breast cancer cells , ERX-11 blocks the proliferation and induces apoptosis . ERX-11 has no activity against ER-negative breast cancer cells .
The peptidomimetic D2 blocks the interaction between the androgen receptor ( AR ) and NR-box containing coregulators , such as PELP1 , with an IC50 of 40 nM ( Ravindranathan et al . , 2013 ) . String analyses of the PELP1 interactome suggested an equally robust interaction between PELP1 and ER as that between PELP1 and AR ( Figure 1—figure supplement 1A ) . However , D2 was unable to block the interaction between PELP1 and ER ( data not shown ) and required much higher concentrations ( µM range ) to block the proliferation of ER-driven MCF-7 breast cancer cells ( Figure 1—figure supplement 1B ) . Since sequences flanking the NR-box may influence the affinity and selectivity of coregulator interactions ( McInerney et al . , 1998 ) , we hypothesized that a longer oligo-benzamide scaffold may more effectively target the interaction between ER and coregulators . This strategy generated a series of tris-benzamides ( see Appendix 1—chemical structures 1—22 ) ( Figure 1A ) that added a functional group ( R ) to the D2 bis-benzamide , corresponding to the amino acid side chain groups found at the i-3/4 or i + 7 position surrounding the NR-box sequences ( Figure 1A , Figure 1—figure supplement 2G ) . Importantly , each of these small molecules were named as ER coregulator binding modulators ( ERXs , where X refers to their multiple potential and unknown targets ) had differential activities in ER-positive breast cancer cells ( Figure 1A , Figure 1—figure supplement 1C ) , confirming our hypothesis that the sequences flanking the core LXXLL motif could determine specificity and activity . The ERXs maintained the structural requirement for mimicking helices ( confirmed by molecular modeling ( MacroModel , Schrodinger , NY ) ( Figure 1—figure supplement 2A , shown for ERX-11 ) . 10 . 7554/eLife . 26857 . 003Figure 1 . Derivation and characterization of ERX-11 . Structure of ERX-11 , as a derivative of the D2 peptidomimetic with a hydroxyethyl moiety in the flanking position to mimic a Serine ( A , left panel ) . Effect of 500 nM of each peptidomimetic on the growth of MCF-7 cells using MTT cell viability assay is shown as percentage inhibition of the growth of E2-treated control cells ( A , right panel ) . Effect of increasing doses of ERX-11 on the cell viability of ER-positive ( B ) and ER-negative ( C ) breast cancer cells using the MTT cell viability assay . Molecular docking studies on the interactions between ERX-11 and ER using AutoDock Vina . Superimposition of the docked ERX-11 ( green ) on the crystal structure ( PDB code 1L2I ) of the LXXLL motif ( orange ) ( D ) . Purified full-length ER was incubated with biotin-control or biotin-ERX-11 in the presence of E2 . ERX-11 interaction with purified ER was analyzed using avidin bead pulldown and western blotting ( E ) . Purified full-length ER was incubated with biotin-ERX-11 in the presence of E2 ± free ERX-11 ( 1 µM ) . ERX-11 ability to compete with the binding of biotin ERX-11 with ER was analyzed using avidin pulldown assay ( F ) . Purified full-length ER was incubated with biotin-ERX-11 in the presence of E2 ±LXXLL peptides ( 1 µM ) from various coregulators SRC1 , SRC2 , AIB1 , and PELP1 . LXXLL peptides ability to compete with the binding of biotin ERX-11 with ER was analyzed using avidin pulldown assay ( G ) . Purified full-length ER was incubated with biotin-ERX-11 in the presence of E2 ± GDC0810 , AZD-9496 , ICI , and Tam ( 1 µM ) and their ability to compete with the binding of biotin ERX-11 with ER was analyzed using avidin pulldown assay ( H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 00310 . 7554/eLife . 26857 . 004Figure 1—figure supplement 1 . Derivation of the α-helix mimetic ERX-11 and structural design , synthesis and activity of the α-helix mimetic . ( A ) String analyses depicting the interactome of PELP1 , especially with ER and AR using string-db . org software v 10 . 4 . ( B ) Effect of increasing concentrations of D1 and D2 peptidomimetic on the cell viability of C4-2 prostate cancer cells and MCF-7 breast cancer cells . ( C ) Effect of ERXs on the growth inhibition of ZR-75 , BT474 , T47D and C4-2 . ( D ) Waterfall graph displaying the effect on growth by 1 μM ERX-11 in a number of ER-positive and ER-negative cell lines . ( E ) Waterfall graph comparing the activity of ERX-11 in ER-positive cell lines to fulvestrant ( ICI ) or tamoxifen . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 00410 . 7554/eLife . 26857 . 005Figure 1—figure supplement 2 . Characterization of ERX-11 activity . ( A ) The lowest energy conformation of ERX-11 ( left ) , Superimposition of the energy-minimized ERX-11 ( green ) on an α-helix ( orange ) ( middle ) and an α-helical LXXLL motif containing a Ser at its flanking sequence ( orange ) ( right ) . ( B ) Effect of increasing doses of ERX-11 on the cell viability of MDA-MB-231 and ZR-75 cell lines in the presence of E2 . ( C ) ZR-75 cells were stimulated with E2 ( 10−8M ) for 7 days in the presence of ERX-11 ( 1 µM ) or tamoxifen ( 1 µM ) or ICI ( 1 µM ) and cell viability was measured by MTT assay . ( *p<0 . 05 , ****p<0 . 0001 ) . ( D , E ) ZR-75 and MCF-7 cells were stimulated with E2 ( 10−8M ) for 7 days in the presence of indicated concentrations of ICI or ERX-11 and cell viability was measured by MTT assay . ( F ) ZR-75 cells were stimulated with E2 ( 10−8M ) for 7 days in the presence of indicated concentrations of ERX-11 or tamoxifen or in combination and cell viability was measured by MTT assay ( ****p<0 . 0001 ) . ( G ) Relative frequency of various amino acid residues at the i-3/4 or i + 7 position flanking the core LXXLL domain in a large number of proteins with known LXXLL domains . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 00510 . 7554/eLife . 26857 . 006Figure 1—figure supplement 3 . Derivation of the α-helix mimetic ERX-11 and synthesis of tris-benzamide peptidomimetics . ( A ) Synthesis of 3-alkoxy-4-nitrobenzoic acids . ( B ) Solid-phase synthesis of tris-benzamdies ( Reagents and conditions: ( a ) 3d , HATU , DIEA , DMF , rt , 24 hr; ( b ) SnCl2⋅2H2O , AcOH/HCl/THF , rt , 24 hr; ( c ) 3 , HATU , DIEA , DMF , rt , 24 hr; ( d ) Piperidine for 7i; ( e ) N , N-Di-Boc-1H-pyrazole-1-carboxamide , DIEA , DMF , rt , 12 hr for 7 j; ( f ) TFA/TIS/H2O ( 95:2 . 5:2 . 5 ) , rt , 1 hr ) . ( c ) Synthesis of ERX-11 ( Reagents and conditions: ( a ) HATU , DIEA , DMF , rt , 24 hr; ( b ) TFA , rt , 30 min ) ( D ) Synthesis of biotinylated ERX-11 . ( E ) Pull down of ER from Tamoxifen-resistant MCF-7 cells using biotinylated ERX-11 was not affected by the presence of Tamoxifen . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 006 Within this series of tris-benzamides , one compound ( ERX-11 ) with a hydroxyethyl functional group mimicking a serine residue ( Figure 1A ) was consistently able to block 17-β-estradiol ( E2 ) -induced proliferation in 8/8 ER-positive breast cancer cell lines ( Figure 1B , Figure 1—figure supplements 1C , D and 2B ) , with an IC50 , ranging from 250 to 500 nM . In contrast , no effect of ERX-11 was noted in 11/12 ER-negative cells , with modest activity in the MDA-MB-157 cell line ( Figure 1C ) . ERX-11 was as effective as tamoxifen or ICI in reducing the growth of ZR-75 and MCF-7 cells ( Figure 1—figure supplements 1E and 2C , D , E ) , and the combination of ERX-11 and tamoxifen was not additive ( Figure 1—figure supplement 2F ) . While several compounds , including selective estrogen receptor modulators , have been shown to have similar activity against ER-positive breast cancers , the novel chemical structure , potential for a unique mechanism of action led to the designation of ERX-11 as a lead compound . We then established a protocol for large-scale batch synthesis ( Figure 1—figure supplement 3A–C ) . Since ERX-11 mimics the NR-box , we expected its direct binding to ER . Modeling studies using Autodock ( The Scripps Research Institute , La Jolla , CA ) indicated that ERX-11 could bind to the AF-2 domain of ER ( Figure 1D ) . Using biotinylated ERX-11 ( synthesis described in Figure 1—figure supplement 3D ) , we showed that ERX-11 interacts in vitro with purified ER protein ( Figure 1E ) . Addition of ‘cold’ ERX-11 efficiently competed for the binding of purified ER protein to biotinylated ERX-11 ( Figure 1F ) . Further , short 15mer peptides , corresponding to NR box sequences within the SRC1 , SRC2 , AIB1 and PELP1 proteins , efficiently disrupted the interaction between biotinylated ERX-11 and purified ER ( Figure 1G ) . However , not all LXXLL peptides interfere ERX-11 interaction with ER , as only peptide surrounding the third PELP1 LXXLL motif , but not the first PELP1 LXXLL motif blocked ERX-11 interaction suggesting ERX-11 can only block some LXXLL interactions ( Figure 1G , last two panels ) . Further , pre-incubation of the purified ER protein with selective estrogen receptor degraders ( SERDs ) GDC-0810 or AZD-9496 , or fulvestrant ( ICI ) was unable to block the interaction between ER and ERX-11 ( Figure 1H ) . In contrast , tamoxifen was able to block the interaction between purified ER and ERX-11 ( Figure 1H ) , suggesting similarities in the ER-binding pockets of ERX-11 and tamoxifen . We then demonstrated that biotinylated ERX-11 could pull down endogenous ER in ZR-75 nuclear extracts ( Figure 2A ) . These data indicate that ERX-11 directly interact with ER , both as purified protein and within a cellular context . Unbiased evaluation , using immunoprecipitation mass spectrometric ( IPMS ) analyses of the biotinylated ERX-11 pulldown in MCF-7 cells , identified ER as one of the top ERX-11 interactors ( Figure 2B , Table 1 ) . Pathways analysis revealed that ERX-11-binding proteins were involved in the activation of transcriptional regulation ( Figure 2—figure supplement 1 ) . Importantly , ERX-11 pulldown included a number of proteins other than ER , including a weak affinity for the progesterone receptor ( PGR ) and several ER-associated proteins ( Table 1 ) . Immunoprecipitation analyses in MCF-7 cells validated the strong affinity of ERX-11 for ER , and weak affinity for the PR-A isoform but not GR , AR or PR-B isoforms ( Figure 2C ) . 10 . 7554/eLife . 26857 . 007Figure 2 . ERX-11 interacts with ER and blocks its interactome . Interaction with endogenous ER was evaluated in nuclear lysates prepared from ZR-75 cells stimulated with E2 , incubated with biotin-control or biotin-ERX-11 and analyzed by avidin pull-down assay ( A ) . Nuclear lysates from MCF-7 cells were incubated with biotinylated ERX-11 for 2 hr and then subject to a streptavidin column . The bound proteins were eluted and subjected to analyses by mass spectroscopy . The fold-enrichment in binding over basal is depicted for two independent replicates ( x axis = replicate one and y axis = replicate 2 ) . Relative binding of ER and PGR are shown ( B ) . MCF-7 nuclear lysates were incubated with biotinylated ERX-11 and then subject to a streptavidin column . The bound proteins were eluted and evaluated by western blotting compared to equivalent amount of input ( C ) . ZR-75 cells were incubated with tamoxifen and its ability to interfere ERX-11 binding to ER was analyzed by immunoprecipitation followed by western blotting ( D ) . To confirm ERX-11 binding to ER-ligand-binding domain ( AF2 ) , a GST pull-down assay was performed . Biotinylated ERX-11 interacted with the GST-AF2 domain of ER but not with the GST-AF1- or GST- DNA-binding domain of ER ( E ) . Purified ER-AF2 domain was incubated with biotin-ERX-11 in the presence of E2 ±ICI or tamoxifen ( 1 µM ) . ICI and tamoxifen ability to compete with binding of biotin ERX-11 with ER-AF2 domain was analyzed using avidin pull-down and western blotting ( F ) . Venn diagram shows the overlap between the ER-binding proteins immunoprecipitated from nuclear lysates from E2-stimulated MCF-7 cells following treatment with vehicle or ERX-11 or tamoxifen ( G ) . Co-immunoprecipitation analyses show the effect of ERX-11 on the interaction of ER with coregulators PELP1 , SRC1 , TIF1α in MCF-7 cells ( H ) . Proximity ligation assay validated the ability of ERX-11 and tamoxifen to disrupt the interactions between ER and coregulators such as PELP1 , SRC1 , SRC3/AIB1 and ARID1B in MCF-7 cells ( I ) . NanoBiT assay: expression plasmids were created to express either ER or PELP1 in conjunction with the large bit or the small bit of the NanoBiT luciferase enzyme . If the proteins directly interact within the cell , the two parts of the NanoBiT luciferase enzyme come together and create a quantifiable luminescent signal . The effect of ERX-11 on the interaction between the two sets of ER and PELP1 constructs is shown ( J ) . Validation of the binding of ERX-11 to the ER-AF2 domain was further explored using AF2-domain mutants of ER stably transfected in ER-negative MDA-MB-231 breast cancer cell lines . Biotinylated ERX-11 was then used to pull-down ER from these cell lines ( K ) . Data shown are the means of ±SEM performed in triplicate wells . *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 00710 . 7554/eLife . 26857 . 008Figure 2—figure supplement 1 . Pathways analysis in terms of either biological processes or molecular functions revealed that ERX-11-binding proteins were involved in the activation of multiple pathways leading to transcriptional regulation . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 00810 . 7554/eLife . 26857 . 009Figure 2—figure supplement 2 . Characterization of ERX-11 interactions with ER . ( A ) Nuclear lysates from E2-stimulated MCF-7 cells treated with vehicle or ERX-11 were subjected to immunoprecipitation with ER antibody . The immunoprecipates were analyzed by mass spectroscopy . The blue box represents the 222 proteins bound to ER in MCF-7 cells , whereas the orange box shows the 91 proteins whose binding to ER is disrupted by ERX-11 ( B ) Venn diagram shows the overlap between proteins binding with biotinylated ERX-11 in MCF-7 and ZR-75 cells . ( C ) Nuclear lysates from MDA-MB-231 cells treated with vehicle or biotin-ERX-11 were subject to immunoprecipitation using avidin beads . The avidin-biotin-ERX-11 precipitates were analyzed by mass spectroscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 00910 . 7554/eLife . 26857 . 010Figure 2—figure supplement 3 . Analyses of the ER-binding proteins blocked by ERX-11 or Tamoxifen . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 01010 . 7554/eLife . 26857 . 011Figure 2—figure supplement 4 . Effect of ERX-11 on inhibition of ER -coregulators interactions . ( A ) Quantification of the effect of ERX-11 on the proximity ligation between ER and PELP1 or SRC1 or SRC3 in MCF-7 and ZR-75 cells . ( B ) Quantification of the effect of ERX-11 on the proximity ligation between ER and ARID1B in MCF-7 cells . ( C ) ZR-75 cells were co-transfected with ERα-VP16 and ERE-Luc reporter constructs . After 24 hr , cells were treated with vehicle or ERX-11 or Tam or ICI and ERE-Luc activity was measured . ( D ) Ishikawa cells were transfected with ERα and ERE-Luc reporter constructs and after 24 hr treated with indicated concentrations of E2 , Tam or ERX-11 . After 24 hr , reporter activity was measured . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 01110 . 7554/eLife . 26857 . 012Figure 2—figure supplement 5 . Effect of SERDs or tamoxifen on ERX-11 interactions with ER . ( A ) Effect of 1 μM of a SERD , GDC-0810 , on the ability of ERX-11 to pulldown ER in MDA-MB-231 cells stably transfected with WT-ER , L540Q ER or ▲12 ER . Input is shown on the right . ( B ) Effect of 1 μM of a SERD , AZD-9496 , on the ability of ERX-11 to pulldown ER in MDA-MB-231 cells stably transfected with WT-ER , L540QER or ▲12 ER . Input is shown on the right . ( C , D ) MDA-MB-231-expressing WT-ER or ▲12 ER were treated with tamoxifen 1 µM , 5 µM , 10 µM for 30 min , E2 ( 1 nM ) for 90 min , and subjected to biotin-ERX-11 pull down followed by western blotting . ( E ) ER was introduced into ER-negative breast cancer model MDA-MB-231 by transfecting WT-ER , or L540QER , or ▲12ER and treated with ERX-11 and cell viability was measured using MTT assays . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 01210 . 7554/eLife . 26857 . 013Figure 2—figure supplement 6 . Models showing the putative interactions of ERX-11 with residues in the ER protein . ( A ) Docking studies of ERX-11 on different ER crystal structures shows that ERX-11 ( green ) can interact with ER , either in the presence of agonist diethylstilbesterol ( red ) ( A: 3ERD . pdb ) , an antagonist , tamoxifen ( red ) ( B: 3ERT . pdb ) , a corepressor peptide ( red ) ( C: 2JFA . pdb ) or a fulvestrant analog ( ICI164 , 384 ) ( red ) ( D:1HJ1 ) . Importantly , tamoxifen displaces ERX-11 from its preferred binding site on ER to a secondary site . The AF2-binding pocket is outlined in the red dotted box . ( B ) Model showing the side chains and interactions of ERX-11 with residues in the ER protein using 1L2I . pdb structure . ( C ) Docking structure of ERX-11 on ER crystal structures of 3ERT and 5ACC with the deletion of the helix 12 . ( A ) ERX-11 ( green ) binds to the tamoxifen-binding site , away from the AF2 domain ( boxed in red ) . ( B ) Superimposition of ERX-11 ( green ) and tamoxifen ( red ) showing the overlap in their binding sites . ( C ) ERX-11 ( green ) still binds to the AF2 domain with the deletion of the helix 12 of 5ACC . ( D ) Superimposition of ERX-11 ( green ) and SERD ( red ) showing no overlap in their binding sites . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 01310 . 7554/eLife . 26857 . 014Figure 2—figure supplement 7 . Model describing interaction between ER ( purple ) and ERX-11 ( green ) in presence of agonist ( yellow ) ( A ) , SERD ( orange ) ( B ) or tamoxifen ( red ) ( C ) . Note that in the presence of tamoxifen , ERX-11 binds to a secondary weaker affinity site on ER . Similarly , the interaction between ER▲12 ( blue ) and ERX-11 is modeled in the absence of agonist ( D ) , SERD ( E ) or tamoxifen ( F ) . Note that in the presence of tamoxifen , ERX-11 does not bind to a secondary site on ER▲12 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 01410 . 7554/eLife . 26857 . 015Table 1 . Top proteins pulled down by biotinylated ERX-11 in MCF-7 cells , as identified by IP-MS . The column marked E2 represents spectral counts for the protein bound to biotinylated control eluted from avidin column , under conditions of E2 stimulation . The column marked E2 +ERX-11 represents spectral counts for proteins bound to biotinylated ERX-11 eluted from an avidin column with E2 stimulation . The column marked E2 + ERX-11/E2 represents the ratio of binders . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 015ProteinDescriptionLength ( AA ) Mw ( Da ) PSMsPeptide seqs% CoverageE2E2 + ERX-11E2 + ERX-11/E2 P38117ETFB_HUMAN Electron transfer flavoprotein subunit beta OS = Homo sapiens GN = ETFB PE = 1 SV = 325537501 . 1012734 . 1017 . 005 . 00 Q96PZ0PUS7_HUMAN Pseudouridylate synthase seven homolog OS = Homo sapiens GN = PUS7 PE = 1 SV = 266175186 . 30201426 . 301 . 007 . 9810 . 99 O953366 PGL_HUMAN 6-phosphogluconolactonase OS = Homo sapiens GN = PGLS PE = 1 SV = 225827601 . 6013841 . 500 . 996 . 006 . 04 Q8TD06AGR3_HUMAN Anterior gradient protein three homolog OS = Homo sapiens GN = AGR3 PE = 1 SV = 116619194 . 9014947 . 000 . 995 . 967 . 03 P18754RCC1_HUMAN Regulator of chromosome condensation OS = Homo sapiens GN = RCC1 PE = 1 SV = 142148241 . 20271355 . 301 . 9911 . 927 . 00 O60506HNRPQ_HUMAN Heterogeneous nuclear ribonucleoprotein Q OS = Homo sapiens GN = SYNCRIP PE = 1 SV = 262369739 . 70202048 . 501 . 9710 . 823 . 50 P03372ESR1_HUMAN Estrogen receptor OS = Homo sapiens GN = ESR1 PE = 1 SV = 259566335 . 20361530 . 801 . 839 . 885 . 93 E9PCR7E9PCR7_HUMAN 2-oxoglutarate dehydrogenase , mitochondrial OS = Homo sapiens GN = OGDH PE = 2 SV = 11038115728 . 00392532 . 803 . 9818 . 884 . 00 O43488ARK72_HUMAN Aflatoxin B1 aldehyde reductase member 2 OS = Homo sapiens GN = AKR7A2 PE = 1 SV = 335939653 . 80221139 . 801 . 998 . 965 . 51 O95994AGR2_HUMAN Anterior gradient protein two homolog OS = Homo sapiens GN = AGR2 PE = 1 SV = 117522277 . 70291265 . 702 . 9711 . 924 . 68 P19338NUCL_HUMAN Nucleolin OS = Homo sapiens GN = NCL PE = 1 SV = 371076766 . 50993548 . 0014 . 0050 . 982 . 43 O43148MCES_HUMAN mRNA cap guanine-N7 methyltransferase OS = Homo sapiens GN = RNMT PE = 1 SV = 147657831 . 9016929 . 401 . 996 . 973 . 50 Q562R1ACTBL_HUMAN Beta-actin-like protein 2 OS = Homo sapiens GN = ACTBL2 PE = 1 SV = 237642084 . 00141439 . 102 . 007 . 003 . 00 Q9Y5A9YTHD2_HUMAN YTH domain family protein 2 OS = Homo sapiens GN = YTHDF2 PE = 1 SV = 257962457 . 80151223 . 102 . 006 . 993 . 00 P16152CBR1_HUMAN Carbonyl reductase [NADPH] 1 OS = Homo sapiens GN = CBR1 PE = 1 SV = 327730427 . 90201156 . 702 . 999 . 982 . 33 Q9UBS4DJB11_HUMAN DnaJ homolog subfamily B member 11 OS = Homo sapiens GN = DNAJB11 PE = 1 SV = 135840578 . 70211235 . 203 . 0010 . 002 . 67 The interaction between ER and ERX-11 within the cells was partially disrupted by high doses of tamoxifen ( Figure 2D ) . Further , in the tamoxifen-resistant cell line , MCF-7-TamR , even high doses of tamoxifen could not disrupt the interaction between ERX-11 and ER ( Figure 1—figure supplement 3E ) . The differences between these results and the in vitro results may be attributed to the context in which ER is presented within the cell . Using GST-fused ER domain constructs , we validated that ERX-11 interact with the GST-AF2 domain of ER but not with the GST-AF1 or GST-DNA-binding domain of ER ( Figure 2E ) . Further , ER-AF2 interaction with ERX-11 was disrupted by tamoxifen but not ICI ( Figure 2F ) . These data clearly establish the interaction between ER and ERX-11 through the AF-2 domain . Using an unbiased approach with IPMS , we showed that ERX-11 significantly disrupted the interactions of 91 nuclear ER-binding proteins with ER in MCF-7 cells ( Figure 2—figure supplement 2A ) , including well-characterized ER coregulators , such as SRC1 , SRC3 , and PELP1 . Global analyses revealed that these proteins may be involved in a number of critical cellular pathways including transcription , cell cycle and regulation of cell death ( Table 2 ) . These findings were validated by IPMS studies in ZR-75 cells , which showed a significant overlap with MCF-7 cells in the coregulators disrupted by ERX-11 ( Figure 2—figure supplement 2B ) . Of the top 10 coregulators , whose interactions with ER were negatively influenced by ERX-11 , five contained LXXLL motifs with serine at i-3/4 and i+7/8 flanking position of the LXXLL motifs Table 3 . Interestingly in the MDA-MB-231 TNBC model cells , we found that biotinylated ERX-11 was able to stringently interact only with a small number of proteins ( n = 8 ) ( Figure 2—figure supplement 2C ) . 10 . 7554/eLife . 26857 . 016Table 2 . Top biological processes of coregulators , whose interactions with ER are disrupted by ERX-11 in MCF-7 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 016Biology processesGenesRNA processingCD2BP2 CHERP CPSF1 CPSF2 CPSF3L CSTF3 DDX17 DDX20DDX23 DHX15 DHX9 DKC1 GEMIN5 HNRNPA3 HNRNPK HNRNPLL HNRNPLL HNRNPR INTS2 INTS4 INTS5 NCBP1 PCF11 POLR2A PPP2R1A PRPF31 PRPF40a PUF60 RBM10 RBM14 SART1 SF1 SF3A3 SF3B1 SF3B3 Sfrs15 SKIV2L2 SMC1A SRRM1 SSB SYNCRIP THOC2 TRNT1 U2AF2 XRN2 ZCCHC8TranscriptionADNP CCNL1 CSDA CTNND2 DIDO1 DMAP1 EIF2S2 ERCC2 FOXA1 GTF2I GTF3C1 GTF3C KDM3B KDM5B LRPPRC MCM2 MED1 MED24 NCOA3 PELP1 POLR2A POLR3C PSIP1 PUF60 RBM14 RFX1 SAP130 SF1 SMARCA2 SMARCA4 SMARCC2 SMARCD2 THRAP3 Th1l TRIM33 UHRF1 XRN2 ZBTB7A ZMYM2 ZNF217 ZNF512BProtein transport , protein localizationAP2A2 CLTC COG1 COG3 COG5 COG8 COPB1 COPB2 COPG2 CSE1L EXOC2 EXOC3 EXOC4 EXOC5 EXOC8 IPO4 KPNA4 KPNB1 NUP153 NUP155 NUP93 RANBP2 SEC16A SEC23A SEC24B SEC24C SRP72 STXBP2 TNPO1 TRAM1 TRNT1 VCP VPS11 VPS18 VPS39RNA splicingCD2BP2 CPSF1 CPSF2 CSTF3 DDX20 DDX23 DHX15 DHX9 GEMIN5 HNRNPA3 HNRNPL HNRNPR LUC7L3 NCBP1 PCF11 POLR2A PPP2R1A PRPF31 PRPF40A PUF60 RBM10 SART1 SF1 SF3A3 SF3B1 SF3B3 SKIV2L2 SMC1A SRRM1 SYNCRIP THOC2 U2AF2 ZCCHC8Macromolecular complex subunit organizationCSE1L DARS DDX20 DDX23 EPRS ERCC2 FKBP4 GEMIN5 GTF2I GTF3C4 HSP90AA1 IPO4 KPNB1 LONP1 MCM2 MED1 MED24 NCBP1 PFKL PFKM PFKP POLR2A PPP2R1A PREX1 PRPF31 SF1 SF3A3 SF3B3 THRAP3 TNPO1 TUBA1B TUBB VCP XRN2Cell cycleCUL1 CUL2 CUL3 CUL4B Dmc1 DNM2 DYNC1H1 EIF4G2 LIG3 MCM2 MRE11A NUMA1 PDS5B PHGDH PPP3CA PSMC1 PSMC4 PSMD3 PSMD5 RAD50 SART1 SF1 SMC1A SMC3 SMC4 TUBB UHRF1Chromosome organizationPDS5B KDM5B RBM14 RAD50 MRE11A CHD1L DMAP1 DKC1 EP400 KDM3B MCM2 SAP130 SMC1A SMC3 SMC4 SMCHD1 SMARCA2 SMARCA4 SMARCC2 SMARCD2Regulation of cell deathACTN1 ADNP CSDA CUL1 CUL2 CUL3 DDX20 DNM2 ERCC2 PPP2R1A PREX1 SART1 SCRIB TUBB UACA VCP10 . 7554/eLife . 26857 . 017Table 3 . Selected pathways modulated by ERX-11 treatment . Differentially expressed genes were subjected to pathway analysis using IPA software and the selected top canonical pathways modulated by ERX-11 are shown . This data is related to Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 017Pathwayp-ValueRatioGenes Retinoic-acid-Mediated Apoptosis Signaling2 . 44E + 001 . 25E-01PARP12 , ZC3HAV1 , TNFSF10 , PARP9 , PARP14 , CRABP2 , RARG , CRABP1 ERK/MAPK Signaling1 . 77E + 007 . 49E-02SRC , MKNK2 , PLA2G4F , DUSP2 , BAD , ELF5 , PPM1J , PPP1R14B , STAT1 , RAC3 , ELF4 , PPP2R1A , RRAS , RPS6KA4 Cyclins and Cell Cycle Regulation1 . 46E + 008 . 97E-02HDAC5 , TGFB1 , PPM1J , PPP2R1A , E2F1 , HDAC11 , HDAC7 Death Receptor Signaling1 . 53E + 008 . 70E-02PARP12 , ACTG1 , ZC3HAV1 , LIMK1 , TNFSF10 , PARP9 , PARP14 , BIRC3 Inhibition of Matrix Metalloproteases2 . 40E + 001 . 54E-01HSPG2 , MMP10 , TIMP1 , MMP13 , MMP15 , SDC1 Estrogen Receptor Signaling1 . 19E + 007 . 09E-02KAT2B , ERCC2 , SRC , G6PC3 , TAF6 , MED24 , TAF6L , RRAS , MED15 Breast Cancer Regulation by Stathmin14 . 63E-014 . 71E-02ADCY1 , ARHGEF19 , PPM1J , PPP1R14B , LIMK1 , PPP2R1A , E2F1 , TUBA4A , RRAS In MCF-7 cells , a majority of ER-binding proteins disrupted by ERX-11 were also blocked by tamoxifen ( 55/88 proteins or 62 . 5% ) ( Figure 2G and Figure 2—figure supplement 3 ) . Importantly , a significant number of ER-binding proteins were disrupted by ERX-11 but not tamoxifen ( 33/88 or 37 . 5% ) ( Figure 2G and Figure 2—figure supplement 3 ) . The combination of tamoxifen and ERX-11 had significant overlap with ERX-11 in its ability to block ER-binding proteins ( Figure 2—figure supplement 3 ) . Co-immunoprecipitation studies validated that endogenous complexes containing ER and coregulator PELP1 do form in MCF-7 cells and that ERX-11 disrupts the formation of these complexes ( Figure 2H ) . Proximity ligation assays confirmed the disruption of the endogenous interaction between ER and several coregulators including PELP1 , SRC1 and SRC3 ( Figure 2I , quantitation in Figure 2—figure supplement 4A ) . In contrast , ERX-11 has no effect on ER interaction with ARID1B ( Figure 2I , quantitation in Figure 2—figure supplement 4B ) . Using an in vivo structural complementation NanoBiT assay , we found that the direct interaction between ER and PELP1 was enhanced by E2 treatment and that ERX-11 significantly reduced the interaction between ER and PELP1 ( Figure 2J ) . To confirm the specificity of ERX-11 for the ER AF-2 domain , we demonstrated using reporter-based assays , that ERX-11 failed to reduce the ERE-Luc reporter activity driven by a ERα-VP16 chimera that does not require AF-2 ( Figure 2—figure supplement 4C ) . As expected , tamoxifen , did not affect the ERα-VP16 chimera-induced reporter activity , while ICI reduced the ERE-Luc reporter activity . Evaluation with an endometrial cancer cell line Ishikawa , which exhibits agonist activity via AF1 , revealed that ERX-11 lacks AF1 agonist activity ( Figure 2—figure supplement 4D ) . Collectively , these results confirm that the ERX-11 block signaling driven by functional AF2 domain but not by AF1 domain . We then specifically evaluated whether interactions through the ER LXXLL motif was responsible for ERX-11 activity . Biotinylated ERX-11 was able to pull down both the wild-type ER and the L540Q point mutant ER ( which retains E2 binding and does not interact with SRC1 ) and these interactions were not affected by tamoxifen ( Figure 2K ) . Using ER L540Q point mutant , we showed that the mutant ER still interacts with biotinylated ERX-11 ( Figure 2K ) . Interestingly , the ERX-11 binds strongly to ER▲12 mutant ( helix 12 deleted ER ) and this binding is blocked efficiently by tamoxifen ( Figure 2K ) . These data suggest that the presence of helix 12 may regulate the conformation of the binding pocket and account for differences in the binding of ERX-11 and tamoxifen to ER . Our data would suggest that removal of helix 12 enables access of ERX-11 to the same binding pocket as tamoxifen and may reflect the in vitro data , where tamoxifen and ERX-11 compete efficiently for ER binding . In contrast , neither GDC-0810 nor AZD-9496 were able to block ERX-11 binding to ER or its mutants , suggesting that their binding to ER occurs through distinct pockets ( Figure 2—figure supplement 5A and B ) . In competition assays , tamoxifen fails to dislodge ERX-11 from ER ( Figure 2—figure supplement 5C , Figure 1—figure supplement 3E ) . Increasing concentrations of tamoxifen is only able to dislodge ERX-11 from ER▲12 mutant at higher concentrations , suggesting that ERX-11 interaction with ER is through a second binding site ( Figure 2—figure supplement 5D ) . Further , using an ER restoration model , MTT cell viability assays revealed that introduction of either ER , ER▲12 or ER-L540Q into MDA-MB-231 cells , restored ERX-11 growth inhibitory activity in non-responsive MDA-MB-231 cells . These results further underscore the importance of ER in ERX-11 mode of action ( Figure 2—figure supplement 5E ) . We have modeled ERX-11 interaction with ER using docking simulations of ERX-11 using known crystal structures ( Figure 2—figure supplement 6 ) . In the agonist-bound conformation ( 3ERD . pdb ) , helix 12 is relocated and forms a hydrophobic cleft ( i . e . AF2-binding site ) that is surrounded by helices 3 , 4 , 5 and 12: ERX-11 can be modeled to make hydrophobic contact with the AF2 site with its two isobutyl side chain groups ( Figure 2—figure supplement 6A ( A ) ) . In addition , the hydroxyl group of ERX-11 may interact with a residue near AF2 domain . These data suggest that ERX-11 interacts with ER LBD differently than the agonist . In the antagonist-bound conformation ( 3ERT . pdb ) , 4-hydroxytamoxifen induces conformational change and makes helix 12 occupy the AF2-binding site , blocking both coactivator and corepressor binding: here , ERX-11 may interact with an alternate pocket formed by helices 5 , 11 and 12 , as shown in the Figure 2—figure supplement 6A ( B ) . These data could explain why the interaction between ERX-11 and purified ER could be blocked by tamoxifen . However , in a cellular context , tamoxifen cannot block ERX-11 binding to ER suggesting that the secondary binding site of ERX-11 on ER may be stabilized by coregulators . Docking simulation of ERX-11 on human ERα with affinity-tagged corepressor peptide ( Figure 2—figure supplement 6A ( C ) ) ( 2JFA . pdb ) or rat ERβ crystal structure with ICI bound ( Figure 2—figure supplement 6A ( D ) ) ( 1HJ1 . pdb ) showed that ERX-11 can still bind to the AF2 domain , in a similar manner as it does when the ligand is bound . These data further support our biochemical findings that ICI does not block ERX-11 interaction with ER . Detailed evaluation showed that ( 1L2I . pdb ) two isobutyl groups of ERX-11 may dock into the AF2 binding site ( black dashed box ) ( Figure 2—figure supplement 6B ) . The hydroxyl group of ERX-11 may interact with Gln375 of the helix five through a hydrogen bond . The carboxamide group of ERX-11 was docked into a pocket formed with Gln542 , Asp545 and Ala546 on the helix 12 . Additional docking experiment on ER crystal structures without helix 12 ( 3ERT . pdb and 5ACC . pdb ) showed that ERX-11 was found to bind to the tamoxifen-binding site in the ER▲12 ( Figure 2—figure supplement 6C ( A ) ) . The superimposition of tamoxifen ( red ) and ERX-11 ( green ) clearly shows the overlap of their binding sites ( 3ERT . pdb ) ( Figure 2—figure supplement 6C ( B ) ) . This may explain our experimental results showing the competition of tamoxifen with ERX-11 on ER▲12 . In the presence of SERDs , ERX-11 can bind to the AF2 domain ( Figure 2—figure supplement 6C ( C ) , which do not overlap with the binding site of SERDs ( Figure 2—figure supplement 6C ( D ) ) ( 5ACC . pdb ) . A model to explain the potential interactions between ER and ERX-11 or between ER▲12 mutant and ERX-11 in the absence or presence of agonist , SERDs or tamoxifen is included ( Figure 2—figure supplement 7 ) . RNA-seq analyses revealed that ERX-11 altered the expression of 880 E2-regulated genes ( p<0 . 01 ) in ZR-75 cells compared to vehicle control ( Figure 3A ) . Using stringent cutoffs ( p<0 . 01 and RPKM FC > 1 . 5 ) , ERX-11 down-regulated more genes ( 669 ) than upregulated ( 211 ) as evidenced by the volcano plot ( Figure 3—figure supplement 1A and B ) ( complete list at GEO database , accession number GSE75664 ) . RT-qPCR analyses validated the expression of the top down-regulated ( Figure 3B and Figure 3—figure supplement 1C ) and up-regulated genes ( Figure 3H ) . Gene set enrichment analysis ( GSEA ) confirmed the correlation between ERX-11–regulated genes and both the tamoxifen-responsive and E2-responsive genes set ( Figure 3C and D ) . In addition , ERX-11-regulated genes correlated well with signatures of early and late response E2 targets as well as genes driven by consensus ER motifs ( Figure 3E and F ) . 10 . 7554/eLife . 26857 . 018Figure 3 . ERX-11 globally disrupts ER-mediated transcriptome . Total RNA was isolated from the ZR-75 cells that were treated with either vehicle or ERX-11 for 48 hr and subjected to RNA sequencing . The heat map of differentially expressed genes between vehicle and ERX-11 is shown ( A ) . ZR-75 cells were treated with either vehicle or ERX-11 for 48 hr , and the selective genes representing each pathway were validated using RTqPCR ( B ) . Gene set enrichment analysis ( GSEA ) testing correlation between ERX-11–regulated gene and both the tamoxifen-responsive ( M3283 ) and estradiol-responsive genes set ( M2234 and M2230 ) ( C , D ) . GSEA analysis testing the correlation of ERX-11-regulated genes with signatures of early ( M5906 ) and late ( M5907 ) response estrogen targets as well as genes driven by consensus ER motifs ( M17968 and M6101 ) ( E , F ) . GSEA analysis of correlation of ERX-11 regulated gene set with an apoptotic gene set ( M15912 ) ( G ) . ERX-11 upregulated apoptotic genes were validated by RT-qPCR analyses ( H ) . Data are represented as mean ±SEM . **p<0 . 01; ***p<0 . 001 . Effect of indicated doses of Tam , ICI and ERX-11 on cell viability ( CellTiter-Gloassay , Promega ) and caspase3/7 activity ( Caspase-Glo 3/7 Assay , Promega ) in ZR-75cells ( I ) . Data shown are the means of ±SEM performed in triplicate wells . ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 01810 . 7554/eLife . 26857 . 019Figure 3—figure supplement 1 . ERX-11 treatment has potential to promote apoptosis . ( A ) Volcano plot of statistical significance ( adjusted p value ) against RPKM fold-change ( FC ) between vehicle and ERX-11 treated ZR-75 cells . Only significantly changed genes ( p<0 . 01 and FC > 1 . 5 ) were shown in the graph . ( B ) Panel B shows the volcano plot in more detail . ( C ) RT-qPCR validation of the expression of some of the top down regulated genes by ERX-11 treatment ( *p<0 . 05; **p<0 . 01; ***p<0 . 001 ) . ( D ) Effect of indicated doses of Tam , ICI , ERX-11 on cell viability ( right panel ) and caspase 3/7 activity ( Caspase-Glo 3/7 Assay , Promega ) in T-47D cells ( left panel ) . ( E ) Effect of ERX-11 treatment ( 24 hr , 48 hr ) on cell viability ( bottom panel ) and caspase 3/7 activity in ZR-75 cells ( upper panel ) . p<1***p<0 . 001****p<0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 019 Ingenuity pathway analysis ( IPA ) revealed that ERX-11 significantly down-regulated genes involved in ER-signaling , breast cancer , cell cycle , and MAPK signaling ( Table 3 ) . ERX-11 upregulated gene set positively correlated with apoptotic genes , on GSEA analyses ( Figure 3G ) . Importantly , ERX-11 but not tamoxifen or ICI-treatment-induced apoptosis in ZR-75 ( Figure 3I ) and in T-47D cells ( Figure 3—figure supplement 1D ) as shown by induction of caspase 3/7 activity . Apoptosis can be seen as early as 24 hr , however , the effect is more pronounced at 72 hr ( Figure 3—figure supplement 1E ) . These results suggest that ERX-11 both reduces the expression of genes involved in proliferation and enhances expression of genes that promote apoptosis . Using ERE-Luc reporter based transcription assays , we found that ERX-11 significantly reduced the E2-induced ERE-Luc reporter gene activity in ZR-75 cells in a similar fashion as tamoxifen ( Figure 4A ) . In HEK-293T cells , expression of AIB1 and SRC1 enhanced ER-driven ERE-Luc reporter activity and was blocked by both tamoxifen and ERX-11 in both a ligand-dependent ( Figure 4B ) and ligand-independent manner ( Figure 4C ) . Further , ZR-75 cells stably overexpressing either PELP1 or AIB-1 were responsive to ERX-11 ( Figure 4D ) . These results indicate that ERX-11 interferes with both ligand-dependent and ligand-independent transcriptional function of ER . 10 . 7554/eLife . 26857 . 020Figure 4 . ERX-11 affects ER ligand-dependent and independent transcriptional activity . ZR-75 cells stably transfected with ER and ERE-Luc vectors were treated with E2 ( 10−8M ) in the presence of indicated concentrations of ERX-11 or tamoxifen . After 24 hr , the reporter gene activity was measured ( A ) . HEK-293T cells stably transfected with ERE-Luc vector were transiently transfected with control or coregulator-expressing vector along with ER expression vector and after 24 hr treated with indicated doses of ERX-11 or tamoxifen along with E2 ( 10−8M ) . After 24 hr , the reporter gene activity was measured ( B ) . HEK-293T cells stably transfected with ERE-Luc vector were transiently transfected with control or coregulator-expressing vectors along with ER expression vector and after 24 hr treated with ERX-11 . After 24 hr , the reporter gene activity was measured ( C ) . Cell proliferation of ZR-75 model cells stably expressing PELP1 or AIB1 was measured using Cell Titer Glo assay ( D , left panel ) . ZR-75 model cells stably expressing PELP1 or AIB1 were transfected with ERE-Luc reporter vector . After 48 hr , the cells were treated with ERX-11 and the reporter activity was measured 24 hr later ( D right panel ) . The effects of ERX-11 on ER recruitment at ER target genes were examined using a ChIP assay in MCF-7 cells ( E ) . The effect of ERX-11 on ER dimerization as evaluated by the NanoBiT luciferase assay is shown ( F ) . ZR-75 cells were treated with E2 with or without ERX-11 ( 1 µM ) for the indicated time , and the stability of ER was determined using western blotting . Quantitation of ER levels compared to control ( E2-treated cells ) was shown after normalizing to the levels of GAPDH ( G ) . ZR-75 cells were treated with ERX-11 for 5 days , and the status of ER was determined using Western blotting ( H ) . Data shown are the means of ±SEM performed in triplicate wells . *p<0 . 05; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 02010 . 7554/eLife . 26857 . 021Figure 4—figure supplement 1 . Effect of ERX-11 on AR functions , E2 mediated non-genomic actions and ER stability . ( A ) Evaluation of the effect of ERX-11 on AR DNA binding to its target sequences . ( B ) The effect of ERX-11 on AR dimerization as evaluated by the NanoBiT luciferase assay . ( C ) Evaluation of the effect of ERX-11 on estrogen dendrimer conjugates ( EDCs , that uniquely localize in the cytoplasm ) on the activation of estrogen-mediated non-genomic signaling . ( D ) T-47D cells were treated with E2 with or without ERX-11 for indicated times and the stability of ER was determined using western blotting . Quantitation of ER fold change over control ( E2 treated ) was shown and is normalized with GAPDH . ( E ) T-47D cells were treated with ERX-11 or Tam or ICI for seven days and level of ESR1 was measured using RTqPCR . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 021 Using chromatin immunoprecipitation studies , we found that ERX-11 treatment significantly blocked ER recruitment to canonical ER target gene promoters following E2 treatment ( Figure 4E ) . In contrast , ERX-11 did not affect AR recruitment to AR target genes ( Figure 4—figure supplement 1A ) . Since ERX-11 binds to ER , we hypothesized that the effect of ERX-11 on ER DNA binding may be mediated via disruption of ER dimerization . Using the NanoBiT assay , we demonstrated that ERX-11 efficiently blocks the dimerization of ER ( Figure 4F ) . In contrast , ERX-11 did not affect AR dimerization ( Figure 4—figure supplement 1B ) . Using E2 dendrimer conjugates ( EDC ) , that are potent in activating ER non-genomic signaling but not ER genomic signaling ( Chakravarty et al . , 2010 ) , we showed that ERX-11 was unable to influence the EDC-mediated non-genomic activation of the Src , AKT and MAPK pathways ( Figure 4—figure supplement 1C ) . A detailed time course evaluation showed that ERX-11 treatment only modestly altered the stability of ER within 24 hr in ZR-75 and T-47D cells ( Figure 4G , Figure 4—figure supplement 1D ) . However , similar to its inhibitory activity on ER transcription , after several days of ERX-11 treatment , decreased ER levels were detected ( Figure 4H ) . Accordingly , RTqPCR results showed that ERX-11 reduce the ER transcript levels under conditions of long-term treatment ( 7 days ) . These results reflect the inhibition of ER signaling indirectly affected autoregulation of ER transcript by E2-ER signaling ( Figure 4—figure supplement 1E ) . Our prior studies indicated that our peptidomimetics are orally bioavailable ( Ravindranathan et al . , 2013 ) . We detected no overt signs of toxicity after 14 days of treatment of C57BL/6 mice ( n = 3 ) with 10 , 50 or 100 mg/kg/day of ERX-11 via oral gavage . ERX-11 treatment neither caused weight loss nor have uterotrophic effects or any observable hematologic , liver and kidney abnormalities ( data not shown ) . We designated our highest dose ( 100 mg/kg ) as the maximum tolerated dose and used 10% of this dose ( 10 mg/kg/day ) for testing as therapeutic dose , so that we would have at least a 10:1 therapeutic to toxicity ratio . Established ZR-75 xenografts ( n = 8 tumors/group ) in the mammary fat pad of nude mice were randomized to feed via oral gavage 5 days/week with either 10 mg/kg ERX-11 or vehicle ( 30% Captisol ) . ERX-11 treatment resulted in significantly smaller tumors ( 63% reduction compared to control ) ( Figure 5A ) . ERX-11–treated tumors exhibited less proliferation ( Ki67 staining ) , and more apoptosis ( TUNEL and caspase-3 staining ) than controls ( Figure 5A , B ) . Further , ERX-11 treatment group had lower ER but similar PELP1 protein expression levels within the tumor , compared to control ( Figure 5B ) . The mice body weights in the control and ERX-11 treated groups were similar ( Figure 5—figure supplement 1A ) . Mice treated with ERX-11 exhibited no uterotrophic effects , no changes in ovary , liver and kidney gross morphology on H and E staining , or acute phase injury to liver and kidney ( Figure 5—figure supplement 2A–D ) . These data indicate that ERX-11 is a potent inhibitor of the growth of ER-positive breast tumors in vivo with no overt signs of toxicity in mice . 10 . 7554/eLife . 26857 . 022Figure 5 . ERX-11 inhibits the growth of ER-positive , syngeneic and coregulator-driven breast tumors in vivo . ER-positive ZR-75 cells were injected into the mammary fat pads of nude mice implanted subcutaneously with E2 pellet . After 2 weeks , mice with xenografts were treated with vehicle or 10 mg/kg/day of ERX-11 ( n = 8 ) by oral gavage . Tumor growth was measured at the indicated time points . Tumor volume is shown in the graph ( A ) . The weights of the control or ERX-11-treated tumors at the time of necropsy are shown . Ki-67 expression as a marker of proliferation was analyzed by IHC and quantitated . Apoptosis was measured using Caspase3 activation and by using TUNEL assay , and the number of TUNEL-positive and cleaved caspase 3 cells were counted in five different fields and plotted as histogram . DAPI was used to visualize the nuclei ( A , B ) . Representative IHC analysis of ER and PELP1 performed on xenograft tumors that were treated with or without ERX-11 ( B ) . Effect of ERX-11 on the growth of ER-positive D2A1 syngeneic tumors . Small pieces of D2A1 syngeneic tumors were implanted subcutaneously into the BALB/c mice . After 1 week , mice ( n = 8 ) were treated with vehicle or ERX-11 ( 20 mg/kg/day ) . Tumor growth was measured at indicated time points . The body weights and extirpated tumor weights are shown . Ki-67 expression was analyzed by IHC and quantitated ( C ) . The effect of ERX-11 on the coregulator-driven cell survival was measured by MTT assay using ZR-75 cells stably expressing SRC3/AIB1 or PELP1 ( D ) . MCF-7-PELP1 cells were injected into the mammary fat pad of nude mice ( n = 5 ) implanted with an estrogen pellet . After 3 weeks , mice were treated with vehicle or ERX-11 ( 10 mg/kg/day ) . Tumor volume , status of Ki-67 and apoptosis was shown ( E ) . Data shown are the means of ±SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 02210 . 7554/eLife . 26857 . 023Figure 5—figure supplement 1 . Characterization of ERX-11 treated tumors . ( A ) Body weights of mice implanted with ZR-75 xenograft tumors treated with vehicle or ERX-11 . ( B ) Characterization of D2A1 model cells: ( A ) Western blot analysis of D2A1 cell lysates . Murine mammary lysate and E0771 lysate was used as positive control . ( B ) Effect of tamoxifen on the growth of D2A1 cells was measured by MTT assay . ( C ) IHC analysis of ER on D2A1 xenografts . ( C ) D2A1 tumors treated with or without ERX-11 were analyzed for Ki-67 expression as a marker of proliferation and apoptosis using TUNEL assay . ( D ) MCF-7-PELP1 tumors treated with or without ERX-11 were analyzed for Ki-67 expression ( as a marker of proliferation ) and apoptosis using TUNEL assay . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 02310 . 7554/eLife . 26857 . 024Figure 5—figure supplement 2 . Normal tissues collected from mice that were treated with vehicle or ERX-11 were examined for toxicity . ( A ) Effect of ERX-11 on various tissues as seen by H and E . ( B ) Effect of ERX-11 on Ki67 staining . ( C ) Effect of ERX-11 on Ki67 with quantitation . ( D ) Effect of ERX-11 on ER staining in the ovary . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 024 To address the potential immunogenicity of ERX-11 , D2A1 ER-positive breast cancer xenografts were established in a syngeneic BALB/c model system with an intact immune system . In D2A1 model cells , cellular gene int-5/aromatase in BALB/c mammary alveolar hyperplastic nodule ( D2 HAN/D2 tumor cells ) is activated as a result of mouse mammary tumor virus integration within the 3' untranslated region of the aromatase gene . Thus , these models also have ability to synthesize local estrogen via aromatase induction . Further , this model express ER , and represent a model of intra-tumoral estrogen-driven mammary cancer ( Figure 5—figure supplement 1B ( A , C ) . D2A1 cells are responsive to antiestrogen treatment ( Figure 5—figure supplement 1B ( B ) . Oral administration of ERX-11 dramatically limited the proliferation of these rapidly progressing tumors ( Figure 5C ) . The proliferative indices of ERX-11-treated tumors were significantly lower than controls ( Figure 5C , Figure 5—figure supplement 1C ) , while the apoptotic indices were higher than control ( Figure 5—figure supplement 1C ) . Again , no overt signs of toxicity was noted in these mice; specifically , no enlargement of the spleen or evidence of immune complex deposition within the kidneys was detected ( data not shown ) . These data further support the potential clinical translatability of ERX-11 . To evaluate the effect of ERX-11 on coregulator-driven proliferation , we used ZR-75 cells stably overexpressing AIB-1 and PELP1 . While these modified ZR-75 cells are highly proliferative ( Figure 4D ) , ERX-11 was potent in blocking their proliferation ( Figure 5D ) . ERX-11 was potent ( 73% reduction in tumor volume compared to control ) against the growth of MCF-7-PELP1 xenografts , which overexpress PELP1 ( 3-fold higher than parental MCF-7 ) ( Figure 5E ) . IHC analysis of ERX-11 treated tumors showed decreased Ki-67 staining ( Figure 5E , Figure 5—figure supplement 1D ) . Importantly , ERX-11 had activity against ER-driven breast cancer cell lines that were either resistant to tamoxifen ( MCF-7-TamR , Figure 6A , or MCF-7-HER2 , Figure 6B ) or to letrozole ( MCF-7-LTLT , Figure 6C ) . In these cell lines , ERX-11 was still able to interact with the ER , both in the absence and presence of tamoxifen ( Figure 6D ) . In contrast to SERD ( ICI ) , which had limited activity on the tamoxifen resistant cell lines , ERX-11 had potent activity ( Figure 6A ) ( Figure 6—figure supplement 1A ) . ERX-11 was potent against the growth of MCF-7-LTLT xenografts ( Figure 6E ) . IHC analysis of ERX-11-treated tumors showed decreased Ki-67 staining ( Figure 6—figure supplement 1B ) . 10 . 7554/eLife . 26857 . 025Figure 6 . ERX-11 reduces the growth of ER positive and ER-MT endocrine-therapy-resistant tumors . Cell viability assays evaluated the effect of ERX-11 on Tamoxifen-resistant MCF-7-TamR cells ( A ) , tamoxifen-resistant MCF-7/HER2 cells ( B ) and letrozole-resistant MCF-7-LTLT cells ( C ) . ICI was used as control . Results are represented as mean ± SEM . *p<0 . 05; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . MCF-7-TamR and MCF-7-LTLT cells were cultured in 1 µM tamoxifen or 1 µM letrozole , respectively , and nuclear lysates were subject to biotin-ERX-11 pull down followed by western blotting with ER antibody ( D ) . Following implantation and growth of ER-positive letrozole-resistant xenografts in nude mice ( n = 8 ) , mice were treated with control or ERX-11 ( 20 mg/kg/day ) . Tumor volume , tumor weight and Ki-67 status of control and treated tumors was shown ( E ) . Nuclear extracts prepared from HEK-293T cells transiently transfected with WT- or MT-ER expression plasmids and analyzed for interaction between WT- and MT-ER to the biotin-ERX-11 using avidin pulldown followed by western blot analysis ( F ) . Effect of ERX-11 on the cell viability of ZR-75 ESR1-KO cells stably expressing ESR1-WT or ESR1-Y537S mutant or ESR1-D538G mutant ( G ) was measured using MTT assay . ER-negative MDA-MB-231 cells were co-transfected with ERE reporter along with WT-ESR1 and MT-ESR1 plasmids . After 48 hr , the cells were treated with ERX-11 ( 500 nM ) and the reporter activity was measured 24 hr later ( H ) . Effect of ERX-11 and tamoxifen on the cell viability of ZR-75 cells stably expressing ER-Y537S mutant was measured using MTT assays ( I ) . ZR-75 cells stably expressing ER-Y537S mutant were injected into the mammary fat pads of nude mice implanted subcutaneously with an estrogen pellet . After 2 weeks , mice with xenografts were treated with vehicle or ERX-11 ( 20 mg/kg/day , n = 6 ) . Tumor growth was measured at indicated time points ( J ) . Ki-67 expression was analyzed by IHC and quantitated ( K ) . Data shown are the means of ± SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 02510 . 7554/eLife . 26857 . 026Figure 6—figure supplement 1 . ERX-11 reduces the growth of ER-positive and ER-MT endocrine-therapy-resistant tumors . ( A ) MCF-7-LTLT xenografts were treated with vehicle or ERX-11 or Fulvestrant . Tumor volume , tumor weights and body weights are shown****p<0 . 0001 . ( B ) MCF-7-LTLT tumors treated with or without ERX-11 were analyzed for Ki-67 expression as a marker of proliferation . ( C ) ESR1 was knocked out in ZR-75 cells using CRISPR/Cas9 system and then stably transfected with WT-ESR1 or MT-ESR1 ( 537S , and 538G ) and cell proliferation was measured ***p<0 . 001; ****p<0 . 0001 . Mutant-ESR1 expressing cells showed higher rate of proliferation compared to WT-ER expressing cells . Expression of WT and mutant ESR1 in the model cells was analyzed using western analysis . ( D ) ZR-75-ESR1-MT Y537S tumors treated with vehicle or ERX-11 were analyzed for Ki-67 expression as a marker of proliferation . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 026 We then evaluated the effect of ERX-11 against two prevalent ER mutants ( MT-ESR1-Y537S , MT-ESR1-D538G ) ( Toy et al . , 2013; Robinson et al . , 2013; Jeselsohn et al . , 2014; Merenbakh-Lamin et al . , 2013 ) . Using biotinylated ERX-11 , we showed that ERX-11 interacts directly with ESR1-MT ( ESR1-MT-D538G , ESR1-MT-Y537S ) , with high affinity comparable to the affinity to WT-ER ( Figure 6F ) . Using CRISPR/Cas9 , we knocked down ER in ZR-75 cells and then stably transfected with WT-ESR1 or MT-ESR1 ( Y537S , and D538G ) . While ESR1-MT expressing cells showed higher rates of proliferation than WT-ESR1-expressing cells ( Figure 6—figure supplement 1C ) , they were still inhibited by ERX-11 ( Figure 6G ) . Further the ability of these ESR1-MT to drive ligand-independent transcription from an ERE-Luc reporter was also efficiently blocked by ERX-11 ( Figure 6H ) . Further , these ESR1-MT expressing cells were resistant to tamoxifen , however , were sensitive to ERX-11-mediated growth inhibition ( Figure 6I ) . Further , oral ERX-11 administration had significant activity against the growth of ZR-75-ESR1MT-Y537S xenografts in vivo ( Figure 6J ) , with significant reduction in proliferative indices ( Figure 6K , Figure 6—figure supplement 1D ) . These data support the efficacy of ERX-11 against breast tumors driven by mutant ESR1 . We recently developed an ex vivo culture model of primary breast and prostate tumors , which allows for the evaluation of drugs on breast tumors while maintaining their native tissue architecture ( Dean et al . , 2012; Schiewer et al . , 2012 ) . In brief , surgically extirpated de-identified breast tissues are sliced into small pieces and grown ex vivo for a short term on a gelatin sponge in the absence or presence of ERX-11 ( Figure 7A ) . Incubation of ERX-11 with ER-positive breast tumor samples ( patient characteristics detailed in Table 4 ) dramatically decreased their proliferation in 11/12 patients ( Ki67 staining ) compared to untreated controls ( Figure 7B ) . Further , ERX-11 treatment significantly reduced the ER staining ( Figure 7C ) , but not PELP1 staining ( Figure 7—figure supplement 1 ) in 12/12 ER-positive tumors . Importantly , ERX-11 treatment had no effect on the proliferation on 6/6 triple negative breast cancer ( TNBC ) tumors ( Figure 7D ) . These results suggest that ERX-11 has the potential to selectively influence the growth of human breast tumors expressing ER . 10 . 7554/eLife . 26857 . 027Figure 7 . ERX-11 decreases the growth of patient-derived explants ( PDEx ) : Schematic representation of ex vivo culture model is shown . ( A ) The explants were treated with ERX-11 for 48 hr . Effect of ERX-11 on Ki67 expression in ER-positive tumors with representative sections from three individual tumors and overall trend are shown ( B ) . Effect of ERX-11 on ER expression in three representative ER-positive tumors is also shown ( C ) . Effect of ERX-11 on Ki67 expression in three representative ER-negative tumors is shown ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 02710 . 7554/eLife . 26857 . 028Figure 7—figure supplement 1 . Effect of ERX-11 treatment on the status of ER . Effect of ERX-11 treatment on ER + PR + patient-derived explants from three individual patients , as assessed by Ki67 , ER and PELP1 immunohistochemistry . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 02810 . 7554/eLife . 26857 . 029Table 4 . Clinicopathologic characteristics of the 12 patients , whose ER+ , PR + status in breast tumors were analyzed by Ki67 and ER staining . This data is related to Figure 7B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 029Case #TumorER%PR%HER2 1IDC10080Negative 2IDC9090Negative 3Papillary10070Non-amplified 4IDC9090Negative 5IDC10095Non-amplified 6IDC10030Negative 7IDC10010–40Negative 8IDC8050Negative 9IDC50–6080–90Negative 10IDC100100Non-amplified 11IDC10095Non-amplified 12IDC9595Negative
The majority of breast cancer is ER positive . Therapeutic agents that suppress oncogenic ER activation by depletion of hormone-driven growth signaling or by blocking synthesis of hormone have become the mainstay of systemic treatment for breast cancer . However , both de novo and acquired therapy resistance are major clinical challenges . Importantly , ER signaling is intact in these therapy-resistant tumors . ER interaction with critical coregulator proteins appears to mediate ER signaling in these therapy-resistant and ER-positive metastatic tumors . While AIs and AEs may disrupt some of these ER-coregulator interactions , their ability to target these interactions is limited in therapy-resistant cells . In this study , we described the development of a novel inhibitor that targets ER interactions with coregulators . Using in vitro and in vivo assays , we demonstrated that ( Musgrove and Sutherland , 2009 ) ERX-11 blocks both ligand-dependent and ligand-independent ER signaling , ( Ma et al . , 2015 ) ERX-11 effectively blocks ER signaling in both therapy-sensitive and therapy-resistant cells and ( McDonnell and Norris , 2002 ) disruption of ER signaling by ERX-11 has biological activity against both therapy-sensitive and therapy-resistant cells , with minimal overt signs of toxicity in vitro and in vivo . Until recently , protein–protein interactions have been viewed as undruggable . However , the recent advent of a transformative class of compounds ( peptidomimetics ) allowed us to rationally design and synthesize small organic molecules that structurally emulate target protein sequences in defined conformations . We have developed a novel oligo-benzamide scaffold for mimicking helical protein segments ( Ravindranathan et al . , 2013; Ahn and Han , 2007 ) . The rigid framework of the oligo-benzamide scaffold can present functional groups mimicking amino acid residues in a helical conformation ( e . g . ones in i , i-3/4 , and i + 7 positions ) . Furthermore , we have established efficient synthetic routes to make bis- and tris-benzamides as alpha-helix mimetics ( Lee and Ahn , 2011; Marimganti et al . , 2009 ) . Based on the helix mimicking tris-benzamide scaffold , we designed ERX-11 to block the interactions between ER and a subset of its coregulator proteins containing the NR box in a helical structure . The unbiased IPMS analyses suggest that ERX-11 interacts with ER and blocks the interactions of ER with multiple coregulators . While the broad effect of ERX-11 protein–protein interactions raises potential concerns about off-target activities , they are largely mitigated by the lack of overt signs of toxicity in cell culture models and multiple animal models tested to date . In addition , the multiplicity of targeted protein-protein interactions makes the development of resistance to the ERX-11 less likely . Thus , ERX-11 represents an exciting new mechanism to attenuate ER oncogenic functions . Like tamoxifen , ERX-11 potently blocks the proliferation of therapy-sensitive cells . Unlike tamoxifen , ERX-11 has activity in multiple therapy-resistant models , including those driven by ER ligand-binding domain mutants . Unlike a classic SERD , ERX-11 does not cause immediate ER degradation , but appears to affect ESR1 levels over several days , by blocking its transcription . Our data clearly indicate that ERX-11 binds to AF2 domain of ER . However , the exact interface between ERX-11 and ER has not been established . The ability of tamoxifen to compete efficiently for ERX-11 binding to purified ER in vitro and to ER within the cell suggests a significant overlap between the tamoxifen and ER-binding site on ER . In addition , the inability of the SERDs and ICI to disrupt the interaction between ER and ERX-11 suggests that ERX-11 may also interact with a secondary tamoxifen-binding site within ER . As shown for ER beta , tamoxifen has two distinct binding sites- one in the consensus ligand-binding pocket , and another in the hydrophobic groove of the coactivator recognition surface ( Wang et al . , 2006 ) . Published studies using LXXLL peptide probes showed ER undergo distinct conformational changes as a result of binding with different ligands and such changes expose distinct surfaces on ER facilitating interaction with various coregulators ( Paige et al . , 1999 ) . These studies also reported that tamoxifen binding can create unique surface on ER that facilitate binding of unique LXXLL-binding peptides ( Tamrazi et al . , 2002 ) . Further investigations of the ERX-11- ER interface with co-crystallization studies are ongoing and are likely to clarify the precise nature of the interaction . Our studies also suggest that ERX-11 can interact with ER , even in therapy-resistant cells . While therapy resistance can be attributed to multiple mechanisms , structural changes in ER via post-translational modifications or point mutations may create new binding surfaces on ER for coregulator interactions and potentially for ERX-11 interactions . The interaction between ERX-11 and ER-MTs in therapy-resistant cells supports a model for ERX-11 interaction with ER at an alternate secondary site distinct from the site of tamoxifen binding . These findings may explain the differences between ERX-11 and tamoxifen in their activity in both therapy-sensitive and therapy-resistant breast cancer cells . The ability of ERX-11 to block ER dimerization may be responsible for its disruption of ER DNA binding and these findings are supported by published reports that LXXLL peptides may affect ER dimerization ( Tamrazi et al . , 2002 ) . We have noted that ERX-11 has a weak affinity for the A-isoform of PGR but not for the B-isoform . We have also found that ERX-11 does not interact with other steroid receptors like GR or AR . In addition , ERX-11 failed to show activity on AR-expressing prostate cancer cells . Disruption of multiple protein–protein interactions by ERXs may result in significant toxicity . To evaluate this , we have initially tested toxicity using a dose of 100 mg/kg in immune competent C57BL/6 mice for 14 days . We did not observe any uterotropic activity or immune effects . Further , no overt signs of toxicity was found in various E2-responsive organs including the liver , lung , heart , and kidney . Similarly , in five separate studies using a dose of 10 mg/kg in tumor-bearing mice , we did not find any overt signs of toxicity; however , ERX-11 treatment significantly limited ER expression in the tumors and reduced tumor growth . To address potential issues with immunogenicity of these tris-benzamides , we also evaluated ERX-11 treatment in syngeneic models and found no overt signs of toxicity . While the ER coregulator protein levels are tightly regulated under normal conditions , many coregulators are over-expressed in breast cancer ( Lonard et al . , 2007 ) , substantially contribute to ER-signaling , drive disease progression ( Singh and Kumar , 2005 ) and correlate with a poor prognosis ( List et al . , 2001; Azorsa et al . , 2001; Tamrazi et al . , 2002 ) . The differential coregulator milieu within breast tumors and the dependence of breast tumors on ER and coregulator-driven signaling may explain why ERX-11 has potent antiproliferative activity within the tumor but does not have any overt signs of toxicity . Since ERX-11 blocked some but not all ER coregulator interactions , ERX-11 may function as a polypharmacology agent , and its activity may depend on the concentration and repertoire of coregulators present in a tumor cell . Importantly , using ex vivo culture of patient-derived tumor tissues , we demonstrated that ERX-11 is effective in limiting proliferation of ER-positive but not ER-negative tumors . We also discovered that ERX-11 treatment of primary tumors within their native microenvironment also reduced ER levels within the tumor . None of the primary ER-negative tumors responded to ERX-11 therapy . These explant studies represent the first evaluation of a drug effect using tissues from primary breast cancer patients and are likely to show biologically relevant outcomes . In response to hormone binding , ER interacts with multiprotein complexes containing coregulators and transcriptional regulators to activate transcription ( McKenna et al . , 1999; Collingwood et al . , 1999; Tsai and O'Malley , 1994; Torchia et al . , 1998 ) . Even though coregulators modulate ER functions , each coregulator protein appears to play an important but not overlapping function in vivo ( Han et al . , 2006; Xu et al . , 1998 ) . Accordingly , the ERX-mediated blockage of coregulator interactions with ER resulted in both inactivation and activation of unique sets of genes and pathways modulated by ER oncogenic signaling , leading to tumor suppression . The pathways/genes modulated by ERX-11 can be used to correlate with the outcome of its therapy , and they may serve as biomarkers that prognosticate response to these agents . The biology of E2- ER signaling is complex and context dependent . Elegant studies have shown that ER , depending on the ligand and presence of unique set of coregulators can promote apoptosis ( Jordan , 2015 ) . In that scenario , antiestrogen and SERM are shown to inhibit apoptosis , hence , many of these drugs exhibit cytostatic response . Estrogen-induced apoptosis is shown to cause an increase in Fas receptor associated with the extrinsic pathway of apoptosis ( Lewis-Wambi and Jordan , 2009 ) . Similarly , a recent study found that inhibition of SRC family coregulators using small molecule inhibitor promote significant apoptosis ( Song et al . , 2016 ) . We consistently observed activation of apoptosis by ERX-11 but failed to observe any apoptosis by tamoxifen or ICI in our assays . We believe that activation of apoptosis is not due to elimination of ER rather due to unique mechanism of action of ERX-11 . Specifically , we predict that changes in the ER signaling is due to alterations in coregulator binding to ER . Accordingly , our RNAseq data showed that alterations in genes that contribute activation of apoptosis . However , further studies are needed to clearly identify the mechanisms by which ERX-11 promotes apoptosis . Currently used drugs ( AEs and AIs ) are associated with an initial period of clinical response; however , most patients develop resistance with cancer progression . Recent studies suggested that selective estrogen receptor downregulators ( SERDs ) , molecules that eliminate ER expression , may have utility for treating breast cancers that have progressed on AE and/or AIs ( McDonnell et al . , 2015 ) . Several orally available SERDs ( GDC-0810 , AZD9496 ) were recently developed and shown to have utility in treating resistant tumors using preclinical models ( Lai et al . , 2015; Weir et al . , 2016 ) . However , it is important to note that each of the SERDs are only in phase I clinical trials ( clinical trials . gov ) and none of them are either FDA approved or have proven efficacy in patients . The only FDA-approved agent is Fulvestrant ( ICI ) , which has known pharmacologic limitations as evidenced by its dosing regimen of intramuscular injections every 14 days . Recent studies also showed ESR1 mutations lead to constitutive activity and reduced sensitivity to ER antagonists , and mutations such as Y537S contribute to fulvestrant resistance in vivo ( Toy et al . , 2017 ) . Since ERXs are small , stable , orally bioavailable small molecular inhibitors , their use as an alternative therapeutic approach may decrease therapy resistance and reduce side effects , which are the current limitations of AEs or AIs . In summary , we have developed and tested the utility of ERX-11 as a novel therapeutic agent for ER-positive , therapy-sensitive and therapy-resistant breast cancers . Since ERX-11 is orally available and is well tolerated with fewer side effects , ERX-11 can be readily extended to clinical use as a therapeutic agent , and may enhance the survival of advanced breast cancer patients .
Human breast cancer cells MCF-7 , ZR-75 , T-47D , MDA-MB-231 , BT474 , BT549 , BT453 , SUM159 , 4T1 , MM468 , HCC1937 , HCC1187 , HCC70 , MDA-MB-157 , MDA-MB-453 , MDA-MB-468 and HEK293T cells were either obtained from American Type Culture Collection ( ATCC , Manassas , VA ) or a kind gift from Dr . John Minna at UT Southwestern . Ishikawa cells were purchased from Sigma ( St . Louis , MO ) . All of these cells were passaged in the user's laboratory for fewer than 6 months after receipt or resuscitation . Validation experiments were performed using a number of luminal , basal and TNBC cells with extensive prior available molecular profiles were a kind gift of Dr . John Minna . All the model cells utilized are free of mycoplasma contamination . Additionally , STR DNA profiling of the cells was used to confirm the identity using UTHSA and UT Southwestern core facilities . MCF-7-PELP1 cells ( Vallabhaneni et al . , 2011 ) , MCF-7-HER2 cells ( Nabha et al . , 2005 ) , MCF-7-TamR cells ( Nabha et al . , 2005 ) , MCF-7-LTLTca cells ( Macedo et al . , 2008 ) and D2A1 cells ( Tekmal and Durgam , 1997 ) were described earlier . MCF-7-LTLTca and MCF-7-TamR cells were cultured in Phenol red-free RPMI medium containing 10% dextran charcoal-treated serum supplemented with either 1 μmol/L of letrozole or 1 μmol/L of tamoxifen , respectively . 17-β-Estradiol ( cat#E2257 ) , and ( Z ) −4-Hydroxytamoxifen ( cat# H7904 ) were purchased from Sigma ( St . Louis , MO ) . CRISPR/Cas9 plasmids targeting ESR1 gene were obtained from Horizon Discovery ( Cambridge , MA ) . The anti-PELP1 ( cat# 300-180A ) and anti-AIB1 ( cat# A300-347A ) antibodies were purchased from Bethyl Laboratories ( Montgomery , TX ) . Cleaved caspase 3 antibody was purchased from Cell signaling technology ( cat# 9661S , Danvers , MA ) . TUNEL kit ( cat# 11684795910 ) for apoptosis detection was purchased from Roche ( Mannheim , Germany ) and Ki-67 ( 1:150 ) anti-human clone MIB-1 antibody ( cat#M7240 ) was purchased from Dako ( Carpinteria , CA ) . The designed tris-benzamides were constructed by iterative amide bond formation of a 3-alkoxy-4-nitrobenzoic acid with a 3-alkoxy-4-aminobenzamide ( Figure 1—figure supplement 3 ) . A 4-nitrobenzoic acid containing a trityl-protected hydroxyethoxy group 3 g was coupled to bis-benzamide 8 that was synthesized by following the previously reported procedure ( Ravindranathan et al . , 2013 ) , making tris-benzamide 9 ( Figure 1—figure supplement 3C ) . See the chemistry supplement for detailed synthetic procedures and characterization . The effects of ERX analogues on cell viability were measured using the MTT Cell Viability Assay in 96-well plates . Breast cancer cells were seeded in 96-well plates ( 1 × 103 cells/well ) in phenol red-free RPMI medium containing 5% dextran-coated charcoal-treated fetal bovine serum ( DCC-FBS ) serum . After an overnight incubation , cells were treated with varying concentrations of the ERX analogues in the presence or absence of E2 ( 1 × 10−8 M ) for 7 days . For some experiments viability was also measured using Cell Titer-Glo Luminescent Cell Viability Assay ( Promega ) in 96-well , flat , clear-bottom , opaque-wall micro plates according to manufacturer’s protocol . For some experiments , apoptosis was measured using Caspase-Glo 3/7 Assay ( Promega ) using manufacturer’s protocol . Western blotting and immunoprecipitation were performed as described previously ( Nair et al . , 2010 ) . Biotin-ERX-11 pull-down assays were done using a previously established protocol using avidin beads ( Mann et al . , 2013 ) . Pull-down assays using ER-AF2-GST were performed as described previously ( Nair et al . , 2010 ) . Purified ER full-length proteins and ER LBD ( AF2 ) protein was purchased from Thermo Fisher Scientific , Waltham , MA . The sequences of LXXLL peptides used in the competition assays . SRC1- LXXLL peptide: LTARHKILHRLLQEGSPSD; SRC2- LXXLL peptide: DSKGQTKLLQLLTTKSDQM; SRC3/AIB1- LXXLL peptide: ESKGHKKLLQLLTCSSDDR; PELP1-LXXLL-1 peptide: GLSAVSSGPRLRLLLLESVSG; PELP1-3 LXXLL peptide: SIKTRFEGLCLLSLLVGESPT All animal experiments were performed after obtaining UTHSA IACUC approval and using methods in the approved protocol . For xenograft tumor assays , 2 × 106 ZR-75 , or ZR-75- ER MT-Y537S , or MCF-7- PELP1 cells were mixed with an equal volume of matrigel and implanted in the mammary fat pads of 6-week-old female athymic nude mice as described ( Cortez et al . , 2012 ) . Based on our previous data as well as published findings , the number of mice needed were chosen to demonstrate differences in tumor incidence or treatment effect . Calculations are based on a model of unpaired data power = 0 . 8; p<0 . 05 . Once tumors reached measurable size , mice were divided into control and treatment groups ( n = 5–8 tumors per group ) . The control group received vehicle and the treatment groups received ERX-11 ( 10 mg/kg/day ) in 30% Captisol orally . Dose were selected based on pilot MTD study of 10 , 50 and 100 mg/kg of ERX-11 for 14 days using C57BL/6 mice . The mice were monitored daily for adverse toxic effects . For MCF-7-LTLT xenograft studies , MCF-7-LTLT model cells were first injected into the mammary glands of nude mice implanted with androstenedione pellets . When the tumor was established , it was dissected into small pieces and they were again implanted subcutaneously into nude mice implanted with androstenedione pellets . Fulvestrant treatment was used as a positive control and MCF-7-LTLT xenografts were treated with 200 mg/kg/ 2 days a week/sc . For syngeneic mice studies , D2A1 cells were first injected into the mammary glands of BALB/c mice . When the tumor was established , it was dissected into small pieces and they were again implanted subcutaneously into the BALB/c mice . After three days of tumor tissue implantation , mice were randomly selected to receive control ( n = 7–8 ) and treatment ( n = 7–8 ) with 20 mg/kg/day of ERX-11 orally . Tumor growth was measured with a caliper at 3–4 day intervals . At the end of each experiment , the mice were euthanized , and the tumors were removed , weighed and processed for IHC staining . UTSW Patients provided written consent allowing the use of discarded surgical samples for research purposes according to an institutional board-approved protocol . De-identified patient tumors were obtained from the UTSW Tissue Repository after institutional review board approval ( STU-032011–187 ) . Excised tissue samples were processed and cultured ex vivo as previously described ( Mohammed et al . , 2015 ) . Briefly , tissue samples were incubated on gelatin sponges for 48 hr in culture medium containing 10% FCS , followed by treatment with either vehicle or E2 ( 10 nM ) in the absence or presence of 10 μM ERX-11 for 48 hr ( see Table 4 for clinicopathologic characteristics of these tumors ) . Representative tissues were fixed in 10% formalin at 4°C overnight and subsequently processed into paraffin blocks . Sections were stained with hematoxylin and eosin and examined to confirm and quantify the presence/proportion of tumor cells . Immunohistochemistry was then performed . String analyses were performed for human PELP1 using the http://string-db . org website , with the evidence view at the highest stringency for no more five interactors . A Monte Carlo conformational search was performed using the torsional sampling method ( MCMM ) implemented in MacroModel ( version 9 . 0 , Schrödinger , New York , NY ) with automatic setup options . The calculation was done with the maximum number of steps set to 5000 using 100 steps per rotatable bond and an energy cutoff of 21 kJ/mol above the global energy minimum . The searches were done using MM3 force field ( chosen for its accuracy with organic molecules ) combined with the GB/SA water solvation model with standard settings and the following cut-offs: van der Waals , 8 . 0 Å; electrostatic , 20 . 0 Å; and hydrogen bond , 4 . 0 Å . The observed conformations were minimized by 500 iterations of Polak-Ribiere Conjugate Gradient ( PRCG ) algorithm ( a conjugate gradient minimization scheme that uses the Polak-Ribiere first derivative method with restarts every 3N iterations ) ( 0 . 05 kJ/mol ) . AutoDock 4 . 2 software package , as implemented through the graphical user interface called AutoDockTools ( ADT ) , was used to create input PDBQT files of a receptor and a ligand . The input file of ER was prepared using the published coordinates ( PDB 1L2I ) . Water molecules were removed from the protein structure and hydrogen was added . All other atom values were generated automatically by ADT . The docking area was assigned visually around the peptide ligand . A grid box of 24 Å x 20 Å x 24 Å was calculated around the docking area using AutoGrid . The x , y , z coordinates of the center of the grid box were set to x = −9 . 0 , y = 14 . 0 and z = 26 . 0 , respectively . The input file of ERX-11 was created from its energy-minimized conformation using ADT . Docking calculations were performed with AutoDock Vina 1 . 1 . 2 . A search exhaustiveness of 16 was used and all other parameters were left as default values . Briefly , cells were transiently co-transfected with 200 ng of ERE-Luc reporter with 100 ng of ER-WT , ER-MT , PELP1 , SRC1 , SRC2 , SRC3 or control vectors using Turbofect transfection reagent ( Thermo Scientific , Waltham , MA ) . After 24 hr , cells were treated with either vehicle or ERX-11 for an additional 24 hr . β-galactosidase reporter ( 50 ng ) plasmid was co-transfected and used for data normalization . Cells were lysed in Passive Lysis Buffer , and luciferase activity was measured using the luciferase assay system ( Promega , Madison , WI ) in a luminometer . RNA-seq was performed using the UTHSA core–established protocol . Briefly , ZR-75 cells were treated with either vehicle or ERX-11 for 48 hr , and total RNA was isolated using RNAesy mini kit ( Qiagen ) according to the manufacturer’s instructions . Differential expression analysis was performed by DEseq and significant genes with at least 1 . 5-fold change with p<0 . 01 were chosen for analysis . The interpretation of biological pathways using RNA-seq data was performed with IPA software using all significant and differentially expressed genes . RNA-seq data have been deposited in the GEO database under accession number GSE75664 . To validate the selected genes , reverse transcription ( RT ) reactions were performed by using SuperScript III First Strand kit ( Invitrogen , Carlsbad ) , according to manufacturer’s protocol . Real-time PCR was done using SybrGreen on an Illumina Real-Time PCR system , using primers listed in Table 5 . 10 . 7554/eLife . 26857 . 030Table 5 . Primer sequences used for RTqPCRDOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 030Gene namePrimer sequenceSRC-325FGAGCGGCTCCAGATTGTCAASRC-410RCTGGGGATGTAGCCTGTCTGTE2F1-378FACGCTATGAGACCTCACTGAAE2F1-626RTCCTGGGTCAACCCCTCAAGERCC2-68FGGAAGACAGTATCCCTGTTGGCERCC2-169RCAATCTCTGGCACAGTTCTTGALIMK1-276FCAAGGGACTGGTTATGGTGGCLIMK1-367RCCCCGTCACCGATAAAGGTCMMP15-149FAGGTCCATGCCGAGAACTGMMP15-305RGTCTCTTCGTCGAGCACACCDUSP2-491FGGGCTCCTGTCTACGACCADUSP2-574RGCAGGTCTGACGAGTGACTGRCOR2-118FCACTCGCACGACAGCATGATRCOR2-285RCATCGCAATGTACTTGTCAAGCDKK1-95FCCTTGAACTCGGTTCTCAATTCCDKK1-232RCAATGGTCTGGTACTTATTCCCGPDGFB-63FCTCGATCCGCTCCTTTGATGAPDGFB-301RCGTTGGTGCGGTCTATGAGPGLYRP2-23FTCCTACTCGGATTGCTACTGTGPGLYRP2-206RAAGTGGTAGAGGCGATTGTGGELF5-357FTAGGGAACAAGGAATTTTTCGGGELF5-519RGTACACTAACCTTCGGTCAACCTNFSF10-46FTGCGTGCTGATCGTGATCTTCTNFSF10-126RGCTCGTTGGTAAAGTACACGTASTAT1-368FATCAGGCTCAGTCGGGGAATASTAT1-553RTGGTCTCGTGTTCTCTGTTCTXAF1-297FGCTCCACGAGTCCTACTGTGXAF1-403RGTTCACTGCGACAGACATCTCIFI6-257FGGTCTGCGATCCTGAATGGGIFI6-401RTCACTATCGAGATACTTGTGGGT Chromatin immunoprecipitation ( ChIP ) analysis was performed using antibodies specific for the ER ( Santa Cruz ) . Briefly , MCF-7 ( 7 × 106 ) or T-47D ( 2 × 107 ) cells were plated in 150 mm dishes , starved in unsupplemented phenol red-free DMEM for 24 hr and then treated for 2 hr with either ethanol or E2 after prior incubation with either DMSO or ERX-11 . Relative recruitment was determined by qPCR of purified ChIP and input DNA in triplicate . The results presented are representative of two independent experiments . The NanoBiT assay utilizes a structural complementation-based approach to monitor protein–protein interactions within living cells . Large BiT ( LgBiT; 18 kDa ) and Small BiT ( SmBiT; 1 kDa ) subunits of NanoLuc Luciferase were optimized for the analysis of protein interaction dynamics . When LgBiT and SmBiT subunits are separated , the Large BiT part loses the majority of luciferase activity . However , when the direct interaction between fusion proteins on LgBiT and SmBiT occurred , the interaction promotes structural complementation between LgBiT and SmBiT and results in full luciferase activity . Protein–protein interaction are then monitored in living cells following addition of the Nano-Glo Live Cell Reagent , a non-lytic detection reagent containing the cell-permeable furimazine substrate and observed luminescent signals . To generate different NanoBiT fusion constructs , human ER and PELP1 coding sequences were amplified by PCR and separately subcloned into NB-MCS vectors ( Promega ) . To test the protein–protein interaction between ER and PELP1 by using the NanoBiT assay , C-LgBit-ESR1 paired with C-SmBit-PELP1 or C-LgBit-PELP1 with C-SmBit-ESR1 constructs were transiently transfected into HEK-293T cells by using Fugene HD transfection reagent ( Promega ) . To test the ER dimerization , C-SmBit-ESR1 were cotransfected with either N-LgBit-ESR1 or C-LgBit-ESR1 constructs . On the day after transfection , the medium for the HEK-293T cells was changed to phenol red-free DMEM containing 1% charcoal-stripped FBS . After a 24 hr incubation , the cells were treated with DMSO or ERX-11 ( 10 μM ) for 2 hr and then treated cells with EtOH or E2 ( 10 nM ) for 30 mins . After treatment , Nano-Glo live cell reagents were added into cells and luminscence was measured after 10 mins . MCF-7 cells were cultured on collagen-coated cover slips . After treated with vehicle or 10 µM ERX-11 for 30 min , cells are treated with vehicle or 10 nM E2 for 30 min . After the treatment , cells were washed with PBS and then fixed with 10% buffered formalin for 20 min and permeabilized with ice cold methanol at −20°C for 5 min . The cells were then blocked with blocking solution ( provided with Duolink In Situ PLA probe , Sigma ) for 30 min at 37°C followed by incubating with primary antibodies for ER from different species at 37°C for 2 hr . After washing twice with Wash Buffer A ( Duolink In Situ Reagents ) at room temperature , cells were then incubated with the appropriate anti-species secondary antibodies to which oligonucleotides had been conjugated ( anti-rabbit PLA probe PLUS and anti-mouse PLA probe MINUS ) for 1 hr at 37°C , followed by treatment with Duolink ligation-ligase solution for 30 min at 37°C . Finally , the cells were incubated with the Duolink amplification-polymerase solution for 60 min at 37°C , followed by washing and mounting on slides with Duolink Mounting Medium with DAPI . Images were taken using a Nikon Fluorescence Microscope . MCF-7 or ZR-75 cells were grown in RPMI-1640 medium supplemented with 10% FBS . After cells reach 90% confluence , the medium was changed to phenol red-free RPMI1640 containing 1 . 5% charcoal stripped serum for 48 hr to starve the cells . After starvation , the cells were treated with vehicle or 10 µM ERX-11 for 2 hr followed by a 2 hr treatment with vehicle or 10 nM E2 . Then , the cells were washed using PBS and incubated with Pierce IP Lysis Buffer ( Thermo Fisher Scientific Inc . ) at 4°C for 20 min . The cell lysates were centrifuged at 14 , 000 rpm at 4°C for 10 min , and the supernatants were used for IPMS analysis . Dynabeads Protein G ( Invitrogen ) are pre-coupled with anti- ER antibody ( Santa Cruz , sc-8002 ) and then incubated with 1 . 5 mg cell lysate over night at 4°C . Then the Dynabeads were washed using PBS and PBST ( 0 . 01% Tween 20 ) and eluted in 30 µL NuPAGE LDS sample buffer at 95°C for 15 min . The final eluents containing ER and associated proteins were separated by using SDS-PAGE . The obtained proteins were proteolytically digested and subjected to mass spectrometry ( MS ) analysis . Gel band samples were digested overnight with trypsin ( Promega ) following reduction and alkylation with DTT and iodoacetamide ( Sigma ) . The samples then underwent solid-phase extraction cleanup with Oasis HLB plates ( Waters ) and the resulting samples were analyzed by LC/MS/MS , using an Orbitrap Fusion Lumos ( Thermo Electron ) coupled to an Ultimate 3000 RSLC-Nano liquid chromatography systems ( Dionex ) . Samples were injected onto a 75 μm i . d . , 50 cm long Easy Spray column ( Thermo ) and eluted with a gradient from 0% to 28% buffer B over 60 min . Buffer A contained 2% ( v/v ) ACN and 0 . 1% formic acid in water , and buffer B contained 80% ( v/v ) ACN , 10% ( v/v ) trifluoroethanol , and 0 . 08% formic acid in water . The mass spectrometer operated in positive ion mode with a source voltage of 2 . 2 kV and capillary temperature of 275°C . MS scans were acquired at 120 , 000 resolution and up to 10 MS/MS spectra were obtained in the ion trap for each full spectrum acquired using higher-energy collisional dissociation ( HCD ) for ions with charge 2–7 . Dynamic exclusion was set for 25 s . Raw MS data files were converted to a peak list format and analyzed using the central proteomics facilities pipeline ( CPFP ) , version 2 . 0 . 3 . Peptide identification was performed using the X ! Tandem and open MS search algorithm ( OMSSA ) search engines against the human protein database from Uniprot , with common contaminants and reversed decoy sequences appended . Fragment and precursor tolerances of 20 ppm and 0 . 1 Da were specified , and three missed cleavages were allowed . Carbamidomethylation of Cys was set as a fixed modification and oxidation of Met was set as a variable modification . Label-free quantitation of proteins across samples was performed using SINQ normalized spectral index Software . For immunohistochemical analysis sections were incubated overnight with the ER ( 1:50 ) or PELP1 ( 1:200 ) or Ki67 ( 1:100 ) antibody in conjunction with proper controls . The sections were then washed three times with 0 . 05% Tween in PBS for 10 min , incubated with secondary antibody for 1 hr , washed three times with 0 . 05% Tween in PBS for 10 min , visualized by DAB substrate and counterstained with hematoxylin QS ( Vector Lab , Burlingame , CA ) . A proliferative index was calculated as the percentage of Ki-67-positive cells in five randomly selected microscopic fields at 40X per slide . TUNEL analysis was done using the in situ Cell Death Detection Kit ( Roche , Indianapolis , IN ) as per the manufacturer’s protocol , and five randomly selected microscopic fields in each group were used to calculate the relative ratio of TUNEL-positive cells . For the PDEx samples , 5-µm sections were de-waxed , rehydrated and endogenous peroxidases were blocked with hydrogen peroxide . Sections were then boiled in citrate and blocked in 5% serum for 1 hr . Primary antibodies were incubated overnight at 4°C at 1:200 for Ki67 ( Vector Laboratories Burlingame , CA ) and at 1:400 for ER ( Santa Cruz Biotechnology , CA ) . Biotinylated anti-rabbit secondary antibodies ( DAKO Carpentaria , CA ) were incubated for 60 min at room temperature after slides were washed for 1 hr in PBS . Slides were incubated in ABC-HRP complex ( Vector Laboratories Burlingame , CA ) for 30 min . Bound antibodies were then visualized by incubation with 3 , 3’ diaminobenzidine tetrahydrochloride ( liquid DAB , DAKO ) . Slides were then rinsed in tap water and counterstained with hematoxylin , and then cover slide were mounted . Tumor cells with nuclear staining were recorded as positive manually per tissue core , by a reviewer who was blinded to the clinical data . The numbers of Ki67-positive tumor cells were counted in three high-power fields ( x40 ) . The ER immunostaining was registered semi-quantitatively in two ways . Staining intensity ( 0 , no staining; 1 , weak staining; 2 , moderate staining; and 3 , intense staining ) and the proportion of stained cells ( 0 , no staining; 1 , 1–25% staining; 2 , 26–50%; 3 , 51–75%; and 4 , if more than 75% of the tumor cells were positive . GraphPad Prism 6 software ( GraphPad Software , SanDiego , CA ) was used to analyze all data . Data represented in the bar graphs is shown as mean ± SEM . t-test was performed for all pairwise comparisons . A value of p<0 . 05 was considered as statistically significant . The multiple groups’ statistical data were analyzed with one-way ANOVA . RNA-seq data were analyzed using IPA software . 10 . 7554/eLife . 26857 . 031Table 6 . Analyses of the amino acids at the flanking sequences of top ER binders whose interactions are blocked by ERX-11 in MCF-7 and ZR-75 cells , as determined by unbiased IP-MS . DOI: http://dx . doi . org/10 . 7554/eLife . 26857 . 031Protein/GENE ID# LXXLL motifsLXXLL sequencesSerine at i ± 3/4 Plectin Q15149-36210 GHNLISLLEVL 220 213 LISLLEVLSGDS 224 421 YRELVLLLLQWM 431 659 LRYLQDLLAWV 669 1102 YQQLLQSLEQG 1112 4006 TGQLLLPLSDA 4016yes FAM83H Q6ZRV22816 AAQLLDTLGRS 826 966 SLRLRQLLSPK 976yes AHNAK Q096660 CLTC: cQ006105563 TAFLLDALKNN 573 854 RNRLKLLLPWL 864 1001 PNELIELLEKIV 1011 1021 HRNLQNLLILT 1031 1418 PLLLNDLLMVLS 1429yes FAS P493271076 DPQLRLLLEVT 86 418 HATLPRLLRAS 428 560 QIGLIDLLSCM 570 691 APPLLQELKKV 701 1175 QQELPRLLSAA 1185 1211 EDPLLSGLLDSP 1221 1346 GFLLLHTLLRGH 1358 1470 RCVLLSNLSST 1480 2216 QLNLRSLLVNP 2226 2381 NRVLEALLPLKG 2391yes TRG P276351116 VNRLLDSLEPP 126no ACTN4 O43707181 GLKLMLLLEVIS 92yes TFG Q927340 Coatomer subunit alpha P53621183 RRCLFTLLGHLDYI 96no Q9NVI7-2199 ALSLLHTLVWA 109no | Around 70% of breast cancers in women need one or both of the female hormones ( estrogen and progesterone ) to grow . To treat these 'hormone-dependent' cancers , patients receive drugs that either block the production of estrogen or directly target a receptor protein that senses estrogen in the cancer cells . Unfortunately , many breast cancers develop resistance to these drugs . This resistance is often caused by genetic mutations that alter the estrogen receptor; for example , the receptor may develop the ability to interact with other proteins in the cell known as coregulators to promote tumor growth . Developing new drugs that prevent estrogen receptors from interacting with coregulators may provide more options for treating hormone-dependent breast cancers . Here , Raj et al . developed a new small molecule named ERX-11 that is able to inhibit the growth of human breast cancer cells that are sensitive to existing drugs as well as cells that have become drug-resistant . For the experiments , hormone-dependent breast cancer cells from humans were transplanted into mice . This procedure usually causes the mice to develop tumors , but giving the mice ERX-11 by mouth stopped estrogen receptors from interacting with coregulators and blocked the growth of tumors . Furthermore , ERX-11 does not appear to have any toxic effects on the mice , indicating that it may also be safe for humans . The findings of Raj et al . suggest that ERX-11 is a promising new drug candidate for treating some breast cancers . The next steps are to examine the effects of ERX-11 on mice and other animals in more detail before deciding whether this molecule is suitable for clinical trials . In the longer term , molecules similar to ERX-11 could also be developed into drugs to treat other types of cancer that are also caused by abnormal interactions of coregulator proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2017 | Estrogen receptor coregulator binding modulators (ERXs) effectively target estrogen receptor positive human breast cancers |
One key bottleneck in understanding the human genome is the relative under-characterization of 90% of protein coding regions . We report a collection of 1200 transgenic zebrafish strains made with the gene-break transposon ( GBT ) protein trap to simultaneously report and reversibly knockdown the tagged genes . Protein trap-associated mRFP expression shows previously undocumented expression of 35% and 90% of cloned genes at 2 and 4 days post-fertilization , respectively . Further , investigated alleles regularly show 99% gene-specific mRNA knockdown . Homozygous GBT animals in ryr1b , fras1 , tnnt2a , edar and hmcn1 phenocopied established mutants . 204 cloned lines trapped diverse proteins , including 64 orthologs of human disease-associated genes with 40 as potential new disease models . Severely reduced skeletal muscle Ca2+ transients in GBT ryr1b homozygous animals validated the ability to explore molecular mechanisms of genetic diseases . This GBT system facilitates novel functional genome annotation towards understanding cellular and molecular underpinnings of vertebrate biology and human disease .
Analyses of genomic sequences from over 100 vertebrate species ( Meadows and Lindblad-Toh , 2017 ) have revealed that we need more than nucleic acid sequence alone to comprehend the vertebrate genome . A more complete understanding of any genetic locus requires knowledge of its expression pattern and its function ( s ) in subcellular , cellular , and organismal contexts—the compendium of information that can be described as a gene ‘codex’ . Despite their importance , the expression patterns and functions of most protein coding genes remain surprisingly uncharacterized . The number of these genes linked to human disease without functional insights into their gene-disease relationships highlights the significance of this knowledge gap ( Kettleborough et al . , 2013 ) . In recent estimates , 80% of rare , undiagnosed diseases are thought to have genetic underpinnings ( Robe , 2005; Varga et al . , 2018; Wangler et al . , 2017 ) . Tools are therefore needed to identify and annotate the expression and function ( s ) of these poorly characterized gene products in both biological and pathological processes . Zebrafish ( Danio rerio ) has emerged as an outstanding model to bridge the gap between sequence and function in the vertebrate genome . Investigations of gene function in zebrafish , from organismal to subcellular , are amenable to both forward and reverse genetic approaches ( Stoeger et al . , 2018 ) . Additionally , the natural transparency of developing zebrafish enables live , non-invasive collection of gene expression data at a subcellular resolution on an organismal scale . Therefore , the zebrafish facilitates parallel discovery of gene expression and function towards a comprehensive codex of the vertebrate genome . To begin constructing this vertebrate codex , we previously developed a unique , revertible mutagenesis tool called the gene-break transposon ( GBT ) with elements that cooperate to report gene sequence , expression pattern , and function ( Clark et al . , 2011a ) . Specifically , when integrated in the sense orientation of a transcriptional unit , the GBT protein trap overrides endogenous splicing and creates a fusion between upstream exons and its start-codon deficient monomeric RFP ( mRFP ) reporter . Then , the strong internal polyadenylation site and putative border element following the mRFP truncate the gene product . Finally , the GBT construct is flanked by loxP sites on either side to enable excision and subsequent rescue with Cre-recombinase ( Clark et al . , 2011a ) . Visualization of the start-codon deficient mRFP reporter requires an in-frame integration . In the original GBT protein trap construct , RP2 . 1 ( Figure 1A ) , this in-frame requirement restricts mRFP expression to a single reading frame and leaves the potential to truncate genes without reporting their expression with mRFP ( Clark et al . , 2011a ) . We therefore developed a new series of GBT protein trap constructs , including versions to trap expression in each of the three potential reading frames ( Figure 1A ) . Alongside the original , we employed these new vectors in zebrafish to generate and catalog over 800 additional GBT protein trap lines with visible mRFP expression at 2 days post-fertilization ( dpf ) ( end of embryogenesis ) or four dpf ( larval stage ) . 147 of these additional GBT lines were cloned , and candidate genes were identified for another 144 GBT lines . mRFP expression in cloned GBT lines showcased novel expression patterns for a population of genes encoding diverse proteins in function and localization , including 64 implicated in human disease . Further , animals homozygous for the GBT allele in ryr1b displayed severely dampened skeletal muscle Ca2+ transients , demonstrating the ability to elucidate molecular mechanisms of genetic disorders . Since detailed investigations of mutant phenotypes are vital to functional annotation of the vertebrate genome , the mutagenic reporters in our GBT system provide the basis for this functional annotation to better understand normal biology and human disease .
We previously reported the intron-based gene-break transposon ( GBT ) as an effective and revertible loss-of-function tool for zebrafish ( Clark et al . , 2011a ) . The original GBT construct called RP2/RP2 . 1 contains the following key features ( Figure 1A ) : 1 ) flanking miniTol2 sequences for transposase-mediated random integration ( Balciunas et al . , 2006; Kawakami et al . , 2004; Urasaki et al . , 2006 ) , 2 ) a 5’ protein trap containing a strong splice-acceptor ( SA ) and a start codon-free mRFP reporter to detect 5’ sequence and visualize in vivo expression of the trapped locus with the endogenous promoter ( Clark et al . , 2011a; Ding et al . , 2013; Ding et al . , 2017; Liao et al . , 2012; Petzold et al . , 2009; Westcot et al . , 2015; Xu et al . , 2012 ) , 3 ) a mutagenic transcriptional terminator containing both a polyadenylation signal ( pA ) and a putative border element to truncate the trapped locus in conjunction with the protein trap ( Sivasubbu et al . , 2006 ) , 4 ) a 3’ exon trap with a β-actin promoter driving expression of GFP to report 3’ sequence and detect lines with weak ( or absent ) mRFP expression or with the integration in other frames of the mRFP reporter . ( Clark et al . , 2011a; Petzold et al . , 2009; Sivasubbu et al . , 2006 ) , 5 ) a second mini-intron within the GFP expression cassette that can further contribute to loss of wild-type transcripts , and 6 ) flanking loxP sites for Cre-mediated excision and restoration of trapped locus function using both germline ( Petzold et al . , 2009 ) and somatic approaches ( Clark et al . , 2011a; Ding et al . , 2013; Westcot et al . , 2015 ) . Initial experiments with the RP2 . 1 construct , however , revealed some limitations . First , effective transcript trapping does not always generate mRFP reporter expression because the RP2 . 1 plasmid is designed for a single reading frame . Molecular cloning of GFP+/mRFP- lines demonstrated the requirement to capture an appropriate reading frame to visualize the mRFP reporter . Even though RP2 . 1 is designed to use one main reading frame , some lines with mRFP expression used an alternate ‘CAG’ five nucleotides downstream of the main splice acceptor which offered a second chance at creating a functional mRFP reporter ( Clark et al . , 2011a ) . Therefore , to maximize genome coverage of our mutagenesis vectors in this study , we created a series of RP2 constructs to encode functional mRFP in each of the three reading frames ( Figure 1A ) . Second , the 3’ exon trap in the RP2 series uses the nearly ubiquitous β-actin promoter to drive expression of GFP ( detectable around the seven- to eight-somite-stage similar to ubiquitous GFP expression driven under the EF1α enhancer/promoter [Davidson et al . , 2003] ) which could interfere with another GFP-based reporter system in future studies ( Clark et al . , 2011a ) . To overcome this limitation , we engineered a novel , next-generation GBT series called RP8 . RP8 constructs possess a new 3’ exon trap cassette that uses the γ-crystalline promoter to drive expression of lens-specific tagBFP instead of the ubiquitous expression GFP with RP2 series vectors . ( Figure 1B–C ) . Additionally , all RP8 series for three reading frames reporting mRFP constructs are built on a smaller vector backbone and include new restriction enzyme sites that render these vectors modular for subsequent genetic engineering . Using all five of these new GBT constructs in zebrafish , we conducted an initial screen for expression of protein trap mRFP and observed that all RP2 and RP8 series constructs readily produced mRFP fusion proteins expressed from their endogenous promoters ( Figure 1—figure supplement 1 ) . We then deployed all of these GBT vectors to generate over 800 additional zebrafish GBT lines . A key feature of the protein trap in these lines is the ability to non-invasively image the spatial and temporal expression patterns of the trapped loci . Our initial mRFP+ lines demonstrated that , using standard methods , this imaging was going to be a major bottleneck ( Figure 2A ) . Consequently , we utilized both SCORE imaging—a capillary tube placed in a refractive index-matched medium for efficient sample rotation—on an ApoTome ( Petzold et al . , 2010 , and see Materials and methods ) or a Zeiss Lightsheet Z . 1 SPIM microscope ( see Materials and methods ) to enable high throughput fluorescence imaging . To date , we have now cataloged over 1 , 200 GBT lines with robust mRFP expression in heterozygous F2 animals at two dpf and/or four dpf according to our screening pipeline and made all imaging data freely accessible on zfishbook ( www . zfishbook . org ) ( Clark et al . , 2012; Figure 2A ) . We have cryopreserved these 1 , 200 GBT lines and have retained them at the Mayo Clinic Zebrafish Facility ( MCZF ) with a copy also sent to the Zebrafish International Resource Center ( ZIRC ) ( Figure 2A ) . Traditional molecular methods , such as inverse PCR ( iPCR ) and thermal asymmetric interlaced ( TAIL ) PCR and 5’ and 3’ rapid amplification of cDNA ends ( RACE ) , are labor-intensive and represent a functional bottleneck in identifying randomly integrated loci in GBT lines . To overcome this , we employed a rapid cloning process based on methods used to isolate retroviral integrations that leverage the massive parallel sequencing technology of the Illumina MiSeq and a custom bioinformatics pipeline that involves both mapping and annotation ( Figure 2B; Varshney et al . , 2013a; Varshney et al . , 2013b ) . First , fin-clips from four male animals per GBT line were obtained during sperm cryopreservation and used as a source of DNA for cloning the integrated locus . Next , high-throughput sequencing amplified reads with barcodes linked to the source of DNA from the sperm-cryopreserved males . Mapping reads to the genome indicated potential GBT integration loci in each individual . A shared integration locus in multiple individuals from a single GBT line was considered a candidate integrated locus , and we termed a GBT line with at least one candidate integrated locus a ‘GBT-candidate line’ . After a candidate was determined using the sequencing pipeline or other manual molecular approaches , such as 5’ RACE , 3’ RACE , iPCR , or TAIL PCR , we used standard PCR to test if the candidate integration locus segregates with mRFP expression . At the end of this pipeline , 204 GBT-candidate lines met the highest stringency of confirmed expression linkage and were classified as ‘GBT-confirmed lines’ ( Figure 2A ) . While 57 of these GBT-confirmed lines have been previously published ( Clark et al . , 2011a; Ding et al . , 2013; Ding et al . , 2017; El-Rass et al . , 2017; Ma et al . , 2020; Westcot et al . , 2015 ) , 147 of these GBT-confirmed lines are newly characterized in this manuscript and were selected for confirmation based upon their expression pattern and/or homozygous phenotype ( Supplementary file 1 ) . A small subset of these GBT-confirmed lines mapped to areas in the genome without annotated transcripts . Publicly available RNA-sequencing data ( White et al . , 2017 ) revealed reads flanking a majority of these mapped integrations . Some of these reads contained evidence of splicing in the sense orientation of the mRFP reporter ( Supplementary file 1 ) . Finally , another 144 GBT-candidate lines from this pipeline have yet to be confirmed ( Figure 2A ) . Integration locus annotation of both GBT-candidate and GBT-confirmed lines is available on zfishbook ( www . zfishbook . org ) ( Clark et al . , 2012 ) . We wanted to determine the knockdown efficacy of the GBT system as a quantitative assessment of mutagenicity . We therefore compiled qRT-PCR data to compare wild-type and truncated , mRFP-fused transcript levels for all 26 RP2 . 1-derived GBT-confirmed lines that we and others have tested ( Clark et al . , 2011a; Ding et al . , 2013; Ding et al . , 2017 , GBT0235—this manuscript ) . This compilation determined at minimum 97% knockdown in animals homozygous for the RP2 . 1 alleles . We next directly compared the transcriptional effects of RP2 . 1 with those of other published transposon-based protein trap systems ( Figure 3 ) . The FlipTrap system produced a range of 70–96% knockdown in six tested fish alleles ( Trinh et al . , 2011 ) , similar to our initial R-series protein trap vectors ( R14-R15 ) that contained a single splice acceptor and a simple transcriptional terminator ( Liao et al . , 2012; Petzold et al . , 2009 ) . The pFT1 , which contains a single splice acceptor , but a tandem array of five simple polyadenylation sites , appears to be an improvement over these systems with 89–94% knockdown from four tested fish alleles ( Ni et al . , 2012 ) . The 97% minimum knockdown observed with RP2 . 1 is quantitatively higher than these other systems and could be deployed with other insertional genome engineering approaches . We expect the RP8 series vectors to have similar transcriptional knockdown to RP2 . 1 , but to date we have not quantified the effect to measure their mutagenicity . In parallel to knockdown efficacy , we wanted to know the GBT construct effectiveness at mutagenizing its integrated locus . To assess this mutagenic efficiency , we conducted an initial phenotypic screen on F2 embryos and early larvae . In 179 mRFP+ GBT lines , we identified 12 recessive phenotypes visible during the first five days of development that , among others , included lethal , cardiac , muscular , and integumentary defects as reported in previous studies ( Supplementary file 2 ) . Molecular analyses revealed that a subset of these phenotypes stem from GBT integrations in genes with established loss of function mutant phenotypes including ryr1b , tnnt2a , fras1 , hmcn1 , and edar ( Clark et al . , 2011a; Westcot et al . , 2015; Hatzold J et al . , unpublished ) . In accord with their mRNA knockdown potency , these five GBT-confirmed lines ( ryr1b , tnnt2a , fras1 , hmcn1 , and edar ) phenocopied their respective homozygous mutants generated with other strategies and thereby validated the mutagenicity of the GBT constructs . To date , 17 of our GBT-confirmed lines have been published with homozygous phenotypes ranging from embryonic lethal , to reduced adult viability , to differences in pharmacological susceptibility ( Supplementary file 2 ) . Additional GBT-confirmed lines with homozygous phenotypes will continue to be identified and characterized in future studies . We next sought to validate the ability of GBT constructs to functionally annotate genes . During the initial GBT library creation , we identified a GBT line with skeletal muscle-specific mRFP expression and a slow swimming phenotype ( Clark et al . , 2011a ) . We confirmed that this line possesses an RP2 . 1 integration between exon 81 and exon 82 of ryr1b and designated it ryr1bmn0348Gt . Animals homozygous for this ryr1bmn0348Gt allele show 97% knockdown of wild-type ryr1b mRNA levels ( Clark et al . , 2011a ) . Due to the well-characterized nature of ryr1b from the relatively relaxed mutant ( Hirata et al . , 2007 ) , this line was ideal for a proof of concept experiment in this study to validate functional genome annotation with GBT constructs . ryr1b orthologs are known across species to encode calcium-activated calcium channels that release sarcoplasmic reticulum Ca2+ stores to facilitate excitation-contraction coupling in skeletal muscles ( Hernández-Ochoa et al . , 2015; Hirata et al . , 2007 ) . Therefore , we set out to test whether loss of ryr1b in ryr1bmn0348Gt/mn0348Gt animals dampens skeletal muscle Ca2+ transients and may explain their previously reported slow swimming phenotype ( Clark et al . , 2011a ) . To address this , we injected the skeletal muscle-targeted construct p-mylpfa:GCaMP3 ( Baxendale et al . , 2012 ) into both ryr1b+/+ and ryr1bmn0348Gt/mn0348Gt animals , treated these animals at 2 dpf with 20 mM pentylenetetrazole ( PTZ ) to maximize the probability of recording muscle activity , and assayed individual skeletal muscle Ca2+ transients associated with PTZ-induced convulsions ( Figure 4A ) . p-mylpfa:GCaMP3 injection at the single-cell stage resulted in mosaic-labeled , GCaMP3+ myocytes in both ryr1b+/+ and ryr1bmn0348Gt/mn0348Gt animals ( Figure 4B , F ) . Likewise , PTZ-treated ryr1b+/+ and ryr1bmn0348Gt/mn0348Gt animals showed spontaneous , convulsion-associated Ca2+ transients in their myocytes at two dpf ( Figures 4C–E , G–I ) . However , PTZ-induced Ca2+ transients in myocytes of ryr1b+/+ animals had higher peak amplitude when averaged within fish ( Figure 4J–K ) or within myocytes ( Figure 4—figure supplement 1A ) and shorter rise time ( Figure 4J , M ) than those in ryr1bmn0348Gt/mn0348Gt animals . PTZ-induced Ca2+ transient peak-width ( Figure 4J , L ) and decay time ( Figure 4J , N ) were not significantly different between ryr1b+/+ and ryr1bmn0348Gt/mn0348Gt animals . In contrast , myocytes in ryr1b+/+ animals had more Ca2+ transients during the imaging period than myocytes in ryr1bRP2 . 1/RP2 . 1 animals ( Figure 4—figure supplement 1B ) . We were thus able to use a GBT-confirmed line to demonstrate that a smaller peak amplitude ( consistent with relatively relaxed [Hirata et al . , 2007] ) , slower upstroke , and lower frequency of Ca2+ transients in skeletal muscle likely provide the basis for the slow swimming phenotype in ryr1bmn0348Gt/mn0348Gt animals . The consistency of our findings in ryr1bmn0348Gt/mn0348Gt with those in relatively relaxed mutants ( Hirata et al . , 2007 ) validates the functional genome annotation available with the GBT mutagenic system . This functional genome annotation available with the GBT system in zebrafish is powerful for understanding the genetic causes of human disease . For instance , mutations in RYR1 , the human ortholog of ryr1b , are well-associated with a rare genetic neuromuscular disorder called central core disease . Central core disease commonly presents with mild to severe muscle weakness ( Jungbluth et al . , 2018 ) which is analogous to the slow swimming phenotype we saw in our ryr1bmn0348Gt/mn0348Gt animals ( Clark et al . , 2011a ) and likely arises from similar disruptions to skeletal muscle Ca2+ transients ( Figure 4 ) . A subset of this GBT collection consequently represents a potential library of human disease models . Intriguingly , 82% of OMIM listed human disease-associated genes ( 2601 genes ) can be related to at least one zebrafish ortholog ( Howe et al . , 2013 ) . We therefore took a new angle and investigated whether any GBT-confirmed lines represent potential human disease models . Within the set of GBT-confirmed lines that match a human ortholog ( n = 177 ) , 64 ( 36% ) are integrated in genes associated with human diseases , including those of the nervous , circulatory , endocrine , metabolic , digestive , musculoskeletal , immune , and integumentary systems ( select genes listed in Figure 5A , all 64 genes with disease-associated human orthologs listed in Supplementary file 3 ) . 40 of these human disease-associated GBT-confirmed lines represent potential novel genetic disease models as we failed to find a description for any established disease models in mice or zebrafish for orthologs of these genes ( Figure 5B and Supplementary file 3 ) . Functional genome annotation with GBT constructs is equally powerful in detecting roles for genes in basic cellular processes . We and others have previously used imaging to investigate effective protein trapping in GBT-confirmed lines . This imaging has revealed diverse cellular and subcellular protein expression patterns ( Clark et al . , 2011a; Ding et al . , 2013; Ding et al . , 2017; El-Rass et al . , 2017; Liao et al . , 2012; Petzold et al . , 2009; Westcot et al . , 2015; Xu et al . , 2012 ) . In this study , we imaged GBT lines using a Lightsheet microscope for the first time . Multi-area tiling with a 20 × objective enabled rapid acquisition of 3-dimensional mRFP fusion protein localization across the entire organism ( Figure 6A ) . Confocal imaging in areas of interest revealed even more detail with subcellular resolution ( Figure 6A–B ) . Further confocal imaging demonstrated diverse subcellular localizations in GBT-confirmed lines , ( Figure 6B–C ) GBT-candidate lines , and GBT lines ( Figure 6—figure supplement 1 ) . This subcellular protein localization data from GBT lines can provide crucial information in piecing together gene function . We therefore wanted to assess the subcellular diversity of all gene products trapped in our current collection of GBT-confirmed lines . As an approach to complement our imaging assessments , we focused on computational approaches to explore subcellular protein diversity in current GBT-confirmed lines . 177 of the GBT-trapped genes were annotated to their human orthologs in at least one public database ( ZFIN , Ensembl , Homologene , and InParanoid version 8 ) ( Supplementary file 1 ) . Several of these genes were provisionally annotated using BLASTP or a synteny analysis tool , SynFind in Comparative Genomics ( CoGe ) database ( https://genomevolution . org/CoGe/SynFind . pl ) ( Lyons and Freeling , 2008 ) . We assessed the functional diversityhuman orthologs of human orthologs of these 177 GBT-confirmed loci with data from the PANTHER version 14 . 1 database on protein class ontology ( http://www . pantherdb . org/ ) ( Mi et al . , 2019 ) , the Human Protein Atlas on genome-wide experimental proteomics ( www . proteinatlas . org . August 27 , 2019 ) ( Uhlén et al . , 2015 ) , and the UniProtKB on knowledge-based proteomics ( UniProtKB , https://www . uniprot . org/ , UniProt Consortium , 2018 ) . PANTHER protein classification revealed that 105 human orthologs ( 60% , n = 176: see Materials and methods ) are classified to at least one of 21 protein classes , and 19 human orthologs ( 11% ) belong to transcription factors ( Figure 6D and Figure 6—source data 1 ) . Human Protein Atlas and UniProtKB subcellular localization data likewise showed diverse classifications of expression with a large group of nuclear localized human orthologs of these 177 GBT-confirmed loci ( Figure 6E and Supplementary file 4 ) . We then asked whether this if our computational analysis corresponded to the patterns seen in our imaging data . We found that LRPPRC ( human ortholog of lrpprc ( GBT0235 ) —Figure 6A–B ) was not annotated to a protein class in PANTHER but mapped to mitochondria in Human Protein Atlas , consistent with its puncta expression pattern ( Figure 6A–B ) . RYR1 ( human ortholog of ryr1b ( GBT0348 ) —Figure 6C ) was annotated as a transporter in PANTHER and was mapped to the cytosol , Golgi apparatus , and vesicles in Human Protein Atlas , consistent with its more uniform expression pattern ( Figure 6C ) . Overall , protein class ontology and known subcellular localizations of cloned GBT genes suggest that the GBT system traps and enables functional annotation for a rich diversity of proteins . Additionally , the GBT-confirmed lines in orthologs of human genes without a known subcellular localization potentiate the discovery of their subcellular expression pattern in the context of a living animal . We next asked if the mRFP expression patterns in our GBT-confirmed lines unveiled novel cellular expression data . To address this , we focused on the GBT-confirmed lines that were non-redundant and mapped to a known protein coding gene . Importantly , these GBT-confirmed lines exhibited expression patterns that are tissue specific and include assorted brain regions , heart , skin , muscle , vasculature , and blood ( Figure 7A–R ) . We analyzed publicly available expression data of these 193 tagged genes in wild-type fish ( downloaded from ZFIN on August 28th , 2019 ) . Our GBT-confirmed lines revealed expression patterns ( available on www . zfishbook . org ) for 67 genes at 2 dpf and 174 genes at four dpf without publicly available expression data in ZFIN ( Figure 7S–T ) .
GBT technology represents the first method for revertible allele generation in vertebrates outside of the mouse model ( Clark et al . , 2011a ) . In this manuscript we broadened GBT genomic coverage through the development of an RP2 construct series to encode functional mRFP in each of the potential reading frames ( Figure 1A ) . While each individual construct still only integrates in-frame in a subset of introns , the RP2 series potentiates in-frame mRFP for any intron with Tol2-mediated integration . We also desired to increase GBT utility for subsequent genomic engineering applications . As RP2 series constructs were not modularly designed , we iteratively developed a next generation RP8 GBT series . All RP8 constructs include new restriction enzyme sites that render them modular for custom engineering . Additionally , RP8 series constructs use a smaller backbone designed to enhance transgenic efficiency . RP8 vectors most notably possess a new 3’ exon trap cassette that uses the γ-crystalline promoter to drive expression of lens-specific tagBFP instead of the ubiquitous expression of GFP delivered from RP2 vectors ( Figure 1B ) . We found this lens-specific tagBFP to be equally useful in screening founders with GBT integrations . While we did not explicitly validate that the RP8 series provides transcriptional effects equivalent to the RP2 series , the major functional change in RP8 lies in its 3’ exon trap . We therefore expect the knockdown to be similar between RP8 and RP2 . Two additional zebrafish transposon-based protein trap vectors have been established . Dr . Fraser’s group developed the FlipTrap system ( Trinh et al . , 2011 ) that is mutagenic in the presence of Cre-recombinase . However , this FlipTrap system is primarily focused on imaging fusion proteins in vivo and addressing cellular dynamics . The Chen lab developed a complementary flipping system called the FT1 system that uses either Cre or Flp recombinase to regulate its alleles depending on the original insertion orientation ( Ni et al . , 2012 ) . Our GBT system is highly complementary and non-redundant with these alternative transposon-based protein trap methods . The 5’ protein trap in the GBT system is terminated through an enhanced polyadenylation signal in conjunction with a putative border element and a second splice acceptor in the 3’ exon trap helps eliminate any pass-through . Together these elements achieve higher gene-specific mRNA knockdown than the single splice acceptor and the basic polyadenylation signal in FlipTrap and FT1 ( Figure 3 ) . In addition to a 5’ protein trap , our GBT system also possesses a 3’ exon trap that serves as a means for screening integrations , reports 3’ sequence , and possesses the ability to trap ( without mRFP expression ) non-coding RNAs that undergo splicing ( Figure 1 ) . Consistent with its mRNA knockdown abilities , the mutagenic efficacy of the GBT system is functionally similar to other genome-wide forward genetic approaches at identifying critical early developmental loci . The GBT system achieved 7% recovery of visible early ( through five dpf ) developmental phenotypes during an initial forward genetic screen . This recovery is comparable to the 5% recovered visible phenotypes from the Sanger TILLING consortium analysis of truncated zebrafish genes ( Kettleborough et al . , 2013 ) . Our 7% phenotype recovery is also comparable to prior retroviral ( Amsterdam and Hopkins , 2004 ) and ENU ( Haffter et al . , 1996 ) zebrafish mutagenesis works that estimated between 1400 and 2400 genes ( ~5–9% of the genome ) would result in a visible embryonic phenotype when mutated . The connections between gene , expression pattern , function , and phenotype ( or human disease ) can be elucidated using our GBT system . During the initial GBT library creation , we identified a GBT-confirmed line with an RP2 integration between exon 81 and exon 82 of ryr1b ( ENSDART00000036015 . 9 , Ensembl Release 100 on April 2020 ) , a zebrafish ortholog of human RYR1 . Mutations in RYR1 are well-linked to a rare genetic neuromuscular disorder known as central core disease that presents with mild to severe muscle weakness ( Jungbluth et al . , 2018 ) . Indeed , homozygous animals in this ryr1b GBT-confirmed line possess skeletal muscle-specific mRFP expression and a slow swimming phenotype ( Clark et al . , 2011a ) . Previously , a spontaneous mutant called relatively relaxed ( ryr1bmi340/mi340 ) was shown to have a slow swimming phenotype , truncated ryr1b protein with a pre-mature stop involved in an insertional mutagen , and defective skeletal muscle Ca2+ transients ( Hirata et al . , 2007 ) . Our ryr1b GBT-confirmed line was therefore ideal to validate the functional genome annotation abilities of GBT constructs . Similar to the relatively relaxed mutants which carry an insertion that introduces a premature stop codon between exons 48 and 49 of ryr1b ( Hirata et al . , 2007 ) , we noted severely dampened skeletal muscle Ca2+ transients in animals homozygous for the ryr1b GBT allele ( Figure 4 ) . In addition to previously reported decreases in peak amplitude ( Hirata et al . , 2007 ) we found that ryr1bmn0348Gt/mn0348Gt animals also displayed a slower upstroke and lower frequency of skeletal muscle Ca2+ transients than wildtypes , functionally annotating these roles for the C-terminal region of ryr1b gene in vivo . Including this ryr1b line , we generated GBT-confirmed lines with integrations in 64 zebrafish orthologs of human disease-associated genes ( Figure 5 , Supplementary file 1 , Supplementary file 3 ) in this study . GBT-confirmed lines with integrations in 40 zebrafish orthologs of human disease-associated genes may represent novel potential disease models , as we failed to find any description of existing zebrafish or mouse models for these genes or diseases . Our GBT system importantly possesses a built-in cure due to its revertible nature . Therefore , these GBT potential disease models will allow direct comparison of tissue-specific gene restoration with any therapeutic approach . To achieve genomic representation , unbiased protein trapping is an important consideration . Tol2 transposase-mediated systems are known to facilitate near-random integration , but we wanted to explore this in the context of the GBT protein trapping constructs . We utilized PANTHER , Human Protein Atlas , and UniProtKB to explore the protein class and subcellular localization of the human orthologs of the genes trapped in GBT-confirmed lines ( Figure 6 ) . While nuclear localized proteins , such as transcription factors , represented the largest class of GBT-trapped genes , we identified a diverse set of proteins in our GBT-confirmed lines that localize all the way from the nucleoli to the extracellular space . The reason for the enrichment of nuclear genes is unknown . The rich diversity of proteins observed in our GBT-confirmed lines still supports that the entire collection has high diversity and is consistent with the random nature of Tol2-mediated genome integration events ( Clark et al . , 2011a ) . Completion of the zebrafish reference genomes also has enabled many new discoveries to be made with regards to the position of hundreds of genes that affect embryogenesis , behavior , and physiology . However , poorly assembled regions remain in both the zebrafish and the human genome ( Howe et al . , 2013 ) . We indeed found that 10 GBT integrations in the confirmed lines ( with mRFP expression ) failed to map to any predicted genes . However , RNA sequencing reads in public datasets identified potential unannotated coding sequences aligned with these GBT integration loci ( Supplementary file 1 ) . While 5’ and 3’ RACE are necessary to confirm the mRNA fusion products , these unannotated coding sequences represent the possibility to annotate novel , protein-coding transcripts in these GBT lines . Therefore , GBT protein trapping can find , illuminate expression , and elucidate in vivo functions of novel genes and/or gene variants in poorly annotated regions of reference genomes . GBT-based mRFP fusion proteins represent a notable advance over traditional techniques for probing endogenous gene expression ( e . g . , immunohistochemistry , in situ hybridization ) that have yielded very little gene expression data at later developmental stages . The truncated mRFP fusion proteins in both RP2 and RP8 series constructs exhibited distinct cellular localizations in our GBT lines throughout development , including two dpf and beyond . Approximately 90% of GBT-confirmed lines showcased novel expression patterns of their annotated genes at four dpf ( Figure 7 ) . These GBT-based mRFP fusion proteins allow investigation of subcellular localization of these tagged-gene products ( Figure 6 , Figure 6—figure supplement 1 ) , with the exception of cases where the protein localization signal is contained in the C-terminal domain ( Clark et al . , 2011a; Trinh and Fraser , 2013 ) . We indeed observed mRFP accumulation in the kidney tubules , white blood cells or developing bones in some GBT lines , likely based upon the remaining signal sequences at the N-terminus of the endogenous protein . Still , visualizing protein expression dynamics of GBT-trapped proteins in most lines should facilitate important initial annotations regarding subcellular localization of uncharacterized proteins to investigate molecular functions in vivo . With ever-improving fluorescence-based imaging tools ( Liu et al . , 2018 ) , our GBT lines have the potential to annotate both cellular and subcellular gene expression at diverse stages on an organismal scale . Taken together , GBT-based mRFP-reporters demonstrate how much we still have left to understand about the expression patterns of the overall proteome and , ultimately , the complex codex that is our genome . Even at the relatively well-studied 2-dpf stage , nearly 40% of GBT-confirmed lines elucidated novel gene expression data ( Figure 7 ) . Cataloging these expression patterns enables investigators to make collections of lines with expression in their cell/tissue of interest and/or a phenotype . The remaining 144 GBT-candidate lines and over 800 GBT lines represent a rich resource for genomic discoveries . For any GBT-candidate or GBT lines of interest , a similar cloning pipeline ( Figure 2 ) can be employed to identify the GBT integration locus . In addition , the refinement of the zebrafish genome will enhance our ability to complete the annotation from GBT-line to GBT-confirmed line for any given line with a desired expression profile and/or phenotype . Together , this 1 , 200+ GBT-line collection is a new contribution for using zebrafish to annotate the vertebrate genome . Although our GBT lines were made with random integration , new targeted integration tools , such as GeneWeld ( Wierson et al . , 2020 ) , that employ gene editing techniques will empower labs to build custom GBT lines for their gene of interest ( El Khoury et al . , 2018 ) . The three reading frames and modularity of the RP8 series are especially well suited to targeted integration approaches . Further , a combination of targeted and random integration may best facilitate discovery . For instance , a targeted approach could integrate a GBT cassette into a well-characterized , process-associated gene . Then , random integration could be used to probe for genes that potentiate or abrogate the disruption in the original process-associated gene . With this approach , GBTs are a powerful tool to investigate multigenic processes , including human disease . Finally , targeted mutagenic technology , such as CRISPR and TALEN systems , has become the gold standard reverse genetic approach . However , engineered mutant animals using approaches generate targeted indel mutation frequently fail to display overt phenotypes , often explained by genetic compensation ( Balciunas , 2018; El-Brolosy and Stainier , 2017 ) . One mechanism includes cellular increases in transcripts of genes in the same family that can functionally substitute when activated in a mutant background ( Balciunas , 2018 ) . Recently , a number of studies using reverse genetics tools have revealed phenotype differences between knockouts ( indel mutants ) , and knockdowns ( antisense-treated animals ) in multiple model systems including Arabidopsis , Drosophila , zebrafish , mouse , and human cell lines . This discrepancy is attributed to transcriptomic changes in mutant but not in knockdown animals ( Reviewed from El-Brolosy and Stainier , 2017 ) . For example , knockdown of egfl7 , an endothelial extracellular matrix ( ECM ) gene , induces severe vascular defects , whereas most egfl7 mutants exhibit no obvious defect , resulting from upregulation of other ECM proteins , especially emilins in egfl7 mutants , but not in egfl7 morphants ( Ronzitti , 2019 ) . As another mechanism of genetic compensation , mRNA processing—including nonsense-associated exon skipping and the use of alternative start or splice sites to escape nonsense-mediated decay—has been recently demonstrated to hinder loss-of-function approaches in zebrafish ( Anderson et al . , 2017; Prykhozhij et al . , 2017 ) , in human cell lines ( Lalonde et al . , 2017; Winter et al . , 2019 ) and in the human population ( Jagannathan and Bradley , 2016 ) . In contrast , the molecular mechanism of GBT mutagenesis can normally avoid this genetic compensation effect seen with small indel mutations because the strong poly ( A ) -trapping element in the 5’ exon trap domain of RP cassettes can reduce mRNA to below 1% of the complete wild-type transcript level . This reduction eliminates many sources of transcriptional adaptations triggered by a loss of function mutation , such as alternative transcriptional start sites , splicing , or alternative translation initiation . The GBT can therefore act as a useful validation tool when targeted mutations with other technologies fail to display any phenotype .
All zebrafish ( Danio rerio ) were maintained according to the procedures described previously ( Leveque et al . , 2016 ) . pGBT-RP8 . 2 and -RP8 . 3 were made by combining three restriction endonuclease fragments of pGBT-RP8 . 1 , a 2 . 2 kb AflII to AgeI , a 0 . 7 kb EcoRI to SpeI , and a 3 . 0 kb SpeI to AflII , with a short adapter to close the space between AgeI and EcoRI that effectively removed one or two thymine nucleotides just following the splice acceptor prior to the AUG-less mRFP cassette . For pGBT-RP8 . 2 , Adapter-GBT ( +2 ) was made by annealing oligos adapter-GBT ( +2 ) -a [CCGGTTTTCTCATTCATTTACAGTCAGCCGG] and adapter-GBT ( +2 ) -b [AATTCCGGCTGACTGTAAATGAATGAGAAAA] . For pGBT-RP8 . 3 , Adapter-GBT ( +3 ) was made by annealing oligos adapter-GBT ( +3 ) -a [CCGGTTTTCTCATTCATTTACAGCAGCCGG] and adapter-GBT ( +3 ) -b [AATTCCGGCTGCTGTAAATGAATGAGAAAA] . pGBT-RP2 . 2 and -RP2 . 3 were made by combining three restriction endonuclease fragments of pGBT-RP2 . 1 ( Clark et al . , 2011a ) , a 3 . 6 kb BlpI to AgeI , a 1 . 9 kb EcoRI to AvrII , and a 3 . 55 kb AvrII to BlpI , with a short adapter to close the space between AgeI and EcoRI that effectively removed one or two thymine nucleotides just following the splice acceptor prior to the AUG-less mRFP cassette . For pGBT-RP2 . 2 , Adapter-GBT ( +2 ) was made by annealing oligos adapter-GBT ( +2 ) -a and adapter-GBT ( +2 ) -b . For pGBT-RP2 . 3 , Adapter-GBT ( +3 ) was made by annealing oligos adapter-GBT ( +3 ) -a and adapter-GBT ( +3 ) -b . pGBT-RP8 . 1 was made by cloning a mini-intron derived from carp beta actin intron one into pGBT-RP7 . 1 . The 234 bp SalI to XhoI mini-intron fragment was isolated from pCR4-bactmIntron following digestion . The pGBT-RP7 . 1 plasmid was digested with XhoI so that the SalI to XhoI fragment was cloned between the gamma-crystallin promoter and nls tagBFP . pCR4-bactmIntron was made by removing a 1 . 1 kb internal portion of the carp beta actin intron one by digestion of pCR4-bact_I1 with BstBI and BssHII , followed by filling in 5’ overhangs and ligating remaining vector fragment . pCR4-bact_I1 was cloning a PCR product containing the carp beta-actin intron into pCR4-TOPO ( 450030 , Invitrogen , Thermo Fisher Scientific , Waltham , MA ) . The intron was amplified from pGBT-RP2 . 1 ( Clark et al . , 2011a ) using MISC-bact_exon-F1 [CAGCTAGTGCGGAATATCATCTGCC] and MISC-bact_intron-R1 [CTTCTCGAGGTGAATTCCGGCTGAACTGTA] primers . pGBT-RP7 . 1 was made by replacing a 501 bp PstI to PstI fragment of pGBT-RP6 . 1 with a 480 bp PstI to PstI fragment of pRP2 . 1 . This changed the nucleotide sequence between the carp beta-actin splice acceptor to replicate the sequences in pGBT-RP2 . 1 . pGBT-RP7 . 1 was never directly tested in zebrafish . pGBT-RP6 . 1 was made by flipping the internal trap cassette relative the Tol2 inverted terminal repeats in pGBT-RP5 . 1 . To do this , pGBT-RP5 . 1 was cut with EcoRV and SmaI . The 2 . 27 kb EcoRV to SmaI vector backbone fragment , which included the ITRs , was ligated to the 3 . 51 kb EcoRV to SmaI trap fragment . pGBT-RP6 . 1 was then selected based on the right ITR of Tol2 being in front of the RFP trap , which is the same orientation of pGBT-RP2 . 1 . pGBT-RP5 . 1 was made by cloning a PCR product with the AUG-less mRFP into pre ( −1 ) GBT-RP5 . 1 . The 698 bp mRFP* PCR product was obtained by amplification of pGBT-R15 ( Clark et al . , 2011a ) with CDS-mRFP*-F1 [AAGAATTCGAAGGTGCCTCCTCCGAGGATGTCATCAAGG] and CDS-mRFP-R1 [AAACTAGTCTTAGGCTCCGGTGGAGTGGCGG] . Prior to cloning the PCR mRFP* product was digested with EcoRI and SpeI to prepare the ends for subcloning into pre ( −1 ) GBT-RP5 . 1 that was opened between the carp beta actin splice acceptor and the ocean pout terminator . pre ( −1 ) GBT-RP5 . 1 was made by cloning 1 . 2 kb SpeI to AvrII fragment from pGBT-PX ( Sivasubbu et al . , 2006 ) that contained the ocean pout terminator into the SpeI site of pre ( −2 ) GBT-RP5 . 1 . The resulting products were screened for the proper orientation of the ocean pout terminator relative to the carp beta actin splice acceptor . pre ( −2 ) GBT-RP5 . 1 was made by inserting an expression cassette to make a 3’ poly ( A ) trap that makes blue lenses . A 1 . 15 kb SpeI to BglII fragment from pKTol2gC-nlsTagBFP was cloned into pre ( −3 ) GBT-RP5 . 1 that had been cut with AvrII and BglII . This moved the Xenopus gamma crystallin promoter driving a nuclear-localized TagBFP in front of the carp beta actin splice donor within pre ( −3 ) GBT-RP5 . 1 to create a localized BFP poly ( A ) trap signal replacing the ubiquitous GFP signal that was in pGBT-RP2 . 1 . pre ( −3 ) GBT-RP5 . 1 was made by cloning a 492 bp XmaI to NheI scaffold fragment from pUC57-I-SceI_loxP_splice into pKTol2-SE ( Clark et al . , 2011b ) opened with XmaI and NheI . pUC57-I-SceI_LoxP_Splice contains a synthetic sequence ( see below ) cloned into pUC57 ( SD1176 , Genscript , Piscataway , NJ ) . The scaffold contains an I-SceI site; loxP site; carp beta actin splice acceptor; cloning sites for mRFP , ocean pout terminator , and BFP lens cassettes; carp beta actin splice donor; loxP site; and an I-SceI site . The synthetic sequence described above is: cccgggatagggataacagggtaatataacttcgtatagcatacattatacgaagttat cgttaccacccactagcggtcagactgcagattgcagcacgaaacaggaagctgac tccacatggtcacatgctcactgaagtgttgacttccctgacagctgtgcactttctaaa ccggttttctcattcatttacagttcagcctgttacctgcactcaccgacaagctgttacc ctggaattcgtttaaacactagtcaccggcgttcctaggttataagatctacctaaggtg agttgatctttaagctttttacattttcagctcgcatatatcaattcgaacgtttaattagaat gtttaaataaagctagattaaatgattaggctcagttaccggtcttttttttctcatttacact gagctcaagacgtctgataacttcgtatagcatacattatacgaagttattaccctgttatccctatggctagc . Generation of the GBT collection was based on the prior described protocols ( Clark et al . , 2011a; Ni et al . , 2016 ) . RP2 constructs were injected and sorted by low mosaicism of GFP expression . RP8 constructs were sorted on the basis of strong tagBFP expression in the eyes . Overall ~30% of injected fish met these criteria . ~ 25% of these F0 fish gave RFP offspring . Larvae were treated with 0 . 2 mM phenylthiocarbamide ( P7629 , Sigma-Aldrich , St . Louis , MO ) at one dpf to inhibit pigment formation . The anesthetized fish were mounted in 1 . 5% low-melt agarose ( BP1360 , Fisher Scientific , Hampton , NH ) prepared with 0 . 017 mg/ml tricaine ( Ethyl 3-aminobenzoate methanesulfonate salt , A5040 , Sigma-Aldrich ) solution in an agarose column in the imaging chamber . The protocol of ApoTome microscopy was described in previous publication . ( Clark et al . , 2011a ) For Lightsheet microscopy , larval zebrafish were anesthetized with 0 . 017 g/ml tricaine in 1 . 5% low-melt agarose ( BP1360 , Fisher Scientific ) and mounted in glass capillaries . To capture RFP expression patterns of 2 dpf and four dpf larval zebrafish , LP 560 nm filter as excitation and LP 585 nm as emission was used for Lightsheet microscopy . The sagittal- , dorsal- , and ventral- oriented z-stacks of the mRFP expression were captured at either 50x magnification using an ApoTome microscope ( Zeiss , Oberkochen , Germany ) with a 5x/0 . 25 NA dry objective ( Zeiss ) or 50x magnification using a Lightsheet Z . 1 microscope ( Zeiss ) 5x/0 . 16 NA dry objective . Each set of images were obtained from the same larva and the images shown are composites of the maximum image projections of the z-stacks obtained from each direction . For confocal microscopy , larval zebrafish were anesthetized with 0 . 017 g/ml tricaine ( Ethyl 3-aminobenzoate methanesulfonate salt , A5040 , Sigma-Aldrich ) in 1 . 0% low-melt agarose ( BP1360 , Fisher Scientific ) and mounted 35 mm glass-bottom dishes ( P35G-1 . 5–14 C , MatTek Life Sciences , Ashland , MA ) . Imaging was performed on an LSM-780 ( Zeiss , Oberkochen , Germany ) using either a C-Apochromat 63x/1 . 2 NA or a C-Apochromat 40x/1 . 2 NA water immersion objective . RFP was excited at 561 nm and emissions 570–750 nm were collected . Sperm collection and cryopreservation was initially based on the protocol described in Draper and Moens , 2009 and moved to the Zebrafish International Resource Center ( ZIRC ) protocol described in Matthews et al . , 2018 . Genomic DNA was isolated from F2 fish tail biopsies to conduct next generation sequencing and from both wild-type and heterozygous larva to manually perform PCR-based analysis for the identification of GBT-insertion site . 60 µl of lysis buffer containing with 10 mM Tris ( pH 8 . 0 ) , 100 mM NaCl , 10 mM EDTA , 0 . 4% SDS and 5 µg/ml proteinase K ( 03115879001 , Roche ) was loaded into each well of a 96-wells plate and clipped adult fins/larvae were individually placed into each well and incubated at 50°C for 3 hr . The solution with lysed tissue were suspended with a multichannel micropipette to dissolve the tissue , mixed with 60 µl of isopropanol and centrifuged at ~3 , 000 rpm/ 15 , 000 × g for 20 min at 4°C . After removing the supernatant , 100 µl of 70% ethanol were added , centrifuged for 20 min at 4°C . After discarding ethanol , the pellets were dried and re-suspended in 50 µl of water/TE . As the alternative protocol to quickly extract genomic DNA from zebrafish larvae , the specimens were individually placed to 0 . 2 ml PCR tubes and lysed with 30–50 µl of 50 mM NaOH and incubated at 95 Co for 20 min . The solutions with lysed specimens were vortexed and neutralized with 1 of 10 vol 1M Tris-HCl . In addition to the broad next gen sequencing approach used to identify GBT integration sites , we used a combination of several a la carte methods including 5' and 3' rapid amplification of cDNA ends ( RACE ) ( Clark et al . , 2011a ) , inverse PCR ( Clark et al . , 2011a ) , and Thermal Asymmetric Interlaced PCR ( TAIL-PCR ) . The protocol used for TAIL-PCR was designed to amplify and clone junction fragments from Tol2-based gene-break transposons was based on a protocol from Parinov et al . , 2004 with some modifications . The following primer mixtures ( containing 0 . 4 μM GBT specific primer and 2 µM degenerate primers ( DP ) ) were prepared: for primary PCR: 5R-mRFP-P1/DP1 , 5R-mRFP-P1/DP2 , 5R-mRFP-P1/DP3 , 5R-mRFP-P1/DP4 , 3 R-GM2-P1/DP1 , 3 R-GM2-P1/DP2 , 3 R-GM2-P1/DP3 , 3 R-GM2-P1/DP4 , 3R-tagBFP-P1/DP1 , 3R-tagBFP-P1/DP2 , 3R-tagBFP-P1/DP3 , 3R-tagBFP-P1/DP4; for secondary PCR: 5R-mRFP-P2/DP1 , 5R-mRFP-P2/DP2 , 5R-mRFP-P2/DP3 , 5R-mRFP-P2/DP4 , 3 R-GM2-P2/DP1 , 3 R-GM2-P2/DP2 , 3 R-GM2-P2/DP3 , 3 R-GM2-P2/DP4 , 3R-tagBFP-P2/DP1 , 3R-tagBFP-P2/DP2 , 3R-tagBFP-P2/DP3 , 3R-tagBFP-P2/DP4; for tertiary PCR: TAIL-bA-SA/DP1 , TAIL-bA-SA/DP2 , TAIL-bA-SA/DP3 , TAIL-bA-SA/DP4 , Tol2-ITR ( L ) -O1/DP1 , Tol2-ITR ( L ) -O1/DP2 , Tol2-ITR ( L ) -O1/DP3 , Tol2-ITR ( L ) -O1/DP4 , Tol2-ITR ( L ) -O3/DP1 , Tol2-ITR ( L ) -O3/DP2 , Tol2-ITR ( L ) -O3/DP3 , Tol2-ITR ( L ) -O3/DP4 . A total of 1 μl of primer mixtures were added to PCR reaction ( total volume 25 µl ) . Cycle settings were as follows . Primary: ( 1 ) 95°C , 3 min; ( 2 ) 95°C , 20 s; ( 3 ) 61°C , 30 s; ( 4 ) 70°C , 3 min; ( 5 ) go to ‘cycle 2’ five times; ( 6 ) 95°C , 20 s; ( 7 ) 25°C , 3 min; ( 8 ) ramping 0 . 3°/sec to 70°C; ( 9 ) 70°C , 3 min; ( 10 ) 95°C , 20 s; ( 11 ) 61°C , 30 s; ( 12 ) 70°C , 3 min; ( 13 ) 95°C , 20 s; ( 14 ) 61°C , 30 s; ( 15 ) 70°C , 3 min; ( 16 ) 95°C , 20 s; ( 17 ) 44°C , 1 min; ( 18 ) 70°C , 3 min; ( 19 ) go to ‘cycle 10’ 15 times; ( 20 ) 70°C , 5 min; Soak at 12°C . A total of 5 μl of the primary reaction was diluted with 95 µl of 10 mM Tris-Cl or TE buffers and 1 μl of the mixture was added to the secondary reaction . Secondary: ( 1 ) 95°C , 2 min ( 2 ) 95°C , 20 s; ( 3 ) 61°C , 30 s; ( 4 ) 70°C , 3 min; ( 5 ) 95°C , 20 s; ( 6 ) 61°C , 30 s; ( 7 ) 70°C , 3 min; ( 8 ) 95°C , 20 s; ( 9 ) 44°C , 1 min; ( 10 ) ramping 1 . 5°/sec to 70°C; ( 11 ) 70°C , 3 min; ( 12 ) go to ‘cycle 2’ 15 times; ( 13 ) 70°C , 5 min; Soak at 12°C . A total of 5 μl of the primary reaction was diluted with 95 µl of 10 mM Tris-Cl or TE buffers and 1 μl of the mixture was added to the tertiary reaction . Tertiary: ( 1 ) 95°C , 2 min; ( 2 ) 95°C , 20 s; ( 3 ) 44°C , 1 min; ( 3 ) ramping 1 . 5°/sec to 70°C; ( 4 ) 70°C , 3 min; ( 5 ) go to ‘cycle 2’ 32 times; ( 6 ) 70°C , 5 min; Soak at 12°C . Products of the secondary and tertiary reactions were separated by using 1–1 . 5% agarose gel . The individual bands from the ‘band shift’ pairs were cut from the gel and purified by using QIAquick Gel Extraction Kit ( 28704 , QIAGEN , Hilden , Germany ) , and sequenced by the sequencing service in the Medical Genome Facility at Mayo Clinic . Alternatively , 5' RACE , 3' RACE , or inverse PCR were used to identify the interrupted gene as previously described ( Clark et al . , 2011a ) . Public datasets of zebrafish wildtype at 36 hpf , 48 hpf and four dpf were downloaded from Ensembl database ( Downloaded datasets: GRCz10 . WTSI . 36hpf . 1 . bam , GRCz10 . WTSI . 48hpf . 1 . bam and GRCz10 . WTSI . 4dpf . 1 . bam , URL: ftp://ftp . ensembl . org/pub/data_files/danio_rerio/GRCz10/rnaseq/ ) ( White et al . , 2017 ) to browse mapping RNA sequence ( RNA-seq ) reads around the integration loci . With Galaxy ( https://usegalaxy . org/ ) as a web-based platform for next generation sequencing data analysis ( Afgan et al . , 2018 ) , these downloaded BAM files of these datasets were converted FASTQ file format using BAMtools ( Barnett et al . , 2011 ) . TopHat created a new BAM file and re-aligned RNA-seq reads in the FASTQ file to identify splice junctions between exons in each dataset ( Kim et al . , 2013 ) . These re-mapped BAM files were used to predict candidate transcripts integrated with RP2/RP8 constructs with Integrative Genomics Viewer ( version 2 . 4 . 19 ) ( Thorvaldsdóttir et al . , 2013 ) . Our calcium imaging protocols were modified from Baxendale et al . , 2012 . Calcium data are compiled from two similar methods with independent experimenters and equipment . L . E . G . performed two independent runs of ‘Method 1’ , and A . J . T . performed two independent runs of ‘Method 2’ . Zebrafish embryos at the single cell ( single-cell to 4 cell in Method 1 ) stage were injected with 80–100 pg ( Method 1 ) or 2 nL of pEXPR-mylpfa:GCaMP3 plasmid ( a gift from Dr . Cunliffe ) diluted in water ( Method 1 ) or to 50 ng/µL in 200 mM KCl/0 . 05% phenol red ( Method 2 ) . On day 2 , GCaMP3+ embryos were de-chorionated and allowed to rest for 30 min . For Method 1 , embryos were singly incubated in E2 medium containing 20 mM pentylenetetrazole ( PTZ ) ( P6500 , Sigma-Aldrich ) for approximately 5 min and mounted with a 3% methylcellulose solution containing 20 mM PTZ in a PPT tube and viewed laterally ( similar to SCORE imaging [Petzold et al . , 2010] ) . After mounting , Ca2+ transients in muscles were assessed using an Axio Scope . A1 ( Zeiss ) equipped with a D3 DSLR camera ( Nikon , Minato , Tokyo , Japan ) and a HXP 120 fluorescent lamp ( Zeiss ) using a 10x , 0 . 45 NA objective . Images were acquired at a rate of 2 Hz over 30 s . For Method 2 , four embryos at a time were transferred into E2 medium containing 5 µM ( S ) - ( - ) -blebbistatin ( 1852 , Tocris , Bristol , United Kingdom ) and incubated for at least 30 min until paralyzed . Once paralyzed , we embedded pairs of embryos into glass bottom dishes ( MatTeK Corporation , Ashland , MA ) with 1% low melting agarose ( BP1360 , Fisher Scientific ) in E2 embryo medium with 5 µM ( S ) - ( - ) -blebbistatin and 20 mM PTZ . After an incubation period of 10–20 min , Ca2+ transients in muscles were assessed using an inverted IX70 microscope ( Olympus , Shinjuku , Tokyo , Japan ) equipped with an OrcaFlash4 . 0 V2 sCMOS camera ( Hamamatsu , Hamamatsu City , Shizuoka , Japan ) and a pE-300ultra LED light source ( CoolLED , Andover , United Kingdom ) using a 20x , 0 . 75 NA objective . Images were acquired for 3 min at 5 Hz as a stream using Metamorph software ( Molecular Devices , San Jose , CA ) , while the camera and LED were triggered through TTL output through a Digidata 1440A ( Molecular Devices ) with Clampex 10 . 3 software ( Molecular Devices ) . During the acquisition process , experimenters were blind to the genotype and mRFP expression pattern of the GCaMP3+ animals in both Method one and Method 2 . After acquisition , images were de-identified using a random number generator to blind the analysis . For Method 1 , images were stitched together as TIFF series in NIH ImageJ/FIJI ( https://fiji . sc/ ) . Due to the lack of paralysis in Method 1 , some contractions resulted in axial motion that temporarily removed the cells from the focal plane . These frames with cells outside of the focal plane were manually removed from the image series . For Method 2 , image series were filtered to 2 . 5 Hz and exported from Metamorph in TIFF format . The Template Matching plugin ( https://sites . google . com/site/qingzongtseng/template-matching-ij-plugin#description ) for NIH ImageJ/FIJI was used to adjust for lateral motion during contractions and drift . After alignment , regions of interest ( ROIs ) were drawn around each cell using the magic wand tool and the background ( an area in each animal devoid of GCaMP3+ cells ) using the rectangle tool in NIH ImageJ/FIJI . The average gray values of these ROIs were measured over the time series using the multi measure tool in NIH ImageJ/FIJI . Raw data were exported to Excel ( Microsoft , Redmond , WA ) and fluorescence time series were converted using background subtraction to ΔF/F0 ( ΔF/F0 = ( F − F0 ) /F0 ) , where F0 was the baseline fluorescence for each trial . Kinetic measurements for individual peaks ( rise time ( 10%–90% ) , decay time ( 90%–50% ) , and peak-width at half max ) were made on the data acquired with Method two using Clampfit 10 . 7 ( Molecular Devices ) . Due to the mosaic nature of injections , each field contained 1–8 ( median = 2 , 25% quartile = 2 , 75% quartile = 5 ) GCaMP3+ myocytes . Further , individual cells exhibited 0–9 ( median = 1 , 25% quartile = 0 , 75% quartile = 2 . 75 ) calcium transients within the imaging window . For ΔF/F0 quantitation , a unique F0 was determined for each Ca2+ transient event . In the case where a cell exhibited multiple Ca2+ transient events , these events were treated as technical replicates and were averaged to give a single peak ΔF/F0 for each cell . In the case where a field contained multiple cells , the peak ΔF/F0 values for each cell were treated as technical replicates and averaged to give an average response for that animal . Cells or animals were considered biological replicates for analyses of peak ΔF/F0 and number of responses . For kinetic measurements , only the first peak in from each cell was analyzed and each peak was considered a biological replicate . For cells with only ‘-ΔF/F0’ ( photo-bleaching ) recorded over the course of the trial , ‘0’ was denoted as the peak ΔF/F0 . Otherwise , the maximum numerical value of ΔF/F0 for each transient was assigned as peak ΔF/F0 . For analysis of transient numbers , any transient with ΔF/F0 ≥0 . 05 was counted as a response . PCR genotyping was used to determine ryr1b alleles and assign data to its respective group . Only ryr1b+/+ and ryr1bmn0348Gt/mn0348Gt animals were included in analyses due to the variable mRNA expression seen in ryr1b+/mn0348Gt animals ( Clark et al . , 2011a ) . Isolated genomic DNA ( 300–500 ng ) was digested with MseI , and BfaI in parallel for 3 hr at 37°C and heat inactivated for 10 min at 80°C . The digested samples from each enzyme were pooled with pre-aliquoted barcoded linker in individual wells . The T4 DNA ligase ( M0202S , New England Biolabs , Inc , Ipswich , MA ) was added , and the reaction mix was incubated for 2 hr at 16°C . The linker-mediated PCR was performed in two steps . In the first step , PCR was done with one primer specific to the 3’- ITR ( R-ITR P1 , 5’- AATTTTCCCTAAGTACTTGTACTTTCACTTGAGTAA-3’ ) and the other primer specific to linker sequences ( LP1 , 5’- GTAATACGACTCACTATAGGGCACGCGTG- 3’ ) using the following conditions: 2 min at 95°C , 25 cycles of 15 s at 95°C , 30 s at 55°C and 30 s at 72°C . The PCR products were diluted to 1:50 in dH2O , and a second round of PCR was performed using ITR ( R-ITR P2 , 5’-TCACTTGAGTAAAATTTTTGAGTACTTTTTACACCTC-3’ ) and linker specific ( LP2 , 5’ - GCGTGGTCGACTGCGCAT-3’ ) nested primers to increase sensitivity and avoid non- specific amplification using the following conditions: 2 min at 95°C , 20 cycles of 15 s at 95°C , 30 s at 58°C and 30 s at 72°C . The nested PCR products from each 96-well plate are pooled and processed for Illumina library preparation as per manufacturer’s instructions . To quantify knockdown efficiencies for a GBT-confirmed lines , GBT0235 carrying the RP2 . 1 insertion into the lrpprc gene locus , quantitative reverse transcription-PCR ( qRT-PCR ) was performed by using the following protocol . Embryo collections were obtained from in-crossed lrpprc+/mn0235Gt adults . The larvae ( six dpf ) were sorted by mRFP expression to separate them based upon GBT allele and visible dark liver phenotype which has been characterized as a specific abnormality in lrpprc mn0235Gt/mn0235Gt previously . Larvae with both RFP expression and dark liver phenotype were used as the experimental group ( lrpprc mn0235Gt/mn0235Gt ) against larvae without either the RFP expression or the liver phenotype as a control ( lrpprc +/+ ) . After the initial sorting , individual zebrafish larvae were placed into a 1 . 7 ml tube with 350 µl RLT buffer within RNeasy Micro Kit ( 74004 , QIAGEN ) with β-mercaptoethanol ( M6250 , Sigma-Aldrich ) . Embryos were homogenized at max frequency ( 30 shakes/s ) for 5 min using ~30 of 0 . 5 mm stainless steel beads , RNase free ( SSB05 , Next Advance , Troy , NY ) and Tissue Lyser II ( QIAGEN ) . Homogenized samples were replaced to MaXtract High Density tubes ( 129056 , QIAGEN ) to separate between the organic solvent ( phenol/chloroform ) and the nucleic acid-containing aqueous . Total RNA was purified from the nucleic acid-containing aqueous using the RNeasy Micro Kit ( QIAGEN ) . 250 ng of total RNA from the individual larva was used for cDNA synthesis with the SuperScript II Reverse Transcriptase ( 18064014 , Thermo Fisher Scientific ) using random hexamer primers . The 16-folds diluted cDNA with deionized water were used as templates and no reverse transcriptase ( RT ) controls were run parallel to test for genomic DNA contamination . To analyze transcript levels , quantitative PCR was performed using SensiFAST SYBR Lo-ROX kit ( BIO-94005 , Bioline ) and the CFX96 Touch Real-Time PCR Detection System ( Bio-Rad , Hercules , CA ) . eef1a1l1 levels were used as reference . Three technical replicates were run for each sample and three biological replicates for each group and two no RT controls were also run . Data were analyzed through calculation of Delta Ct values . Quantitative reverse transcription-PCR was repeated for four individual clutches from the pair of lrpprc+/mn0235Gt adults . Primer sequences are the following information; lrpprc FP: 5’-TGATAATGCTGAGGAAGCTCTCAAACTG-3’ , lrpprc RP: 5’-CCTTCATCTCCTTCAGTATGTCTAACGC-3’ , eef1a1l1 FP: 5’-CCGTCTGCCAACTTCAGGATGTGT-3’ , eef1a1l1 RP: 5’-TTGAGGACACCAGTCTCCAACACGA-3’ . Source data can be found in Figure 3—source data 1 . The human orthologues of 177 cloned zebrafish genes were mainly collected from Zebrafish Information Network ( ZFIN , University of Oregon , Eugene , OR 97403–5274; URL: http://zfin . org/ ) In some cases , the candidates of human orthologues unlisted in ZFIN database were manually searched by using both Ensembl released 98 ( https://useast . ensembl . org/index . html , ( Zerbino et al . , 2018 ) and InParanoid8 ( http://inparanoid . sbc . su . se/cgi-bin/index . cgi , ( Sonnhammer and Östlund , 2015 ) databases . In parallel , the candidates were manually identified by the result of Protein BLAST ( https://blast . ncbi . nlm . nih . gov/Blast . cgi ? PAGE=Proteins ) assembled with human proteins and by the result of an online synteny analysis tool , SynFind in Comparative Genomics ( CoGe ) database ( https://genomevolution . org/CoGe/SynFind . pl ) ( Lyons and Freeling , 2008 ) . If the candidate multiply hit in those manual assessments , it was annotated as a human orthologue . The human phenotype data caused by mutations of 64 human orthologues were collected by using another data mining tool , BioMart provided by Ensembl ( http://useast . ensembl . org/biomart/martview/ ) ( Kinsella et al . , 2011; Smedley et al . , 2015 ) . The 177 human orthologs of cloned GBT-tagged genes were analyzed using PANTHER14 . 1 ( http://www . pantherdb . org/ ) ( Mi et al . , 2019 ) . With those gene symbols , 176 gene were identified in the PANTHER system ( the exception was NRXN Entrez Gene ID: 9378 ) , and 105 genes were classified at least one PANTHER protein class ( details are listed in Figure 6—source data 1 ) . 8282/20996 human genes have been annotated with 214 protein classes in PANTHER14 . 1 ( April , 2018 , http://www . pantherdb . org/panther/summaryStats . jsp ) . Experimentally validated subcellular localization data of 177 human orthologous genes tagged by GBT were manually collected from the Human Protein Atlas Subcellular Localization data downloaded on August 27th , 2019 ( Courtesy of Human Protein Atlas , www . proteinatlas . org ) ( Uhlén et al . , 2015 ) and knowledge-based subcellular localization data for the 49 genes un-validated in the Human Protein Atlas was acquired from UniProtKB on Oct 2nd , 2019 ( https://www . uniprot . org/ ) ( UniProt Consortium , 2018 ) . This data ( Supplementary file 4 ) contains 271 entries because some human orthologous genes were annotated to multiple subcellular localizations . Mouse models were found by using descriptions of animal models in both Online Mendelian Inheritance in Man ( OMIM . Johns Hopkins University , Baltimore , MD: October 28th , 2019 . URL: https://omim . org/ ) ( Amberger et al . , 2019 ) and in Mouse Genome Database ( MGD ) at the Mouse Genome Informatics ( MGI ) website , The Jackson Laboratory , Bar Harbor , Maine: October 28th , 2019 ( URL: http://www . informatics . jax . org ) ( Bult et al . , 2019 ) . MGI provided the details of mouse models of human disease , such as the number of models that have been established . Zebrafish model were also found by using OMIM and ZFIN; August 28 , 2019 . ZFIN provided all data of fish strains listed in this database ( Ruzicka et al . , 2019 ) . The area proportional Venn diagram were created using BioVenn ( Hulsen et al . , 2008 ) to visualize the number of human orthologs of the cloned genes associated with human genetic disorders which have at least one established disease model in zebrafish or mouse ( Figure 5B ) . The cloned genes with published expression data were isolated by using the wild-type expression data retrieved from ZFIN; August 28 , 2019 ( Ruzicka et al . , 2019 ) . In parallel , some published expression data were also manually searched from ZFIN . The mRFP reporter expression patterns of the cloned genes 2 and 4 dpf were manually searched using zfishbook database ( Clark et al . , 2012 ) . The comparison with the number of genes with description about expression in both ZFIN and zfishbook ( https://zfishbook . org/ ) was presented using BioVenn ( Hulsen et al . , 2008 ) to create the area proportional Venn diagrams in Figure 7S and Figure 7T . Knockdown efficiency and calcium imaging graphs were made in JMP 14 ( SAS , Cary , NC ) and in GraphPad Prism 8 ( GraphPad Software , San Diego , CA ) , respectively . All other statistical analyses were performed with R ( www . R-project . org ) using R-Studio ( www . rstudio . com/ ) . Code used for analysis in R can be found in Supplementary file 6 . Sample sizes for calcium imaging studies were estimated using peak ΔF/F0 data from Method one and the ‘pwr’ package ( https://CRAN . R-project . org/package=pwr ) . Statistical analyses for calcium imaging data were performed using the ‘wilcox . test’ function or the ‘coin’ package ( Hothorn et al . , 2006; Hothorn et al . , 2008 ) due to the non-normality visualized in the data . Outliers ( determined by Grubb’s test for one outlier in the ‘outliers’ package ( https://CRAN . R-project . org/package=outliers ) were included in overall analysis , although each statistical analysis was also performed without them as a proxy for the sensitivity of our conclusions to the outliers . All statistical tests supported the same conclusion with and without the outliers . Therefore , plots and p-values in the figures include all data points . p-values calculated excluding outliers can be found in Supplementary file 6 . Effect size was measured by Cohen’s d using the ‘effsize’ package ( https://CRAN . R-project . org/package=effsize ) . For ease of reporting , all p-values less than 0 . 0001 were reported in figures as ‘p<0 . 0001’ , but exact p-values are reported in Supplementary file 6 . All reagents are available upon request and all protein trap vectors in each reading frame will be deposited to Addgene ( http://www . addgene . org ) . Zebrafish lines are available either from Zebrafish International Resource Center ( ZIRC , http://zebrafish . org/ ) or the Mayo Clinic Zebrafish Facility , respectively . | The human genome counts over 20 , 000 genes , which can be turned on and off to create the proteins required for most of life processes . Once produced , proteins need move to specific locations in the cell , where they are able to perform their jobs . Despite striking scientific advances , 90% of human genes are still under-studied; where the proteins they code for go , and what they do remains unknown . Zebrafish share many genes with humans , but they are much easier to manipulate genetically . Here , Ichino et al . used various methods in zebrafish to create a detailed ‘catalogue’ of previously poorly understood genes , focusing on where the proteins they coded for ended up and the biological processes they were involved with . First , a genetic tool called gene-breaking transposons ( GBTs ) was used to create over 1 , 200 strains of genetically altered fish in which a specific protein was both tagged with a luminescent marker and unable to perform its role . Further analysis of 204 of these strains revealed new insight into the role of each protein , with many having unexpected roles and localisations . For example , in one zebrafish strain , the affected gene was similar to a human gene which , when inactivated , causes severe muscle weakness . These fish swam abnormally slowly and also had muscle problems , suggesting that the GBT fish strains could ‘model’ the human disease . This work sheds new light on the role of many previously poorly understood genes . In the future , similar collections of GBT fish strains could help researchers to study both normal human biology and disease . They could especially be useful in cases where the genes responsible for certain conditions are still difficult to identify . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"tools",
"and",
"resources"
] | 2020 | Building the vertebrate codex using the gene breaking protein trap library |
Homeostatic systems that rely on genetic regulatory networks are intrinsically limited by the transcriptional response time , which may restrict a cell’s ability to adapt to unanticipated environmental challenges . To bypass this limitation , cells have evolved mechanisms whereby exposure to mild stress increases their resistance to subsequent threats . However , the mechanisms responsible for such adaptive homeostasis remain largely unknown . Here , we used live-cell imaging and microfluidics to investigate the adaptive response of budding yeast to temporally controlled H2O2 stress patterns . We demonstrate that acquisition of tolerance is a systems-level property resulting from nonlinearity of H2O2 scavenging by peroxiredoxins and our study reveals that this regulatory scheme induces a striking hormetic effect of extracellular H2O2 stress on replicative longevity . Our study thus provides a novel quantitative framework bridging the molecular architecture of a cellular homeostatic system to the emergence of nonintuitive adaptive properties .
Homeostatic systems are ubiquitous in biology and function to restore internal physiological variables to a given set point following fluctuations in the internal or external environment . The accuracy of such control mechanisms ( i . e . its ability to reach an equilibrium state that is as close as possible to the pre-existing state ) , is thought to be essential to ensure robust physiological adaptation . Therefore , understanding the mechanisms underlying accurate control in regulatory networks has emerged as a central question in Control Theory applied to biological systems ( Khammash , 2016; Ma et al . , 2009; Whitacre , 2012 ) . Several seminal studies have pointed to the existence of ‘perfectly adapting’ systems in which exact restoration of the pre-existing state is observed , including bacterial chemotaxis ( Barkai and Leibler , 1997; Berg and Tedesco , 1975 ) , calcium signaling ( El-Samad et al . , 2002 ) , and yeast hyperosmolarity response ( Muzzey et al . , 2009 ) , all of which are based on a regulatory scheme referred to as ‘integral feedback’ ( Yi et al . , 2000 ) . However , high control accuracy alone is insufficient to protect against the potentially damaging effects of environmental challenges , and other dynamical properties of homeostatic systems may determine the cell’s ability to adapt: indeed , most stress regulatory pathways feature transcriptional responses , which are intrinsically slower processes than the other biochemical effects of stress exposure ( Muzzey and van Oudenaarden , 2009; Young et al . , 2013 ) . Because of this limiting response time , it is expected that a transient peak in the internal stress level ( output peak of magnitude Omax , Figure 1A ) may occur in response to stepwise exposure to an external stressor ( input , Figure 1A ) . Such a transient overshoot may trigger irreversible deleterious effects leading to cell death , irrespective of the ability of the homeostatic system to accurately restore the pre-existing steady state ( Oeq on Figure 1A ) . In this case , interestingly , the rate at which stress is applied ( while keeping constant the overall magnitude of stress ) should directly control Omax and hence determine cellular stress resistance ( Figure 1A ) . Remarkably , this hypothesis has neither been formally addressed theoretically nor been tested experimentally . Yet , whether the adaptation range of a homeostatic system is set only by the overall stressor level , or alternatively , depends on the kinetics of the input stress pattern remains a question of fundamental importance . 10 . 7554/eLife . 23971 . 003Figure 1 . A single-cell microfluidics assay to monitor yeast adaptation to H2O2 . ( A ) Schematic representation of ‘training’ and ‘stress tolerance’ phenomena in a simple negative feedback-based system . ( B ) Schematic of the microfluidic device setup for live-cell imaging in H2O2-containing media . ( C ) Decline in extracellular H2O2 concentration comparing the single-cell assay and bulk experiment ( starting at cellular OD600 = 0 . 5 ) . ( D ) Sequence of phase-contrast and fluorescence images of cells at the indicated time points after addition of 0 . 4 mM H2O2 at t = 300 min . The red and green channels represent the Htb2-mCherry ( nuclear marker ) and Yap1-GFP signals , respectively . Orange and white lines represent the cellular and nuclear contours obtained after automated segmentation . The white bars represent 5 µm . ( E ) Top: Mean growth rate per cell as a function of time after addition of different H2O2 concentrations at t = 300 min , as indicated in the bottom panel . Middle: Mean nuclear Yap1-GFP localization as a function of time , with the same color coding as in the top and bottom panels . Yap1-GFP scoring is not possible at 0 . 6 mM due to metabolic arrest/cell death and GFP signal decline . ( F ) Increase in mean Yap1-GFP localization ( relative to the pre-stress level ) at steady state ( measured for t > 800 min ) as a function of H2O2 concentration added at t = 300 min during a step experiment . Lines represent the best fits to mathematical models of integral ( blue ) , linear ( red ) and nonlinear ( magenta ) feedback with different sets of assumptions ( see Materials and methods ) . ( G ) Comparison of growth rate at steady state as a function of H2O2 concentration for the wild-type ( WT ) strain and the Δyap1 mutant . Error bars and shaded regions are SEM ( C , N = 6; E , N > 100 for most time points; F and G , N > 100 for each H2O2 concentration ) . See also Figure 1—figure supplement 1 and Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 00310 . 7554/eLife . 23971 . 004Figure 1—figure supplement 1 . Medium diffusion properties in the microfluidics device . ( A ) Fluorescein diffusion kinetics in an empty trapping cavity . Top left: phase-contrast image centered on a trapping cavity with two supply channels on the sides . The blue circle corresponds to the region of interest ( ROI ) of the supply channel in which the GFP signal is scored over time ( bottom left: blue line ) . The red circle corresponds to the ROI of the trapping cavity in which the GFP signal is scored over time ( bottom left: red line ) . Right: fluorescence images were taken at the indicated time points . The white bar represents 5 µm . ( B ) Same experiment as ( A ) for a crowded cavity . The magnification region on the right bottom corner ( GFP and phase contrast ) shows that the fluorescein doesn’t enter the cells during the experiment . The white bar represents 5 µm . ( C ) Scoring of the mean nuclear Yap1-GFP localization as a function of time for cells located at the edge of the trapping cavity ( 30 s time resolution ) . Error bars are SEM ( N > 100 for all time points ) . ( D ) Phenotypic distribution across the cells depending of their position in the trapping cavity . p-Value is calculated using a Chi-squared test of independency . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 00410 . 7554/eLife . 23971 . 005Figure 1—figure supplement 2 . Principle of growth rate measurements . ( A ) Growth rate measurements in the absence of stress . Top: cell volume of individual cells . Each color corresponds to a single cell followed over its successive divisions . The colored filled circle indicates a budding event . Middle: mean growth rate per cell as defined in Materials and methods . The error bars indicate the standard error of the mean ( +/- SEM , N > 100 cells by the end of the experiment ) . Below: temporal profile of H2O2 concentration used during the experiment . ( B ) Same as ( A ) , but following the switch from 0 to 0 . 4 mM H2O2 at t = 300 min . ( C ) Scatter plot showing the absence of correlation between cell growth rate and cell volume of individual cells . ( D ) Top: Evolution of mean ( +/- SEM , N > 100 for most time points ) cell size during the switch from 0 to 0 . 4 mM H2O2 at t = 300 min . Below: temporal profile of H2O2 concentration used during the experiment . ( E ) Similar experiment as in Figure 1E , but with the Δyap1 mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 005 Interestingly , however , there is one prominent example of stress pattern that confers improved cellular adaptation despite fast environmental changes , known as acquired stress resistance , or stress tolerance: in this case , a mild stress preconditioning increases resistance to subsequent , acute exposure to large doses of the same stressor . This effect has been observed in a broad spectrum of species , from unicellular organisms to mammals , in response to diverse environmental challenges ( Davies et al . , 1995; Hecker et al . , 2007; Kandror et al . , 2004; Kensler et al . , 2007; Lewis et al . , 1995; Lindquist , 1986; Lou and Yousef , 1997; Lu et al . , 1993; Scholz et al . , 2005 ) and is considered to be an anticipation strategy to overcome potentially harmful environmental conditions in the future ( Mitchell et al . , 2009 ) . Stress resistance is thus itself an adaptive trait reflecting an intrinsic plasticity of the homeostatic machinery . However , the mechanisms underlying such adaptive homeostasis- the robustness of which is improved following stress exposure ( Davies , 2016 ) , remain to be elucidated . In particular , it is not known how this acquired stress resistance bypasses/overcomes the intrinsically slow response time of the stress response ( e . g . through faster transcriptional response , or initially higher stressor degradation rate- analogous to stress buffering ) and how this might be mechanistically achieved ( Figure 1A ) . Here , we used the response of budding yeast to hydrogen peroxide ( H2O2 ) stress as a model system to investigate the driving principles that govern the adaptation to arbitrary stress patterns and to decipher the mechanisms at work in adaptive homeostasis . We developed a quantitative framework based on mathematical modeling , live-cell imaging , and microfluidics to apply controlled H2O2 stress patterns and to measure the adaptive stress response with a high level of accuracy . We demonstrate that cell survival is highly stress-rate dependent , validating the paradigm that adaptation is limited by the response time of the cellular homeostatic machinery and thus revealing an unprecedented ‘trainability’ of the cells to stress . We also show that the acquisition of stress resistance is distinct from cellular training and originates in the nonlinear scavenging of H2O2 by peroxiredoxins ( Prx ) . Along with this , our study unravels an important Prx-dependent replicative lifespan extension in the presence of very low doses of H2O2 . This reveals an unprecedented direct hormetic effect that can be quantitatively accounted for by the nonlinear Prx feedback model . Our study provides the first quantitative and mechanistic analysis that establishes the link between the architecture of a fundamental stress homeostatic system and the emergence of distinct nonintuitive properties , namely cellular training to stress and adaptive homeostasis .
We first sought to characterize the kinetics of the cellular response to stepwise exposure to increasing H2O2 concentrations and the emergence of cellular adaptation . Adaptation to H2O2 is usually measured by quantifying the fraction of surviving cells after the addition of a bolus of H2O2 to cells in culture . Under these conditions , cells rapidly degrade the H2O2 in the medium and , therefore , adaptation is partly due to the removal of the stressor . To circumvent this problem and to allow precise control of the stressor level , we developed custom microfluidic device to monitor the divisions of individual cells in trapping cavities while controlling external H2O2 concentrations by continuous replenishment of H2O2-containing medium ( Figure 1B and C and Materials and methods ) . Using a fluorescent dye , we checked that diffusion in the trapping cavities was not impaired by the presence of a large cell cluster in the cavity ( Figure 1—figure supplement 1A and B ) and we also verified that cell growth was not affected despite the increasing cell confinement during a typical experiment ( up to 1000 min , see Figure 1—figure supplement 2A ) . Using this technique , therefore , we could follow cellular adaptation by monitoring activation of the H2O2 stress response and real-time cellular growth rate . The signaling network activated in response to H2O2 is centered on the Yap1 regulon ( Kuge and Jones , 1994; Lee et al . , 1999 ) . Yap1 is a transcription factor that , in the absence of H2O2 , shuttles between the cytoplasm and nucleus . Upon exposure to H2O2 , Yap1 is oxidized and nuclear export is prevented ( Azevedo et al . , 2003; Kuge et al . , 1997; Lee et al . , 1999 ) . Hence , nuclear accumulation of Yap1 is a sensitive reporter of internal H2O2 levels ( Toledano et al . , 2004 ) . In the nucleus , Yap1 activates the expression of genes involved in redox homeostasis and H2O2 scavenging ( Gasch et al . , 2000; Godon et al . , 1998; Lee et al . , 1999 ) , leading to a negative feedback regulatory loop ( Toledano et al . , 2004 ) . We monitored the dynamics of individual cells expressing Yap1-GFP and a nuclear marker ( histone Htb2-mCherry ) with a 3-min interval . Following the switch to medium containing 0–0 . 4 mM H2O2 ( at t = 300 min; Figure 1D and Video 1 ) , the cells experienced a transient stress level-dependent reduction in growth rate ( see Materials and methods for the principle of the measurement ) followed by a complete recovery , revealing intrinsic cellular adaptation ( Figure 1E ) . In parallel , the switch to H2O2 induced an abrupt burst in Yap1-GFP nuclear localization , which saturated above 0 . 2 mM H2O2 . Similar experiments performed with higher temporal resolution ( 30 s interval ) revealed that the nuclear relocation time is around 120 s ( see Figure 1—figure supplement 1C ) , which is much faster that the overall adaptation timescale ( ~45–100 min , see Figure 1E: Yap1-GFP quantification ) . However , this stands significantly higher than the timescale of stressor diffusion across cavity ( see Figure 1—figure supplement 1A and B ) , therefore revealing that the activation of Yap1 is set by the diffusion of the stressor across the cell membrane rather than by limited diffusion in the device . 10 . 7554/eLife . 23971 . 006Video 1 . Growth rate monitoring upon H2O2 stress ( refers to Figure 1 ) Movie showing a time-lapse experiment where cells are exposed to sudden step stress of 0 . 4 mM H2O2 at t = 300 min . Left: phase contrast , right: growth rate evolution graph . The white bar represents 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 006 The burst of nuclear Yap1-GFP relocation was followed by partial recovery to a steady-state level that was also H2O2 concentration-dependent ( Figure 1D–F and Video 2 ) , suggesting that Yap1 activity is still required in the adapted state ( i . e . following growth recovery ) . Consistent with this , deletion of Yap1 strongly decreased the cells’ capacity to adapt , in agreement with previous findings ( Inoue et al . , 1999 ) ( Figure 1G and Figure 1—figure supplement 2E ) . Last , the fact that a 0 . 1 mM H2O2 stress induces a partial nuclear relocation of Yap1-GFP but has no effect on growth rate indicates that Yap1 signaling is more sensitive to H2O2 than is the overall cellular physiology ( Figure 1E ) . 10 . 7554/eLife . 23971 . 007Video 2 . Yap1 nuclear relocation upon H2O2 stress ( refers to Figure 1 ) Movie showing the nuclear enrichment of Yap1 in cells exposed to 0 . 4 mM H2O2 at t = 300 min . Left: Phase contrast and mCherry ( Htb2-mCherry ) channels . Right: GFP ( Yap1-GFP ) channel . The white bar represents 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 007 Interestingly , while the cells adapted to 0 . 3 mM H2O2 ( complete mean growth rate recovery ) , exposure to 0 . 6 mM H2O2 induced full growth arrest ( Figure 1E and Video 3 ) , revealing that cell fate is controlled by a sharp threshold in external H2O2 level . To determine the physiological changes accompanying this switch from adaptation to arrest , we focused on the behavior of individual cells exposed to a sublethal dose ( 0 . 5 mM ) of H2O2 ( see Figure 2A and B ) . Notably , while all cells experienced at least a transient growth arrest ( at t = 390 min , Figure 2B ) following exposure to the stressor ( at t = 300 min ) , we observed a high degree of heterogeneity in cell fate across the population at steady-state: 22% of cells present at the time of H2O2 addition recovered a normal growth and division rate at t > 600 min ( referred to as the ‘adapted’ phenotype in the following ) ; 36% experienced a prolonged slow-down of cell cycle progression characterized by an extended budded period , but continued to increase in size over time ( ‘prolonged cell cycle arrest’ phenotype ) ; and 42% stopped growing and failed to divide ( ‘permanent growth arrest’ phenotype ) ( Figure 2A–C and Figure 2—figure supplement 1 and Video 4 ) . We checked this heterogeneity in cell fate was not dependent on the position of cells with respect to the border of the cavity ( see Figure 1—figure supplement 1D ) . Interestingly , performing the same analysis of phenotypic distribution at various H2O2 levels revealed that the cell cycle arrest phenotype emerged at lower concentrations of H2O2 than the growth arrest phenotype ( typically ~0 . 2 and 0 . 5 mM H2O2 , respectively; Figure 2C ) . 10 . 7554/eLife . 23971 . 008Figure 2 . DNA damage checkpoint and metabolic arrest during exposure to H2O2 . ( A ) Lineage of cells after addition of 0 . 5 mM H2O2 at t = 300 min . Each colored line corresponds to a single cell , and cell budding is indicated by the vertical black lines . The plot shows examples of cells with three distinct growth phenotypes , as indicated . ( B ) Quantification of the growth rate and division rate of the cellular phenotypes displayed in ( A ) . Error bars are SEM . ( C ) Relative distribution ( % ) of cellular phenotypes as a function of H2O2 concentration added at t = 300 min . N = 100 for each concentration . ( D ) Top: Upregulation of cytoplasmic Rnr3-GFP expression at 300 min after addition of 0 . 4 mM H2O2 . Bottom: Quantification of the Rnr3-GFP signal after addition of 0 . 4 mM H2O2 . ( E ) Top: Upregulation of Ddc2-GFP expression at 300 min after addition of 0 . 4 mM H2O2 . Bottom: Percentage of cells with durable Ddc2-GFP foci measured before and at 300 min after addition of 0 . 4 mM H2O2 for the cell phenotypes identified in ( A ) . Error bars are 95% confidence intervals ( CI ) . Two-proportions Z test . The whites bar represent 5 µm . See also Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 00810 . 7554/eLife . 23971 . 009Figure 2—figure supplement 1 . Lineage of cells following the exposure to 0 . 5 mM H2O2 . Top: Pedigree analysis of individual cells growing in the microfluidic during a typical H2O2 stress assay at 0 . 5 mM . Each colored line represents a single cell , and vertical black lines indicate cellular parentage . The vertical magenta line indicates the timing of switch to media containing H2O2 . Each color corresponds to the phenotype acquired by the cells following the addition of stress . Bottom: same experiment as in the top panel but displaying the evolution of individual cell volume over time using the indicated color-coding . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 00910 . 7554/eLife . 23971 . 010Figure 2—figure supplement 2 . Cell death quantification during high ( 0 . 6 mM H2O2 ) step stress . ( A ) Propidium iodide ( PI ) -based microfluidics essay detecting mortality in cells exposed to 0 . 6 mM H2O2 at t = 300 min . Red-colored ( PI-positive ) cells correspond to dead cells with loss of cell membrane integrity . The white bar represents 5 µm . ( B ) Top: Quantification of the PI incorporation by the cells displayed in ( A ) as a function of time . Bottom: Temporal profile of the H2O2 concentration . The orange filled circles correspond to the time points displayed in ( A ) . ( C ) Reversibility of the growth arrest . Top: mean growth rate of cells exposed to 0 . 6 mM H2O2 for various durations , as indicated in the bottom panel . Middle: mean cellular expression of TRX2pr-GFP-deg; Bottom: schematics of the temporal profile of the pulse . Error bars are SEM ( N > 100 for most time points ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 01010 . 7554/eLife . 23971 . 011Figure 2—figure supplement 3 . Cell death markers for mild ( 0 . 4 mM H2O2 ) step stress . ( A ) Cells exposed to 0 . 6 mM H2O2 ( lethal stress ) doesn’t induce Trx2-GFP and show bright vacuoles . Top: Quantification of Trx2-GFP level and the mean cellular intensity of the phase-contrast highest decile ( note the increase in the intensity after stress addition revealing the presence of bright vacuoles , see Materials and methods ) . Middle: Temporal profile of the H2O2 concentration . Bottom: Phase-contrast and fluorescence images of individual cells right before ( 300 min ) and 500 min after 0 . 6 mM H2O2 addition ( 801 min ) . ( B ) After addition of 0 . 4 mM H2O2 , the fraction of cells with no growth recovery show low Trx2-GFP level and bright vacuoles . Top: Quantification of Trx2-GFP level and the mean cellular intensity of the phase-contrast highest decile . Middle: Temporal profile of the H2O2 concentration . Bottom: Phase-contrast and fluorescence images of individual cells 500 min after 0 . 4 mM H2O2 addition . The red contours show permanently arrested cells . ( C ) Summary of scoring of death markers in lethal ( 0 . 6 mM H2O2 ) and permissive ( 0 . 4 mM H2O2 ) step stress experiments . The white bars represent 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 01110 . 7554/eLife . 23971 . 012Video 3 . Cell metabolic arrest at high H2O2 dose ( refers to Figure 1 ) Movie showing a permanent cell growth arrest in cells exposed to 0 . 6 mM H2O2 . Left: phase contrast , right: growth rate evolution graph . The white bar represents 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 01210 . 7554/eLife . 23971 . 013Video 4 . Phenotypic variability upon H2O2 stress ( refers to Figure 2 ) Movie showing different cell fates in cells exposed to 0 . 5 mM H2O2 at t = 300 min . Blue cell contours: Adapted cells; Yellow cell contours: Prolonged cell cycle arrest; Red contours: Permanent growth arrest . The white bar represents 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 013 To go beyond the characterization of growth and division impairments in cells submitted to H2O2 , we asked whether these defects were accompanied by the activation of DNA damage response and/or ultimately lead to cell death . To this end , we examined the response of cells expressing Rnr3-GFP or Ddc2-GFP fusion proteins . Rnr3 is a subunit of ribonucleotide reductase and is upregulated during the DNA damage response ( DDR ) , and Ddc2 is a DNA damage checkpoint protein that forms foci upon recruitment to DNA lesions . Exposure to 0 . 4 mM H2O2 substantially increased cytoplasmic Rnr3-GFP levels in cells displaying the prolonged cell cycle arrest phenotype , but not in adapted cells ( Figure 2D ) . Similarly , the vast majority of cell-cycle-arrested cells displayed bright foci of Ddc2-GFP fluorescence , unlike adapted cells ( Figure 2E ) . These experiments indicated that H2O2-induced DNA damage and subsequent DDR checkpoint activation were responsible for prolonged cell cycle arrest . In contrast , cells experiencing a permanent arrest showed no detectable increase in Rnr3-GFP ( not shown ) . We considered that these cells may have a compromised physiological state preventing any response to the oxidative stress threat , and may ultimately die . To test this hypothesis , first , we used a vital stain ( propidium iodide , PI ) to monitor the onset of death in cells abruptly exposed to 0 . 6 mM H2O2 ( see Figure 2—figure supplement 2A and B and Video 5 ) . We found that all cells ( N = 123 ) eventually became fluorescent , therefore demonstrating that a stress exposure to this H2O2 concentration ultimately induces cell lysis . Next , we wondered to which extent this growth arrest phenotype could be reverted by stress removal . To this end , we monitored the mean growth rate and the expression of the thioredoxin promoter fused to both sfGFP and a destabilizing degron sequence , TRX2pr-sfGFP-deg ( TRX2 encodes a Yap1-regulated thioredoxin ) , after exposure of cells to 0 . 6 mM H2O2 for varying periods . H2O2 addition induced rapid growth arrest and irreversible decay of TRX2pr-sfGFP-deg levels ( black lines on Figure 2—figure supplement 2C upper and middle panels , respectively ) . However , removal of the stress by switching back to H2O2-free medium at various times after H2O2 addition led to recovery of the mean cellular growth rate and induced reactivation of the Yap1 regulon if the duration of exposure was less than 4 hr ( Figure 2—figure supplement 2C ) . This indicated that a few hours of exposure at 0 . 6 mM H2O2 were necessary to induce an irreversible growth arrest phenotype that ultimately lead to cell death . 10 . 7554/eLife . 23971 . 014Video 5 . Cell mortality at high H2O2 step stress ( refers to Figure 2—figure supplement 2 ) Movie showing the incorporation of PI in cells exposed to 0 . 6 mM H2O2 at t = 300 min . The white bar represents 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 014 Last , we checked that the behavior of permanently arrested cells , even at sublethal H2O2 concentrations ( 0 . 4 mM H2O2 ) , was similar to the irreversible growth arrest phenotype at 0 . 6 mM H2O2 . For this , we compared the expression of the TRX2-GFP fusion protein following either a 0 . 4 mM or a 0 . 6 mM H2O2 step . In both cases , we observed that the subpopulation of permanently arrested cells showed very low TRX2-GFP levels ( Figure 2—figure supplement 3A–C ) . In addition , we found that , unlike adapted and cell cycle arrested cells , cells with a permanent growth arrest displayed the same bright and large vacuole phenotype ( Figure 2—figure supplement 3A–C ) in both conditions ( 0 . 4 mM and 0 . 6 mM H2O2 ) . Therefore , these results indicate that the permanent growth arrest phenotype observed in step experiments ( lethal or sublethal ) reflects the inability of the cells to defend against the stressor . Collectively , our analysis reveals the existence of distinct cell fates following the exposure to acute sublethal doses of H2O2 , and demonstrates that these phenotypes occur in a H2O2 concentration-dependent manner . The observation that increasing H2O2 external levels resulted in a sharp and partially reversible transition from adaptation to growth arrest suggested that the transient internal stress level Hmax reached closely after exposure to H2O2 may exceed a toxic concentration Htox beyond which cellular function is impaired , as hypothesized in the introduction . If so , exposure of cells to a gradual increase in H2O2 concentration should allow the cells more time to activate the antioxidant response and should thus dramatically improve cellular adaptation ( see Figure 1A ) . To understand quantitatively how the kinetics of a stress pattern may influence cellular adaptation , we first developed a mathematical description of the homeostatic machinery based on the negative feedback regulation in the Yap1 network . This model was then used throughout this study to help identify and formalize the emergent properties of this system through iterative cycles of predictions and experimental challenges , rather than to perform exhaustive data fitting aimed at retrieving individual parameter values . The model assumed that nuclear relocation of Yap1 increases the production of antioxidants ( referred to as ‘A’ in the model , see Figure 3A ) , which then scavenge intracellular H2O2 ( ‘H’ in Figure 3A ) . For the sake of simplicity , we first developed a linear version of this model ( f ( H ) =1 and g ( H ) = H; Figure 3A ) that could successfully recapitulate the limited accuracy of the homeostatic system; the internal H2O2 level at steady-state Heq increases with the magnitude of H2O2 steps , unlike a system based on an ‘integral’ feedback regulatory scheme ( Figures 1E , F and 3B , and Materials and methods ) . This property is a direct consequence of the assumption that antioxidants are not infinitely stable but must be diluted in growing cells ( μ’≠0 , see Materials and methods and Figure 3—figure supplement 1A and B ) . However , the growth rate does not affect the kinetics of the internal H2O2 burst during the transient response to H2O2 steps ( see Figure 3—figure supplement 1 and Materials and methods ) . Therefore , the observation that the growth rate undergoes a transient slowdown during the regime that precedes adaptation to sublethal H2O2 steps should not impact the overall internal H2O2 kinetics nor the cellular adaptation capacity . Therefore , for the sake of simplicity , we neglected the variations in growth rate in the model . 10 . 7554/eLife . 23971 . 015Figure 3 . A negative feedback-based model to describe adaptation to H2O2 . ( A ) Schematic of the regulatory network involved in H2O2 scavenging: external H2O2 ( represented by I ) , internal H2O2 ( represented by H ) , and antioxidants ( represented by A ) . A linear set of differential equations is used to describe the evolution of this system over time . ( B ) Response of the linear system described in panel ( A ) to sudden exposure to external H2O2 ( step of amplitude I ) . Each colored line corresponds to a given concentration of H2O2 . Heq is the steady-state internal H concentration , Hmax is the maximum H concentration reached during the transient regime . Htox is the threshold concentration beyond which growth/division is assumed to stop ( obtained for I = 0 . 6 mM ) . ( C ) Phase diagram showing Hmax as a function of the amplitude of the step I and the rate of the H2O2 ramp δ = ΔI/ΔT . Inset shows a graphical representation of these parameters . The solid black line indicates the contour given when Hmax = Htox , assuming the general assumptions of the linear model described in panel A . ( D ) Sequence of phase-contrast and fluorescence images of individual cells at the indicated times after initiation ( t = 100 min ) of a linear ramped increase in H2O2 concentration at a rate δ of 2 . 2 μM/min . The red and green channels represent the Htb2-mCherry and Yap1-GFP signals , respectively . The white bars represent 5 µm . ( E ) Top: Mean growth rate per cell as a function of time after initiation ( t = 100 min ) of linear ramps in H2O2 concentration . The line colors correspond to the indicated ramp slopes in the bottom panel . Middle: Mean nuclear Yap1-GFP localization . Error bars and shaded regions are SEM , N > 100 for most time points . ( F ) Phase diagram recapitulating the mean growth rate of cells during adaptation to steps ( Figure 1E ) and linear ramps at various rates δ . The gray shading delimits the regions of adaptation and arrest , as expected from the linear feedback model . See also Figure 3—figure supplement 1 and Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 01510 . 7554/eLife . 23971 . 016Figure 3—figure supplement 1 . Linear model: response to step , ramps , and training capabilities . ( A ) Dynamics of model variables under the assumptions of a linear integral feedback model , when the system is submitted to a step concentration . Each color line represents the temporal evolution of the variables H , AmRNA and A following the switch to the indicated H2O2 concentration . Parameter values: ε = 0 , α = 1 min−1 , β = 5 min−1 , γ = 0 . 53 min−1 , γ’=0 . 01 min−1 , μ= log ( 2 ) /2 . 5 min−1 , μ’=0 min−1 . ( B ) Similar as in ( A ) , but in the general case of the linear model . Parameter values: ε = 0 , α = 1 min−1 , β = 5 min−1 , γ = 0 . 53 min−1 , γ’=0 . 01 min−1 , μ= log ( 2 ) /2 . 5 min−1 , μ’ = log ( 2 ) /100 min−1 . ( C ) Similar as in ( A ) , but during linear stress ramp . ( D ) Similar as in ( B ) , but during linear stress ramp . ( E ) Phase diagram ( similar to Figure 3C ) for the case of an integral feedback model ( μ’=0 min−1 ) . ( F ) Similar as ( E ) for the general case of the linear model ( μ’ = log ( 2 ) /100 min−1 ) . ( G–J ) Phase diagram obtained with different values of mRNA decay rates ( μ ) , as indicated , while keeping the value of Iabs = 7 . 3 mM . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 01610 . 7554/eLife . 23971 . 017Figure 3—figure supplement 2 . Experimental setup to generate H2O2ramps and scoring of the permanent growth arrest phenotype . ( A ) Schematics of the setup used to generate linear ramps . ( B ) Dosage of H2O2 concentration during the ramp experiments performed using the setup described in ( A ) . Each colored line corresponds to the expected ramp slope , whereas the actual H2O2 measurements are indicated as colored filled circles . Error bars are SEM , N = 3 . ( C ) Top: Scoring of the fraction of cells able to grow as a function of time during ramped H2O2 stress . Only the cells present at the time of ramp initiation ( t = 100 min ) are scored . The red and black lines correspond to 4 . 4 µM/min and 8 . 8 µM/min H2O2 ramps , respectively . Cells are considered as not able to grow ( permanent growth arrest ) when their growth stops and does not recovers until the end of the experiment . Bottom: Temporal profile of the H2O2 concentration for both ramps . ( D ) Same experiment as ( C ) , but the fraction of cells able to grow is represented as a function of the absolute H2O2 concentration . ( E ) Cells with permanent growth arrest during 8 . 8 µM/min H2O2 ramp show bright vacuoles . Top: the mean cellular intensity of the phase-contrast highest decile ( see Materials and methods ) is displayed for individual cells during the ramped stress . Middle: temporal profile of the H2O2 concentration . Bottom left: MatLab boxplot showing the quantification of the mean cellular intensity of the phase-contrast highest decile for the subpopulation of cells with permanent growth arrest during the ramp . The quantification is done at the moment of the cell growth arrest as well as 200 min before and after the growth arrest ( N = 50 , total of 64 cells but only 50 were scored because not all cells can be followed 200 min after growth arrest ) . Bottom right: phase-contrast image of cells 611 min after the initiation of 8 . 8 µM/min H2O2 ramp . The red contours show permanently arrested cells . Two-means Z test . The white bar corresponds to 5 µm . ( F ) Determination of hard limit H2O2 concentration allowing adaptation . Mean cell growth rate per cell as a function of time after initiation ( t = 100 min ) of linear ramps at a rate δ of 2 . 2 μM/min during 140 hr . Expected and measured ( N = 6 ) H2O2 concentrations are displayed on the bottom panel . Error bars are SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 017 Importantly , the linear model also predicted that , whereas steep ramps would quickly lead to growth arrest ( since , eventually Hmax > Htox; green line on Figure 3C and inset ) , as observed experimentally beyond 0 . 6 mM , slower stress ramps may allow the cells to adapt to much higher stress levels ( keeping Hmax < Htox; blue lines on Figure 3C and inset ) . The phase diagram in the ( δ , I ) space recapitulated these predictions and , additionally , clearly delimited the region in which adaptation to high-stress magnitude I is permitted ( indicated by the black line Hmax = Htox on Figure 3C ) . The extent of this region is limited by the overall accuracy of the homeostatic system , which is mostly set by the degradation rate μ’ of antioxidant enzymes ( Figure 3—figure supplement 1E and F and Materials and methods ) and the response time of the antioxidant system , which derives from the degradation rate μ of corresponding mRNAs ( Figure 3—figure supplement 1G–J and Materials and methods ) ; μ was set to log ( 2 ) /40 min−1 , according to previous measurements ( Geisberg et al . , 2014 ) . To test the prediction of the linear model , we developed a protocol to generate a linear increase in H2O2 concentration over time in the microfluidic chip , with rates from 1 . 1 μM/min to 17 . 6 μM/min ( Figure 3—figure supplement 2A and B and Materials and methods ) . Under these conditions , we observed no decline in growth rate up to >4 mM H2O2 when the slope δ was ≤4 . 4 μM/min ( Figure 3D and E ) , whereas the growth rate decreased progressively at δ ≥8 . 8 μM/min ( Figure 3E ) . We checked that this decay in growth rate observed at the population level mainly resulted from a progressively increasing number of individual cells undergoing a permanent growth arrest , consistently with the phenotype described in step experiments ( Figure 3—figure supplement 2C ) . Similarly , these arresting cells displayed a high-phase-contrast intensity due to large vacuoles , suggesting that these cells are unable to adapt and ultimately die ( Figure 3—figure supplement 2E ) . In addition , we verified that the growth rate decay was not due to the higher absolute H2O2 concentration reached during a 8 . 8 μM/min ramping experiment compared to a ramp of 4 . 4 μM/min . Indeed , we observed that the onset of occurrence of the growth arrest phenotype at 8 . 8 μM/min occurred at a lower H2O2 level that the one reached at 4 . 4 μM/min , whereby no growth arrest was observed ( Figure 3—figure supplement 2D ) . Altogether , these results confirmed the prediction of the model that adaptation is strongly stress-rate dependent and validated the hypothesis that the transient internal H2O2 peak level reached with stepwise addition limits the ability to adapt . In agreement with this , nuclear localization of Yap1 was lower during ramping ( Figure 3E ) than during the step experiments ( Figure 1E ) . Interestingly , the model also predicted the existence of an absolute H2O2 level Iabs beyond which no adaptation is possible , even with extremely slow ramping , due to the dilution of antioxidants ( Figure 3C , Figure 3—figure supplement 1 , and Materials and methods ) . To estimate this threshold experimentally , we monitored the growth rate of cells upon exposure to a slow H2O2 ramp ( δ = 2 . 2 μM/min ) for >5000 min and found that the onset of growth decline occurred at a H2O2 concentration of 7 . 2 mM ( Figure 3—figure supplement 2F ) . From a theoretical viewpoint , this concentration is a fundamental constant that characterizes the overall buffering capability of the homeostatic system and integrates most of the parameters of the model ( see Materials and methods ) . Based on this estimate , combining all step and ramp experiments in the ( δ , I ) phase space ( using growth rate as a readout of adaptation ) provided good agreement between the experiments and the model ( Figure 3C and F ) . Overall , this analysis identified a > 10-fold increase in the H2O2 adaptation limit observed when using slow versus fast stress ramping , which could be explained by a linear negative feedback model in which the response time of antioxidant expression plays a critical role . The unprecedented analysis therefore revealed the ‘training’ capabilities of individual yeast cells , which can be progressively acclimated to increasingly high levels of stress . Next , we considered how our framework based on a linear feedback model could explain the phenomenon of acquired stress tolerance , in which mild ( I0 = 0 . 1–0 . 4 mM ) H2O2 pretreatment increased by several orders of magnitude the fraction of cells surviving a subsequent challenge with a more severe stepwise stress of magnitude ΔI ( Davies et al . , 1995 ) . To transpose these observations using our methodology , we first verified that pretreating cells with I0 = 0 . 2 mM H2O2 shifted the adaptation threshold to ΔI = 1 mM ( Figure 4A and B ) , contrasting with the ΔI = 0 . 6 mM threshold obtained for I0 = 0 mM ( as shown in Figure 1 ) . Here again , the large vacuole phenotype obtained with ΔI = 1 mM ( as measured using phase-contrast intensity of cells ) strongly suggested that these permanently arrested cells failed to adapt , as in steps and ramp experiments ( Figure 4—figure supplement 1A ) . In addition , we found that this effect was clearly dependent on the Yap1 regulon , since yap1Δ mutants did not display acquisition of tolerance ( Figure 4—figure supplement 1B and C ) . However , this increased resistance to stepwise stress exposure could not be explained by the linear feedback model , which predicted that both pretreated and naive cells should experience a similar internal peak stress during the subsequent stress challenge ( Figure 4C ) . Mathematically , this results from an additivity principle , according to which the response to a perturbation ΔH ( i . e . stress challenge ) is independent of the response triggered by a preceding input fluctuation ( i . e . preconditioning ) . 10 . 7554/eLife . 23971 . 018Figure 4 . Mechanism of acquisition of tolerance to stress . ( A ) Mean cell growth rate ( top panel ) of cells exposed to H2O2 steps of the magnitude indicated in the bottom panel at t = 400 min after a 0 . 2 mM pretreatment at t = 0 min . ( B ) Fraction of cells without permanent growth arrest at different concentrations of H2O2 in the presence ( yellow ) or absence ( blue ) of pretreatment . N = 100 for each concentration . ( C ) Response of the linear system to simple H2O2 step of amplitude ΔI = 0 . 6 mM for naive ( magenta: I0 = 0 mM ) or pretreated ( bleu: I0 = 0 . 2 mM ) cells . ( D ) Top: Mean cell transcriptional dynamics ( top panel ) of the TRX2 promoter ( Trx2-sfGFP-degron ) for the H2O2 treatments shown in the middle panel . Bottom: Quantification of the maximum transcription rate of the TRX2 promoter during the indicated steps . Two-means Z test . ( E ) Mean cell Yap1-GFP nuclear localization upon a 0 . 2 mM H2O2 step for cells with ( green line ) or without ( magenta line ) a 0 . 2 mM pretreatment , as indicated in the bottom panel . ( F ) Quantification of the amplitude of the burst in Yap1 nuclear localization during a 0 . 1 mM H2O2 challenge step . The lines indicate the fit of the linear ( red ) and nonlinear ( magenta ) models . ( G ) Numerical phase diagram indicating the region in which adaptation is permitted as a function of the overall stress magnitude ΔI and stress rate δ for the linear ( left ) and nonlinear ( right ) models . The solid black line indicates the contour given when Hmax = Htox ( survival threshold ) as in Figure 3C . The vertical dashed line represents the basal stress resistance , as observed in step experiments . The green color represents the region in which cells can be trained to resist higher stress levels through a slow ramping process . The magenta region highlights the shift in survival threshold obtained following a pretreatment according to the nonlinear model . A , D–F , error bars and shaded regions are SEM ( N > 100 for most time points ) . B , error bars are 95% CI . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 01810 . 7554/eLife . 23971 . 019Figure 4—figure supplement 1 . Acquisition of tolerance: comparison of the linear and nonlinear models . ( A ) The mean cellular intensity of the phase-contrast highest decile is quantified for a population of cells exposed to critical ΔI = 1 mM ( black line ) after pretreatment of I0 = 0 . 2 mM and for control population of cells only exposed to the pretreatment of I0 = 0 . 2 mM ( magenta line ) . ( B–C ) Same as Figure 4A–B , but with the yap1Δ mutant . ( D ) Same as Figure 4E , but with various levels of H2O2 during pretreatment and with a challenging step of ΔI = 0 . 1 mM H2O2 . The Yap1 maximal amplitudes from this graph are represented in Figure 4F . ( E ) Numerical simulation of the response of the H2O2 homeostatic machinery to a sequence of two consecutive stress steps of indicated amplitude ( top ) , using the linear feedback model . The first step ( I0 ) corresponds to the pre-treatment , while the second represents the challenging step ( ΔI ) described in Figure 4 . Each colored line corresponds to a particular temporal profile of H2O2 concentration . H ( Middle ) and A ( Bottom ) are two variables of the model described in Figure 3 and Materials and methods . ( F ) Same as ( E ) , but for the nonlinear feedback model . ( A , B , D ) Error bars and shaded regions are SEM ( N > 100 for most time points ) . ( C ) error bars are 95% CI . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 019 To explain this phenomenon of acquired stress tolerance , we first hypothesized that the pretreatment may switch the cell to an activated/adapted state capable of a much quicker transcriptional response to the subsequent challenge , as proposed previously in the context of salt cross-tolerance ( Guan et al . , 2012 ) . To test this hypothesis , we monitored the rate of TRX2pr-GFP-deg accumulation upon exposure to a range of H2O2 steps and found a lower transcription rate of the TRX2 promoter during the challenging step of magnitude ΔI = 0 . 6–0 . 8 mM than during the pretreatment of magnitude I0 = 0 . 2 mM ( Figure 4D ) , thus ruling out the hypothesis of quicker and/or stronger transcriptional reactivation of the homeostatic machinery . As an alternative hypothesis , we reasoned that stress preconditioning might increase the H2O2-buffering efficiency ( through higher scavenging rate ) leading to a lower transient internal H2O2 level upon exposure to the subsequent stress challenge . In line with this , we found that the amplitude of the burst in Yap1 nuclear relocation decreased as a function of the pretreatment level I0 ( Figure 4E and F , and Figure 4—figure supplement 1D ) , consistent with the lower transcriptional activation of its effector gene TRX2 . This indicated that the pretreated cells perceive a lower internal stress level than do naive cells . To quantitatively account for these observations , we sought to refine the mathematical description of the homeostatic system . According to the linear model , the scavenging rate depends only on the concentration of antioxidant enzymes A , meaning that H2O2-scavenging enzymes would always be saturated by the H2O2 substrate following stress exposure . If , instead , we consider that the enzymes are sufficiently abundant or the internal H2O2 level is sufficiently low that enzyme saturation does not systematically occur , then the scavenging rate in the model becomes a nonlinear function of the two variables A and H ( see Materials and methods ) . Consequently , stress pretreatment may drive the homeostatic system to an equilibrium state in which the upregulated enzymes not only function to counteract the existing H2O2 flux but may also contribute with no delay ( i . e . before any transcriptional response ) to the scavenging of a future stepwise H2O2 exposure . Thus , unlike the linear model , this nonlinear model was able to quantitatively recapitulate the clear I0-dependent reduction in peak internal H2O2 during the challenge step ( Figure 4—figure supplement 1E and F ) , the magnitude of which was similar to the experimentally observed Yap1-GFP nuclear relocation ( Figure 4F and Figure 4—figure supplement 1D ) . Finally , computing the phase diagram for the nonlinear model revealed that , whereas buffering of slow external fluctuations in H2O2 levels ( i . e . through cellular training ) is a generic property of homeostatic systems based on negative feedback loops , the adaptation to fast fluctuations in the external stressor levels following stress preconditioning ( i . e . through acquisition of stress tolerance ) , is a distinct property that requires a specific nonlinear scavenging model ( Figure 4G ) . Thus far , our framework has made no assumptions regarding the nature of the scavenging enzyme ( s ) responsible for H2O2 degradation . Therefore , we next sought to identify which of the Yap1 regulon effectors ( Godon et al . , 1998 ) are critical for H2O2 homeostasis . During the step experiments , adaptation could result from parallel protective and repair mechanisms ( DDR , protein quality control , metabolic control , H2O2 scavenging ) . However , the ramp experiments , by eliminating the cellular response triggered by high transient H2O2 levels , provided a unique framework to specifically decipher the core genes of the H2O2 homeostatic machinery . To address this , we examined the growth rates of various mutants at 300 and 800 min after initiation of a stress ramp of δ = 1 . 1 μM/min ( Figure 5A ) . Deletion of Yap1 abolished adaptation ( Figure 5B ) with onset of growth arrest occurring at ~0 . 1 mM H2O2 ( Figure 5C ) , similar to the threshold observed in step experiments ( Figure 1G ) . The complete absence of ‘trainability’ of the Yap1 mutant contrasted with the efficient adaptation of the msn2Δmsn4Δ mutant ( Figure 5B ) , which lacks the transcription factors involved in the general stress response . Similarly , mutants lacking enzymes involved in membrane lipid biosynthesis , erg3Δ and erg6Δ , adapted perfectly ( Figure 5B ) , thus ruling out the possibility that reduced membrane permeability is responsible for adaptation to H2O2 ramps ( Branco et al . , 2004 ) . 10 . 7554/eLife . 23971 . 020Figure 5 . Genetic determinants of adaptation to ramped increases in H2O2 . Quantification of mean growth rate upon exposure to a linear ramp ( δ = 1 . 1 μM/min ) starting at t = 100 min in various genotypes . ( A ) Illustration of the H2O2 ramp experiment indicating the timing of the measurements . ( B ) Stress response and membrane permeability mutants . ( C ) Details of the ramp experiment in the Δyap1 mutant . The dashed blue lines indicate the adaptation threshold obtained in step experiments ( Figure 1G ) . ( D ) Yap1 effectors mutants . ( E ) Step experiment performed with Yap1 effectors mutants exposed to 0 . 4 mM H2O2 . ( F ) Prxs mutants . ( G ) Schematic of a negative feedback control showing the essential role of Prxs in the H2O2 homeostasis . ( H ) Mutants affecting the peroxidatic cycle of Prxs . Error bars are SEM ( N > 100 ) . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 02010 . 7554/eLife . 23971 . 021Figure 5—figure supplement 1 . Quantification of Tsa1 expression from the ACT1 promoter . ( A ) Phase and GFP fluorescence image samples of cells carrying a Tsa1-GFP fusion ( top ) or an ACT1pr-TSA1-GFP fusion ( bottom ) . ( B ) Quantification of mean cell cytoplasmic fluorescence for strains described in ( A ) Error bars are SEM ( N > 100 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 021 Interestingly , we found that deletion of known H2O2 scavengers , such as the mitochondrial cytochrome c peroxidase Ccp1 or the cytosolic and peroxisomal catalases Ctt1 and Cta1 , did not contribute to adaptation ( Figure 5D ) . Since these mutants were previously described to be hypersensitive to H2O2 ( Jiang and English , 2006 ) , we wondered whether they are required for adaptation to H2O2 steps . Indeed , growth recovery was delayed in the ccp1Δ mutant ( but not the ctt1Δcta1Δ mutant ) compared with wild-type cells following exposure to a single dose of 0 . 4 mM H2O2 ( Figure 5E ) . This finding indicates that some H2O2 scavengers may actively contribute to the transient stress response even if they are not implicated in overall H2O2 homeostasis . In contrast , simultaneously deleting the three peroxiredoxin genes TSA1 , TSA2 , and AHP1 , which encode the two yeast 2-Cys Prxs and the atypical Prx Ahp1 ( tsa1Δtsa2Δahp1Δ ) , abolished cell growth whether the cells were exposed to H2O2 as ramps or as steps ( Figure 5D and E , respectively ) . Additionally , the extent of adaptation was proportional to the number of cytosolic Prxs present ( Figure 5F ) . Constitutive expression of Tsa1 under the ACT1 promoter ( albeit less effective than the endogenous promoter , Figure 5—figure supplement 1A and B ) in a tsa1Δ tsa2Δ background did not complement Tsa1 function , indicating the importance of the H2O2-dependent transcriptional induction of the corresponding effector genes ( Figure 5F ) . Altogether , these observations demonstrate that Prxs are the essential antioxidants ensuring H2O2 homeostasis ( Figure 5G ) . 2-Cys Prxs are moonlighting enzymes that reversibly switch their function from H2O2 scavengers to chaperones upon hyperoxidation of their peroxidatic cysteine ( CP ) and reduction of this form by sulfiredoxin ( Srx1 ) ( Biteau et al . , 2003; Jang et al . , 2004 ) . To determine which of the two Tsa1 functions is involved in adaptation to stress ramps , we first tested a mutant lacking the two cytosolic Trxs ( trx1Δtrx2Δ ) that assist Prxs in H2O2 scavenging , but not in protein quality control ( PQC ) . We observed that adaptation was severely compromised in this strain ( Figure 5H ) . Next , we examined a strain lacking Tsa2 and carrying a TSA1 mutation that impairs the peroxidatic cycle of Prx but not its function in PQC ( tsa1C171Stsa2Δ ) ( Hanzén et al . , 2016 ) , and found that this strain also adapted poorly to H2O2 ( Figure 5H ) . However , a strain lacking Tsa2 and carrying a TSA1 mutation that prevents hyperoxidation and specifically impairs the enzyme’s PQC function ( tsa1ΔYFtsa2Δ ) displayed a wild-type adaptation response ( Figure 5H ) . Lastly , we tested the Δsrx1 strain , in which the defective reduction of hyperoxidized Tsa1 and Tsa2 severely impairs H2O2 scavenging in classical techniques ( Biteau et al . , 2003 ) and which Prx-mediated PQC is also defective ( Hanzén et al . , 2016 ) . Surprisingly , the Δsrx1 strain displayed wild-type adaptation to H2O2 ( Figure 5H ) . Taken together , this genetic analysis indicated that the peroxidatic , not the chaperone , function of 2-Cys Prxs is required for adaptation to stress ramps . The dispensability of Srx1 for adaptation suggests that the ramp protocol allows the cell to maintain internal H2O2 at low levels , thereby preventing Tsa1 and Tsa2 hyperoxidation . The mathematical model predicts that if Prx enzymes are the essential mediators of adaptation to H2O2 , we would expect to observe strong and stable upregulation of these proteins upon exposure to H2O2 stress ( Figure 4—figure supplement 1F ) . Indeed , a sustained increase in cytoplasmic Tsa1-GFP levels was observed upon exposure to a 0 . 4 mM H2O2 step ( Figure 6A ) . This upregulation was accompanied by formation of fluorescent foci , as noted in previous studies with Tsa1-GFP ( Hanzén et al . , 2016; Weids and Grant , 2014 ) . However , from the quantitative analysis of the dynamics of Tsa1-GFP protein upregulation ( Figure 6B and C ) and transcriptional activation ( Figure 6—figure supplement 1A and B ) upon H2O2 stress , we found that steady-state Tsa1 levels did not scale linearly , especially at low H2O2 concentrations ( Figure 6C inset ) . Similarly , the scaling of Tsa1-GFP expression during a ramp experiment ( δ = 1 . 1 μM/min ) was sublinear ( Figure 6D ) . These observations were in very good agreement with the nonlinear model , further ruling out the linear model ( Figure 6B–D ) , and the steady-state Yap1-GFP level was also best fit using the nonlinear model ( Figure 1F ) . These results suggest that the H2O2 scavenging capacity of the homeostatic system becomes increasingly more efficient with the accumulation of Prxs as the stress level increases , so that Prxs do not need to be upregulated in proportion to the external H2O2 concentration . 10 . 7554/eLife . 23971 . 022Figure 6 . Tsa1 scaling properties . ( A ) Sequence of phase-contrast and fluorescence images of individual cells at the indicated times after initiation ( t = 300 min ) of a 0 . 4 mM step in H2O2 concentration . The green channel represents the Tsa1-GFP signal . The white bars represent 5 µm . ( B ) Left: Mean cell expression ( top ) of Tsa1-GFP following the H2O2 steps indicated in the bottom panel . Right: Dynamics of antioxidant level with increasing stress , as expected from the nonlinear model . ( C ) Quantification of mean cell expression of Tsa1-GFP at steady state as a function of H2O2 concentration ( from experiments in ( B ) ) . Colored lines indicate the fit of the linear ( red ) and nonlinear ( magenta ) models . Inset: log-log representation of Tsa1-GFP level with H2O2 level . Green lines indicate lines of slope one on a log-log scale , to emphasize the nonlinearity of Tsa1-GFP expression . ( D ) Mean Tsa1-GFP expression ( top ) during a ramp experiment , as indicated in the bottom panel . Colored lines indicate the fit of the linear ( red ) and nonlinear ( magenta ) models . ( E ) Top: Mean cellular transcription of the TRX2 promoter for cells exposed to the temporal H2O2 profiles described in the middle panel ( with corresponding color coding ) . Bottom: Quantification of maximal transcriptional output during a step experiment performed at t = 200 min ( early stress ) or 500 min ( late stress ) with or without pretreatment . Student’s t-test . ( F ) Mean cellular expression of Tsa1-GFP for cells exposed to the temporal H2O2 profiles described in the bottom panel . Error bars and shaded regions are SEM ( B , D-F: N > 100 for most time points , C: N > 100 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 02210 . 7554/eLife . 23971 . 023Figure 6—figure supplement 1 . Analysis of the TSA1 promoter activity during H2O2stress . ( A ) Dynamics of the mean transcriptional response of TSA1 promoter ( TSA1pr-sfGFP-deg ) in step experiments at indicated H2O2 concentrations . ( B ) Quantification of mean cell expression of TSA1pr-sfGFP-deg at steady state as a function of H2O2 concentration ( from experiments in ( A ) ) . ( C ) Quantification of the maximal transcription rate of TSA1 promoter during a step experiment , as a function of the amplitude of the stress . The transcription rate was calculated by fitting a line over a 18-min time window to the data reported in ( A ) and by determining the maximal slope . Error bars and shaded regions are SEM ( A ) N > 100 for most time points , ( B and C ) N > 100 . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 023 We reasoned that , in this scenario , the stress-resistant state following pretreatment should quickly disappear as Prx enzymes are diluted upon removal of H2O2 . To test this , we pretreated the cells with a 100 min pulse of 0 . 2 mM H2O2 and examined the transcriptional response ( TRX2pr-sfGFP-deg ) upon challenge 100 or 400 min later ( Figure 6E ) . Notably , the transcriptional output of cells challenged 400 min after pretreatment was identical to that of cells exposed to H2O2 for the first time , indicating that the cells had returned to a naïve state by 400 min ( Figure 6E upper panel: green vs red lines , and histogram ) . In contrast , TRX2pr-sfGFP transcription was lower in cells challenged 100 min after pretreatment compared with naive cells ( Figure 6E , blue vs pink lines , and histogram ) . In parallel , we also observed that Tsa1-GFP levels were comparable to the basal level after a 400-min recovery period , but not after 100 min ( Figure 6F ) , suggesting that Prx levels are tightly associated with the H2O2 buffering efficiency . The nonlinear scaling of Tsa1 expression upon exposure to H2O2 may prove beneficial to cellular physiology in general , particularly during replicative aging . In support of this , the extension of both chronological and replicative longevity by caloric restriction has been shown to be mediated , at least in part , by activation of H2O2-dependent genes ( Mesquita et al . , 2010; Molin et al . , 2011 ) . Furthermore , recent work has shown that Tsa1 plays a role in processing of age-related protein aggregates , and overexpression of Tsa1 alone increases longevity through a mechanism involving the PQC machinery ( Hanzén et al . , 2016 ) . To explore how activation of the H2O2 homeostatic machinery affects longevity , we measured the replicative lifespan ( RLS ) of cells exposed to various doses of H2O2 . For this , we developed a microfluidic device that allows individual cells with different genetic backgrounds to be tracked microscopically from birth to death in separate channels ( Figure 7A and B and Video 6 ) . This is similar to our previously described device ( Fehrmann et al . , 2013 ) , except that the large increase in capacity allows cells from up to 10 mutant strains to be tracked in parallel . 10 . 7554/eLife . 23971 . 024Figure 7 . Hormetic effect of H2O2 on replicative longevity . ( A ) Sketch of the microfluidic device used for replicative aging experiments . ( B ) Sequence of overlaid phase-contrast and fluorescence images ( Htb2-sfGFP , used as a nuclear marker to score cell division ) of mother cells growing in individual cavities . Numbers indicate the timing ( white ) and number of cell divisions ( yellow ) for the mother cell at the tip of the cavity . ( C ) Survival curves for wild-type cells growing in media containing H2O2 at the indicated concentrations . ( D ) RLS as a function of H2O2 concentration . Box represents median and 95% CI . U test . Red line shows the median RLS for 0 mM H2O2 . ( E ) Survival curves of Δyap1 and Δtsa1 mutants in the presence or absence of 10 μM H2O2 . ( F ) Frequency of specific fluorescence foci ( Ddc2-GFP , Hsp104-GFP ) as a function of H2O2 concentration . Error bars are 95% CI . ( G ) ( Top to bottom ) Recapitulation of measurements of RLS ( blue ) , Tsa1-GFP steady-state upregulation ( green ) , and frequency of damage ( DDC2 foci in brown , Hsp104 foci in purple ) . Bottom: Conceptual sketch showing the contributions of protective ( black ) and deleterious ( red ) effects of H2O2 on RLS . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 02410 . 7554/eLife . 23971 . 025Video 6 . Replicative lifespan monitoring ( refers to Figure 7 ) Movie showing the entire lifespan of a cell from birth to death in a cavity of PDMS device . The white bar represents 5µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23971 . 025 Using this technique , we observed that H2O2 had a biphasic effect on RLS . Exposure to 10 and 25 μM H2O2 significantly increased RLS up to ~25% compared with unstressed cells ( median RLS generations of 26 , 33 , and 32 at 0 , 10 , and 25 μM H2O2 respectively ) , whereas concentrations > 50 μM H2O2 decreased RLS ( median RLS of 25 and 20 generations at 50 and 100 μM H2O2 , respectively; Figure 7C and D ) . This suggested that H2O2 had a hormetic effect , in which low doses of H2O2 were beneficial and improved RLS , whereas higher doses had deleterious consequences on longevity ( Ristow and Schmeisser , 2011 ) . We next determined whether this effect was dependent on activation of the H2O2 homeostatic core machinery . Deletion of Yap1 or Tsa1 abolished the increase in longevity at low H2O2 doses ( Figure 7E ) , implicating these proteins in the hormetic effect . We hypothesized that hormesis results from an unbalance between the substantial upregulation of Tsa1 observed at low H2O2 levels and the production of deleterious cellular damages . Indeed , the fraction of cells with foci of Ddc2-GFP ( DDR marker ) and Hsp104-GFP ( protein aggregate marker ) was similar to the unstressed control up to 25 μM H2O2 ( Figure 7F and G ) , while Tsa1 level experienced a ~ 25% increase in this range of concentration ( Figure 6B and C ) . Above 25 μM H2O2 , however , the increase in Tsa1 expression was relatively lower than the increase in foci ( Figure 7F and G ) . Collectively , these data suggest that nonlinear activation of the homeostatic machinery in response to H2O2 leads to a biphasic effect on replicative longevity as recapitulated on Figure 7G .
In the present study , we combined microfluidics technology with yeast genetics and live-cell imaging to perform a comprehensive analysis of the complex adaptive properties of budding yeast to oxidative stress . The decisive element of our methodology was the precise temporal control of oxidative stress levels , without which the detailed mechanisms of adaptation to H2O2 stress could not have been quantitatively addressed . Indeed , it is likely that many of the discrepancies observed in studies of cellular resistance to H2O2 using classical bulk techniques originate from both the lack of environmental control and the population size-dependent consumption of the stressor by the cells growing in a test tube or on a plate . Our methodology provides a rationalization of H2O2 resistance assays , in which the final H2O2 concentration and its rate of increase share equal importance in determining cell fate . To date , only a few studies have used linear ramps ( Muzzey et al . , 2009; Sorre et al . , 2014; Young et al . , 2013 ) or other time-varying stimuli ( Castillo-Hair et al . , 2015 ) to decipher how temporal patterns of specific inputs govern activation of a regulatory network and the determination of cell fate . We anticipate that such precise dynamic environmental control will become increasingly important for refining our understanding of information processing by signaling networks . The use of customized temporal stress patterns appears to be a unique methodology for unraveling the specific functional role of genes involved in H2O2 homeostasis . Our analysis has not only identified the peroxiredoxins coding genes TSA1 , TSA2 , and AHP1 as key elements in ensuring adaptation to stress ramps but also revealed that other antioxidant genes , such as CCP1 , CTT1 , and CTA1 , are not required under these conditions . However , CCP1 does contribute to growth rate recovery in the step experiments . These data , therefore , suggest that Prx activity is the major determinant of H2O2 homeostasis and is essential regardless of the temporal stress pattern , whereas CCP1 may only contribute to detoxification during the high transient H2O2 peak that accompanies acute stress . Future functional studies will be instrumental in dissecting the relative contributions of all H2O2-scavenging enzymes . Beyond the necessary accuracy of a homeostatic control mechanism , our study stresses that ‘trainability’ is another essential functional property intimately linked to the response time of the underlying regulatory mechanism . Training has important physiological implications; namely , that cells can be acclimated to much higher stress levels ( >10 fold ) when the rate of increase is low compared with acute exposure . The quantitative framework developed here formalizes the key parameters , such as enzyme dilution rate , that control the extent of trainability in a homeostatic system based on transcriptional regulation . In addition , the framework suggests that there is an optimal ( i . e . , fastest ) stress pattern ( following the Hmax = Htox contour line on Figure 3C ) for driving a homeostatic system to its maximal stress-buffering capability . Interestingly , the ability to resist progressively increasing environmental insults resembles the concept of mithridatism , which was originally described as a defense strategy against poisoning , but which has lacked overall biological relevance due to the absence of similar mechanistic evidence ( Valle et al . , 2012 ) . In the context of H2O2 stress , our study identifies a new form of mithridatism at the cellular scale that could be quantitatively explained by a mathematical description of a generic negative feedback-based regulatory network . Therefore , we envision that this framework could be transposed to other homeostatic systems and , in particular , may find potential technological applications in which improved cellular adaptation is desired without any genetic modification , such as environmental detoxification and chemotherapy . An important outcome of our study is the establishment of a clear distinction between training and acquired stress resistance . Although both mechanisms provide a way to increase basal cellular stress resistance , their functional properties and their underlying mechanisms are completely different . Unlike training , which enables increased adaptation to slow stress fluctuations , acquired stress resistance elicits improved survival of preconditioned cells to rapid environmental changes . This latter phenomenon cannot be described using a linear model but can be explained by nonlinearity in the control loop , in which the H2O2 degradation rate depends on both the level of scavenging enzymes and internal H2O2 concentrations . Even at basal levels , Prxs are very abundant proteins ( ~1% of the total dry cellular mass in mammals , [Low et al . , 2007] ) , and our data show that their expression increases up to fivefold upon exposure to H2O2 . Therefore , it is likely that the cellular concentration of the reduced form of these enzymes always surpasses the internal H2O2 concentration , and , consequently , the main hypothesis of our nonlinear model is that the H2O2 scavenging rate of Prxs is limited by the internal H2O2 concentration . Further biochemical studies of the entire peroxidatic cycle will be necessary to understand how and why the Prx scavenging rate depends on internal H2O2 levels in vivo , since other Yap1-dependent effectors , such as thioredoxins and sulfiredoxins , control the pool of reduced Prxs upon stress exposure . In the meantime , we speculate that the reason redox homeostasis relies on these extremely abundant stoichiometric enzymes rather than on catalases ( which catalyze the dismutation of H2O2 ) , as shown here , is that nonlinearity in the scavenging rate has important physiological consequences , such as stress tolerance . The phenomenon of acquired stress resistance illustrates the plasticity of the homeostatic machinery and shows that cell survival threshold is not static but may be improved by previous environmental conditioning , following the principles of adaptive homeostasis ( Davies , 2016 ) . A unique feature of our study is to provide a framework to quantitatively explain the origin of this anticipatory behavior , as well as to link this phenomenon to the nonlinearity of stress doses responses ( Zhang and Andersen , 2007; Zhang et al . , 2010 ) . Previous work has pointed to the existence of cross-protective effects , whereby exposure to a particular stress ( e . g . heat , ethanol ) increases tolerance to subsequent oxidative stress ( Mitchell et al . , 2009 ) . Since Prx genes are also regulated by the transcription factors of the general stress response , Msn2/Msn4 , we envision that the induction of tolerance to H2O2 by other stressors may be mediated by the same principles as those described here . More generally , it will be interesting to see how the proposed mechanism of tolerance applies to other homeostatic systems in which nonlinear stressor degradation might be conserved ( Zhang and Andersen , 2007 ) . The last facet of nonlinear feedback regulation that emerges from our quantitative study is related to the concept of hormesis ( Calabrese et al . , 2015 ) . Due to the substantial upregulation of Prxs in response to very low doses of H2O2 , the tradeoff between beneficial and deleterious effects of H2O2 leads to a biphasic effect on cellular longevity as a function of H2O2 level . The effect of H2O2 on longevity provides a striking model of lifespan extension induced by environmental perturbation , similar to the effects of caloric restriction . Interestingly , this observation not only complements the previously described hormesis-like effect of H2O2 on chronological aging ( Mesquita et al . , 2010 ) but also suggests a simple mechanistic explanation in which the H2O2 homeostatic machinery plays a central role . Therefore , our study emphasizes that complex functional properties , such as acquired stress tolerance and hormetic lifespan effects , can only be understood by developing specific approaches that bridge the molecular architecture of regulatory pathways to its quantitative dynamical properties .
All strains were congenic to S288C ( Sikorski and Hieter , 1989; Huh et al . , 2003 ) unless specified . Strains generated in this study were made using standard genetic techniques or classical PCR-mediated genome editing . See Supplementary file 1 for the list of strains used in this study . The transcriptional reporter strains TSA1pr-sfGFP-deg , TRX2p-sfGFP-deg were generated by a one-step cloning-free method ( Huber et al . , 2014 ) . The promoters of TSA1 , TRX2 were duplicated in strain SY992 and a sfGFP-degron ( superfolder GFP fused to CLN2 PEST sequence ) tag was simultaneously inserted downstream of them . The genes TSA1 , TRX2 remain fully intact . The sequences of homology ( used to design the primers ) of the PCR products that duplicated the promoters at their genomic locations in vivo are listed below: Sequences of homology used in PCR primers for promoter duplication in vivo Target locusSequence of homology to the target locusTSA1pr-1TTCCCCTCGTTCAATTGCTCACAACCAACCACAACTACATACACATACATACACATSA1pr-2CCTTGATCTGGCTAAACTTGACTTCGTCAATTTCATTCAAGTGGAGATAGTCTCGTRX2pr-1TTATACACGCACACATACACGAGAGTCTACGATATCTTTAAATAACACATCAATATRX2pr-2TGGATCATGGGCGCATGTGAACCGTACCCACCGAATTGCGCTTGAAGTGTGTCCA The strain tsa1Δ; tsa2Δ; ACT1pr::TSA1-GFP was generated by substituting the promoter region of TSA1 by the promoter of ACT1 in strain TSA1-GFP ( BY4741 ) . The cassette kanMX4::ACT1pr was amplified by PCR and then transformed into the strain TSA1-GFP ( BY4741 ) using standard lithium acetate protocol . The primers contained sequences homologous to the TSA1 genomic locus that are listed below: Sequences of homology used in PCR primers for promoter substitution Target locusSequence of homology to the target locusTSA1-1CCACCGCCACAGGTGCGCAACCTCATCTCTACATTCCTGATGAAGACTAATSA1-2ACGGCAGTTTTCTTAAAAGTTGGAGCTTGCTTTTGAACTTGAGCGACCAT To introduce tsa2Δ mutation , the transformed strain was crossed with strain Y14287 . The H2O2 ( Hydrogen peroxide solution 35wt . % in H2O , 349887–500 ML , Sigma ) was mixed in the SCD media at the suitable concentration prior the microfluidic experiment . In order to increase the stability of the H2O2 during the microfluidic experiments , the SCD media were kept on ice . Temperature measurements showed that this procedure does not affect media temperature in the microfluidic device since media reached room temperature before entering the device ( data not shown ) . The H2O2 concentration was measured using a colometric H2O2 assay kit ( OxiSelect Hydrogen Peroxide Assay Kit ( Colorimetric ) , STA-343 , EUROMEDEX , France ) in media samples taken from the outlet of the chip . Dosage experiments were run in triplicates . Over-night SCD-media culture ( strain WT BY4742 ) was diluted in SCD to OD600 = 0 . 1 and incubated at 30°C , 220 rpm . At OD600 = 0 . 5 , H2O2 was added to final concentration of 0 . 4 mM . Consumption of the H2O2 was determined by measuring the H2O2 concentration in the cell media over time . Bulk H2O2 consumption experiments were performed six times . Microfluidic chips were designed and made using standard techniques as previously described ( Fehrmann et al . , 2013 ) . The microfluidic master used to assess the adaptation to oxidative stress was made using standard SU-8 lithography process at the ST-NANO facility of the IPCMS ( Strasbourg , France ) . The microfluidic master for aging studies was made using similar techniques in the FEMTO-ST nanotechnology platform of the French Renatech network ( Besançon , France ) . Prototypic molds were replicated in epoxy to ensure long-term preservation . The micro-channels were cast by curing PDMS ( Sylgard 184 , 10:1 mixing ratio ) and then covalently bound to a 24 × 50 mm coverslip using plasma surface activation ( Diener , Germany ) . Microfluidic chips were connected using Tygon tubing and media flows were driven by a peristaltic pump ( Ismatec , Switzerland ) with a 30 μL/min flow rate . Media switches -synthetic complete 2% dextrose ( SCD ) , with or without H2O2 at the appropriate concentration- were performed using a computer-controlled electro-valve ( Biochemfluidics ) or using a media multiplexer ( Elvesys ) . Linear ramps of medium containing H2O2 were achieved using a custom setup allowing a progressive increase in H2O2 concentration in the main medium tank . For aging experiments , cells were maintained in a chip during typically 140 hr with constant medium perfusion ( flow rate 10 μL/min ) . Cells started to enter the cavity about 20 hr following their loading in the device . In case H2O2 was used throughout the assay , fresh medium was prepared every 24 hr to prevent any decay in H2O2 concentration over time . The diffusion of the media in the trapping cavities was tested using a fluorescein dye assay . To do this , fluorescein ( Sigma ) was flown trough the microfluidics device either in the absence or in the presence of a dense microcolony of cells and fluorescence images were snapped every 2 s . The fluorescence was measured both in the supply channel at the border of the trapping cavity and in the middle of the trapping cavity . As expected , the fluorescence in the supply channel increased almost instantaneously as shown in Figure 1—figure supplement 1 , whereas the increase of the fluorescence in the middle of the empty trapping cavity displayed a delay ( half rising time t1/2 = 21 s ) . A comparable time ( t1/2 = 24 s ) was obtained in the presence of a dense microcolony of cells , hence showing that diffusion is not impaired in a crowded environment . All time-lapse experiments have been performed at least two times . Freshly thawed cells were grown overnight at various final cell densities . In the morning , log phase cells were allowed to grow a few divisions and were transferred into the microfluidic device . Cells were imaged using an inverted Zeiss Axio Observer Z1 ( adaptation assay ) or a Nikon Tie ( aging experiments ) . Focus was maintained using dedicated hardware throughout the assays . Fluorescence illumination was achieved using LED light ( precisExcite , CoolLed or Lumencor ) and light was collected using a 100× N . A . 1 . 4 objective and an EM-CCD Luca-R camera ( Andor; adaptation experiments ) or an Hamamatsu Orca Flash 4 . 0 ( Aging experiments ) . We used automated stages in order to follow up to 20 ( adaptation experiments ) or 60 ( aging experiments ) positions in parallel over the course of the experiment . Images were acquired every 3 min ( adaptation experiments ) or 10 min ( aging experiments ) . Temperature control was achieved using custom sample holder with thermoelectric modules and an objective heater with heating resistors . Temperature control was achieved using a PID controller ( 5C7-195 , Oven Industries ) . Raw images were processed using custom software , called phyloCell , based on MATLAB and the image-processing toolbox ( Fehrmann et al . , 2013; Paoletti et al . , 2016 ) . This software features a comprehensive graphical user interface to perform segmentation/tracking and to introduce manual error corrections . The software is available for download on GitHub ( Charvin , 2017 ) . A copy is archived on https://github . com/elifesciences-publications/phyloCell . In this study , the software was used to segment cell contours based on phase-contrast images; to track cells over time; and to measure the fluorescence within the cells , including nuclear localization of fusion proteins . After segmentation of cell contours from time-lapse data , the volume V of individual cells was estimated by spherical approximation of the cellular shape ( Figure 1—figure supplement 2A ) . The volume increase rate ( which we refer to as cellular growth rate ) was then determined for each time point as the increase in the cell volume in consecutive frames per unit of time ( cell growth rate at t1 = ( V ( t2 ) -V ( t1 ) ) / ( t2-t1 ) , where t2-t1 = 3 min ) . More specifically , the measurements of mean growth rate per cell used throughout this study ( which reflects the metabolic capacity of a population of cells , see below ) were measured by averaging the volume increase rate of individual budded cells over large micro-colonies of cells , whereas unbudded cells were not considered in this analysis . This is motivated by two reasons: first , the increase in cell volume is higher during the budded phase of the cell cycle , and therefore , it provides a more robust estimate of cell growth . Second , the purpose of this measurement is to provide a readout of cellular metabolic activity , independently of cell cycle progression . When cells are exposed to H2O2 , the fraction of budded cells increases due to the activation in some cells of a G2/M checkpoint , but this cell cycle arrest is not necessarily accompanied by a change in metabolic activity . Therefore , if we were pooling the growth rate of both unbudded ( slow volume increase ) and budded cells ( fast volume increase ) , the change in the repartition of each type of cells upon H2O2 exposure would induce a misleading change in the mean growth rate per cell , which would be unrelated to a potential change in metabolic activity . Instead , selecting only the budded cells provides a relevant assessment of cell growth rate , which is unaffected by the inherent variations in cell cycle progression during stress exposure ( Figure 1—figure supplement 2B ) . Noteworthy , the definition of growth rate used in our study ( dV/dt ) differs from the classical one , which is given by μ= ( 1/V ) x dV/dt . The reason why we used the simple first time derivative of cell volume is because the exponential growth model appeared to be irrelevant to assess the change in metabolic activity during stress exposure . Indeed , the growth rate of cells ( defined as dV/dt ) appeared to be independent of cell volume ( Figure 1—figure supplement 2C ) . In addition , using dV/dt as readout of growth rate , we observed a complete recovery following the exposure to stress , suggesting that cells had recovered a normal metabolic activity . However , using μ= ( 1/V ) x dV/dt as the definition of growth rate , we would have missed the recovery in metabolic activity , since the mean cell volume of adapted cells following the recovery was higher than that before the stress ( Figure 1—figure supplement 2D ) , due to cell cycle arrested – yet metabolically active-cells . Following the segmentation of the nucleus using the Htb2-mCherry signal ( Figure 1D ) , nuclear Yap1-GFP localization was measured at each frame by subtracting the average cytoplasmic fluorescence level from the mean nuclear level in order to remove the background . In addition , we noticed that there was a drift in the fluorescence level over time , which , for unclear reasons , appeared to mostly affect the cytoplasmic level . Therefore , we used a no-stress control experiment , in which Yap1-GFP nuclear localization was supposedly constant , to measure the extent by which background subtraction modifies the quantification of the nuclear localization over time . This measurement was then used to correct the drift observed in experiments in which a stress was applied . Media containing 5 µg/ml PI ( Sigma , Saint-Louis , MO , USA ) was flown through the microfluidics device during a 0 . 6 mM H2O2 step experiment , and images were recorded continuously . Dead cells , which experienced a loss of cell membrane integrity , incorporated the dye and displayed red fluorescence , as reported in Figure 2—figure supplement 1 . We used the presence of persistent bright vacuoles as a readout to characterize the cells experiencing a permanent growth arrest following a H2O2 stress . To quantify this , we measured the mean intensity of the highest decile in the phase-contrast channel of the images . In order to generate ramps of H2O2 stress , we used an extra peristaltic pump driving the flow of a H2O2 stock solution to gradually increase the H2O2 concentration in the medium tank used to feed the microfluidic device ( Figure 3—figure supplement 2A ) . Assuming C1 and V1 , the concentrations and volume of the medium tank , respectively , μ1 the flow rate used to perfuse the microfluidic device from the tank , C0 and μ0 the concentration and flow rate from the H2O2 stock solution , using the conservation of mass , one can derive the evolution of V1 and C1 with time:dV1dt= μ0−μ1dC1dt= μ0V10+ ( μ0−μ1 ) t ( C0−C1 ) In the particular case of identical flow rates ( μ0= μ1 ) , V1 is constant and the evolution of C1 with time is given by:C1=C0 ( 1− eδt ) where δ=μ0V10 . When t << 1/δ , then the concentration C1 increases linearly as:C1= C0 δ t With μ0 = 30 μL/min and V10 = 1L , δ = 3 . 10−5 min−1 . Therefore , we should get a linear slope of 3 μM/min using C0 = 100 mM , provided t << 105 min , which is much higher than the duration of experiments ( typically 103 min ) . However , H2O2 dosages during the calibration of stress ramp experiments showed that the actual concentrations were systematically 21 . 5 ± 2% lower than expected for all tested time points ( 0–1000 min ) and ramp slopes ( 1 . 4–22 . 4 µM/min ) ( data not shown ) . To explain the origin of this deviation from the expected measurements ( which stands higher than the 10% decay observed during step experiments , see Figure 1C ) , we hypothesize that it can be attributed to a higher H2O2 degradation rate in the medium tank during ramp experiments because the tank was not kept on ice due to the need to perform constant mixing using a magnetic stirrer . Therefore , the actual evolution of concentration in the medium tank is given by:C1= 0 . 785 C0 δ t For instance , a 1 . 1 µM/min ramp slope is obtained with μ0 = 30 μL/min and V10 = 1L and C0 = 46 . 67 mM . To understand the mechanism that underlies H2O2 homeostasis , we developed a mathematical framework based on the negative feedback regulation in the Yap1 network . The model was used throughout this study to help identify and formalize the emergent properties of this system through iterative cycles of predictions and experimental challenges , rather than to perform exhaustive data fitting to retrieve individual parameter values . In the following , we describe in details the principles of the mathematical model . | Harmful external conditions , such as extreme heat or radiation , can cause stress to cells that may lead to permanent damage and even death . Cell stress is responsible for some cancers and degenerative diseases , and is involved in the process of aging . Cells respond to stress by modifying their activities in order to prevent damage from occurring . Some studies have suggested that the ability of cells to survive a stressful situation might depend both on the severity of the stress and also on the way in which the stress is applied . For example , the stress might start suddenly or develop more gradually . Cells exposed to a mild level of stress develop a tolerance that enables them to survive stronger doses of the same stress in the future . However , it is not clear how cells acquire such tolerance , and whether mild levels of stress can have more general benefits to cells , such as increased lifespan . Hydrogen peroxide and other “oxidative” compounds play important roles in cells , but they are also capable of causing damage so their levels must be tightly controlled . Goulev et al . developed a “microfluidic” device to study the effects of oxidative stress on yeast cells . The device made it possible to precisely control the level of hydrogen peroxide in the cells’ environment while monitoring the cells’ stress responses . The experiments show that exposing yeast cells to gradually increasing levels of hydrogen peroxide can train the cells to be able to survive when they are exposed to high levels of this compound . This ability depends on the activity of specific enzymes called peroxidases that are known to be able to destroy hydrogen peroxide inside the cells . The experiments suggest that gradually increasing levels of hydrogen peroxide trigger increases in the production of peroxidases that protect the cells against future oxidative stress . Further experiments show that even a very low dose of hydrogen peroxide is sufficient to activate the production of the enzymes , leading to an increase in the lifespan of the cells . A future challenge will be to investigate whether the principles identified in this work also apply to other stress responses in yeast . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2017 | Nonlinear feedback drives homeostatic plasticity in H2O2 stress response |
Neurotransmitter release is mediated by the fast , calcium-triggered fusion of synaptic vesicles with the presynaptic plasma membrane , followed by endocytosis and recycling of the membrane of synaptic vesicles . While many of the proteins governing these processes are known , their regulation is only beginning to be understood . Here we have applied quantitative phosphoproteomics to identify changes in phosphorylation status of presynaptic proteins in resting and stimulated nerve terminals isolated from the brains of Wistar rats . Using rigorous quantification , we identified 252 phosphosites that are either up- or downregulated upon triggering calcium-dependent exocytosis . Particularly pronounced were regulated changes of phosphosites within protein constituents of the presynaptic active zone , including bassoon , piccolo , and RIM1 . Additionally , we have mapped kinases and phosphatases that are activated upon stimulation . Overall , our study provides a snapshot of phosphorylation changes associated with presynaptic activity and provides a foundation for further functional analysis of key phosphosites involved in presynaptic plasticity .
In chemical synapses , neurotransmitters are stored in synaptic vesicles and released by Ca2+-dependent exocytosis upon stimulation . After exocytosis , the components of the synaptic vesicle membrane are retrieved , mainly by clathrin-dependent endocytosis , and then used to regenerate synaptic vesicles for another round of exo-endocytotic recycling ( Sudhof , 2013; Rizzoli , 2014 ) . During the past decades , the steps of the synaptic vesicle cycle have been studied in great detail . Before fusion , synaptic vesicles attach to specialized sites at the presynaptic plasma membrane ( active zones ) , which is associated with activation ( priming ) of the exocytotic apparatus . Furthermore , functionally distinct pools of synaptic vesicles coexist in each nerve terminal , with only a small proportion of them participating in recycling under normal physiological conditions ( Denker et al . , 2011 ) . Major progress has also been made in our understanding of the underlying molecular events . Synaptic vesicles contain proteins mediating neurotransmitter uptake ( vesicular neurotransmitter transporters and a vacuolar ATPase ) as well as proteins involved in vesicle docking ( Rab proteins ) , calcium sensing ( synaptotagmins ) and fusion ( SNAREs ) ( Takamori et al . , 2006 ) . Active zones contain both scaffolding proteins such as piccolo , bassoon , liprins as well as proteins that mediate vesicle docking and priming ( e . g . RIM and Munc13 ) ( Südhof , 2012 ) . Exocytotic fusion is executed by SNAREs , controlled by synaptotagmin , and regulated by additional proteins recruited from the cytoplasm such as Munc18 ( Toonen and Verhage , 2007 ) and complexins ( Brose , 2008 ) . Clathrin-Mediated Endocytosis ( CME ) is mediated by a whole array of adaptor proteins , many of which containing curvature-sensing BAR and F-BAR domains ( McMahon and Boucrot , 2011 ) such as dynamin that is crucial for pinching off of coated pits . Actin dynamics plays a critical role in vesicle movement and endocytosis ( Cingolani and Goda , 2008 ) , and additional proteins are involved in other functions such as vesicle clustering synapsins ( Cesca et al . , 2010 ) . Much less is known about how the synaptic vesicle cycle is regulated . Chemical synapses display profound plasticity and adaptation , and they are known to rapidly respond to a multitude of signals by changing their release properties . A large body of evidence shows that many presynaptic proteins are reversibly phosphorylated , suggesting that protein phosphorylation plays a key role in output regulation . In fact , synapsin1 was the first presynaptic protein to be discovered because it undergoes rapid phosphorylation upon depolarization of the presynaptic nerve terminal ( Ueda et al . , 1981; Dolphin and Greengard , 1981 ) . Additionally , the key protein in membrane fission , dynamin , is rapidly dephosphorylated by a calcium-dependent phosphatase ( calcineurin ) ( Liu et al . , 1994 ) . Dephosphorylation alters the affinity of dynamin for some of its binding partners involved in endocytosis ( Robinson et al . , 1994; Anggono et al . , 2006; Clayton and Cousin , 2009 ) . In addition to dynamin , several other proteins involved in clathrin-mediated endocytosis , termed 'dephosphins' are dephosphorylated upon depolarization , in calcium-dependent manner , including the scaffold proteins amphiphysin1 and amphiphysin2 , the PtdIns ( 4 , 5 ) P phosphatase synaptojanin , the adapter proteins epsin , eps15 , and AP180 ( Cousin and Robinson , 2001 ) . Proteins mediating exocytosis also appear to be regulated by phosphorylation-dephosphorylation reactions . For example , phosphorylation of the SNARE SNAP-25 at Ser187 accelerates vesicle recruitment ( Nagy et al . , 2002 ) , regulates formation of SNARE complex ( Gao et al . , 2016 ) and consequently promotes exocytosis in PC12 cells , and it is also induced by stress in affected brain regions ( Yamamori et al . , 2014 ) . Additionally , it was reported that phosphorylation of SNAP-25 at Thr138 interferes with assembly of SNARE complex and exocytosis ( Gao et al . , 2016 ) . Moreover , phosphorylation of syntaxin1 at Ser14 and Ser188 regulates its interaction with Munc18-1 ( Tian et al . , 2003; Rickman and Duncan , 2010 ) . Dephosphorylation of N-ethylmaleimide-sensitive factor ( NSF ) at Tyr 83 and Thr645 was reported to promote secretory vesicle fusion ( Huynh et al . , 2004 ) , probably by enhancing disassembly of SNARE complexes ( Belluzzi et al . , 2016 ) . Certain presynaptic proteins have multiple phosphorylation sites that can simultaneously undergo opposing changes in their phosphorylation states . Best characterized are the synapsins that are phosphorylated at multiple sites by several different kinases . For example , synapsin1 is phosphorylated at Ser9 by Ca2+/calmodulin-dependent protein kinase I ( CaMKI ) . It is also phosphorylated by mitogen-activated protein kinase ( MAPK ) and CdK5 at both N and C termini ( Ser62 , Ser67 , Ser549 and Ser551 ) and by CaMKII at its C terminus ( Ser566 and 603 ) ( Cesca et al . , 2010 ) . Moreover , the tyrosine-kinase Src phosphorylates synapsin1 at Tyr301 ( Onofri et al . , 2007 ) . Intriguingly , the functional consequences of synapsin phosphorylation/dephosphorylation at various phosphosites are different . While phosphorylation of Ser9 , Ser566 and Ser603 upon stimulation decreases actin binding and increases exocytosis of synaptic vesicles , phosphorylation of Tyr301 has the opposite effect ( Cesca et al . , 2010 ) . Moreover , phosphorylation at Ser62 , Ser67 , Ser549 and Ser551 is downregulated upon stimulation of neurotransmitter release and decreases the binding of synapsin to actin filaments , leading to the suggestion that phosphorylation by Cdk5 at Ser549 and 551 defines the ratio between resting and recycling synaptic vesicle pools ( Verstegen et al . , 2014 ) . In the past , synaptic phosphoproteins were mostly identified by in vitro assays using cell/tissue extracts or purified proteins . Based on this classical approach , we know that for example syntaxin1A , synaptobrevin ( VAMP ) and SNAP25 are phosphorylated by calcium/calmodulin dependent protein kinase type II ( CaMKII ) ; synaptobrevin and SNAP25 by protein kinase C ( PKC ) , and synaptotagmin1 and syntaxin1A by casein kinase II ( CK2 ) ( Bennett et al . , 1993; Nielander et al . , 1995; Hirling and Scheller , 1996; Shimazaki et al . , 1996 ) . While these classical approaches typically focus on individual proteins , modern screening methods have considerably enlarged the coverage of phosphoproteins . The introduction of chip arrays and synthetic peptides as substrates for kinases has allowed the mapping of consensus motifs for individual kinases ( Ptacek et al . , 2005 ) , which were later used to develop algorithms for predicting phosphosites in silico ( Linding et al . , 2008 ) . Commercially available kinase libraries combined with in vitro synthesis of peptides permit extended screening of predicted phosphorylation sites although it needs to be borne in mind that this method is prone to both false negatives and false positives . Most importantly , advanced mass spectrometry ( MS ) methodologies have completely changed the field as they permit the identification of phosphorylation sites in a robust , global and quantitative manner ( Huttlin et al . , 2010; de Graaf et al . , 2014; Sharma et al . , 2014; Giansanti et al . , 2015 ) . Over the last few years , progress in MS-based techniques for phosphopeptide identification and quantification has allowed global and in-depth analysis of thousands of phosphosites in a single analytical run ( Lemeer and Heck , 2009; Kelstrup et al . , 2014; Ludwig et al . , 2015 ) . These novel techniques have triggered a rapid increase in studies dealing with global phosphoproteome maps of tissues and organs , resulting , for instance , in an atlas of mouse phosphorylations ( Huttlin et al . , 2010 ) and quantitative maps of phosphorylations in rat organs ( Lundby et al . , 2012 ) . Several studies have dealt with global phosphoproteomics in the brain ( Ballif et al . , 2004; 2008; Trinidad et al . , 2008; Tweedie-Cullen et al . , 2009; Palmisano et al . , 2012; Goswami et al . , 2012; Corradini et al . , 2014; Tagawa et al . , 2015 ) or in isolated nerve terminals ( synaptosomes ) ( Munton et al . , 2007; Collins et al . , 2008; Filiou et al . , 2010; Trinidad et al . , 2012 ) . Nevertheless , we still do not have a systematic overview of phosphorylation changes in the presynaptic terminal that are associated with the triggering of neurotransmitter release . To close this gap , we have investigated changes in the presynaptic phosphoproteome during stimulation of exocytosis by taking advantage of synaptosomes that constitute pinched off and resealed nerve terminals enriched from brain homogenate . Synaptosomes retain the ability to synthesize ATP , maintain the resting potential and respond to depolarization with Ca2+-influx and Ca2+-dependent exocytosis of synaptic vesicles ( Nicholls , 2003 ) . Synaptosomes respond not only to external signals but also to activators and inhibitors of various steps of the synaptic vesicle cycle such as clostridial neurotoxins ( Blasi et al . , 1994 ) , black widow spider venom ( Nicholls et al . , 1982 ) , or dynamin inhibitors such as dynasore ( Daniel et al . , 2012 ) . Indeed , synaptosomes were used to monitor stimulation-dependent changes in the phosphorylation of individual proteins such as synapsin and dynamin . Here we have used synaptosomes as a tool for the study of the presynaptic phosphoproteome . We observed many hitherto unknown changes in the phosphorylation pattern of , in particular , active zone proteins that exhibit concurrent changes in phosphorylation and dephosphorylation at multiple locations/positions . Our results show that protein phosphorylation plays an even more important role than previously anticipated and appears to be critically involved in regulating active zone function .
For the analysis of the presynaptic phosphoproteome , we prepared isolated nerve terminals ( synaptosomes ) from brain cortices of 5–6 weeks old rats using standard subcellular fractionation procedures ( see Materials and methods ) . To examine whether the synaptosomes are responsive to stimulation , we performed two independent quality checks . First , we measured glutamate release upon depolarization by increasing the external potassium concentration . While this treatment clamp-depolarizes the membrane , calcium-overload is prevented by inactivation of the calcium channels ( Simms and Zamponi , 2014 ) . As expected , addition of KCl led to a fast increase in glutamate release that persisted for several minutes after depolarization and that was reduced in the absence of Ca2+ ( Figure 1A ) . It has been shown that the difference between glutamate release in the presence and absence of calcium represents exocytotic release whereas the increase upon depolarization in the absence of Ca2+ is due to non-exocytotic release caused by the reversal of glutamate transporters in the plasma membrane that depend on the membrane potential ( Nicholls and Sirha , 1986 ) . Second , we monitored the phosphorylation status of two well-characterized phosphosites in the nerve terminal , synapsin1 ( Ser603 ) and dynamin1 ( Ser774 ) , using phosphospecific antibodies . These two sites are known to be regulated antagonistically by Ca2+-influx , with synapsin being phosphorylated and dynamin being dephosphorylated ( Jovanovic et al . , 2001; Graham et al . , 2007 ) . As shown in Figure 1B and C , both sites changed their phosphorylation status as expected , confirming that the synaptosomes are functional . Both tests for glutamate release and alteration of marker phosphorylation sites were carried out for all preparations before mass spectrometry ( LC-MS/MS ) analysis . For phosphoproteomic analysis , we compared three different incubation conditions of synaptosomes: ( i ) depolarization with ( K+ , Ca2+ ) or ( ii ) without calcium ( K+ , EGTA ) , and ( iii ) control ( no depolarization , EGTA ) . Stimulation of all samples were stopped after 2 min by the addition of ice-cold lysis buffer and processed as described in Materials and methods . While rapid and reversible phosphorylation changes may not be detectable anymore at this time point , both exo- and endocytosis are at an activated steady-state level . Equal amounts of protein were digested by trypsin . For quantification , we used stable isotope dimethyl labeling ( Boersema et al . , 2009 ) . To elucidate the changes in phosphorylation status of phosphosites , we compared the phosphoproteome of depolarized synaptosomes in the presence of Ca2+ against depolarized synaptosomes in the absence of Ca2+ and synaptosomes that were not depolarized . In each set of comparisons , one of the incubation conditions was labeled as heavy and the other one as light . The differentially labeled peptides from two conditions were then mixed , and fractionated by strong cation exchange ( SCX ) chromatography . This step was introduced since it reduced sample complexity and resulted in higher recovery rates of phosphopeptides ( data not shown ) . In the next step , each fraction obtained from the SCX-column was separately enriched for phosphopeptides , and analyzed by MS ( see Materials and methods for details and Figure 2—figure supplement 1 for an overview of the workflow ) . Figure 2—figure supplement 5 shows exemplary MS spectra obtained for phosphopeptides of synapsin ( Ser603 ) and dynamin ( S774 ) after mixing heavy and light labeled samples , and the corresponding MS/MS spectra used for sequence determination . In parallel , we determined the entire proteome of synaptosomes ( three biological replicates ) using standard procedures ( see Materials and methods for details ) . We identified 4961 proteins ( for list of identified proteins see Supplementary file 1 , available at Dryad [Kohansal-Nodehi et al . , 2016] ) in the proteome and 1257 phosphoproteins in the phosphoproteome . We compared the biological functions enriched in our proteome and phosphoproteome datasets using Ingenuity Pathway Analysis ( IPA , http://www . ingenuity . com ) . The analysis revealed that most of the functions related to neurotransmission and synaptic vesicle recycling are enriched in both proteomes , with a higher enrichment being observed in the phosphoproteome compared to the proteome ( Figure 2—figure supplement 2 ) . Note that due to redundancies and multiple overlaps in the functional categories , this type of analysis constitutes only a rough estimate . However , it documents that even though synaptosomes are contaminated with many other organelles and post synaptic density proteins , the phosphorylation events occur predominantly , and perhaps exclusively , inside the presynaptic nerve terminal under our experimental conditions . To quantify changes of phosphorylation sites under different incubation conditions , we performed pairwise comparisons of the phosphoproteome in the respective synaptosome samples . All data are based on three biological replicates . Phosphosites that were quantified in at least two out of three replicates were filtered and subjected to downstream data analysis to determine whether they are differentially ( up or down ) regulated ( see Materials and methods ) . Due to the rigorous selection criteria , the coverage of phosphosites that were reliably quantified in the depolarization with Ca2+ versus depolarization without Ca2+ comparison and depolarization with Ca2+ versus no depolarization comparison corresponded , respectively , to 67 . 9% and 62 . 5% of the total phosphosites that were initially identified . The complete list of quantified phosphosites in all three comparisons is reported in Supplementary file 2 , available at Dryad ( Kohansal-Nodehi et al . , 2016 ) . To assess the reproducibility of the biological replicates , the Pearson correlation coefficients between dimethyl ratios and intensities of phosphosites in three biological replicates were determined ( Figure 2—figure supplement 3 ) . High correlation between the replicates in both comparison were confirmed , with the coefficients greater than 0 . 72 for intensities and greater than 0 . 65 for dimethyl ratios ( Figure 2—figure supplement 3A–D ) . The overall results of the comparative quantitative analysis are shown in Figure 2 . Several important conclusions can be drawn from this analysis . First , only a small fraction ( 12–15% ) of all quantified phosphosites show significant changes upon depolarization in the presence of Ca2+ ( Figure 2A–C ) . Both phosphorylation and dephosphorylation events are detectable , with phosphorylation events being slightly more prominent ( Figure 2C ) . The overlap among all quantified phosphosites is very high between the two comparisons ( Figure 2D ) . There is also substantial overlap in regulated phosphosites between the comparisons ( 37% in upregulated and 35% in downregulated phosphosites , Figure 2E–G ) with identical tendency of regulation and similar fold changes ( Figure 2H ) . Taken together , these data show that only a limited set of proteins undergo rapid phosphorylation changes that are specifically triggered by Ca2+ , most probably involving Ca2+-dependent kinases and phosphatases ( see below ) . As expected , under stimulation conditions , synapsin1 ( Ser603 ) and dynamin1 ( Ser774 ) ( Figure 1A ) were identified as up- and downregulated phosphosites , respectively ( Figure 2A and B ) . This result agrees with our western blot analyses using phospho-specific antibodies , confirming the high quality of the datasets . 10 . 7554/eLife . 14530 . 003Figure 1 . Isolated synaptosomes are responsive to stimulations . ( A ) Glutamate release by synaptosomes , monitored by enzymatic conversion of glutamate by glutamate dehydrogenase ( GluDH ) ( see Materials and methods ) . The functionality of isolated synaptosomes was also assessed by immunoblot detection of phosphorylated Ser603 of synapsin1 and Ser774 of dynamin1 ( B and C ) . ( A ) Synaptosomes were preincubated in buffer containing either Ca2+ ( final concentration 1 . 3 mM ) or EGTA ( final concentration 0 . 5 mM ) , followed by sequential addition of GluDH ( 200 U ) and KCl ( K+ ) for depolarization ( final concentration 50 mM ) or not stimulated ( no K+ was added ) . Data represented as mean of three replicates , with the bars indicating the range of values . ( B and C ) Immunoblot analyses of phosphorylation changes in synapsin1 ( Ser603 ) and dynamin1 ( Ser774 ) . Synaptosomes monitored and treated 2 min after addition of KCl as described in ( A ) were analyzed by immunoblotting using phosphospecific antibodies . ( B ) Representative immunoblots . To ensure equal loading , all samples were also blotted using antibodies insensitive to phosphorylation change , and tubulin as a loading control . ( C ) Quantification of the blot signals obtained from three independent experiments shown as the mean of the three replicates , with the error bars indicating the range of values . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 00310 . 7554/eLife . 14530 . 004Figure 2 . Overview of the quantified phosphosites identified in resting and stimulated synaptosomes . Distribution of quantified phosphosites from pair-wise comparisons of ( A ) depolarization with Ca2+ versus depolarization without Ca2+ ( K+ , Ca2+ versus K+ , EGTA ) ( B ) depolarization with Ca2+ versus no depolarization ( K+ , Ca2+ versus EGTA ) . Quantifications were done by dimethyl labeling and significant outliers were determined using the Perseus module based on two-tailed ‘Significant A’ test ( p-value ≤0 . 05 ) ( see Materials and methods ) . Red dots represent upregulated phosphosites , green dots represent downregulated sites when comparing depolarization in the presence to absence of Ca2+ . Positions represented by black dots did not change ( C ) Number of quantified , up- or down-regulated phosphosites and phosphoproteins in the two comparisons . ( D–G ) The proportional Venn diagrams show the overlap of quantified phosphosites ( D ) , regulated phosphosites ( E ) , upregulated phosphosites ( F ) and downregulated phosphosites ( G ) between the two conditions indicated above . ( H ) Plot comparing the extent of stimulation-dependent phosphorylation changes with respect to non-depolarized and depolarized control samples . All regulated phosphosites shared between K+ , Ca2+ versus K+ , EGTA and K+ , Ca2+ versus EGTA comparisons are depicted . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 00410 . 7554/eLife . 14530 . 005Figure 2—figure supplement 1 . Workflow for large-scale quantitative phosphoproteomics of synaptosomes . Synaptosomal proteins were extracted by lysis of synaptosomes and acetone precipitation . The protein pellets were dried , resuspended in 1% RapiGest buffer and digested by trypsin . Peptides were labeled with heavy and light dimethyl labeling ( Boersema et al . , 2009 ) . To obtain pair-wise comparisons between the three conditions , samples were mixed with a ratio of 1:1 . The mixed peptides were separated by strong cation exchange ( SCX ) chromatography , with the eluate being divided into 12 fractions . Phosphopeptides in each fraction were separately enriched by TiO2 microbeads ( Larsen et al . , 2005 ) and subjected for mass spectrometry analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 00510 . 7554/eLife . 14530 . 006Figure 2—figure supplement 2 . Comparison of functional categories for proteins identified in the phosphoproteome versus the proteome of synaptosomes . The comparison analysis function of Ingenuity Pathway Analysis tools was used to determine the biological functions that are significantly enriched between proteome and phosphoproteome datasets . The functions with p-value more than 0 . 01 were included in the analyses . Redundant functions and additional functions related to neuronal disease and neuronal development related functions were manually filtered . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 00610 . 7554/eLife . 14530 . 007Figure 2—figure supplement 3 . Correlation of phosphoproteome between biological replicates . Correlation plots depicting correlation of phosphosite between three biological replicates at Log10 intensity level and Log2 ratio of H/L level in depolarization with Ca2+ versus depolarization without Ca2+ ( K+ , Ca2+ versus K+ , EGTA ) comparison ( A , C ) , depolarization with Ca2+ versus no depolarization ( K+ , Ca2+ versus EGTA ) ( B , D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 00710 . 7554/eLife . 14530 . 008Figure 2—figure supplement 4 . Phosphorylation status and its change upon membrane depolarization . Distribution of quantified phosphosites in the comparison between depolarized and non-depolarized synaptosomes in the presence of EGTA . Quantification was performed by dimethyl labeling and significant outliers were determined based on two-tailed ‘Significant A’ test; p-value ≤0 . 05 in the Perseus module . Red dots represent phosphosites that were upregulated , and , green dots represent sites that were downregulated during depolarization . Correlation plots are shown depicting correlation of phosphosite between three biological replicates at Log10 intensity ( B ) level and Log2 ratio ( C ) of H/L level . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 00810 . 7554/eLife . 14530 . 009Figure 2—figure supplement 5 . MS1 and MS/MS spectrum of synapsin Ser603 and dynamin Ser774 . Partial spectra of mass spectrometric survey scans in MS1 ( precursor spectra ) containing phosphopeptides of synapsin Ser603 ( A ) and dynamin S774 ( B ) in light ( K+ , Ca2+ ) and heavy ( K+ , EGTA ) isotopic pattern that is used to calculated the relative abundances . MS/MS spectra resulting from HCD fragmentation of phosphopeptides containing synapsin Ser603 m/z 538 . 74782+ ( C ) and dynamin Ser774 m/z 703 . 21232+ ( D ) , respectively , showing the phosphopeptides _QAS ( ph ) QAGPGPR_ and _RS ( ph ) PTS ( ph ) SPTPQR . Both of these phosphopeptides were eluted in the first fraction of SCX . MaxQuant Viewer ( version 1 . 5 . 0 . 25 ) was used to illustrate the spectra . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 009 To check whether depolarization alone leads to changes in the phosphorylation state of proteins , we compared the phosphoproteome of depolarized with non-depolarized synaptosomes in the presence of the calcium chelator EGTA . Although some statistically significant changes were observed , they were less pronounced and clustered close to the threshold ( compare Figure 2A and B with Figure 2—figure supplement 4A ) . Moreover , the correlation of the ratios was low ( Figure 2—figure supplement 4C ) although the Pearson correlation of the intensities was high ( Figure 2—figure supplement 4B ) . For these reasons , this comparison was not considered in our further analysis . Given that the large number of phosphosites detected were unchanged in the relatively large number of quantified phosphorylation sites , the corresponding small fraction of sites that underwent stimulation dependent changes is highly significant and potentially of importance in regulating neurotransmitter release . To gain an overview about the functional properties of the proteins exhibiting phosphosite changes in the two comparisons , we annotated them manually and classified into 15 groups based on their function and localization ( Figure 3 ) . With only a few exceptions ( e . g . translation/transcription or neurotransmitter metabolism related proteins ) , proteins undergoing phosphorylation changes during Ca2+-dependent stimulation are all involved in synaptic vesicle trafficking or in presynaptic signaling ( Figure 3A and B ) , again documenting the specificity of the experimental approach . Note that postsynaptic proteins , particularly proteins associated with the post synaptic density , were largely missing ( Figure 3A and B ) . 10 . 7554/eLife . 14530 . 010Figure 3 . Overview of the regulated phosphosites , based on corresponding grouping of protein functions . Proteins containing regulated phosphosites were individually divided into 15 groups as described previously ( Boyken et al . , 2013 ) . Proteins with function/localization other than listed or uncharacterized proteins were grouped into the 'other' category . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 010 A more detailed representation of the quantified phosphosites is shown in Figure 4 . Here , the size of the circles is proportional to the total number of quantified phosphosites in the two comparisons . Red , green and white areas in the circles are proportional to the number of up , down and non-regulated phosphosites , respectively . The lines connecting the proteins show the interaction of the proteins based on STRING database and literature mining ( Active Zone ( AZ ) proteins ( Ohtsuka et al . , 2002; Schoch et al . , 2002; Wang et al . , 2002; Takao-Rikitsu et al . , 2004; Lu et al . , 2005; Schoch and Gundelfinger , 2006; Wang et al . , 2009 ) , Clathrin Mediated Endocytosis ( CME ) proteins ( Ringstad et al . , 1997; Anggono et al . , 2006; Anggono and Robinson , 2007 ) ) . The figure shows that protein groups involved in synaptic vesicle recycling that exhibit stimulation-dependent changes in phosphorylation include active zone proteins , synaptic vesicle proteins , endocytotic proteins , and few proteins in the presynaptic plasma membrane ( mainly ion channels ) . In addition , it is evident that some of the proteins ( e . g . bassoon , RIM1 , synapsin1 , dynamin1 ) contain multiple phosphorylation sites that can undergo parallel phosphorylation and dephosphorylations . Surprisingly , proteins of the active zone were most heavily phosphorylated , which is discussed further below . 10 . 7554/eLife . 14530 . 011Figure 4 . Changes in the phosphorylation status of proteins involved in synaptic vesicle trafficking following stimulation . Overview of changes in the phosphorylation status of Active Zone proteins ( AZ ) , synaptic vesicle proteins ( SV ) , proteins involved in Clathrin Mediated Endocytosis ( CME ) and Plasma Membrane proteins ( PM ) in the two pair-wise comparisons: ( A ) depolarization with Ca2+ versus depolarization without Ca2+ ( K+ , Ca2+ versus K+ , EGTA ) ( B ) depolarization with Ca2+ versus no depolarization ( K+ , Ca2+ versus EGTA ) . The circles are proportional to the number of quantified phosphosites in the two conditions . The red , green and white area in each circle is proportional to the number of up , down and nonregulated phosphosites , respectively . The dotted lines connecting the circles show the interactions of the proteins based on STRING database and literature mining ( references included in the main text; for a glossary of the abbreviations see supplemental Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 01110 . 7554/eLife . 14530 . 012Figure 4—source data 1 . Protein name , Gene name , Uniprot ID and abbreviation of proteins shown in Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 012 With few exceptions , most of the phosphosites in the Clathrin-Mediated Endocytosis ( CME ) functional group were dephosphorylated upon stimulation in the presence of Ca2+ ( 13 out of 15 regulated phosphosites in the depolarization with Ca2+ versus depolarization without Ca2+ comparison and 12 out 17 in the depolarization with Ca2+ versus no depolarization comparison ) ( Figure 4A and B ) . This is in agreement with the previous report ( Cousin and Robinson , 2001 ) . Phosphorylation/dephosphorylation of CME proteins was proposed to serve two related functions . First , phosphorylation regulates the interactions between different CME proteins . For example , phosphorylation of the adaptor protein AP180 weakens its interaction with the AP-2 adaptor complex , which is required for synaptic vesicle endocytosis ( Hao et al . , 1999 ) . In our dataset , three AP180 phosphosites ( Ser600 , Ser621 and Ser627 ) are dephosphorylated , one of which ( Ser627 ) is located in the reported binding site . Second , phosphorylation generally weakens the affinity towards membranes containing acidic lipids and thus promotes dissociation of coat proteins . For instance , phosphorylation of dynamin1 by PKC at Ser795 prevents its association with phospholipids ( Powell et al . , 2000 ) . Figure 4 also confirms and extends previous reports on the regulation of multiple phosphosites in synapsin1 and dynamin1 . For synapsin1 , regulation of some of the sites has been previously characterized extensively ( Jovanovic et al . , 2001 ) . Such as downregulation of Ser62 and upregulation of Ser556 and Ser603 . However , no stimulus-dependent changes were observable at the phosphorylation sites of CdK5 ( Ser549 and 551 ) that were suggested to regulate the size of the synaptic vesicle pools ( Verstegen et al . , 2014 ) . Components of the presynaptic active zone make up the most conspicuously phosphorylated proteins ( e . g . bassoon , piccolo , RIM1 , liprin-α3 and ERC1-2 ) . The amino acid positions of phosphosites identified on these proteins are shown in Figure 5 , with the well characterized phosphoproteins synapsin1 and dynamin1 included for comparison . Similarly to synapsin and dynamin , we observed many sites of active zone proteins undergoing parallel changes in phosphorylation and dephosphorylation upon Ca2+-dependent stimulation . Active zone proteins regulate synaptic transmission speed , precision , and plasticity ( Kaeser , 2011; Sudhof , 2012 ) . Our results show that , the proteins bassoon and piccolo are massively phosphorylated , with 48 and 31 sites being quantified in our experiments , respectively . Intriguingly , with the exception of 6 phosphosites , most of the piccolo phosphosites remained unchanged following stimulation . In comparison , bassoon exhibited a dramatic decrease in phosphorylation at multiple sites in addition to increased phosphorylation at 5 other positions . While the role of these novel sites is still unclear , it is conceivable that downregulation of the bassoon phosphosites may serve to regulate the global properties of the protein ( e . g . surface charge density ) rather than specific protein-protein interactions . In previous studies it has been reported that in spite of having similar structural motifs , bassoon and piccolo have different protein interaction partners ( Takao-Rikitsu et al . , 2004; Schoch and Gundelfinger , 2006; Wang et al . , 2009 ) . The more pronounced observed dephosphorylation of phosphosites in bassoon compared to piccolo can be an explanation for their different protein interactions . 10 . 7554/eLife . 14530 . 013Figure 5 . Phosphosites identified in proteins of the active zone . Colors indicate phosphosites that are upregulated ( red ) , downregulated ( green ) , or remain unchanged ( black ) in depolarization with Ca2+ versus depolarization without Ca2+ comparison ( K+ , Ca2+ versus K+ , EGTA ) . The underlined phosphosites represent sites previously reported to be physiologically relevant ( for references see Supplementary file 3 , available at Dryad [Kohansal-Nodehi et al . , 2016] ) . Unmarked phosphosites represent sites previously reported only by proteomic discovery-mode mass spectrometry , defined as HTP in the publicly available PhosphoSitePlus database . Phosphosites marked with a star represent phosphosites reported here for the first time . For comparison , phosphosites found in the well-studied proteins dynamin1 and synapsin1 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 013 RIM , considered to be central organizer of active zone , is known to have several interactions with other active zone proteins as well as with calcium channels and SV proteins such as Rab3a ( Südhof , 2012 ) . Here , we identified 7 regulated phosphosites that are distributed along the protein ( Figure 5 ) including Ser413 that was previously reported to be phosphorylated by PKA , and its role in presynaptic long-term plasticity has been discussed extensively ( Lonart et al . , 2003; Kaeser et al . , 2008 ) . A summary of all previously characterized phosphosites that were assigned as significantly regulated phosphosites in our datasets is given in Supplementary file 3 , available at Dryad ( Kohansal-Nodehi et al . , 2016 ) . Next , we tried to find out which kinases and phosphatases are responsible for the phosphorylation changes upon stimulation . First , we analyzed the respective sites for sequence motifs known to be targeted by specific kinases . It is well established that the substrate specificity of Ser/Thr kinases is strongly influenced by the amino acids surrounding the phosphorylated Ser/Thr residues , with specific consensus motifs being described for an array of protein kinases ( Ubersax and Ferrell , 2007 ) . To establish whether there are over-represented consensus sites , the phosphorylation sites were analyzed by the Motif-X tool ( Schwartz and Gygi , 2005 ) . Phosphosites upregulated during Ca2+-dependent stimulation were highly enriched for the RxxS sequence motif ( 7 . 36 fold , p=1 . 0 × 10–16 compared to control and 5 . 98 fold , p=4 . 8 × 10–13 , when compared to samples depolarized in the absence of Ca2+ ) ( Figure 6A and B ) . RxxS is the consensus motif of calmodulin-dependent kinase II ( CaMKII ) but also targeted by protein kinase A ( PKA ) and conventional isoforms of protein kinase C ( PKC ) . This is in agreement with previous studies showing that these kinases are activated upon depolarization of nerve terminals in the presence of Ca2+ ions ( Millán et al . , 2003; Yamauchi , 2005 ) . 10 . 7554/eLife . 14530 . 014Figure 6 . Analysis of sequence motifs and experimental verification of specific kinase phosphorylation on selected phosphosites . ( A–D ) Sequence analysis of regulated phosphosites using the Motif-X tool ( Schwartz and Gygi , 2005 ) . The frequency of residues surrounding the phosphorylated serine is indicated by the size of the letters . Among the upregulated sites , the motif RXXS is overrepresented , whereas the SP motif is conspicuously present among downregulated motifs . ( E ) In vitro kinase assays evaluating the ability of selected kinases to phosphorylated phosphosites identified on active zone proteins . Five phosphosites were selected from four active zone proteins for testing against CaMKII , PKA , and PKC . For each site , two 13-mer peptides were synthesized as substrates ( for sequences of the peptide see Figure 6—source data 1 ) that correspond to the sequence surrounding the phosphorylated serine , with one of them containing an alanine instead of the serine as control . The following peptides were used as positive controls: ( i ) synapsin S603 that is known to be phosphorylated by CaMKII ( Jovanovic et al . , 2001 ) and ( ii ) autocamide-2 , kemptide , and neurogranin were used as known substrates of CaMKII , PKA , and PKC , respectively . Background was determined by omitting peptide from the assay . The data show means of three replicates , with the bars indicating the range of values . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 01410 . 7554/eLife . 14530 . 015Figure 6—source data 1 . Sequence of the peptides used in in vitro kinase assay . DOI: http://dx . doi . org/10 . 7554/eLife . 14530 . 015 A similar analysis using the Motif-X tool was also carried out for downregulated phosphosites . Here , SP sequence motif was enriched significantly ( 8 . 21 fold , p=1 . 0 × 10–16 when compared to control and 8 . 03 fold , p=1 . 0 x 10-16 when compared to samples depolarized in the absence of Ca2+ Figure 6C and D , respectively ) . Phosphatases are not known to have preferred substrate consensus motifs ( Kennelly and Krebs , 1991 ) . However , the SP motif is preferred by proline-directed kinases such as ERK1/2 , GSK-3 and Cdk5 ( Guan et al . , 1991; Lee et al . , 1997; Veeranna et al . , 1998 ) . Between the mentioned kinases , ERK1/2 is less likely to be inactivated upon stimulation , since we have quantified the upregulation of two activating phosphosites in the active loop of ERK1/2 ( T183 , Y185 , Supplementary file 2 , K+ , EGTA vs K+ , Ca2+ ) . Overall , it is conceivable that phosphosites with SP motifs turnover more rapidly , with the decrease reflecting either reduced kinase activity or increased processing by the phosphatases . Net dephosphorylation of SP-sites may have functional consequences for presynaptic function . For instance , CdK5 is known to be major control point of neurotransmitter release ( Kim and Ryan , 2010 ) . The balance between the activation of calcineurin ( PP2B ) and Cdk5 is reported to control the dynamics of resting and recycling pools as well as activity of voltage gated calcium channels ( Kim and Ryan , 2013 ) . Also the coordination of CdK5 activity and other proline directed kinases such GSK-3 and ERK1/2 can result in regulation of neurotransmitter release . For example , Cdk5 and GSK-3 phosphorylate P/Q- and N-type voltage-dependent calcium channels ( Tomizawa et al . , 2002; Zhu et al . , 2010 ) whereas ERK1/2 phosphorylate L-type calcium channels ( Subramanian and Morozov , 2011 ) and phosphorylation decreases the activity of both types of calcium channels . Thus , it is conceivable that Ca2+ influx increases the activity of calcium channels via dephosphorylation . However regulation of calcium channels via Ca2+ influx seems to be more complex since they are also inactivated by calcium dependent mechanisms ( Simms and Zamponi , 2014 ) . Due to the lack of consensus motifs it is not possible to identify the phosphatases that are activated upon stimulation using sequence analysis . However , the increase in the Ca2+ concentration activates phosphatase PP2B ( calcineurin ) ( Sun et al . , 2010 ) . Also , in our dataset we observe that the regulatory subunit delta ( LOC100909464 ) of protein phosphatase 2A shows increased phosphorylation at Ser567 ( Supplementary file 2 , K+ , EGTA vs K+ , Ca2+ , Kohansal-Nodehi et al . , 2016 ) , a site known to be phosphorylated by PKA , resulting in an increase of its enzymatic activity ( Ahn et al . , 2007 ) . This suggests the activation of PP2A upon stimulation . We also observe phosphorylation changes in Neurabin-2 ( S100 ) , regulatory protein of protein phosphatase 1 ( PP1 ) , ( Supplementary file 2 , K+ , EGTA vs K+ , Ca2+ ) . It has been shown previously that phosphorylation of S100 alters PP1 localization by the decrease of binding affinity to actin and the diffusion from the nerve terminal ( Colbran , 2004 ) . In summary , our data suggest that upon Ca2+ influx subset of kinases and phosphatases such as PKA , PKC , CaMKII , PP2A and PP2B are activated while PP1 is less active and the target sites of proline directed kinases show a decrease in phosphorylation , mediated either by decreased activity of the kinases or preferred targeting by the phosphatases . To confirm that hitherto uncharacterized phosphosites containing the RxxS motif are indeed phosphorylated by one or more of the three kinases known to target these sites , five of these sites were selected for experimental verification: RIM1 ( Ser413 and Ser1141 ) , bassoon ( Ser2844 ) , piccolo ( Ser3626 ) and liprin-α3 ( Ser75 ) . For each of these sites , 13-mer peptides were synthesized ( containing 6 residues upstream and 6 residues downstream of the phosphoserine ) and tested by a commercially available kinase assay against CaMKII , protein kinase A , and protein kinase C . As a control , corresponding peptides were synthesized , in which the phosphorylated serine was replaced by alanine . As shown in Figure 6E , each of the peptides was recognized by at least one of the kinases . For instance , RIM1 Ser413 was phosphorylated only by PKA; consistent with a previous report ( Lonart et al . , 2003 ) . RIM1 Ser1141 was phosphorylated mostly by CaMKII and with lower efficiency by PKA . Bassoon Ser2844 was phosphorylated only by CaMKII . Piccolo Ser3626 was phosphorylated by CaMKII and PKA albeit with low efficiency . Liprin-α3 Ser75 was the only phosphosite that was phosphorylated by all three kinases although PKC was less efficient . Together , these findings agree with the notion that activation of these three kinases during depolarization in the presence of Ca2+ leads to the phosphorylation of the active zone proteins at their corresponding phosphosites . Furthermore , it reveals another layer of regulatory diversification since each of the three kinases exhibited selective phosphorylation of the different substrates despite sharing a common core motif .
In the present study we have used quantitative phosphoproteomics to obtain a snapshot of changes that occur in protein phosphorylation during Ca2+-dependent exocytosis of synaptic vesicles . A surprisingly complex pattern of both up- and downregulated sites was observed , frequently within the same protein , with the proteins of the active zone being most conspicuous . Synaptosomes are uniquely suitable for the study of phosphorylation-dephosphorylation events that are associated with synaptic vesicle cycling . Although even the most highly enriched fraction of nerve terminals still contains significant contamination from many other parts of the starting tissue ( as confirmed by the composition of the proteome of this fraction ) , synaptosomes represent the only compartment in the extract that forms membrane enclosed structures retaining basic cellular properties . These include the ability of generating and maintaining a membrane potential and of synthesizing ATP from externally added glucose or ketone bodies ( Nicholls , 2003; Choi et al . , 2011 ) . Thus , physiological ATP-levels can only be maintained inside nerve terminals whereas residual external ATP is likely to be rapidly degraded by ATPases , in particular by uncoupled ion pumps . Consequently , protein kinases can only operate inside the nerve terminal during the experiment . Indeed , our comparison of the phosphoproteome versus the total proteome of the synaptosome fraction revealed a significant enrichment of proteins involved in presynaptic function while other proteins ( including those of the postsynapse ) were reduced or conspicuously absent ( Figure 2—figure supplement 2 ) . Although our analysis of phosphosites is certainly not yet comprehensive , several important conclusions can be drawn from our data . First , only a limited number of phosphosites is regulated by stimulation whereas the majority of the quantified sites did not change . Second , almost an identical pattern of regulated sites was observed regardless of whether the Ca2+-free control was depolarized or not . This agrees well with the observation that depolarization in the absence of Ca2+ resulted only in very limited changes of the phosphosite pattern . This is remarkable when considering that clamp-depolarization by potassium is nonphysiological and may have deleterious side effects ( e . g . drop in ATP levels ) . Consequently , our data show that the phosphorylation changes observed after stimulation are dominated by Ca2+-dependent kinases ( such as CaMKII , PKC ) and phosphatases ( such as calcineurin ) , all of which are implicated in regulating presynaptic function ( Scholz and Palfrey , 1998; Brager et al . , 2003; Leenders and Sheng , 2005 ) . In addition , our data hint at the activation of a network of kinases and phosphatases that operates downstream of the primary Ca2+-dependent enzymes . For instance , calcineurin upregulates adenylate cyclase in nerve terminals ( Ferguson and Storm , 2004; Chan et al . , 2005 ) , increasing cAMP levels and activating protein kinase A . Also , we observed increased phosphorylation of Ser567 in the regulatory subunit of protein phosphatase 2A ( PP2A ) , which is known to enhance its activity ( Ahn et al . , 2007 ) . Moreover , our motif enrichment analysis suggests that CdK5 , ERK1/2 and GSK-3 are downregulated upon stimulation . Another example is contributed by the increased phosphorylation of protein phosphatase 1 regulatory subunit Neurabin-2 at S100 that decreases its activity ( Colbran , 2004 ) . We conclude that the activity of kinases and phosphatases is precisely coordinated: upon Ca2+ entry , CaMKII , PKA , PKC as well as PP2A and PP2B are activated whereas proline directed kinases ( e . g . CdK5 and GSK-3 ) as well as PP1 appear to be downregulated . The panoply of phosphorylation sites and phosphorylation changes detected in many proteins of the presynaptic active zone was surprising . Active zone proteins are thought to structure the site for synaptic vesicle docking and fusion by forming dynamic scaffolds , and several of these proteins such as RIM and Munc13 have been shown to play a crucial role in vesicle docking and priming ( Südhof , 2012 ) . Only a few of these phosphorylation sites have been previously studied , such as phosphorylation dependent binding of 14-3-3 adaptor protein to bassoon at Ser2844 or the cAMP-regulated site in RIM1 ( Ser413 ) that plays a role in long-term potentiation ( Lonart et al . , 2003; Kaeser et al . , 2008 ) . Similar to the well-characterized and multiply phosphorylated synapsin1 , active zone proteins contain sites that exhibit opposing changes in response to Ca2+ . The functional significance of such multiple phosphorylation remains to be explored . There are many examples showing that within such proteins individual phosphosites can have specific functions , e . g . in regulating the activity of enzymes such as CaMKII ( Cohen , 2000; Coultrap and Bayer , 2012 ) or channels such as Kv2 . 1 ( Park et al . , 2006 ) . At present , we do not yet know whether there are individual sites on a given active zone protein that are solely responsible for a functional change ( e . g . in regulating binding to a specific protein ) or whether multiple phosphosites cooperate in changing the overall properties of a protein . Multi-site phosphorylation may modulate the binding affinity of a protein to other proteins or block binding by increasing its overall negative charge leading to repulsion from other proteins and charged membrane lipids ( Powell et al . , 2000; Arbuzova et al . , 2002; Ferreon et al . , 2009 ) . Also , our data only represent a snapshot of phosphorylation changes 2 min after depolarization , i . e . when exo- and endocytosis are both at elevated steady-state levels . Thus it is conceivable that transient changes occurring in the first seconds after stimulations were missed . Intriguingly , in our data no significant changes were observed in the phosphorylation status of SNARE and other proteins that are mediating exocytosis of synaptic vesicles ( e . g syntaxin1A Thr10 , Ser14 ( Rickman and Duncan , 2010 ) , Ser59 , Ser64; VAMP-2 Ser75 , Ser80; SNAP-25 Ser25; Munc18 Ser506 , Ser507 , Ser509 , Ser593 and NSF Ser739 ) ( Supplementary file 2 , Kohansal-Nodehi et al . , 2016 ) . Thus it appears that the rapid feedback loops governed by kinases and phosphatases , which are activated during stimulation , do not target the exocytotic machinery itself but rather upstream processes such as docking , priming , and endocytosis . Further detailed studies will be required to unravel the functional consequences of these changes at the molecular level .
Sucrose , glucose monohydrate , sodium chloride ( NaCl ) , potassium chloride ( KCl ) , magnesium chloride hexahydrate ( MgCl2x6H20 ) , sodium hydrogen carbonate ( NaHCO3 ) , sodium hydrogen phosphate dihydrate ( Na2HPO4x2H20 ) , sodium hydroxide ( NaOH ) were purchased from Merck KGaA ( Darmstadt , Germany ) . Calcium chloride ( CaCl2 ) , nicotinamide adenine dinucleotide phosphate hydrate ( NADPx1H20 ) , ammonium hydrogen carbonate ( NH4HCO3 ) , 2-chloroacetamide ( CAA ) , acetone ( Chromasolv grade ) , glutamate dehydrogenase ( GluDH ) , formic acid ( FA ) , dithiothreitol ( DTT ) , pepstatin-A , trifluoroacetic acid ( TFA ) , Nonidet P40 , tetraethylammonium bromide ( TEAB ) , 2 , 5-dihydroxybenzoic acid ( DHB ) , ammonium formate ( NH4HCO2 ) , Ficoll PM 400 , sodium cyanoborohydride ( NaBD3CN ) , formaldehyde ( CH2O ) were purchased from Sigma-Aldrich ( Steinheim , Germany ) . Pierce bicinchoninic acid protein concentration assay ( BCA ) , albumin standard , Pierce protease and phosphatase inhibitor , acetonitrile ( ACN , LC-MS grade ) were purchased from Thermo Fisher Scientific ( Rockford , IL ) . 2-[4- ( 2-hydroxyethyl ) piperazin-1-yl]ethanesulfonic acid ( Hepes ) , ethylene glycol tetraacetic acid ( EGTA ) were purchased from Gerbu Biotechnik ( Heidelberg , Germany ) . The remaining chemicals and reagents were purchased from individual supplier: sequencing grade modified trypsin ( Promega , Madison , WI ) , formaldehyde ( D2 , 98% , CD2O , Cambridge Isotope Laboratories , Andover , MA ) , ammonium hydroxide ( NH4OH , BAKER ANALYZED , Deventer , the Netherlands ) , PMSF ( AppliChem , Darmstadt , Germany ) , 2-Amino-2-hydroxymethyl-propane-1 , 3-diol ( Tris , VWR International , Leuven , Belgium ) , Rapigest ( Waters , Milford , MA ) , titanium dioxide beads ( TiO2 , GL Sciences Inc . , Tokyo , Japan ) . Sucrose buffer: 320 mM sucrose , 5 mM Hepes , pH=7 . 4; sodium buffer: 10 mM glucose , 5 mM KCl , 140 mM NaCl , 5 mM NaHCO3 , 1 mM MgCl2 , 1 . 2 mM Na2HPO4 , 20 mM Hepes , pH=7 . 4; lysis buffer: 50 mM Tris , 150 mM NaCl , 1% Nonidet P40 , pH=7 . 4 containing Pierce protease and phosphatase inhibitors; 6% , 9% and 13% Ficoll solutions were prepared in sucrose buffer ( pH=7 . 4 ) . The buffers were prepared using nanopure water obtained with an electric resistance of greater that 18 MΩ from a Mili-Q purification system ( MerckMillipore , Darmstadt , Germany ) . Solutions for mass spectrometry analysis were prepared using LiChrosolv water ( Merck KGaA , Darmstadt , Germany ) . Wistar rats originated from the local animal facility were kept until use at a 12:12 hr light/dark cycle with food and water ad libitum . Rats of age between 5 to 6 weeks were sacrificed by cervical dislocation followed by decapitation . The brains were removed from the scull and placed into an ice-cold sucrose buffer . Cerebral cortises and cerebellums were dissected , pooled together and subjected to the synaptosomes enrichment by the non-continuous Ficoll gradient as described previously ( Fischer von Mollard et al . , 1991 ) . Glutamate release was monitored by enzymatic conversion of glutamate as previously reported ( Nicholls and Sihra , 1986 ) with minor modifications . The synaptosomal pellet was resuspended in ice cold sucrose buffer with the addition of protease inhibitors ( PMSF , 200 mM and Pepstatin-A , 1 µg/ml ) . For glutamate release assay 2 . 5 mg of synaptosomes were resuspended in 2 . 5 ml of sodium buffer and incubated with stirring for 5 min at 37°C . Depending on the stimulation , either CaCl2 ( 1 . 3 mM ) or EGTA ( 0 . 5 mM ) and NADP ( 1 mM ) was added and incubation was continued for 3 min , then GluDH ( 200 U ) was added and after 3 min KCl ( 50 mM ) was spiked in . Generation of NADPH was monitored by fluorescence for 2 min . After this time synaptosomes were collected for mass spectrometry and western blot analysis . Synaptosomes were spun down at 6200 × g for 2 min in 4°C , resuspended in 100 µl lysis buffer . Then 900 µl of 100 mM TEAB ( pH=8 ) buffer was added . The samples were homogenized ( RW20-DZM , IKA , Staufen , Germany ) at maximum speed ( 2000 rpm ) by 3 strokes . Global protein concentration was determined by BCA protein concentration assay , using albumin as a standard . 1 mg of proteins per condition was precipitated by acetone ( 100% ) over night at -20°C . Proteins were spun down at 16 , 000 × g for 30 min at 4°C . The pellets were washed once with 200 µl of 80% acetone and centrifuged again at 16 , 000 × g for 30 min at 4°C . The protein pellets were dried briefly in the air at room temperature and later stored at -20°C for further analysis . Dried protein pellets were resuspended in 80 µl of 1% Rapigest ( in 100 mM TEAB , pH=8 ) by vortexing and incubated in thermoshaker at 60°C for 15 min at 1050 rpm . Twenty µl of 50 mM DTT ( in 100 mM TEAB , pH=8 ) were added; samples were incubated at 60°C for 45 min 1050 rpm . Then 20 μl of 100 mM CAA ( in 100 mM TEAB , pH=8 ) were added; samples were incubated in thermoshaker at 37°C for 30 min at 750 rpm . Samples were diluted by 100 mM TEAB ( pH=8 ) to decrease final concentration of Rapigest to 0 . 1% and then 1:20 trypsin to protein ratio was added . Samples were digested overnight at 37°C , at 750 rpm . The digestion was stopped by addition of 4 µl of 100% FA and incubation for 2 hr at 37°C at 750 rpm . Samples were spun down at 13 , 000 rpm for 30 min at 4°C; supernatant was transferred to the new vials , dried in vacuum concentrator and kept in -20°C for further analysis . For determination of the proteome 50 µg of synaptosomes were separated by SDS-PAGE using a 4–12% gradient gel ( NuPAGE , Life Technologies , Carlsbad , CA ) according to Shevchenko et al ( Shevchenko et al . , 1996 ) . For each sample , three biological replicates were carried out . Each lane was cut into 23 pieces . Proteins in each piece were in-gel digested overnight by trypsin and later digested peptides were extracted by 20 µl water followed by addition of 80 µl 100% acetonitrile . Extracted peptides were acidified by formic acid to final concentration of 0 . 05% vol/vol , dried in the vacuum concentrator and analyzed by liquid chromatography ( LC ) coupled to mass spectrometry ( MS/MS ) . Stable isotope dimethyl labeling was performed according to Boersema et al . ( Boersema et al . , 2009 ) . Briefly , digested sample were labeled separately to yield light and heavy isotope tags by adding 4 μl of 4% ( vol/vol ) CH2O followed by 4 μl of 4% ( vol/vol ) CH2O and CD2O , to light and heavy samples respectively . Later 4 μl of 0 . 6 M NaBD3CN and 16 μl of 1% ( vol/vol ) ammonia and 8 μl of 5% FA ( vol/vol ) were added to the labeling reaction of both heavy and light samples . Light and heavy samples were mixed at a volume ratio of 1:1 according to experimental design ( Figure 2—figure supplement 1 ) and were kept on ice for further analysis . Mixed light and heavy labeled peptides were diluted to a final volume of 1 ml with solvent A ( 10 mM NH4HCO2 , 30% ACN ( v/v ) , pH=2 . 7 ) and loaded onto a Mono S column PC 1 . 6/5 ( Pharmacia Biotech , Uppsala , Sweden ) at a flow rate of 100 µl/min . Elution was performed with a gradient of 0–90% solvent B ( 500 mM NH4HCO2 , 30% ACN ( ( v/v ) , pH=2 . 7 ) over 40 min . The first twelve fractions ( 0 . 2 ml ) including the flow-through were collected and subjected separately to phosphopeptide enrichment . Phosphopeptides were enriched using TiO2 chromatography as previously described by Larsen et al ( Larsen et al . , 2005 ) . Digested peptides were dissolved in 60 μl of 200 mg/ml DHB in 80% ACN , 5% TFA and loaded onto TiO2 columns . The columns were washed three times with 60 μl of 200 mg/ml DHB in 80% ACN , 5% TFA and five times with 60 μl of 80% ACN , 5% TFA . Then , phosphopeptides were eluted by three consecutive additions of 40 μl of 0 . 3 N NH4OH , pH ≥ 10 . 5 . Eluted phosphopeptides were dried in the vacuum concentrator for further MS analysis . Enriched phosphopeptides were analyzed on a Q-Exactive hybrid Quadrupole-Orbitrap mass spectrometer ( Thermo Fisher Scientific , Dreieich , Germany ) coupled to a NanoLC pump ( EASY-nLC , Thermo Fisher Scientific , Dreieich ) . The peptides were pre-concentrated on a Reversed Phase-C18 precolumn ( 0 . 15 mm ID × 20 mm self-packed with Reprosil-Pur 120 C18-AQ 5 μm , Dr . Maisch GmbH , Ammerbuch-Entringen , Germany ) and then separated by reversed phase-C18 nanoflow chromatography ( 0 . 075 mm ID × 200 mm self-packed with Reprosil-Pur 120 C18-AQ , 3 μm , Dr . Maisch GmbH ) . Peptides were injected with solvent A ( 0 . 1% FA , 5% acetonitrile in water ) at a flow rate 320 nL/min and eluted by 0–37% solvent B ( 80% acetonitrile , 0 . 1% FA in water ) with an overall run-time of 45 min . Separated peptides were ionized by electrospray ionization ( ESI ) source in a positive ion mode . Full scan MS spectra were acquired in the range of 350–1600 m/z at a resolution of 70 , 000 . The top 12 most intense peaks from the survey scan were selected for fragmentation with Higher-energy Collisional Dissociation ( HCD , 28 of normalized collision energy ) . For proteomics analysis , LTQ Orbitrap XL ( Thermo Fisher Scientific ) coupled to Agilent 1100-LC system ( Agilent Technologies , Waldbronn , Germany ) was used . Peptides were loaded onto a trap column packed in-house ( 0 . 15 mm ID × 20 mm self-packed with Reprosil-Pur 120 C18-AQ 5 μm , Dr . Maisch GmbH ) and separated at a flow rate of 15 µl/min on an analytical column ( 0 . 075 mm ID × 120 mm self-packed with Reprosil-Pur 120 C18-AQ , 3 μm , Dr . Maisch GmbH ) . Peptide were eluted from the column with 3–35% solvent B with an overall run-time of 45 min . Separated peptides were ionized by electrospray ionization ( ESI ) source in a positive ion mode . Full scan MS spectra were acquired in the range of 350–1600 m/z at a resolution of 35 , 000 . The top 10 peaks from survey scan were selected for fragmentation with Collision-Induced Dissociation ( CID ) with 37 . 5 normalized collision energy . Acquired MS spectra were processed using the MaxQuant software package version 1 . 5 . 0 . 25 ( Cox and Mann , 2008 ) . Spectra were searched using the Andromeda search engine ( Cox et al . , 2011 ) against the proteome database of Rattus Norvegicus ( Uniprot complete proteome updated at 2015-01-07 , with 29 , 378 entries ) . MaxQuant search was configured as follows; the mass tolerance was set to 20 and 4 . 5 ppm for first and main peptide search , respectively . The multiplicity was set to two ( DimethLys0 and DimethNter0 was for ‘light’ and DimethLys4 and DimethNter4 for ‘heavy’ samples ) . Trypsin/P was fixed as protease and maximum of 2 missed cleavages were allowed . Carbamidomethylation of cysteine was set as fixed modification and methionine oxidation , acetylation ( on N-terminal ) , phosphorylation on serine/threonine/tyrosine were specified as variable modifications . False discovery rate of 1% was applied and re-quantification was enabled . For proteomics analysis MaxQuant search was configured like phosphoproteomics search except that multiplicity was set to one with methionine oxidation and acetylation ( on N-terminal ) as variable modifications . The phospho ( STY ) output file from the MaxQuant was processed by ‘Perseus version 1 . 5 . 0 . 9’ ( Cox and Mann , 2008 ) for downstream data analysis . In each comparison phosphosites quantified in at least two out of three biological replicates were selected for further analysis . Perseus data processing was configured as follows; reverse hits were removed and minimum localization probability of 0 . 75 was set as criterion to accept phosphosites for further analysis . The normalized ratios and the intensities of phosphosites reported by MaxQuant were log2 and log10 transformed respectively , and were used to plot the distribution of quantified phosphosites ( Figure 2A , B ) . The significantly regulated phosphosites ( significant outliers ) were determined by two-tailed ‘Significant A’ test ( p-value ≤ 0 . 05 ) using the Perseus software ( Cox and Mann , 2008; Cox et al . , 2009 ) . The Motif-X tool ( Schwartz and Gygi , 2005 ) was used remotely from motif-x . med . harvard . edu/motif-x . html for sequence motif analysis . Sequences of upregulated or downregulated phosphorylation sites in the comparisons were independently searched against the IPI Rat Proteome as background . The default suggested setting ( occurrences = 20 and significance = 0 . 000001 ) were used for the analysis . Central character was set as 'S' and sequence length was specified as 31 . The Ingenuity Pathway Analyses software ( IPA , build version 21901358 , Qiagen , Redwood City , CA ) was used to identify significantly enriched biological functions in the synaptosomal proteome and phosphoproteome data sets . The comparison analysis function of IPA was then used to further determine which functional clusters differ significantly between the two datasets . Only clusters having a p-value < 0 . 01 were considered for the analysis . Right-tailed Fisher’s exact test was used to calculate a p‐value determining the probability that each biological function assigned to that data set is due to chance alone . Functionally redundant clusters were manually removed . The heat map was generated using Microsoft Excel . In vitro kinase assays were carried out using the ADP-Glo kinase assay kit ( Promega , Meinheim , Germany ) in three replicates . In all assays 0 . 2 µg/µl of each peptide were used . First , standard curves were obtained by making a series of ADP-ATP ( provided together with kinase assay kit ) dilutions based on the kit protocol , and the luminescence of each dilution was measured . Then , we optimized the concentration of each kinase and of the ATP concentration to achieve 100% conversion of ATP to ADP based on the standard curves , defined as 100% activity . Then the phosphorylation of each investigated phosphosite by the respective kinase was calculated and reported as percent of activity of the enzyme . For PKA ( Promega , Meinheim , Germany ) assay , Kemptide ( provided with the enzyme ) was used as positive control peptide , and 5 µM ATP and 0 . 7 U/µl enzyme were used . For CaMKII ( New England Biolabs , Frankfurt , Germany ) we first activated the enzyme by addition of 5 µM ATP , 1 . 2 µM calmodulin ( provided with the enzyme ) and 2 mM CaCl2 and 30 min incubation at 37°C . Then , for the kinase reactions we used 5 µM ATP and 12 U/µl of kinase . Autocamide ( GenScript , Piscataway , NJ ) was used as positive control peptide . For PKC ( Promega , Meinheim , Germany ) assays , Neurogranin ( Provided with enzyme ) was used as positive control peptide , and the incubation was carried out with 25 µM ATP and 5 ng/µl enzyme . Peptides were purchased from commercial supplier ( Thermo Fisher Scientific , Ulm , Germany ) . The sequences of the peptides are listed in Figure 6—source data 1 . 10 µg of synaptosomal extract was used for immunoblotting according to standard procedures ( Liu et al . , 2014 ) after separation of samples by 4–12% gradient SDS-PAGE gels . The following primary antibodies were used: synapsin1 , dynamin1-3 and α-tubulin ( Synaptic Systems , Göttingen , Germany ) , phospho S774-dynamin1 , phospho S603-synapsin1 ( Rockland , Pottstow , PA ) . All primary antibodies were used at a dilution of 1:1000 . HRP-conjugated mouse , rat or sheep secondary antibodies were used for detection of the respective primary antibodies . Immunoreactive proteins were developed with enhanced chemiluminescence reagent ( SuperSignal West Pico kit , Thermo Fisher Scientific , Rockford , IL ) , and signals were detected with an LAS-1000 imaging system ( Fujifilm , Tokyo , Japan ) . Quantification of immunoblots was performed using ImageJ , and Prism ( version 5 . 04 , GraphPad software ) was used for the statistical analysis . | The human nervous system contains more than a hundred billion neurons that are connected with each other via junctions called synapses . When an electrical impulse travelling along a neuron arrives at a synapse , it triggers bubble-like packages called synaptic vesicles within the neuron to merge with the neuron’s surface membrane . The contents of these vesicles – chemical messengers called neurotransmitters – are then released into the synapse and carry the signal to the next neuron . Complex molecular machines made from many different proteins control the release of neurotransmitters . Quite a few of these proteins are regulated by the addition of phosphate groups at specific sites . However , not all of the proteins involved in the release of neurotransmitters have been studied in detail and it is largely unclear how most of them are regulated . Now , Kohansal-Nodehi et al . have used techniques involving mass spectrometry to find out which proteins have phosphate groups added or removed in neurons that are releasing neurotransmitters . The experiments used pinched-off synapses isolated from rat brains . These structures , referred to as “synaptosomes” , lend themselves to this kind of study because they can be induced to continuously release neurotransmitters for several minutes . Kohansal-Nodehi et al . identified over 250 specific sites on proteins in the synaptosomes where phosphate groups are attached , including many on the key proteins known to operate in neurotransmitter release . Moreover , some proteins were modified at multiple sites , especially the proteins that form a scaffold to capture synaptic vesicles close to the membrane and prepare them for release . The data also revealed important clues about the enzymes that either attach or remove the phosphate groups . Together , these findings provide new insights into the regulatory networks that control many proteins at the same time . The next challenge is to sort out which of these modifications change the interactions between the proteins that control neurotransmitter release , and to understand how these changes influence the trafficking of synaptic vesicles . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Analysis of protein phosphorylation in nerve terminal reveals extensive changes in active zone proteins upon exocytosis |
Ribosome stalling during translation can potentially be harmful , and is surveyed by a conserved quality control pathway that targets the associated mRNA and nascent polypeptide chain ( NC ) . In this pathway , the ribosome-associated quality control ( RQC ) complex promotes the ubiquitylation and degradation of NCs remaining stalled in the 60S subunit . NC stalling is recognized by the Rqc2/Tae2 RQC subunit , which also stabilizes binding of the E3 ligase , Listerin/Ltn1 . Additionally , Rqc2 modifies stalled NCs with a carboxy-terminal , Ala- and Thr-containing extension—the 'CAT tail' . However , the function of CAT tails and fate of CAT tail-modified ( 'CATylated' ) NCs has remained unknown . Here we show that CATylation mediates formation of detergent-insoluble NC aggregates . CATylation and aggregation of NCs could be observed either by inactivating Ltn1 or by analyzing NCs with limited ubiquitylation potential , suggesting that inefficient targeting by Ltn1 favors the Rqc2-mediated reaction . These findings uncover a translational stalling-dependent protein aggregation mechanism , and provide evidence that proteins can become specifically marked for aggregation .
Under various circumstances , translating ribosomes can halt NC elongation and become stalled , such as upon translation of mRNA templates lacking stop codons , containing sequential suboptimal codons , or encoding homopolymeric Lys tracts ( Wang et al . , 2015; Comyn et al . , 2014; Lykke-Andersen and Bennett , 2014 ) . Ribosome stalling poses a problem , as it can both reduce the pool of translation-competent ribosomes and give rise to aberrant—and potentially toxic—nascent polypeptide chains ( NCs ) . To prevent these undesirable consequences from taking place , stalled ribosomes are rescued by factors that split the subunits , releasing the mRNA ( for degradation by the exosome ) , the 40S subunit , and the 60S subunit stalled with a nascent peptidyl-tRNA conjugate , which is then targeted by the RQC complex ( Wang et al . , 2015; Comyn et al . , 2014; Lykke-Andersen and Bennett , 2014 ) . The RQC is minimally composed of the Ltn1 ( Listerin in mammals ) , Rqc1 , and Rqc2/Tae2 ( NEMF in mammals ) subunits ( Bengtson and Joazeiro , 2010; Brandman et al . , 2012; Defenouillere et al . , 2013 ) . How the RQC functions has only begun to be understood . According to the current model , Rqc2 first recognizes the stalled 60S and facilitates binding of the Ltn1 E3 ligase , which , in turn , ubiquitylates the aberrant NC ( Lyumkis et al . , 2014; Shen et al . , 2015; Shao et al . , 2015 ) . Next , in a manner dependent on Rqc1 , the Cdc48/VCP AAA ATPase and its ubiquitin-binding cofactors are recruited to the complex and facilitate NC delivery to the proteasome for degradation ( Brandman et al . , 2012; Defenouillere et al . , 2013; Verma et al . , 2013 ) . It has been recently discovered that Rqc2 can , in addition , recruit Ala- or Thr-loaded tRNA to promote the C-terminal elongation of stalled NCs in a template- and 40S ribosomal subunit-independent manner ( Shen et al . , 2015 ) . Such 'CAT tails' have no defined sequence and are heterogeneous in length , forming a smear extending as much as 5 kDa above the unmodified reporter band in SDS-PAGE . The physiological relevance of this process has remained unclear , however , as so far it has only been reported for Ltn1- or Rqc1-deficient cells ( Shen et al . , 2015 ) . Moreover , the fate of CATylated NCs has remained unknown .
We observed that , in ltn1△ cells , in addition to the previously described increased steady-state levels and 'CATylation' of stalling reporters ( Bengtson and Joazeiro , 2010; Shen et al . , 2015 ) , a fraction of those reporters migrated very slowly in gel electrophoresis , close to the loading well ( Figure 1A; in the experiments presented , smears due to CATylation often run close to the unmodified reporter band , and can be more clearly observed with the smaller molecular weight reporters , PtnA-NS and GRR; see also below ) . The phenomenon was also observed with cells in which only Ltn1’s E3-catalytic RING domain had been deleted ( Ltn1 △R; Figure 1A , left panel ) suggesting that it is prevented by Ltn1-mediated ubiquitylation under normal conditions . 10 . 7554/eLife . 11794 . 003Figure 1 . Stalled translation can lead to the formation of nascent chain aggregates . ( A ) Top panels , diagrams of reporter constructs encoding stalling-prone nascent chains and respective controls . PolyLys-dependent stalling ( left ) : GFP-Flag-HIS3 fusion protein control ( K0 ) , its bona fide nonstop ( NS ) protein derivative , and a derivative fused to 12 lysines ( K12 ) . Endonucleolytic mRNA cleavage-dependent stalling ( middle ) : Protein A ZZ domain-Ribozyme-GFP fusion constructs . A self-cleaving ribozyme ( Rz ) within coding sequence generates a nonstop ( NS ) mRNA encoding stalled Protein A ( PtnA ) . Controls are constructs with a cleavage-defective ribozyme generating a full-length PtnA-GFP fusion ( rz ) , or with a stop codon preceding the Rz cleavage site ( STOP-Rz ) , such that nascent PtnA is not expected to become stalled in ribosomes . Horizontal lines represent the encoded polypeptides . Arg CGN codon-dependent stalling ( right ) : GFP-R12-RFP ( GRR ) , where R12 is encoded by unpreferred Arg codons . Lower panels , reporter protein expression in a wild type strain ( WT; BY4741 ) , a LTN1-deleted strain ( ltn1△ ) or a strain whose endogenous Ltn1 lacks the RING domain ( Ltn1 △R ) . Immunoblots of SDS-boiled cell extracts: anti-Flag , anti-PtnA , or anti-GFP to monitor reporter expression , and anti-Pgk1 as loading control . The migration of CATylated species is indicated . Asterisks indicate bands of unknown identity . ( B ) Stalling reporter slow-migrating species are pelleted upon high speed centrifugation . The extract of a K12 reporter-expressing ltn1△ strain was pre-cleared by centrifugation at 1000 x g for 5 min and its supernatant ( S1 ) was then subjected to 16 , 000 x g for 10 min . The resulting supernatant ( S16 ) and pellet ( P16 ) were analyzed by western blot against Flag tag ( K12 ) , Rpl3 ( a 60S ribosomal protein ) , Pgk1 ( phosphoglycerate kinase 1 , a soluble protein ) and ubiquitin ( high-molecular weight conjugates migrating above 120 kDa are shown ) . ( C ) Translational stalling is required for reporter aggregation . NS and K12 reporter protein expression in strains lacking Ltn1 and/or the translational stalling factor , Hel2 . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 00310 . 7554/eLife . 11794 . 004Figure 1—figure supplement 1 . Stalled translation can lead to the formation of nascent chain aggregates . ( A ) The slow migration of stalling reporters expressed in ltn1△ cells is not due to poly-ubiquitylation . Strains utilized were wild type ( WT ) , ltn1△ cells , and ltn1△ cells expressing the K12 reporter . Cell extracts were treated with the recombinant catalytic core of Usp2 ( Usp2cc ) , which has general deubiquitylase activity , for 1h at room temperature , as described ( Kaiser et al . , 2011 ) . The anti-Flag blot shows a lack of effect of Usp2cc on the migration of high molecular weight stalling reporter species ( compare lanes 3 and 6 ) while the anti-ubiquitin blot confirms that the enzyme was able to efficiently disassemble ubiquitin chains linked to proteins in the extract . ( B ) Loss of LTN1 is not generally associated with increased formation of protein aggregates . Expression of GFP-Huntingtin exon 1 polyglutamine reporters ( Htt-Q25 and Htt-Q72 ) in WT and ltn1△ strain extracts revealed by anti-GFP immunoblot . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 004 We analyzed several stalling reporters that have been previously described . In one set , the reporters consist of a GFP-Flag-His3 fusion followed by a stop codon ( K0 ) , lacking stop codons ( NS , for 'nonstop' ) , or followed by a 12 Lys track and stop codons ( K12 ) ( Figure 1A , left panel , and [Bengtson and Joazeiro , 2010; Ito-Harashima et al . , 2007] ) . Translation of NS and K12 reporters is believed to stall due to polyLys tract synthesis . In a second set of reporters , the Protein A ZZ domain ( PtnA ) is followed by a stop codon ( STOP-Rz ) , by a wild type self-cleaving ribozyme sequence ( NS-Rz ) , or by a mutant ribozyme ( rz ) ( Figure 1A , middle panel , and [Wilson et al . , 2007] ) . In the case of the NS-Rz reporter mRNA , translating ribosomes stall as they reach the 3’end of the cleaved mRNA with no stop codons present; on the other hand , with the mutant rz reporter , the mRNA fails to be cleaved , allowing translation to proceed through an in-frame GFP sequence without stalling . In a third set of experiments , a GFP-12 ( Arg ) -RFP fusion protein ( GRR ) was utilized ( Figure 1A , right panel , and ref [Ito-Harashima et al . , 2007] ) . In this case , stalling occurs as a result of the presence of multiple unpreferred Arg CGN codons ( Letzring et al . , 2013 ) . Despite their unrelated encoded protein sequences and distinct stalling mechanisms , we were able to observe slow-migrating species for all stalling reporters examined , but not their respective parental controls ( e . g . , K0 , STOP-Rz ) . The formation of slow-migrating reporter species thus appears to be translational stalling-dependent . We next investigated the nature of these high-molecular weight species . Slow migration was not due to Ltn1-independent poly-ubiquitylation of stalling reporters , since migration was not shifted after treatment with the deubiquitylating enzyme , Usp2 ( Figure 1—figure supplement 1A; [Kaiser et al . , 2011] ) . We reasoned that those species might instead correspond to insoluble aggregates . Consistent with this possibility , slow-migrating reporter species were efficiently sedimented by centrifugation under conditions normally used to pellet protein aggregates ( see , e . g . , [Fang et al . , 2011; Koplin et al . , 2010] ) , in contrast to a soluble protein ( Pgk1 ) or the bulk of high-molecular weight poly-ubiquitylated proteins in the extract ( Figure 1B ) . The ability to observe aggregates of stalling reporter proteins by western-blot implies that those aggregates are resistant to solubilization by boiling in 1% sodium dodecyl sulfate ( SDS ) , as samples for the experiments above were subjected to this treatment prior to gel running . Resistance to ionic detergents is characteristic of ordered fibrillar structures such as the amyloid formed by yeast prions or by expanded polyglutamine ( polyQ ) tracts ( e . g . , [Toyama and Weissman , 2011; Liebman and Chernoff , 2012] ) . To our knowledge , this is the first report of E3 dysfunction leading to formation of aggregates sharing properties with amyloid . However , in contrast to its effects on stalling reporters , deletion of LTN1 failed to affect levels or stimulate aggregation of a Huntingtin exon 1 polyQ-GFP reporter carrying either a disease-associated expansion ( Htt Q72 ) or a normal length tract ( Htt Q25; Figure 1—figure supplement 1B; [Krobitsch and Lindquist , 2000] ) . Thus , loss of Ltn1 function is not associated with the increased formation of protein aggregates in general , but rather appears to be specifically associated with stalled NC aggregation . To further verify that ribosome stalling is required for NC aggregation , we took advantage of the knowledge that the Hel2 protein is required for polybasic tract-mediated translational stalling ( Brandman et al . , 2012 ) . Thus , in a hel2△ background , ribosomes translating a 12-Lys tract in a stop codon-containing reporter ( K12 ) would be expected to translate through the tract , reach the stop codon , and terminate translation normally , releasing the NC; on the other hand , the stalling of ribosomes translating a bona fide non-stop mRNA ( e . g . , NS ) would not be expected to be prevented by HEL2 deletion ( Brandman et al . , 2012 ) . We thus asked what consequence HEL2 deletion would have on reporter aggregation . As predicted , the results in Figure 1C show that HEL2 deletion efficiently suppressed aggregation of K12—but not NS—in the ltn1△ background . The formation of NC aggregates in Ltn1-deficient cells correlated with NC modification with CAT tails . Furthermore , the low-complexity CAT tail sequences ( Shen et al . , 2015 ) are reminiscent of aggregate-forming polyAla tracts ( Albrecht and Mundlos , 2005 ) . We thus hypothesized that the aggregation of stalled NCs was mediated by CAT tails . Consistent with the hypothesis , stalled NC CATylation and aggregation were also observed in rqc1△ cells ( Figure 2A ) . The observation that RQC1 deletion phenocopied LTN1 deletion with regard to NC aggregation also implies that it is not a defect in Ltn1-mediated ubiquitylation per se—which is functional in the rqc1△ background ( Brandman et al . , 2012; Defenouillere et al . , 2013 ) —that causes aggregation ( consistent with this interpretation , treatment of wild type yeast with the proteasome inhibitor MG132 led to stalling reporter accumulation without producing aggregates; Figure 2—figure supplement 1A ) . Rather , these results suggest that stalled NCs are driven towards CATylation and aggregation as a result of a defect in a step downstream of ubiquitylation but upstream of the proteasome , such as Cdc48/VCP recruitment ( Verma et al . , 2013 ) . 10 . 7554/eLife . 11794 . 005Figure 2 . Rqc2-mediated modification of stalled nascent chains with CAT tails results in their aggregation . ( A ) NC CATylation correlates with aggregation—effects of RQC1 and RQC2 deletion . The indicated strains were transformed with the PtnA NS-Rz reporter . Reporter expression was monitored by immunoblot anti-PtnA . The migration of CATylated species is indicated . ( B ) An Rqc2 mutant defective in CAT tail synthesis fails to promote aggregation of stalled NCs . The ltn1△ rqc2△ strain expressing the GRR reporter was transformed with plasmids encoding Rqc2-Flag wild type ( WT ) or D98Y mutant . ( C ) Endogenous Rqc2 is limiting for NC CATylation and aggregation in ltn1△ cells . The ltn1△ strain expressing the GRR reporter was transformed or not with plasmid encoding Rqc2-Flag wild type ( WT ) . Reporter expression was monitored by immunoblot anti-GFP . ( D ) Fusion of a CAT tail-mimetic sequence to the C-terminus of the K0 reporter protein suffices to promote aggregation independently of stalling or Rqc2 . Top panel , diagram of constructs . Lower panel , as in 'a' . The indicated strains were transformed with plasmids encoding the parental reporter ( K0 , as described in 1a ) or its derivatives fused to a C-terminal tail of 20 Ala , 20 Thr , or 10 Ala-Thr repeats , as indicated . ( E ) Punctae formed by stalling reporters in intact cells correlate with aggregates observed in WCE . Fluorescence microscopy imaging of indicated strains expressing the GRR reporter . GFP-positive punctae can be observed in the ltn1△ strain . ( F ) CAT tail-dependent incorporation of the GRR stalling reporter into punctae . Left , The ltn1△ rqc2△ strain was transformed with plasmids encoding Rqc2-Flag wild type ( WT ) or D98Y mutant as in panel 'B' and examined by fluorescence microscopy . Three different distribution patterns of the GFP signal that are representative for each strain are shown . Arrows point to selected punctae . Scale bar , 2 μm . Right , Quantification of cells harboring GFP-positive inclusions in the ltn1△ rqc2△ strains expressing Rqc2 WT or D98Y mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 00510 . 7554/eLife . 11794 . 006Figure 2—figure supplement 1 . Rqc2-mediated modification of stalled nascent chains with CAT tails results in their aggregation . ( A ) Accumulation of stalling reporters by proteasome inhibition is not sufficient to result in aggregation . Immunoblot analysis of K0 and NS reporter expression in wild type or ltn1△ strains , after treatment ( + ) or not ( - ) with the proteasome inhibitor MG132 for 2 hr . ( B ) The Rqc2 D98Y mutant is competent to support Ltn1 function . Stalling reporter expression in wild type or rqc2△ strains transformed with empty vector , wild type Rqc2-Flag , or Rqc2-Flag D98Y . ( C ) CATylation is required for aggregation of the NS reporter . The ltn1△ rqc2△ strain expressing the NS reporter was transformed with plasmids encoding Rqc2-Flag wild type ( WT ) or D98Y mutant . Expression of NS monomers , NS aggregates , and Rqc2-Flag revealed by anti-Flag immunoblot . Panels with different exposure times shown . ( D ) Loss of RQC2 is not generally associated with the failure to form protein aggregates . Expression of GFP-Huntingtin exon 1 polyglutamine reporters ( Htt-Q25 and Htt-Q103 ) in the indicated strains revealed by anti-GFP immunoblot . ( E ) Stalling reporter aggregates do not form post-lysis . K12 reporter-expressing strains ( labeled in red ) were mixed 1:1 with a second , untransformed strain ( labeled in blue ) before lysis , and extracts were analyzed for aggregate formation by anti-Flag immunoblot . The presence of the ltn1△ strain constituents during lysis was not sufficient to promote aggregation of the K12 reporter expressed in a ltn1△ rqc2△ strain ( lane 2 ) . Conversely , the presence of constituents of the ltn1△ rqc2△ strain did not interfere with aggregates formed by K12 reporter-expressing ltn1△ strain ( compare lane 4 to lane 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 006 Further correlation between CATylation and aggregation was obtained by inspecting the PtnA-Rz reporter in a Ltn1-deficient strain also lacking Rqc2—the results in Figure 2A show that , in the absence of CAT tails , NC aggregates also failed to form . We further examined the CAT tail requirement for NC aggregation by using ltn1△ rqc2△ strains expressing a stalling reporter and transformed with plasmids encoding either wild type Rqc2 , or Rqc2 carrying a mutation in the highly conserved Asp98 residue . Asp98 is required for CAT tail synthesis but not for ribosome binding—in fact , the Rqc2 D98Y mutant is expressed normally and is fully competent to support Ltn1 function in rqc2△ cells ( Figure 2—figure supplement 1B and [Shen et al . , 2015] ) . As predicted by the hypothesis , while overexpression of wild type Rqc2 led to the quantitative conversion of the stalling reporter into aggregated forms , the Rqc2 D98Y mutant was unable to promote stalling reporter aggregation ( Figure 2B ) . Overexpression of wild type Rqc2 also led to more quantitative conversion of monomeric to CATylated and aggregated forms of the stalling reporter in ltn1△ cells , suggesting that endogenous Rqc2 is limiting ( Figure 2C ) . Moreover , differently from what we had observed for Hel2 requirement ( Figure 1C ) , the CAT tail synthesis requirement for aggregation was evident with all stalling reporters examined , including NS ( Figure 2—figure supplement 1C ) . Yet , this requirement appeared to be specific to stalling reporters , as RQC2 deletion did not affect polyQ reporter aggregation ( Figure 2—figure supplement 1D ) . The hypothesis also predicted that hard-coding a CAT tail on a stop codon-containing reporter construct might bypass the requirements for both stalling and Rqc2 for protein aggregation . To test this possibility , we generated a construct in which a tract of 10 Ala-Thr repeats [ ( AT ) 10] was fused to the C-terminus of the control reporter K0 with the intent to mimic a CAT tail , followed by stop codons . The ( AT ) 10 reporter has no stalling sequence , so its encoded protein is not expected to be a target of the RQC . As shown in Figure 2D , ( AT ) 10 , but not homopolymeric constructs with 20 Ala ( A20 ) or 20 Thr ( T20 ) , was indeed able to form aggregates . Given that homopolymeric Ala tracts have been previously implicated in amyloid formation ( Albrecht and Mundlos , 2005 ) , we presume that polyAla sequences may need to be longer in order to form aggregates under the conditions utilized here . Thus , the combination of Ala and Thr as found in CAT tails appears to be more prone to aggregate , which can be observed with a sequence as short as 20 amino acids long . These results suggest that , in addition to being required , CAT tails can also be sufficient for aggregate formation . Arguing against the possibility of reporter aggregation being a post-lysis artifact , mixing ltn1△ cells with K12-expressing ltn1△ rqc2△ cells immediately prior to cell lysis did not support aggregate formation ( Figure 2—figure supplement 1E ) . In order to obtain further evidence for the formation of NC aggregates in intact cells , we examined the distribution of the GRR stalling reporter under fluorescence microscopy ( Figure 2E ) . In the WT strain transformed with GRR , the low GFP signal was evenly distributed throughout the cytoplasm , with 1–3 local accumulations of signal being observed; in contrast , LTN1 deletion led to the appearance of numerous GFP punctae , characterized by foci with intense brightness over a low , diffuse cytosolic fluorescence background , in a third of cells [i . e . , 15 out of 45 cells showing this phenotype; we note that aggregate formation in the ltn1△ strain is limited by endogenous Rqc2 levels ( see Figure 2C above ) ] . GFP punctae formation in ltn1△ cells depended on Rqc2-mediated CATylation , as evidenced by the finding that the phenomenon could be suppressed by RQC2 deletion ( Figure 2E ) , and subsequently rescued by overexpression of Rqc2 wild type but not D98Y ( Figure 2F ) . Thus , the appearance of GFP punctae in living cells correlated with the presence of protein aggregates in cell extracts . The results presented so far indicate that NCs can be assembled into aggregates through a process triggered by ribosome stalling and requiring CAT tail synthesis by Rqc2 . This implies that proteins can become 'tagged' for aggregate assembly , and that this happens while nascent chains are still associated with the 60S subunit . Because molecular chaperones are implicated in handling misfolded and aggregated proteins ( e . g . , [Parsell et al . , 1994; Mogk et al . , 2015; Nillegoda et al . , 2015] ) we next examined the association of candidate chaperones with stalling reporters . Consistent with previous reports indicating that yeast amyloid aggregates are typically bound to the Hsp40/J protein , Sis1 ( e . g . , [Aron et al . , 2007; Park et al . , 2013; Yang et al . , 2013] ) , the binding of Sis1 to stalling reporters was readily apparent in extracts of a ltn1△ strain ( Figure 3A and Figure 3—figure supplement 1A ) . Other chaperones , such as Hsp70 Ssa1 and the Hsp40/J-protein Ydj1 , also co-immunoprecipitated ( co-IP’ed ) with stalling reporters above background level but the differences were less marked and appeared more variable . Consistent with these observations , analysis of chaperones co-IP’ed with the K12 stalling reporter expressed in a ltn1△ strain by mass spectrometry uncovered Sis1 among the most abundant hits , and with the apparent highest signal-to-noise ratio [Supplementary file 1 ( Table SI ) ] . In contrast to the conspicuous co-IP of Sis1 with stalling reporters in ltn1△ cells , markedly less Sis1 was pulled down by anti-Flag antibody-conjugated beads from a ltn1△ strain expressing no Flag-tagged stalling reporters , from a ltn1△ strain expressing the K0 parental reporter , or from rqc2△ or ltn1△ rqc2△ strains ( Figure 3A ) . Thus , the Sis1 co-IP depended on stalling and on Rqc2 , suggesting it binds stalled NCs via the CAT tails and/or aggregates . 10 . 7554/eLife . 11794 . 007Figure 3 . Sis1 association reveals endogenous stalled nascent chain aggregates . ( A ) Stalling reporters co-IP with Sis1 in an Rqc2-dependent manner . Whole cell extracts ( WCE ) of the indicated strains were Flag IP’ed ( to pull down K0 or K12 reporters ) , followed by immunoblotting as indicated to the left of the panels . ( B ) Sis1 associates tightly with Rqc2-dependent aggregates formed by endogenous proteins in cells deficient for Ltn1 or Rqc1 . WCE of the indicated strains were immunoblotted against Sis1 , Ssa1 , and Ydj1 , as indicated . The ~47 kDa band in the Sis1 blot ( asterisk ) is nonspecific . ( C ) The formation of slow-migrating Sis1 species is dependent on Rqc2’s ability to synthesize CAT tails . WCE of the ltn1△ rqc2△ strain expressing Rqc2-Flag wild type or D98Y mutant were analyzed by immunoblotting . ( D ) Sis1 depletion increases NC aggregation . ltn1Δ tetO7-Sis1 cells expressing the K12 stalling reporter were treated or not with doxycycline ( DOX ) . WCE were analyzed by immunoblot against Sis1 ( left panel ) or Flag ( right panel; for K12 detection ) . Asterisk , cross-reacting band . ( E ) Sis1 depletion increases GFP punctae formation in ltn1Δ cells . ltn1Δ WT-Sis1 or ltn1Δ tetO7-Sis1 cells expressing the GRR stalling reporter were grown for 24 hr in the presence ( + ) or absence ( - ) of doxycycline ( DOX ) . Top , Three representative images are presented for each strain and treatment condition . Scale bar , 2 μm . Arrows point to selected punctae . Bottom , Quantification of cells harboring GFP punctae is represented by red bars; among those , the fraction of cells with 1 or 2 punctae is represented in dark gray , and the fraction of cells with 3 or more punctae , in light gray . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 00710 . 7554/eLife . 11794 . 008Figure 3—figure supplement 1 . Sis1 association reveals endogenous stalled nascent chain aggregates . ( A ) The stalling reporters NS and K12 , but not the parental K0 , co-IP with Sis1 . Whole cell extracts ( WCE ) of ltn1Δ strains expressing the indicated reporters were subjected to Flag IP , and analyzed by anti-Sis1 ( or Flag control ) immunoblot . ( B ) The migration of the high molecular weight Sis1 species present in ltn1Δ cells is not affected by treatment with a deubiquitylating enzyme ( see Figure 1—figure supplement 1A ) . Extracts from the indicated strains were treated with Usp2cc and analyzed by anti-Sis1 immunoblot . ( C ) Sis1 aggregates co-IP with stalling reporters . Extracts of Ltn1-deficient cells , untransformed or expressing the K12 reporter , were used for Flag IP ( to pull down K12 ) and immunoblotted against Sis1 or Flag tag , as indicated . Asterisk , nonspecific band . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 00810 . 7554/eLife . 11794 . 009Figure 3—figure supplement 2 . tetO7 promoter-dependent Sis1 depletion . ltn1Δ tetO7-SIS1 and ltn1Δ WT-SIS1 cells were treated with doxycycline as indicated , and analyzed by western blot using antibodies against Sis1 . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 009 Strikingly , a fraction of Sis1 itself exhibited slow-migrating species that were resistant to boiling in 1% SDS in ltn1△ or rqc1△ cells , but not in the ltn1△ rqc2△ strain ( Figure 3B ) . The specificity of this phenomenon is underscored by the failure to observe similar slow migrating species for Ssa1 or Ydj1 ( Figure 3B ) . Like stalled NC aggregates , those Sis1 species were unaffected by treatment with Usp2cc ( Figure 3—figure supplement 1B ) . Moreover , slow-migrating Sis1 species could be pulled down along with an IP’ed stalling reporter ( K12 Flag IP; Figure 3—figure supplement 1C ) , and their formation was dependent on Rqc2’s ability to synthesize CAT tails ( Figure 3C ) . Together , these results suggest that Sis1 tightly associates with stalled NC aggregates . Importantly , given that the formation of these Sis1 species was independent of ectopic expression of stalling reporters , these findings also provide evidence for aggregate formation by endogenous stalled NCs , and suggest that stalling reporter aggregation is not an artifact caused by their overexpression . The above observations raised the question of whether Sis1 plays a role in NC aggregation . Given that SIS1 is an essential gene , to shed light onto this issue , stalling reporter aggregation was examined in a ltn1△ strain in which SIS1 expression is under the tetO7 promoter , so it can be turned off by treatment with doxycycline ( Aron et al . , 2007 ) . The results in Figure 3D show that the levels of Sis1 in both free and aggregated forms were indeed reduced in response to doxycycline ( left panel ) , and that this was accompanied by a marked increase in steady-state levels of NC aggregates ( right panel ) . We also examined the effect of doxycycline in intact cells expressing the GRR stalling reporter ( Figure 3E ) . Consistent with the immunoblot data in Figure 3D , the imaging results show that both the fraction of cells with GFP punctae and the number of punctae per cell were increased in ltn1△ Tet-Sis1 cells ( Figure 3E ) . In contrast , such effects were modest at best in ltn1△ cells in which Sis1 expression is under its endogenous promoter and unaffected by doxycycline treatment ( Figure 3E and Figure 3—figure supplement 2 ) . NC CATylation has so far only been observed with mutant yeast strains harboring inactivating mutations in Ltn1 or Rqc1 , raising the issue of physiological relevance ( Shen et al . , 2015 ) . This observation suggested it might also be possible to observe Rqc2-dependent effects in wild type cells by utilizing stalling reporters expected to be less easily targeted by Ltn1—e . g . , with fewer potential ubiquitylation sites . To test this possibility , we generated reporters with all lysine residues mutated to arginine ( 'K-less' ) and fused or not to the R12 stalling sequence . One set of reporters was based on HA-tagged GFP K-less fused or not to 12 suboptimal Arg codons that cause ribosomal stalling ( K-less R12 ) . Cells transformed with GFP K-less or GFP K-less R12 constructs were analyzed for HA tag expression ( Figure 4A ) . Remarkably , CATylation of the GFP K-less R12 reporter was readily evident in the wild type strain , as indicated by the presence of a smear immediately above the monomeric reporter band ( lane 3 ) , which was dependent both on Rqc2 ( compare lanes 3 and 5 ) and on the R12 stalling sequence ( compare lanes 2 and 3 ) . 10 . 7554/eLife . 11794 . 010Figure 4 . Evidence for stalled nascent chain modification with CAT tails and aggregation in wild type cells . ( A ) Stalling Lys-less reporter modification with CAT tails in wild type yeast . All constructs were HA-tagged . Expression of GFP , GFP-R12 , GFP K-less ( 'K-less' ) , or GFP K-less R12 ( 'K-less-R12' ) reporter proteins in the indicated strains , revealed by anti-HA immunoblot . 'R12' is the stalling signal , consisting of 12 suboptimal Arg CGN codons . GFP-R12 expression in ltn1△ cells is used as a control for aggregate formation . Lower panel , shorter exposure to reveal relative steady-state levels of monomeric reporter species . ( B ) Stalling PtnA-Rz reporter CATylation and aggregation in wild type yeast . All constructs were HA-tagged . Expression of PtnA-STOP-Rz , PtnA-Rz , PtnA-STOP-Rz K-less , and PtnA-Rz K-less reporter proteins in the wild type strain , revealed by anti-HA immunoblot . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 01010 . 7554/eLife . 11794 . 011Figure 4—figure supplement 1 . Stalling NC reporter aggregation in wild type yeast . HA-tagged GFP , GFP-R12 , GFP K-less ( 'K-less' ) , or GFP K-less R12 ( 'K-less R12' ) reporters expressed in the MG132-treated wild type strain ( or MG132 treated rqc2△ strain as a control , lane 5 ) were concentrated by anti-HA IP and analyzed by anti-HA immunoblot . Lower panel , the monomeric forms are shown in a shorter exposure to reveal relative steady-state levels of monomeric reporter species , as well as CAT tails in lane 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 11794 . 011 However , under the above conditions , the formation of aggregates with the GFP K-less R12 reporter was not observed ( Figure 4A ) . We presume this could be due to a detection issue , since GFP K-less was expressed at lower steady-state levels compared to GFP , perhaps due to imperfect folding ( Sokalingam et al . , 2012 ) [see shorter exposure; indeed , Rqc2- and stalling-dependent K-less reporter aggregates became evident in wild type cells treated with proteasome inhibitor and after concentrating the samples by affinity purification ( Figure 4—figure supplement 1 ) ] . The formation of stalling reporter aggregates in wild type cells was more conspicuous when analyzing a different set of reporters , based on Protein A ZZ domain-Rz construct ( described in Figure 1A ) . Wild type cells transformed with stop codon-containing controls or Rz-dependent stalling constructs were analyzed for PtnA expression ( Figure 4B ) . Regardless of whether or not they contained lysines , PtnA-STOP-Rz reporters exhibited a similar expression pattern , consisting of a major band of the expected size of the PtnA ZZ domain ( lanes 1 and 3 ) . As observed in Figure 1A , the PtnA-NS-Rz construct additionally encoded a protein corresponding in size to the product of full-length , uncleaved PtnA- ( Rz ) -GFP ( lane 2 ) . In contrast , in wild type cells expressing the PtnA-NS-Rz K-less construct , a band corresponding to the PtnA ZZ domain was not evident; instead , smears corresponding to its CATylated and aggregated forms were conspicuous ( lane 4; compare to lane 2 ) . We interpret this result as indicating that stalled NC encoded by the truncated PtnA-NS-Rz mRNA are normally targeted for degradation ( compare lanes 1 and 2 ) , but targeted for CATylation and aggregation if lysine ubiquitylation sites are not readily available . Furthermore , we conclude that Rqc2-dependent CATylation and aggregation of stalled NCs can both be observed in wild type cells with a functional RQC complex .
Stalled NC metabolism has recently emerged as a paradigm for the understanding of co-translational quality control , and includes mechanisms for handling aberrant NCs along with their encoding mRNA ( Wang et al . , 2015; Comyn et al . , 2014; Lykke-Andersen and Bennett , 2014 ) . In the context of the RQC complex , while Ltn1 mediates stalled NC ubiquitylation , Rqc2 mediates their CATylation ( in addition to playing a non-essential role in supporting Ltn1 function ) . Rqc2 residues implicated in CAT tail synthesis are conserved in evolution , suggesting an important function , but the fate of CAT tail-modified proteins had remained unknown . CATylation was not essential for Ltn1-mediated NC degradation ( Figure 2—figure supplement 1B ) . Rather , here we report that CATylation promoted formation of NC aggregates—yet another process operating in stalled NC quality control . To our knowledge , this is the first demonstration that proteins can become specifically marked for aggregation . Although physical characterization of NC aggregates remains to be carried out , evidence presented here indicate shared features with amyloid , including their dependency on a low-complexity aminoacid sequence bearing similarity to polyAla tracts , detergent insolubility , and binding to the J protein , Sis1 . With regard to the latter , accumulating evidence suggests that Sis1 may function as an amyloid-recognition factor that enables Ssa- and Hsp104-mediated disassembly or random fragmentation ( Harris et al . , 2014 ) . This proposed function provides a plausible explanation for our observation that Sis1 depletion led to increased NC aggregate levels , although it remains unclear whether and how regulation of presumably antagonistic Rqc2 and Sis1 activities controls steady-state levels of NC aggregates . NC aggregation could have different functions , such as in sequestering aberrant NCs into forms less likely to interfere with cellular activities , or in mediating NC elimination via vacuolar degradation . Furthermore , NC aggregate formation correlates with the CATylation-dependent activation of Heat Shock Factor 1 ( Hsf1 ) signaling observed in Ltn1-deficient cells ( Shen et al . , 2015 ) . As protein aggregation has been previously shown to elicit Hsf1 activation ( Zou et al . , 1998 ) , another conceivable role for NC aggregation is therefore in inducing translational stress signaling . On the other hand , excessive aggregate production certainly has the potential to cause toxicity . With these possibilities in mind , it will be important for future studies to investigate the role of NC aggregation in the neurodegenerative phenotype of Listerin-mutant mice ( Chu et al . , 2009 ) .
Rabbit polyclonal antibodies were: anti-Protein A ( Sigma , St . Louis , MI ) , anti-ubiquitin ( Dako , Carpinteria , CA ) , anti-Sis1 ( Yan and Craig , 1999 ) , anti-Ydj1 ( Yan and Craig , 1999 ) , and anti-Ssa1 ( Lopez-Buesa et al . , 1998 ) . Mouse monoclonal antibodies used were anti-Flag tag ( M2; Sigma ) , anti-HA tag ( 12CA5; Roche , Germany ) , anti-Rpl3 ( a gift of J . Warner ) , anti-GFP ( Roche ) and anti-Pgk1 ( Invitrogen , Carlsbad , CA] ) . Secondary antibody was HRP-conjugated ( Molecular Probes , Eugene , OR ) . Doxycycline hydrochloride was from Fischer Scientific ( Waltham , MA ) . MG132 was from Cayman Chemical ( Ann Arbor , MI ) . The recombinant catalytic core of Usp2cc was a generous gift of R . Kopito ( Kaiser et al . , 2011 ) . All strains used in this work are isogenic to BY4741 ( MATa; his3△1; leu2△0; met15△0; ura3△0 ) or BY4742 ( MATa; his3△1; leu2△0; lys2△0; ura3△0 ) , except for the experiments shown in Figure 1—figure supplement 1B , in which the DS10 strain was used ( MATa; trp1△; lys1; lys2; ura3-52; leu2-3 , 112; his3-11 , 15 ) , and Figures 3D and E , in which the W303 strain was used ( MATa; leu2-3 , 112; trp1-1; can1-100; ura3-1; ade2-1; his3-11 , 15 ) . Derivative strains are shown in Supplementary file 2 ( Table SI ) . Mutant strains carrying single gene deletions were obtained commercially ( Thermo Scientific , Waltham , MA ) . Additional deletions were performed by using cassettes designed to replace genes of interest with selection markers via homologous recombination . Primers used follow: Cells were grown at 30° in SD-media . For Sis1 depletion Sis1-Tet Off strains were grown in SD-Trp media overnight , diluted in SD-Trp containing 10 µg/ml doxycycline to 0 . 05 OD600 and grown for 12 hr . Cells were re-diluted in fresh SD-Trp containing doxycycline and incubated for further 12 hr to complete Sis1 deletion . It was necessary to keep the doxycycline incubation time as short as possible , as extended Sis1 depletion caused reduction in cellular growth and viability . The Protein A constructs were a gift of A . van Hoof ( UT-Houston ) ( Wilson et al . , 2007 ) . The Htt-polyQ constructs were gifts of S . Lindquist ( Whitehead Institute ) ( Alberti et al . , 2009 ) . The GFP-R12-RFP ( GRR ) construct was a gift of O . Brandman ( Stanford Univ . ) ( Brandman et al . , 2012 ) . The K0 , NS , and K12 reporters have been described ( Bengtson and Joazeiro , 2010; Ito-Harashima et al . , 2007 ) . Constructs HA-GFP-stop-R12 ( 'GFP' ) and the stalling derivative , HA-GFP-R12-stop ( 'GFP-R12' ) were made on pRS316 ( URA3 marker; CEN ) with the GPD promoter . These constructs were utilized as the basis for the GFP K-less derivatives: HA-GFP K-less-stop-R12 ( 'K-less' ) and HA-GFP K-less-R12-stop ( 'K-less-R12' ) by replacing Lys codons in the GFP coding sequence with the preferred Arg AGA or AGG codons through gene synthesis ( IDT; sequence below ) . The R12 stalling sequence utilized unpreferred CGN Arg codons: CGG CGA CGA CGG CGC CGC CGG CGA CGA CGG CGC CGC . For Protein A K-less constructs , HA-tagged Protein A ZZ domain ( wild type or K-less; stop codon-containing or non-stop ) were generated by gene synthesis ( IDT; sequence below ) in the same way as the HA-GFP-K-less constructs above , and used to replace the homologous sequence upstream of the hammerhead ribozyme sequence of the PtnA-Rz construct gifted by A . van Hoof ( UT-Houston ) . MG132 treatment of yeast cells was as described ( Liu et al . , 2007 ) . Total soluble extracts were prepared as described ( Bengtson and Joazeiro , 2010 ) . Protein quantitation was performed by the BCA method . 7 . 5–30 μg of protein extract were resuspended in 1% SDS , 0 . 005% bromophenol blue , 5% glycerol , 50 mM dithiothreitol , 50 mM Tris-Cl ( pH 6 . 8 ) and incubated at 100°C for 5 min before fractionation by gel electrophoresis ( 4-20% Tris-Glycine gel , Invitrogen ) and immunoblotting . 30 μg of ltn1△ cell extracts expressing K12 were treated with the recombinant catalytic core of Usp2 ( Usp2cc; 5 μM ) for 1h at room temperature , as described ( Kaiser et al . , 2011 ) . The reaction was stopped by adding DTT-containing SDS-PAGE loading buffer and boiling for 3 min . Aggregate fractionation was performed as described ( Fang et al . , 2011; Koplin et al . , 2010 ) with minor modifications . Briefly , cells were grown to late logarithmic phase ( A600 = 0 . 8 ) and harvested by centrifugation at 2000 x g for 2 min . The cell pellet was resuspended in 750 μL cold non-denaturing lysis buffer [1% Triton , 140 mM NaCl , 1 . 5 mM MgCl2 , EDTA-free protease inhibitor cocktail ( Roche ) , 20 U/mL RNase inhibitor ( RNaseOUT , Invitrogen ) and 10 mM Tris-Cl ( pH 7 . 4 ) ] and combined with 750 μL of 0 . 5-mm diameter glass beads . Suspensions were homogenized ( three times for 45 s in a FastPrep FP120 Savant homogenizer ) and lysates were transferred to clean tubes and centrifuged at 1000 x g , for 10 min at 4°C . Pre-cleared cell extract supernatants ( S1 ) were further fractionated at 16 , 000 x g for another 10 min to pellet aggregates . Cells were grown to late logarithmic phase ( A600 = 0 . 8 ) and harvested by centrifugation at 2000 x g for 2 min . The cell pellet was resuspended in 750 μL of cold non-denaturing lysis buffer [0 . 2% NP40 , 140 mM NaCl , 1 . 5 mM MgCl2 , 1x EDTA-free protease inhibitor cocktail ( Roche ) , 20 U/mL of RNase inhibitor ( RNaseOUT; Invitrogen ) and 10 mM Tris-Cl ( pH 7 . 4 ) ] and combined with 750 μL of 0 . 5-mm diameter glass beads . Suspensions were homogenized ( three times for 45 s in a FastPrep FP120 Savant homogenizer ) and lysates were transferred to clean tubes and centrifuged at 2800 x g , for 10 min at 4°C . 1 mg of soluble extract was incubated with 1 μg of anti-Flag antibody overnight at 4°C . Next , 600 μg of washed protein G magnetic beads ( Life Technologies ) were added and incubated for 2 hr at 4°C . Beads were pelleted and washed 5 times with cold lysis buffer before proteins were eluted by boiling in sample buffer . For the experiment presented in Figure 4—figure supplement 1 , Cells were grown in SD media lacking uracyl , supplemented with 0 . 1% proline and 0 . 003% SDS until OD600 = 0 . 8 , followed by treatment with MG132 ( 75 μM ) or DMSO for 2 hr . Cells were lysed with 2% SDS lysis buffer and extracts were boiled for 3 min . Samples were diluted 20-fold in 0 . 5% triton lysis buffer and IP’ed with HA-agarose ( Sigma ) for 3 hr at 4C before running on SDS-PAGE . For image acquisition and analysis , cells were grown as indicated and harvested by centrifugation and resuspended in PBS . Widefield microscopy was performed with an Olympus xcellence IX81microscope system using a 100x/1 . 45 NA Plan-Apochromat oil objective lens ( Olympus , Japan ) and a single band GFP filter set ( AHF , Germany ) . As fluorescence light source the illumination system MT20 ( Olympus ) with a 150 W Xe arc burner was used . Z-stacks of images were recorded with slice distances of 200 nm and displayed as maximum intensity projections . Deconvolution of widefield images from Z-stacks was performed by using the Wiener filter of the Olympus xcellence software . Image processing for final figure preparation was performed with ImageJ ( Schneider et al . , 2012 ) . For quantification of phenotypes , several areas with cells were randomly acquired and 45 or more cells were used for the analysis . Exponentially growing cells were collected , washed in cold water and lysed using glass beads in 10 mM Tris-HCl pH 7 . 4 , 140 mM NaCl , 1 . 5 mM MgCl2 and 0 . 2% NP40 . 40 mg of total protein were used for IP with 40 μg Flag antibody and Protein A Dynabeads for 6 hr at 4°C . After 4 washes in lysis buffer , proteins were eluted in 200 μl of 300 μg/ml 3X Flag peptide for 30 min , at room temperature . Eluted samples were TCA-precipitated , resuspended in 8 M urea , and treated with ProteasMAX ( Promega , Madison , WI ) per the manufacturer’s instruction . Subsequently , samples were reduced by 20 min incubation with 5 mM TCEP ( tris ( 2 carboxyethyl ) phosphine ) at room temperature , alkylated in the dark by treatment with 10mM Iodoacetamide for 20 min , and quenched with excess TCEP . Proteins were digested overnight at 37 degrees with Sequencing Grade Modified Trypsin ( Promega ) and the reaction was stopped with formic acid . The protein digest was subjected to cation-exchange microcapillary chromatography , and elutates from the column were electrosprayed directly into an LTQ mass spectrometer ( ThermoFinnigan , Palo Alto , CA ) . A cycle of one full-scan mass spectrum ( 400–2000 m/z ) followed by 7 data-dependent MS/MS spectra at a 35% normalized collision energy was repeated continuously throughout each step of the multidimensional separation . Application of mass spectrometer scan functions and HPLC solvent gradients were controlled by the Xcalibur datasystem . Protein identification and quantification analysis were done with Integrated Proteomics Pipeline ( IP2 , Integrated Proteomics Applications , Inc . San Diego , CA ) . Tandem mass spectra were searched against the Saccharomyces Genome Database ( SGD ) protein database ( http://www . yeastgenome . org/download-data/sequence , released on 01-05-2010 ) . LTQ data was searched with 3000 . 0 milli-amu precursor tolerance and the fragment ions were restricted to a 600 . 0 ppm tolerance . The ProLuCID search results were assembled and filtered using the DTASelect program ( version 2 . 0 ) ( Cociorva , 2007; Tabb et al . , 2002 ) with false discovery rate ( FDR ) of 0 . 15 , under such filtering conditions , the estimated false discovery rate was less than 5 . 4% at the protein level in each analysis . | Cells use molecular machines called ribosomes to build proteins by connecting amino acids – the building blocks of proteins – together in a particular sequence . The chain of amino acids gradually lengthens as the protein forms , yet remains attached to the ribosome until the protein is complete . While this process is underway , cells can check that a newly forming chain is not abnormal or damaged . If it is , a cell then essentially ‘decides’ on whether to correct or eliminate it . Such protein quality control processes are important for ensuring the health and fitness of cells and organisms . Recently , a new protein quality control mechanism was discovered that senses when a ribosome becomes jammed as it produces a new protein . This mechanism recycles the ribosome so it can make more new proteins . It also disposes of the stalled protein using a cell complex , called the ribosome-associated quality control complex , which is found in all eukaryotic organisms including yeast and humans . This protein complex consists of three subunits; one of which , called Rcq2 , tags ribosome-stalled proteins with a “tail” that contains the amino acids alanine and threonine . However , the purpose of this tag was not clear . Yonashiro , Tahara et al . now show that the tagging of ribosome-stalled proteins by Rqc2 in yeast cells induces the tagged proteins to clump together . This clumping probably prevents these proteins from inadvertently interfering with other molecules or processes within the cell . The formation of these clumps also correlates with the activation of a stress response in the cell , indicating that these clumps create a signal that prompts the cell to protect itself in response to the accumulation of more abnormal proteins . Mutations in one subunit of the ribosome-associated quality control complex in mice cause a condition that resembles a neurological disease in humans , called amyotrophic lateral sclerosis or ALS for short . A future challenge is therefore to understand how much Rqc2-mediated tagging and clumping of ribosome-stalled protein has a role in this and other neurodegenerative diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2016 | The Rqc2/Tae2 subunit of the ribosome-associated quality control (RQC) complex marks ribosome-stalled nascent polypeptide chains for aggregation |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.